Skip to main content

Full text of "Electrical Engineering Papers"

See other formats


LI Bit All Y 


i.. Williams 

Engineering Papers 

By Benjamin Q. Lamme 

This volume contains a collection of the author's 

more important engineering papers presented 

before various technical societies and 

published in engineering journals 

and elsewhere from 

time to time 


Published by 

Westinghouse Electric & Mfg. Co. 
East Pittsburgh, Pa. 

Permission to reprint these papers has been granted 
by the owners of the copyrights and individual 
credit is given in the foreword of each papei 


The papcis of Beniamm G. Lamme have always interested American 
engineers Distributed in many publications, some quite macessible to 
readers, it is indeed a fortunate circumstance which now makes available a 
collection of these papers for engineers, prcfessors, instructors, and students 
and all those interested in, and able to understand, the progress of electrical 

Besides his achievements in the art of engineering, Mr. Lamme has been 
gifted with the faculty for clear expression and explanation, which is one of the 
rarest to be found in the engineering profession The collection begins with 
his early paper on the Polyphase Induction Motor, which, in its time, was a 
pnmer of the characteristics and operation of such motois in the hands of the 
numerous users of these machines. Then follows a period in which he prepared 
few papers, but which was one of great personal activity Then comes his 
epoch-making paper, in 1902, before the Ameiican Institute of Electrical 
Engineers in New York on the Single-Phase System of the Washington 
Baltimore and Annapolis Railway. Up to this time, the development of the 
electric railway systems, as a whole, was at a point of complete stagnation, 
in the utilization of 600 volts direct current, and this paper represented the 
first great and successful attempt to break away from established practice 
toward materially higher trolley voltages. Its advent gave an impulse to 
the entire subject of the electrification of railroads greater than any other it 
had ever received, leading to the complete abandonment of old and apparently 
well established standards, as well as to later attempts to meet the new con- 
ditions with higher direct-current voltages. 

In 1904, there are two papers, one on a 10,000 Cycle Alternator, and 
another on the Synchronous Motor for Regulation of Power Factor. In the 
same year, he contnbuted a discussion to the subject of Single-Phase Motors 
which ranks as one of the clearest and most suggestive descriptions of this type 
of motor, which owes to him its development and use. 

Recognizing the importance of closer relations with the American Insti- 
tute of Electrical Engineers, we find him contributing, from time to time, 
papers on Commutation, on the Homopolar Dynamo, on Rotary Converters, 
on Turbo Generators, on Losses in Electrical Machinery, and on Engineering 
Education, etc. It is safe to say that these papers will be read in their present 
form by many who enjoyed them when they came out originally, and their 
contents will perhaps be more appreciated today than at the time when they 
were written. 

To all those who have followed the development of electrical engineering 
in America during the past thirty years, and to all those who would like to 
know the histoncal development of electrical apparatus, the senes of papers 
which appeared in the Electric Journal, on the History of the Railway Motor, 
of the Direct-Current Generator and of the Alternating-Current Generator. 

and the History of the Frequencies, now collected for the first time, will form 
most interesting reading Here Mr, Lamme had an opportunity to recount 
the work of himself and of his associates, adding to it the clarity and lucidity 
which have always marked his style. 

I think those who have known Mr. Lamme 's interest in education and in 
his instruction of young engineers will be glad to find reprinted in this volume 
two of his contributions on the subject of engineering education. The whole- 
some, sound sense which permeates these papers cannot fail to appeal tc all, 
and to impress the reader with the sound judgment of their author. 

Although these papers represent a work of thirty years, during which 
time Mr. Lamme has been continuously associated with the great company 
which bears the name of Mr. Westinghouse, yet I believe they do not com- 
plete his whole life work Those of us who have had the good fortune to have 
known him for a score of years, or more, well know that many contributions 
will yet be made by him to the art and science of electrical engineering The 
publication by the Westinghouse Company of this collection of Mr. Lammc's 
engineering papers on the anniversary of his first connection with this com- 
pany, thirty years ago, represents a most dignified appreciation of his services 
to the entire engineering profession. 

Boston Mass, 
April 2, 1919. 





PRESSURE. . .. . .. . S3 




CURRENTS . . . . . .... Ill 






















ERICA 591 






FOREWORD This paper was prepared in the early part of 1897, 
or over twenty-two years ago. It was presented at the twentieth 
convention of the National Electnc Light Association at Niagara 
Falls on June 10, 1897, and was prepared for the purpose of 
illustrating the characteristics and properties of the Westing- 
house Type C motor which, at that time, was beginning to at- 
tract much attention. This motor was radically new in that 
it had a "cage" type secondary winding for large, as well as 
small, sizes, whereas, it was generally believed that the cage 
type was only suitable for small power machines, due to lack 
of starting torque. (ED.) 


THE polyphase motor is usually treated from the theoretical 
standpoint, and the results obtained are of interest mainly to 
designers and investigators. Such treatment has been principally 
of a mathematical nature, the object being to show how the 
various characteristics of the motor may be predetermined. In 
the following treatment of the subject, the general operation of the 
motor will be explained in a non-mathematical way by the use 
of diagrams which illustrate its characteristics under different 
conditions. Only the non-synchronous type of motors will be 
considered, and no distinction will be made between two-and 
three-phase motors ; for, if properly designed, they are practically 
alike in operation. 

It is necessary to understand the characteristics of the poly- 
phase motor in order to consider properly its application to the 
different classes of work to be met with in practice. These 
characteristics can be presented in the most intelligible manner by 
means of curves, which represent the relations between the speed, 
torque or turning effort, horse power expended and developed, 
amperes, etc. The speed-torque curve, which represents the 
speed in terms of the torque, is the most important one, as upon 
this depends the adaptability of the motor to the various kinds of 
work. The starting conditions also depend upon the speed-torque 
characteristics. The other curves that are of importance in prac- 
tice are the current, efficiency aad power factor. As these are 
dependent, to some extent, upon the speed-torque curve, this will 
be considered first. Before treating of its characteristics a short 
description of the motor itself will be given. 



The polyphase motor, like a direct-current motor, consists 
primarily of two parts, one stationary and the other rotating, each 
of which carries windings. The inside bore, or face, of the sta- 
tionery part is generally slotted, and carries windings that resemble 
those of the rotating part, or armature of an ordinary direct-current 
motor without commutator. The rotating part is also slotted on 
its outside face, and there are windings in the slots. Both cores, 
or bodies, are built up of thin iron or steel plates. The general 
arrangement is shown in Fig. 1. One of these windings, generally 


that on the stationary part, receives current from a two or three- 
phase supply circuit. The coils of this winding, although dis- 
tributed symmetrically over the entire face of the core, are really 
connected to form distinct groups which overlap each other. These 
windings form the two or three circuits in the motor. When al- 
ternating electro-motive forces are applied to these circuits, currents 
will flow which set up magnetic fields in the motor. These alter- 
nating fields in turn generate electro-motive forces in the windings. 
Part of the current flowing in the windings represents energy ex- 
pended usefully, or in heating, and part serves merely as magnetiz- 
ing current. The latter, like the magnetizing current of a direct- 
current machine, is dependent upon the dimensions of the mag- 
netic circuit and upon the magnetic density in the various parts. 
Even when the motor is running with no load the magnetizing cur- 
rent is required. 


The second part of the motor, generally the rotating part, 
receives no current from the supply circuit. The magnetic fields 
set up by the first set of windings pass through the second windings, 
and, tinder certain conditions, generate electro-motive forces in 
them. If the second windings are arranged to form closed circuits, 
currents will flow in them. These currents are entirely separate 
from those of the supoly circuits. 


When running, the motor has a maximum speed that is ap- 
proximately equal to the alternations of the supply circuit divided 
by the number of motor poles in each circuit. This is the no-load 
speed. As the motor is loaded, the speed falls off almost in pro- 
portion to the load. The drop in speed is sometimes called the 


"slip." This is usually expressed in percent of the maximum 
speed. If, for instance, a motor has a maximum speed of 1000 
revolutions and drops fifty revolutions below this at full load, 
it has then a slip of five percent. 


With this type of motor, a drop in speed is necessary for de- 
veloping torque. A fairly simple illustration of this action may be 
obtained by considering the operating of an alternating current 
generator tinder certain conditions. We will take a type of alter- 
nator having a stationary armature and a rotatable field magnet, 
which can be driven at various speeds. Leads are carried out from 
the armature to adjustable resistances. To avoid complexity, the 
armature circuits and the resistances are considered as non-induc- 
tive. The field coils are excited by direct current. Fig. 2 shows 
this arrangement. 


When the field is rotated at a certain speed, with the field coils 
charged, there is an alternating electro-motive force set up in the 
armature winding. When the armature circuit is closed through 
a resistance a current will flow and the armature will develop 
power. The power developed by the armature is slightly less 
than the power expended on the field shaft, which is proportional 
to the product of the speed and the turning or driving effort i. e , 
torque on the shaft. Consequently, at a given speed, a driving 
effort is required at the field shaft, corresponding to the power de- 
veloped by the armature If the armature current is increased or 
decreased, the power developed is increased or decreased also, and 
the driving effort will vary in proportion. 

Let the field now be rotated at one-half the above speed. The 
armature electro-motive force becomes what it was before Re- 
ducing the resistance in the armature circuit also to one-half, the 
same current as before will flow. The power developed by the 
armature is now one-half and the speed of the field is one-half, con- 
sequently, the driving effort or torque is the same as before. Re- 
ducing the speed further, and decreasing the resistance in the ar- 
mature circuit in proportion, to keep the armature current con- 
stant, we find the driving effort on the field remains constant. 
Finally, if we reduce the speed so much that the external armature 
resistance is all cut out, and the armature is short circuited on itself 
with the same current as before, the same driving effort is still 

The field is now rotating very slowly, and the alternations in 
the armature are very low, being just sufficient to generate the 
electro-motive force required to drive the armature current against 
the resistance of the windings. Any further reduction in speed 
will diminish the armature electro-motive force, and hence the 
armature current must fall, the power developed be dimmivshed 
and the driving effort also fall in proportion. An increase in 
speed will increase the armature current, and thus increase the 
driving effort required. 

If but one armature circuit is closed, the power developed will 
pulsate as the armature current varies, from zero to a maximum 
value, and the driving effort will also vary, But if the armature 
has two or more circuits having different phase relations, it may 
develop power continuously and the driving effort will then be 



The armature has been considered as stationary and develop- 
ing power while a certain driving effort was applied to the field. 
According to the well known law that any force is met by an op- 
posing force, the armature must have a certain resisting effort. 
The armature really tends to rotate with the field, and the resist- 
ing effort is exerted to prevent this. 

Assume the armature to be arranged for rotation, but locked, 
in the above operations. Release the armature, attach a brake, 
and adjust for a torque equal to the resisting effort of the armature. 
The armature just remains stationary. Speed up the field, and the 
armature will speed up also, keeping a certain number of revolu- 
tions behind the field. This difference in speed is that required 
for generating the electro-motive force necessary for sending the 
current through the armature. The alternations in the arma- 
ture will remain constant for a given armature current, indepen- 
dent of the speed at which the armature is running. 

If the brake be tightened, the armature must drive more cur- 
rent through its windings to develop the required effort, the 
armature alternations must hence increase, and the armature will 
therefore lag behind its field more than before, or the "slip" is 
increased. If the brake be loosened, the armature will* run 
nearer the speed of the field. If the field be driven at a constant 
speed and the brake be released, the armature will run at practic- 
ally the same speed as the field. 

If the winding consist of but one closed circuit, the torque 
developed by the armature varies periodically, and that developed 
by the brake will vary also, but to a less extent, as it is steadied 
by the inertia of the rotating armature. But with two or more 
circuits having different phase relations, arranged for constant 
power developed in the armature windings, the torque developed 
is also constant at all times. Consequently, for constant torque 
at the brake, there should be two or more phases in the armature 


This explanation of the development of torque in the short- 
circuited armature is merely an attempt to illustrate certain of the 
actions in the polyphase motor armature by a comparison with the 
operations of other apparatus, that is, in general, much better 
understood. We cannot infer, from the above illustration, that an 
alternating-current generator would run as a motor tinder" the 


assumed conditions, for, in the above operations, mechanical 
power is supplied to the field shaft, and mechanical power is de- 
livered by the rotating armature to the brake. There is no true 
electro-motor action; that is, there is no transformation of elec- 
trical power supplied to mechanical power developed. 

No I 

Circuit 1 at Maximum 

Circuit 2 at Zero Current 

Circuit 1 Decreasing 

Circuit 2 with Increasing Current 

Circuit l at Zero 

Circuit 2 at Maximum Current 

No 4 

Circuit 1 Increasing m 

Reverse Direction 
Circuit 2 Decreasing 


Circuit 1 at Maximum In 
Reverse Direction 
Circuit 2 at Zero 

Circuit 1 Decreasing 

Circuit 2 Increasing in Reverse 



The action of the short-circuited armature of the above gen- 
erator and that of the polyphase motor are very similar in regard 
to drop in speed for developing torque But in the polyphase 
motor, instead of the mechanically rotated field magnet, there is 
a stationary core provided with two or more windings which cany 
currents having different phase relation. These windings are 
placed progressively around the core, either overlapping or on 
separate poles. When the currents flow in the windings, resultant 
magnetic poles or fields are formed, which are progressively shift- 
ing around the axis of the motor. The closed or short-circuited 
armature, rotating in this field, develops torque by dropping in 


speed, in the same way that It developed torque with mechanically 
rotated field magnets. But electrical power, instead of mechanical, 
is now supplied to produce the shifting of rotating field, and the 
conversion from electrical power supplied to the field windings, to 
mechanical power developed by the armature shaft is a trans- 
former action which does not appear in the above illustration. 


Fig. 3 shows diagrammatically a progressively shifting field, 
with two overlapping windings arranged for two-phase currents. 
Coils 1-1, etc., form one circuit, while coils 2-2, etc., form the other. 
Starting with the instant when the current in 1 is at its maximum 
value, the magnetizing force of this set of coils must be at its maxi- 
mum. The current and magnetizing force of circuit 2 are at zero 
value. Four poles or magnetic fields, alternating N-S-N-S around 
the core, are formed directly over coils 1. As the current in one 
begins to decrease, that in 2 rises. We then have the combined 
magnetizing forces of the two overlapping windings. These two 
magnetizing forces act together at some points and oppose at 
others. The resultant magnetic field shifts to one side of the 
former position. As the current in 1 gradually falls to zero and 
2 rises to its maximum value the magnetic field shifts around until 
it is directly over coils 2. If the current in 1 should next increase 
in the same direction as before, while 2 diminished, the magnetic 
poles would shift back again to their former position. But the 
current in 1, after reaching zero value, rises in the opposite direc- 
tion, while that in 2 falls. This shifts the resultant poles forward 
instead of backward, and they gradually shift ahead until they 
are again directly over coils 1. But the "N" poles have shifted 
around until they now occupy the former position of the "s" 
poles. Thus, with the current in 1 passing from a maximum in 
one direction to a maximum in the opposite, the poles have shifted 
forward the width of one polar space. Current in 2 next rises in a 
reversed direction and the poles shift forward until, when the cur- 
rent in 2 is a maximum, they are over coils 2. 

In the diagrams, Nos. 1, 2, 3, etc , show the positions of the 
shifting field under certain conditions of current in the two circuits. 
In No. 2, the position shown is an arbitrary one, for it depends upon 
the relative values of the currents in the two circuits. With the 
two currents equal, the position of the line N-N would be half-way 
between coils 1 and 2. 


These diagrams show that the magnetic field due to two-phase 
currents in properly arranged windings shifts progressively around 
the axis, just as if the field were rotated mechanically. 


In polyphase motors, the part that resembles the field in the 
above description and which receives the current from the line, is 
usually called the primary, on account of its electrical resemblance 
to the primary of a transformer. The equivalent of the armature 
in the preceding description is called the secondary. If the alter- 
nations of the supply circuit are constant, the reversals of the cur- 
rents in the field or primary will occur at a uniform rate and the 
magnetic field will shift around its center at a definite speed, de- 
pending upon the rate of alternation of the supply circuit and the 
number of poles in each circuit of the motor. If the armature or 
secondary rotates at the same speed as the field shifts, there will be 
no reversals or alternations in its magnetism, and there will be no 
currents and consequently, no torque. If a load is thrown on, the 
speed will drop and the resultant alternations in the secondary 
will generate electro-motive forces which will drive currents 
through the windings, and thus develop torque. The speed will 
continue to fall, and the secondary electro-motive forces will con- 
tinue to increase until a torque sufficient for the load is developed. 



Increasing the load on the motor, the speed should fall and the 
torque increase until zero speed is reached. The speed-torque 
curve would then be of the form shown in Pig. 4, curve "A." But 
the shape of this curve is modified to a great extent in actual 
motors by certain effects which cannot be entirely eliminated. 

In the case of the revolving field, the magnetization was sup- 
posed to remain constant under different conditions. But in the 
motor primary, the magnetism of the primary is not constant 
under all conditions and it does not all pass through the secondary 


circuits. y The primary windings necessarily have some resistance, 
and a certain electro-motive force is required to drive the primary 
current through the windings. With a constant applied electro- 
motive force, the primary counter-electro-motive force will dimin- 
ish as the drop in primary resistance increases, and the magnetic 
field required will diminish also. Consequently, to develop the 
required secondary electro-motive force for driving the secondary 
current through the windings the speed must drop more than 
shown by curve "A" in Pig 4. This gives a speed-torque curve 
as shown by curve "B," in Fig. 4. Instead of being a straight line 
it is somewhat curved. 


But there is a still more important effect in the motor. The 
primary and secondary currents, and their consequent magnetizing 
forces, are opposed to each other. The result is that part of the pri- 
mary magnetism threads across between the primary and second- 
ary windings without passing into the secondary. Thus, the 
electro-motive force of the secondary is reduced, or, for a required 
secondary electro-motive force, the secondary alternations must be 
increased. This means a further drop in speed. 

The secondary currents also tend to form local magnetic fields 
around their own coils. These local fields are alternating and set up 
electro-motive forces in the secondary circuits. In consequence, 
the electro-motive forces generated by the magnetism from the 
primary have to drive currents, not only against the resistance of 
the secondary windings, but also against these local electro-motive 
forces. This necessitates a further drop in speed for the required 
torque. These local electro-motive forces depend upon the 
secondary alternations and, therefore, vary with the drop in speed, 
and are greatest at zero speed. This introduces a very complicated 
condition in the secondary circuits. These magnetic fields which 
thread around only the primary or secondary windings are called 
the magnetic leakages, or stray fields, or the magnetic dispersion. 

If the magnetic leakage is relatively large, that is, twenty to 
twenty-five percent of the total induction, and the secondary re- 
sistance is low, the speed-torque curve will have the peculiar shape 
shown in Fig. 5, This curve shows the torque increasing as the 
speed falls, until a certain maximum is reached. Beyond this 
point the torque diminishes with further drop in speed. If the 
motor is loaded to the maximum torque, a slight increase in load 
causes a further drop in speed, the torque diminishes and the motor 



stops. As a consequence, the normal rating of the motor must be 
considerably below this "pulling-out point. 7 ' The margin neces- 
sary depends upon the nature of the load to be carried. 



The starting torque, speed regulation, etc., of the polyphase 
motor depend upon the form of the speed-torque curves. The 
different methods of varying the form of these curves will be 
considered next. 

As the secondary electro-motive force is that necessary to 
drive the secondary currents through the windings, it follows that 
the electro-motive force required must depend upon the resistance 
of these windings. A larger resistance means a larger electro- 
motive force for the required current, and, therefore, a greater 
number of secondary alternations, or a greater drop in speed. 
The torque being held constant, any variation of the secondary 
resistance requires a proportionate variation in the slip. If the 
slip with a given torque is 10 percent, for instance, it will be 20 
percent with double the secondary resistance, or 50 percent with 
five times the resistance. This is true only with the primary con- 
ditions of constant applied electro-motive force and constant alter- 
nations. The secondary resistance may be in the windings them- 
selves, or may be external to the windings but part of the secondary 
body, or it may be entirely separate from the machine and con- 
nected to the windings by the proper leads. 

Fig. 6 shows the speed-torque curves for a motor with different 
resistances in the secondary circuit. In curve " A" the secondary 
resistance is small In curve "B" the secondary resistance is 
doubled. The maximttm torque remains the same but the slip 
for any given torque is doubled. This motor starts much better 



than that in curve "A." In curve "c," the resistance is again 
doubled and the slip is also doubled. The starting torque is in- 
creased but the slip is rather large at the rated torque, "x." In 
curve "D," the slip is again doubled. In this case the torque is 
high at start and falls rapidly as the speed increases. In curve 
"E," the maximum torque is not yet reached at zero speed. Con- 
tinuing these curves below the zero-speed line, that is, running the 
motor in the reverse direction, we get the general form of these 
different speed-torque curves. They are all of the same general 
shape, and all have the same maximum torque. 




So far as torque is concerned, curve * ' D ' ' is the best for starting. 
But for running, curve "A" gives the least drop in speed. Con- 
sequently, if a resistance is introduced at start that will give the 
speed-torque curve "D," it should be cut out or short-circuited for 
the running condition. This is one method of operation that has 
been much used. 


In determining the best starting condition, the current sup- 
plied to the primary must be considered in connection with the 
speed-torque curves. This current is plotted with the series of 
speed-torque curves shown in Fig, 6. Referring to this figure, 
curve "A*' represents the primary amperes in terms of torque. 
Starting at the point "B," of no-load, or zero torque, it rises at 
a nearly uniform rate until maximum torque is approached; that 
is, below the point of maximum torque the current is nearly pro- 
portional to the torque, but beyond this point the current continues 
to increase and reaches a maximum at the torque represented by 
zero speed. At reversed speed this current is further increased. 
This one current curve holds true for all the speed-torque curves, 

1 A 

'c," "D," etc. 

P f 



Comparing the different curves, we see that "A" takes the 
most current at start, and gives low torque; "B" takes less cur- 
rent than " A, ' ' and gives more torque ; " c " takes less current than 
"B"; "D" takes less current than "c" and gives the maximum 
torque at start ; " E " takes less current than " D, ' ' and develops less 
torque; but the current and torque are very nearly in proportion 
over the whole range. From this we see that a speed-torque curve 
of the form of ' ' D " or ' ' E " is decidedly better for starting than ' ' A ' ' 
or "B." But for running at less than the maximum torque there 
is no advantage, so far as current is concerned, in curve "D" over 
curve "A," and the speed regulation of "D" is poor. 


Fig. 7 represents the conditions of speed, current, etc., when 
a variable secondary resistance is used at start. The motor 
starts at "F" on curve "D," and takes a current "G." The cur- 
rent falls to "H," while the speed rises to "i," which corresponds 
to the normal torque "T" at which the motor will run under the 
given conditions as long as the motor operates on curve "D." 
The speed will remain at this point. If the resistance in the 
secondary is now short-circuited, and the load thus shifted to the 
speed-torque curve "A," the torque at the speed "i," increases to 
"K" on torque curve "A." The current corresponding to this is 
"L." As the torque at "K" is greater than the normal torque 
"T," the motor speed will increase until normal torque is reached 
again at "M," while the current falls from "L" to "H." 

At the moment of cutting out the secondary resistance there 
was a very considerable increase in the current. By arranging the 
starting resistance in the secondary so that the motor will start at 
some curve intermediate between "A" and "D" and thus take 
more current at start, somewhat less would be required upon 
switching to curve "A." If curve "E" is used for starting, and if 
the torque required when speeding up is greater than that at the 
point where curves "A" and "E" cross each other, the motor will 
not pull up because in switching from "E " to "A," the torque falls, 
and the motor will stop. The current on switching over increases 
to " N," and then rises to " o " as the motor stops. In this case the 
resistance that gives curve "E" is too great, and a lower starting 
resistance is required; with a large number of resistance steps 
small variation of current is secured. 

By making several steps of the secondary resistance, so that 
it may be cut out gradually, the motor may be made to pass 



through a series of speed-torque curves with much smaller varia- 
tions of current than shown in the preceding diagrams. This 
method has been used to some extent, but requires collector rings 
or a complicated switching arrangement in connection with the 
motor secondary. 


Fig. 8 shows the conditions for starting and speeding up with 
five speed-torque curves. The motor starts on curve " E " at "p." 
The speed rises to "G." The motor is then switched to curve 
"D," the torque rising to "H." The speed then rises to "i." In 
this way the motor passes successively from "D" to "c," "B" and 
"A," until the full speed is reached. The currents at no time reach 
very high values. 

Plotting the current in, terms of speed, the use of a large 
number of steps is shown to better advantage. This is shown in 
Pigs. 9 and 10. Fig. 9 shows the same starting conditions as Fig. 
7 with curves "D" and "A." The current starts at "A" and falls 
to "B." The resistance is then short-circuited and the current 
rises to " c " and then falls to "D," which is the same as "B," If 



"A" had been higher at start, "c" would have been lowered 
slightly. But as the time required for passing from "A" to "B" 
is generally greater than that from "c" to "D," "c" may be 
higher than "A." If the motor is not required to develop such a 



large torque when pulling up, then " c " may be lowered while "A " 
is left unchanged. 

In Fig 10, the currents in terms of speed are shown for five 
steps with the five speed-torque curves of Fig 8. The starting 
current "A" is low, and none of the currents, when switching from 



one curve to another, is large. The dotted lines show the cor- 
responding currents for two steps, as in Fig. 9. 


For variable speed work, such as cranes, elevators, etc., the 
series of curves in Fig. 6 shows one method of regulating the speed 
By varying the secondary resistance over a wide range, any speed 
from zero to maximum may be obtained with any torque up to 



the maximum. This requires the use of collector rings and ad- 
justable rheostats. The vanations in speed are obtained by 
wasting energy in resistance. For a given torque the same power 
is expended on the motor whether the speed is zero or maximum. 
To obtain a certain torque at start requires as much power as 
when running at full speed. 

An analysis of the motor shows another way in which the speed- 
torque curves may be varied. In Fig. 6, all the curves show a 
certain maximum torque which is the same in all cases; but this 
is with the condition of constant primary electro-motive force. 
By varying the electro-motive force applied to the primary we 
may obtain a quite different series of curves. Taking, for ex- 



ample, a speed-torque curve of the form "A" in Fig. 11, and 
applying a higher electro-motive force to the primary, a curve is 
obtained of the same shape as "A," but with a much higher point 
of maximum torque. Lowering the applied electro-motive force, 
the maximum torque is lowered. The torques at any given speed 
are raised or lowered in the same proportion as the maxima are 
varied. At any given speed the torques are proportional to the 
square of the electro-motive forces applied. This relation holds 
good for any form of the torque curve, whether of the shape "A," 
"D," or "B," shown in Fig. 6. 

The current curves are also shown in Fig. 11. They all have 
the same general shape, but have different maximum values, these 
being proportional to the electro-motive forces applied. The 
speed-torque curve "A" in Fig. 11 has the same shape as "D" in 
Fig. 6, which gave too great a drop in speed. In Fig. 11, curve 
"B," which is the same form as "A," gives less speed drop for the 
same torque. Curve "c" gives less than "B," and has fairly 
good speed regulation from no-load up to normal torque "T." But 


this result is obtained at the expense of increased induction in the 
iron, and large no-load or magnetizing current due to the higher 
electro-motive force, is required. If it is possible to obtain a 
speed-torque curve like "c" in Fig 11 with the normal electro- 
motive force applied, we can obtain good speed regulation from 
no-load up to the rated torque, and shall be able to start the motor 
with the maximum torque it can develop. Then, by lowering the 
applied electro-motive force, the same form of speed-torque curve 
will be retained, but the starting torque and starting current may 
be lowered to any extent desired. 


Returning to Fig 5, it was stated that the peculiar shape of 
this curve, with the torque falling rapidly after reaching a maxi- 
mum value, was due mainly to magnetic leakage between the 
primary and secondary windings. But if the motor is so pro- 
portioned that the leakage is very small compared with the useful 
field, the speed-torque curve takes a quite different shape. The 
maximum torque is increased directly as the magnetic leakage is 
diminished. This is shown in Fig. 12. Here "A" is similar in 
shape to curve "A" in Fig. 6; "B" represents the speed-torque 
curve with the magnetic leakage reduced one-half; "c" repre- 
sents it with about one-half the leakage of "B," and "D" with one- 
half that of "c." 

In comparing Figs 6 and 12, it may be noted that "A" in one 
is the same form as "A" in the other, although drawn to a dif- 
ferent scale. In Fig. 12, "B " has the same shape as in Fig. 6, but 
has a different maximum value. The same is true of curves "c" 
and "D" in the two figures. By lowering the applied electro- 
motive forces for curves "D," " c " and " B " of Fig. 12, so that the 
maximum torques are equal to that of "A," as shown by the dot- 
ted curves, we get practically the same curves as in Fig. 6. 

Curve "D," in Fig. 12, gives as good running conditions as 
curve "A" in Fig. 6, having about the same drop in speed at the 
normal torque "T." We have, then, in "D" a curve which starts 
at the point of maximum torque, and which also has a small drop 
in speed at the normal load. The objection to this curve is that 
the starting current and starting torque, although in the proper 
proportion to each other, are both much greater than is necessary 
or desirable. By reducing the applied electro-motive force at 
start, however, lower torques and currents are obtainable. In this 
way we may combine good starting and running conditions in one 



motor without the use of starting resistances, and with a secondary 
that has no resistance except that of its own windings. Fig. 13 
shows the speed-torque and current curves of such a motor with 
the applied electro-motive force varied over a considerable range. 


If but one electro-motive force is desired for starting and 
speeding up, and the motor is then to be transferred to the work- 
ing electro-motive force, the speed-torque curves should preferably 



have the shape shown in Fig. 14. The motor starts with the de- 
sired torque at reduced e. m. 1, and comes up to almost rated 
speed before switching over. This is suitable for constant speed 
work. In Fig. 14 are shown both the starting and running speed- 
torque curves, and the currents both in the motor and the line. 
The line currents are smaller than the motor currents in the ratio 
of reduction of electro-motive force in the regulating transformers. 



For cranes, elevators, and variable speed work in general, 
curves of the form shown in Fig. 15 are preferable. The line cur- 
sents are also shown in this figure. This series of speed-torque 





curves shows that a wide range of speed may be obtained by proper 
variations of the applied electro-motive force. The line currents 
"A," "B," etc., practically overlap each other. This means that 
the line current required with this method of control is very nearly 



constant for any given torque, independent of the speed. The 
same is true of the method of control by varying the secondary 
resistance. It may be noted that the current for starting, as on 
curve "c," for instance, is slightly greater than that required for 
running at the same torque on "B" or "A." This is due to the 
speed-torque curve being somewhat curved at its outer end. With 
a somewhat higher resistance of the secondary the curves are more 


nearly straight, but the drop in speed is somewhat increased on the 
speed torque for any given electro-motive force In practice, a 
compromise is made between the best possible starting condition 
and a condition of less speed drop. 

A comparison of the methods of control by varying the 
secondary resistance and by varying the applied electro-motive 
force shows that they give practically the same results in regard 
to starting, speed regulation, etc. But a motor that has been 
designed for regulation by varying its secondary resistance, will 
generally give very poor results when an attempt is made to 
operate it by the variable electro-motive force method A motor 
must be especially proportioned for small magnetic leakage when 
this method of control is to be used. The proportions and the ar- 
rangement of the parts are such as may class this as a practically 

distinct type of motor. 


We come now to the other characteristics of the polyphase 
motor, the most important of which are the efficiency and the 
power factor. The importance of efficiency is generally apprec- 
iated, but the question of power factor in most cases appears to be 
not thoroughly understood or else is entirely overlooked. 

The efficiency of a polyphase motor is the ratio of the power 
developed to the true power expended, as in any other kind of a 
machine. The power developed may be obtained from the speed- 
torque curves. If the torques are given for one foot radius, and 
the speed in revolutions per minute, then the product of any given 
torque by the corresponding speed, divided by 5,250, will give the 
power developed in horse-power; or torque multiplied by speed, 
divided by seven, gives the power developed in watts. This 
power, plus the iron, copper and friction losses, gives the true 
power expended. 

The power factor is the ratio of the true power to the apparent 
power expected. This apparent power is proportional to the 
products of the primary currents by the electro-motive forces. 
If there is magnetizing current, and if the motor has magnetic 
leakage, the primary currents are not in phase with their electro- 
motive forces and their products represent an apparent power 
which is greater than the true energy expended. The current of 
each circuit can be considered as made up of two currents, one of 
which is in phase with the applied electro-motive force, represent- 
ing true energy, and the other at right angles to the electro-motive 
force, representing no energy. The right-angled component is" the 


one that has an injurious effect on the regulation of the generator, 
transmission lines, transformers, etc. 

The size of this component, compared with the useful current, 
may be shown by a table : 

Useful 90 Degree 

Power Factor Total Current Component Component 
100 100 100 0. 

99 100 99 14.2 

98 100 98 19.9 

95 100 95 31.2 

90 100 90 43 6 

80 100 80 60.0 

70 100 70 71 4 

60 100 60 80 

50 100 50 86 6 

40 100 40 91 6 


At 90 percent power factor, for instance, the current that is 
lagging 90 degrees behind the electro-motive force is equal to 43 6 
percent of the total current flowing. This lagging current reacts 
on the generator, affecting the regulation. In an alternating- 
current generator, a 90-degree lagging current in the armature 
coils directly opposes the field magnetization. When delivering 
a current at 90 percent power factor there is over 43 percent of this 
current opposing the field, and at 80 percent power factor 60 percent 
is opposing the field. If the armature ampere turns are normally 
20 percent as great as the field ampere turns, then a load of 80 
percent power factor will give an opposing magnetization in the 
armature of about 60 percent of the total armature ampere turns, 
or about 12 percent of the total field, and the armature electro- 
motive force will be lowered approximately that percent more than 
with a load of 100 percent power factor. 

The inductive effects of the lagging current in the transmission 
circuits and transformers are much more serious than those from 
a current that is in phase with the electro-motive force. The 
generator, transformers, lines and motors also have increased 
losses, due to the large current required when the power factor is 
low. An 80 percent power factor in a system means losses due to 
heating of conductors more than 50 percent greater than those 
with 100 percent power factor. These figures indicate the im- 
portance of good power factors in an alternating-current system. 


The lagging, or 90-degree component, of the current in a 
motor depends upon the amount of the no-load, or magnetizing, 
current and upon the magnetic leakage Let this lagging com- 
ponent be expressed in percent of the total current Also express 
the magnetizing current in percent of the total current, and the 
total magnetic leakage in percent of the total primary induction. 
Then the sum of the percents of magnetizing current and magnetic 
leakage represents very closely the percent of the lagging com- 
ponent of the primary current. If, for example, the magnetizing 
current is 30 percent and the leakage is 14 percent, the resulting 
lagging component is about 44 percent. From the preceding 
table, this indicates about 90 percent power factor A low leakage 
and a high magnetizing current may give the same power factor at 
full load as a high leakage and low magnetizing current; but at 
half load, the percent magnetizing current is practically doubled, 
while the percent magnetic leakage is halved. Hence, a low mag- 
netizing current is of great importance in maintaining a high power 
factor If a high value of power factor over a wide range is de- 
sired, then both the leakage and the magnetizing current must be 

The method of control by varying the primary electro-motive 
force is dependent upon the fact that the motor has a low magnetic 
leakage. By using certain proportions and arrangements of the 
windings on the primary and secondary, the magnetizing current 
may be made comparatively low. Thus both conditions for good 
power factor are obtained. 

With the method of control by varying the secondary resist- 
ance, good power factors may be obtained. But the form of 
secondary winding required when variable resistances are used 
tends to reduce both the power factor and the maximum torque. 


An elaborate series of tests was made to determine the best 
type of winding for the secondary of a polyphase motor. First, 
two circuits were arranged to give secondary phases ninety degrees 
apart. The starting, running and maximum load conditions were 
determined. Then a three-phase secondary winding was used. 
This gave a higher pulling-out torque and better power factor than 
the two-phase Pour phases were tried and were bottcr than 
three; and six were better than four. Then twelve phases were 


tried, with a gain over six in maximum torque, but not much gain 
in efficiency. The power factor was somewhat improved. Finally 
the winding was completely short-circuited on itself, all coils being 
connected to a common ring. This gave a further increase in - 
maximum torque and power factor over the preceding arrange- 
ment, but there was very little gain in efficiency. The same 
primary was used in all these tests. Each time the number of 
secondary circuits was increased the power factor was somewhat 
improved. This was due to the fact that the secondary currents 
were able to so distribute themselves that the local electro-motive 
forces in the coils, due to leakage, were diminished; or, the mag- 
netic leakage may be considered to have been diminished. This 
would necessarily give higher pulling-out torques and higher-power 


Very complete tests were also made to determine the best 
form of primary winding, and a certain method of distribution of 
the coils was found to diminish the primary magnetic leakage very 
considerably. This somewhat increased the maximum torque 
and the power factor. Utilizing the arrangements of the primary 
and the secondary windings just described, and otherwise pro- 
portioning for small magnetic leakage, a motor may be obtained 
that has a comparatively low total induction, and yet has a mag- 
netic leakage of but a few percent. The low induction allows a 
small magnetizing current and comparatively low iron losses. The 
low leakage gives a high pulling-out torque, and thus allows a good 
speed regulation, and also good starting conditions, by varying the 
applied electro-motive force. 


Motors that are adapted for operation under the conditions of 
variable applied electro-motive forces with constant secondary 
resistance must have the special forms of speed-torque curves 
shown in Figs. 12 to 15, and they may therefore be considered as 
forming a distinct type. This type has received the name Type C. 
The Type C motor is always characterized by low magnetic leak- 
age and consequent high pulling-out torque. The secondary has 
no adjustable resistance and all regulation is obtained by varying 
the adjustable electro-motive force. The secondary is made the 
rotating part, on account of the type of winding used, which con- 
sists of copper bars placed in tunnels or slots in the core and 
bolted to two end rings. There are no bands, and the question of 


insulation is of very little importance for the maximum secondary 
electro-motive force does not exceed three volts in a 500 horse- 
power motor and is less with smaller sizes. 


This type of motor possesses several distinct advantages over 
other forms of polyphase motors. The method of control, by 
varying the electro-motive forces applied to the motor, leads to 


3000 ALTS.. POLES 


two very important advantages, one of which is mechanical and 
the other electrical. With this method of control there are no 
regulating appliances on the motor and, in consequence, it may be 
of the simplest possible form. The electrical advantage is that 
the motor may be started and controlled from a distance. Thus 
it may be placed entirely out of reach of the operator. On travel- 
ing cranes, for example, this is of special advantage, for in this case 
only the primary wires need be run from the operator's cage to the 
motor. If there are several motors on the crane, there may be one 
wire common to all the motors and but two additional wires per 



motor are required. Thus for the three motors, a minimum of 
eleven trolley wires may be used. 

If the variable electro-motive forces are obtained from trans- 
formers, the switches for operating several motors may be wired 
to one set of transformers and the motors may be started and 
regulated independently. For traveling cranes, only one set of 
transformers is used for the hoisting, bridge and traveling motors, 




4000 5000 COCO 7000 8000 0000 10000 11000 1200^ 
Pqunds Torque at i Foot Radius. 


and this set may supply currents at different electro-motive forces 
to all the motors at the same time. A further advantage possessed 
by this motor lies in the high pulling-out torque. If a heavy over- 
load, or a load having great inertia, is suddenly thrown on a motor 
that has a speed-torque curve like "A" in Fig. 6, the point of 
maximum torque may be passed for an instant, and the motor will 
be stopped unless the load is quickly removed. A Type C motor 
in this case would have its speed pulled down for a moment, but 
this reduction in speed gives an increased torque, thus enabling 
the motor to carry the overload. 


If the electro-motive force of the system Is suddenly lowered, 
the pulling-out torque of the motors is lowered very materially. 
A reduction of twenty percent in the electro-motive force will 
lower the pulling-out torque to about two-thirds of its former value. 
Even with a temporary drop in the electro-motive force, such as 
would be caused by a momentary short-circuit on the lines, this 
may be sufficient to stop the motor But a motor that has a 
pulling-out point several times as large as its normal running 
torque is very rarely in danger of being shut down from this cause. 
This type of motor has a starting torque from two to four times 
as large as the full-load running torque and it is thus able to start 
any kind of load. In practice the starting torque is adjusted to the 
load to be started by applying a suitable electro-motive force, as 
will be explained below. 

A last, but not least, advantage of the Type C motor is its 
adaptability for large sizes. The larger the motor of this type, the 
lower in proportion can be its magnetic leakage and its magnet- 
izing current. In consequence, the power factors are very high. 
The efficiencies are also very good over a wide range of load The 
curves for a seventy-five horse-power, six-pole, 3,000-altcrnntion 
motor are given in Fig. 16, also the curves for a 400-horsc-power, 
2,300-volt, eight-pole, 3,000-alternation motor in Pig. 17 The 
power factors of these motors are good examples of what can be 
obtained on large motors of this type. 


There are six methods of varying the speed of polyphase 
motors, but some of them are applicable only in special cases. 
These methods are: 

(1) Varying the number of poles. 

(2) Varying the alternations applied. 

(3) Motors in tandem, or series-parallel. 

(4) Secondary run as single-phase. 

(5) Varying the resistance of the secondary. 

(6) Varying the electro-motive force of the primary, 

with constant secondary resistance. 

Some of these methods are efficient, while some are very in- 
efficient if the speed is to be varied over a wide range. 


The first method, varying the number of poles, is efficient to a 
certain extent, but is limited in the number of combinations of 


poles obtainable. But if combined with some of the other methods 
it may be made fairly effective over a wide range. It consists in 
varying the arrangement of the primary coils in such a way that 
the number of resulting poles is varied. This may be accom- 
plished by having two or more separate windings on the primary; 
or one winding may be used, it being rearranged for different speed. 
With this method of varying the speed, a secondary of the "cage " 
type is the only practical one. With a "grouped" or "polar" 
winding on the secondary, this would need rearranging for the dif- 
ferent speeds, just as in the case of the primary. But the cage 
winding, being short-circuited on itself at all points, is adapted to 
any number of poles. In general, this method of regulation will 
allow for only two speeds without great complications, and the 
ratio of the two speeds is preferably two to one, although three to 
one may be obtained. The simplest arrangement of winding con- 
sists of two separate primary windings; one for one number of 
poles, and the second for the other. In combination with a var- 
iable primary electro-motive force, the speed-torque curves being 
of such shape that this method may be used, the variable-pole 
method of regulation may be made fairly efficient over a wide 
range of speed. But the two windings considerably increase the 
size of the motor, while the one-winding arrangements are rather 
complicated. Consequently, we may consider that this method 
of speed variation will be used only in special cases. 


The second method, variable alternations, is theoretically the 
ideal method; but it is practically limited to a few special applica- 
tions, for we have as yet no commercial alternation transformer. 

In a few cases, where but one motor is operated, the generator 
speed may be varied. If the generator is driven by a water-wheel, 
its speed may be varied over a wide range, and the motor speed will 
also vary. If the generator field be held at practically constant 
strength, then the motor speed may be varied from zero to a 
maximum at constant torque with a practically constant current. 
This is a convenient method of operating a motor at a distance 
from the generator. The speed of the motor may be completely 
controlled by an attendant at the generating station. 

Fig. 18 shows the speed-torque and other curves of a motor 
when operated at 7200, 3600, 1800 and 720 alternations per minute, 
or^at 100, SO, 25 and 10 percent of the normal alternations. The 
speed-torque curves, corresponding to the above alternations are, 



"a," "b," "c" and "d." The current curves are "A," "&>" "c" 
and "D." This figure shows that for the rated torque "T," the 
current is practically constant for all speeds, but the electro-motive 
force varies with the alternations. Consequently, the apparent 
power supplied, represented by the product of the current by 


electro-motive force, varies with the speed of the motor, and is 
practically proportionate to the power developed 


The third method is to run motors in tandem or series-parallel- 
In this arrangement, the secondary of one motor is wound with a 
grouped or polar winding to give approximately the same electro- 
motive force and number of phases as the primary. The secondary- 
is connected to the primary of a second motor, The secondary 
of the second motor may be closed on itself, with or without a 
resistance, or may be connected to the primary of a third motor, 
etc. The arrangement with two motors is shown in Fig. 19. At 
start, motor No. 1 receives the full number of alternations on its 
primary, and its secondary -delivers the same number to the prim- 
ary of motor No. 2. Both motors will start. As motor No, 1 
speeds up, its secondary alternations fall. At about one-half 
speed, its secondary alternations are about one-half its primary, 
and motor No. 2 receives one-half the alternations of motor No. 1 ; 
it also tends to run at half-speed, Therefore, if both motors ate 


coupled to the same load, this half speed is a position where the 
two motors tend to operate together. By connecting both 
primaries across the line, both motors will be run at full speed. 
Thus, with two motors, two working speeds may be obtained. 
This method always requires at least two motors. Its application 
is limited to a few special cases. 


The fourth method the secondary run with a single circuit 
closed will give a half-speed, and with two or more circuits 
closed, will give full speed. But the power factor at the half -speed 
is very low, and the efficiency is not nearly so good as when run 
at full speed. This may have a few special applications. Fig 20 
shows this arrangement. 



The fifth arrangement is by varying the resistance in the 
secondary. This method was considered before when the speed- 
torque characteristics were shown. This will not give constant 
speed except with constant load, as the speed-torque curve, with a 
relatively large resistance is a falling curve. At heavy torques, the 
motor will run at very low speeds, while with light loads it will run 
at almost full speed. The speed regulation will be similar to that 
of a direct-current shunt motor with a resistance in circuit with the 
armature. To hold constant speed with variable load, this resis- 
tance requires continual adjustment. 


The sixth method that in which the primary electro motive 
force is varied while the secondary resistance is held constant 
gives the same results as the fifth method, as the speed-torque 
curves axe similar. To hold a constant low speed, the electro- 
motive force must be varied continually if the load is changing. 


Like the fifth method, it is not efficient at low speeds, as the reduc- 
tion in speed is obtained by means of a corresponding loss of energy 
in the secondary circuits. / 



For crane work, hoisting, etc , where it is necessary to run at 
reduced speed for but a portion of the time, either of the methods 
five or six is satisfactory, but method five requires the use of a vari- 
able secondary resistance, and there must be a set of secondary 
leads earned out to a rheostat if the speed changes are to be 
gradual. This introduces complication, especially on a crane where 
several motors are to be controlled In this case there must be 
trolley wires for both the primary and the secondary circuits of each 
motor. But by method six, the control is effected in the primary 
circuit and only primary trolley wires are needed, and these may 
be controlled from one pair of transformers, as explained before. 
The sixth method is therefore the simplest and most practical one 
to use for hoisting, etc , and will be found to present many advan- 
tages for all classes of work, whether speed regulation is important 
or not. 


There are several methods of varying the electro-motive force 
applied for starting and varying the speed on the Type C motor. 
These may be classified under three headings : 

(1) Varying the electro-motive force from the genera- 

(2) Varying the electro-motive force by transformers. 
(3) Varying the motor connections. 

A variable electro-motive force may be obtained from the 
generator in several ways. The generator may be run at low 



speed, with the field charged. This gives lower electro-motive 
force and lower alternations at the same time. This is adapted 
only to places where all the motors are to be started at once. 

The generator may be run at normal speed and its field charge 
lowered. This gives the normal alternations with lower electro- 
motive force. This is practicable only where all the motors are to 
be started at once. 

A third method is to so arrange the generator windings that 
two or more electro-motive forces for each phase may be obtained. 
A lower electro-motive force may be used at start, and a higher 
for running. 

The different arrangements of the generator windings for this 
purpose are as follows: 

If the armature has but one winding closed on itself, like a 
direct-current machine, two or three phases may be taken off. For 
two phases four leads are used, Pig. 21 illustrates this Between 






1-3 and 2-4 is the maximum electro-motive force, and between 1-2, 
2-3, 3-4 and 4-1 there is 0.7 the electro-motive force of 1-3. The 
electro-motive force 1-2 is at quarter phase to that of 4-1 and 2-3, 
and the electro-motive force 3-4 is at quarter phase to that of 2-3 
and 4-1. Therefore, across any two adjacent side circuits we have 
quarter phase circuits of 0.7 the electro-motive force of the main 
circuit. A motor may thus be started on any adjacent side cir- 
cuit and then switched to the main circuit. This method is wdl 
adapted for local plants where the generator electro-motive force 
is 200 or 400 volts. If there are many motors to be started, and 


the starts are numerous, it is advisable to wire the starting switches 
so that the various motors are started on different side circuits. 

If the generator winding is of the " open coil" type, a similar 
arrangement may be obtained for two phases The two windings 
may be connected to the middle point, thus giving side circuits of 
0.7 electro-motive force This is shown in Fig 22. 


Three-phase connections do not allow any convenient com- 
binations with the generator winding. A fourth wire may be run, 
however, which will give about 58 electro-motive force for 


The method of varying the electro-motive force by means of 
transformers admits of many different combinations. Several of 
the simpler forms will be given. 

(1) The transformers may be so connected that two or more 
electro-motive forces may be obtained. 

For two-phase circuits, the secondaries may be connected 
together at the centre, as shown in Fig. 23 This gives two main 
circuits, and four side circuits of lower electro-motive force If an 
extra wire be carried out from the point 5, then 1-5, 2-5, will form 
a two-phase combination for 5 voltage, while 1-2, 2-3 form a 


two-phase combination for 0.7 voltage, and 1-3 and 2-4 give full 

Another method is to connect the secondaries at one side of 
the centre, as shown in Fig. 24. Then 3-5 and 4-5 give one electro- 
motive force; 1-5 and 2-5 give a higher electro-motive force, and 
1-3 and 2-4 give full electro-motive force. 




These combinations are useful in certain cases, but are not as 
general in their application as the following method . 

(2) Auto-transformers with loops brought out for lower 
electro-motive forces. 


In this method, no special combinations of the lines, lowering 
transformers or generators are made, but, in connection with each 
motor, a small pair of auto, or one-coil transformers, is used for 


auto-transformers are made larger From these auto- transformers 
several loops or connections are brought out. For regulating the 
speed these are connected to the contact plates or dials of a con- 
troller, as shown in Figs 25 and 26 But for starting purposes only, 
when but one loop from each transformer is used, a pair of switches 





are used in connection with the transformers. With the switches 
open, the motor is disconnected. Throwing one direction starts 
the motor at reduced voltage and brings it up to almost full speed. 
The switches are then thrown over to full electro-motive force. 

Two small transformers in a case with one four- jaw, throw- 
over switch, form what is called an "auto-starter." This is 
readily arranged for either two or three-phase circuits and motors. 
This makes a most flexible arrangement for starting, as the motor 
may be put at any location, and the auto-starter may be put in the 
most convenient position. It also loads all the line wires equally 
at start, and each motor and starter really form a unit separate 
from all the others One pair of transformers may be connected 
to several sets of switches and thus be used for starting several 

Where motors are close to reducing transformers, the second- 
aries of the transformers may have loops brought out, to which 
one or more switches are connected. The primaries of the trans- 
formers may have loops connected to proper switches, and the 
number of primary turns in the circuit may be varied instead of 
the secondary. This is applicable when the transformers supply 
only one motor, or when several motors are started at the same 
time. A regulator with secondary movable with respect to the 
primary may be used Regulators of this type vary the electro- 


motive forces without any "make" or "break" devices, and con- 
sequently have no sparking tendency. But they are in general 
too complicated and costly to compete with the transformer with 


This is not a method for changing the electro-motive force 
applied, but for varying the number of turns in series with a given 
electro-motive force, and the effect is the same as varying the ap- 
plied electro-motive force. This method is rather limited in its 
application owing to the complication involved. The simplest 
case for two-phase motors is a series-parallel combination of the 
windings of each phase. This is equivalent to using 0.5 electro- 
motive force at start. For three-phase motors, series-parallel may 
be used or the winding may be thrown from the star system of 
connection at start to the delta system for running. This is equiv- 
alent to using about 0.6 electro-motive force for start. But, as 
the star connection is preferred for the running condition, this com- 
bination is not advisable. 

There is a fourth method of regulation which may be men- 
tioned, but which is not advisable in general practice. This is the 
use of choke coils or resistance in the primary circuits of the motor, 
to reduce the electro-motive force. These really give varying 
electro-motive forces. With choke coils, the power factor at start 
is lowered, with correspondingly bad effect on the generator and 
system. With ohmic resistance in the primary circuit, the reduc- 
tion of electro-motive force is accompanied by a consumption of 
energy in the primary circuit which in no way represents torque. 


FOREWORD This paper was presented before the American Insti- 
tute of Electrical Engineers, September, 1902. It was the very 
first information given out for publication regarding the single- 
phase alternating-current railway system as developed and in- 
stalled so extensively since that time. 

Before the publication of this paper, it was generally as- 
sumed that the difficulties in the commutation of alternating 
current were so great that only motors of relatively small capac- 
ity could be built. Following its publication, many, of the larger 
companies throughout the world began work on such motors 
and produced operating railway equipments with more or less 
success (ED.) 

THE Washington, Baltimore and Annapolis Railway is a new 
high-speed electric line extending from the suburbs of Wash- 
ington to Baltimore, a distance of about 31 miles, with a branch 
from Annapolis Junction to Annapolis, a distance of about 15 
miles. The overhead trolley will be used, and schedule speeds 
of over 40 miles per hour are to be attained. This road is to be 
the scene of the first commercial operation of an entirely new 
system of electric traction. 

The special feature of this system is the use of single-phase alter- 
nating current in generators, transmission lines, trolley car equip- 
ment and motors It constitutes a wide departure from present 
types of railway apparatus. The standard D. c. railway equipment 
possesses several characteristics which fit it especially for railway 
service. These characteristics have been of sufficient importance to 
overbalance many defects in the system. In fact, a far greater 
ambunt of effort and engineering skill has been required for over- 
coming or neutralizing the defects, than for developing the good 
features possessed by the system. By far the most important 
characteristic possessed by the D. c. system is found in the type 
of motor used on the car. The D. c. railway motor is in all cases 
a series-wound machine. The series motor is normally a variable 
field machine and it is this feature which has adapted the motor 
especially to railway service. Shunt-wound motors have been 
tried and abandoned. All manner of combinations of shunt, 



series and separate excitation have been devised and found want- 
ing, and in many casesthe real cause of failure was not recognized 
by those responsible for the various combinations. They all 
missed to a greater or less extent the variable-field feature of 
the straight series motor. It is true that a variable field can be 
obtained with shunt or separate excitation, but not without con- 
trolling or regulating devices, and the variation is not inherently 
automatic, as in the series motor Polyphase and single-phase 
induction motors do not possess the variable field feature at all, 
as they are essentially constant-field machines They are 
equivalent to direct current shunt or separately excited motors 
with constant field strength, which have been unable to compete 
successfully with the series motor The variable field of the 
series motor makes it automatically adjustable for load and 
speed conditions. It also enables the series motor to develop 
large torques without proportionately increased currents The 
automatically varying field is accompanied by corresponding 
variations in the counter e.m f. of the armature, until the speed 
can adjust itself to the new field conditions This feature is of 
great assistance in reducing current fluctuations, with a small 
number of steps in the regulating rheostat Any increase in 
current, as resistance is cut out, is accompanied by a momentary 
increase in the counter e.rn f , thus limiting the current increase 
to a less value than in the case of constant field motor 

Next to the type of motor, the greatest advantage possessed by 
the D, c. system lies in the use of a single current or circuit, thus 
permitting the use of one trolley wire The advantages of the 
single trolley are so well-known that it is unnecessary to discuss 
them For third rail construction, the use of single current is of 
even greater importance than in the case of overhead trolley 
It is seen, therefore, that it is not to the direct current that 
credit should be given for the great success of the present railway 
system, but to the series type of motor and the fact that up to 
the present time no suitable single-phase A c. motor has been 

Some of the undesirable features of the D c railway system 
should also be considered. The speed control is inefficient. A 
nominally constant voltage is supplied to the car, and speed con- 
trol is obtained by applying variable voltage at the motor ter- 
minals This variation is produced by the use of resistance in 
series with the motors, with a loss proportional to the voltage 
taken up by the resistance By means of the series-parallel 


arrangement, the equivalent of two voltages is obtainable at the 
motor terminals without the use of resistance. Therefore, with 
series-parallel control, there are two efficient speeds with any 
given torque, and with multiple control there is but one efficient 
speed with a given torque All other speeds are obtained 
through rheostatic loss, and the greater the reduction from 
either of the two speeds, series or parallel, the lower will be the 
efficiency of the equipment. At start, the rheostatic losses' are 
always relatively large, as practically all the voltage of the line 
is taken up in the rheostat For heavy railroad service, where 
operation for long periods at other than full and half speeds may 
be necessary, the rheostatic loss will be a very serious matter. 

The controlling devices themselves are also a source of trouble. 
An extraordinary amount of time and skill has been expended' 
on the perfection of this apparatus The difficulties increase 
with the power to be handled The controller is a part of the 
equipment which is subjected to much more than ordinary 
mechanical wear and tear, and it can go wrong at any one of 
many points. The larger the equipment to be controlled, the 
more places are to be found in the controller which can give 
trouble. The best that can be said of the railway controller is 
that it is a necessarv evil. 

Another limitation of the D. c. system is the trolley voltage 
Five hundred volts is common at the car and 650 volts is very 
unusual. By far the larger number of the railway equipments 
in service to-day are unsuited for operation at 600 volts, and 700 
volts in normal operation would be unsafe for practically all 
The maximum permissible trolley voltage is dependent upon 
inherent limitations in the design of motors and controllers 
The disadvantages of low voltage appear in the extra cost of cop- 
per and in the difficulty of collecting current. In heavy railroad 
work the current to be handled becomes enormous at usual 
voltages. A 2400 h p. electric locomotive, for example, will 
require between 3000 and 4000 amperes at normal rated power 
and probably 6000 to 8000 amperes at times. With the 
overhead trolley these currents are too heavy to be collected in 
the ordinary manner, and it is a serious problem with any form 
of trolley or third rail system which can be used. It is evident 
that for heavy service, comparable with that of large steam rail- 
ways, a much higher voltage than used in our present D. c. sys- 
tem is essential, and the use of higher voltage is destined to come, 
provided it is not attended by complications which more than 


overbalance the benefits obtained. A 111 1 her disadvantage of 
the D. c. system is the destructive action known as electrolysis 
This may not be of great importance in interurban lines, chiefly 
because there is nothing to be injured by it. In city work its 
dangers are well-known, and very expensive constructions are 
now used to eliminate or minimize its effects. 

From the above statements it is evident that an A c railway 
system, to equal the D. c., should possess the two principal 
features of the D. c. system, viz A single supply circuit and the 
variable field motor, and to be an improvement upon the D. c. 
system, the A. c. should avoid some of the more important dis- 
advantages incident to the present D. c. railway apparatus, 

The system must, therefore, be single- phase. The importance 
of using single-phase for railway work is well known The diffi- 
culties and complications of the trolley construction are such 
that several A. c. systems have been planned on the basis of 
single-phase supplied to the car, with converting apparatus on 
the car to transform to direct current, in order that the standard 
type of railway motors may be used Such plans are attempts 
to obtain the two most valuable features of the present D c. 
system. The polyphase railway system, used on a few European 
roads, employs three currents, and therefore does not meet the 
above requirement. The motor for the A c railway service 
should have the variable speed characteristics of the series D. c. 
motor. The polyphase motor is not suitable, as it is essentially 
a constant field machine, and rloes not possess any true variable 
speed characteristics. Therefore it lacks both of the good fea- 
tures of the D. c. railway system. A new type of motor must, 
therefore be furnished, as none of the alternating current motors 
in commercial use is adapted for the speed and torque require- 
ments of first-class railway service. Assuming that such a 
motor is obtainable for operation on a single-phase circuit, the 
next step to consider is whether the use of alternating instead 
of direct current on the car, will allow some of the disadvan- 
tageous features of the D. c. system to be avoided. The D. c. 
limits of voltage are at once removed, as transformers can be 
used for changing from any desired trolley voltage to any con- 
venient motor voltage. Electrolysis troubles practically disap- 
pear. As transformers can be used, variations in supply voltage 
are easily obtainable. As the motor is assumed to have the 
characteristics of the direct-current series motor, speed control 
without rheostatic loss is practicable when voltage control is 


obtained. This combination, therefore, allows the motor to 
operate at relatively good efficiency at any speed within the 
range of voltage obtained If the voltage be varied over 
a sufficiently wide range, the speed range may be car- 
ried from the maximum desired down to zero, and there- 
fore, down to starting conditions. With such an arrange- 
ment no rheostat need be used under any conditions, and the 
lower the speed at which the motor is operated, the less the power 
required from the line. The least power is required at start, as 
the motor is doing no work and there is no rheostatic loss. The 
losses at start are only these in the motor and transforming 
apparatus, which are less than when running at full speed with 
an equal torque. Such a system, therefore, permits maximum 
economy in power consumed by motor and control. This 
economy in control is not possible with the polyphase railway 
motor, as this motor is the equivalent of the D. c. shunt motor, 
with which the rheostatic loss is even greater than with the 
series motor. 

The use of alternating current on the car allows voltage control 
to be obtained in several ways. In one method a transformer 
is wound with a large number of leads carried to a dial or con- 
troller drum. The Stillwell regulator is a well-known example 
of this type of voltage control. This method of regulation is 
suitable for small equipments with moderate currents to be 
handled. The controller will be subject to some sparking, as in 
the case of D. c. apparatus, and therefore becomes less satisfactory 
as the car equipment is increased in capacity. Another method 
of control available with alternating current is entirely non- 
sparking, there being no make-and-break contacts. This con- 
troller is the so-called " induction regulator," which is a trans- 
former with the primary and secondary windings on separate 
cores. The voltage in the secondary winding is varied by shift- 
ing its angular position in relation to the primary. With this 
type of voltage controller, very large currents can be handled, 
and it is especially suitable for heavy equipments,such as loco- 
motives It is thus seen that there is o^ie method of control, 
available with alternating current, which avoids the troubles 
inherent to the D. c. controller. The induction regulator is 
primarily a transformer, and all wear and tear is confined to the 
supports which carry the rotor. Therefore the objectionable 
controller of the standard D. c. system can be eliminated, pro- 
vided a suitable A. c. motor can be obtained. This ideal type 


of controller is not applicable to the polyphase railway motor, in 
which speed control can be obtained only through rhcostatic 
loss. The polyphase control system is even more complicated 
than the D. c , as there must be a rheostat for each motor, and 
two or three circuits in each rheostat It is thus apparent that 
by the use of single-phase alternating current with an A. c 
motor having the characteristics of the D c. series motor, the 
best features of the D c. system can* be obtained, and at the 
same time many of its disadvantages can be avoided. 

This portion of the problem therefore resolves itself into the 
construction of a single-phase motor having the characteristics 
of the D c. series motor There are several types of single phase 
A. c. motors which have the series characteristics. One 
type is similar in general construction to a D. c. motor, but with 
its magnetic circuit laminated throughout, and with such pro- 
portions that it can successfully commutate alternating current 
Such a motor is a plain series motor, and can be operated on 
either alternating or direct current and will have the same torque 
characteristics in either case. Another type of motor is similar 
in general construction to the above, but the circuits are ar- 
ranged in a different manner. The field is connected directly 
across the supply circuit, with proper control appliances in series 
with it. The armature is short-circuited on itself across the 
brushes, and the brushes are set at an angle of approximately 
45 from the ordinary neutral point The first of these two types 
of motors is the one best adapted for operation in large units. 

This is the type of motor which is to be used on the Washington 
Baltimore and Annapolis Railway. Several motors have been 
built and tested with very satisfactory results, both on the test- 
ing stand and under a car. The results were so favorable that 
the system was proposed to the Cleveland Engineering Company, 
representing the Washington, Baltimore and Annapolis Railway, 
and after investigation by their engineers, the system was 
adopted. A description of the apparatus to be used on this road 
will illustrate the system to good advantage. 

Single-phase alternating current will bo suppled to the car at 
a frequency of 16 J cycles per second, or 2 T 000 alternations per 
minute. The current from the overhead trolley wire is normally 
fed in by one trolley at approximately 1,000 volts. Within 
the limits of the District of Columbia two trolleys are employed, 
as by Act of Congress the use of rails as conductors is prohibited 
in this District, presumably on account of electrolysis. In this 


case the trouble, of course, will not exist, but the contracting 
company has been unable to obtain permission for the grounded 

The alternating current to the car is carried through a main 
switch or circuit breaker on the car, to an auto-transformer 
connected between the trolley and the return circuit. At 
approximately 300 volts from the ground terminal, a lead is 
brought out from the auto-transformer and passes through the 
regulator to one terminal of the motors. For starting and con- 
trolling the speed, an induction regulator is used with its second- 
ary winding in series with the motors. This secondary circuit 
of the regulator can be made either to add to, or substract from 
the transformer voltage, thus raising or lowering the voltage 


<v*wvw> I j 


Fio. 1 a Auto-Transformer b. Induction Regulator c Reversing Switch, d Fta d 
of Motors e Armature of Motors f Equalizing Transformer. 

supplied to the motors. The regulator therefore does double 
duty. The controller for D. c. motors merely lowers the voltage 
supplied to the motors but cannot raise it, but an A. c. regulator 
can be connected for an intermediate voltage, and can either 
raise or lower the motor voltage. In this way the regulator can 
be made relatively small, as it handles only the variable element 
of the voltage and the maximum voltage in the secondary wind- 
ing is but half of the total variation required. 

In the equipments in question, the range of voltage at the 
motor is to be varied from approximately 200 volts up to 400 
volts or slightly higher. The transformer on the car will supply 
315 volts, and the secondary circuit of the regulator will be 


wound to generate slightly more than 100 volts when turned to 
the position of its maximum voltage. This voltage of the regu- 
lator is about one-fourth of that of the motors at full voltage. 
The regulator can consequently be made relatively small, in 
comparison with the motor capacity of the equipment. It has 
been found unnecessary to use much lower than 200 volts in this 
installation, as this voltage allows a comparatively low running 
speed, and approximately 200 volts will be necessary to start 
with the required torque The greater part of this voltage 5s 
required to overcoine the e m 1. of self-induction in the motor 
windings, which is dependent upon the current through the 
motor and is independent of the speed ot the armature. 

There will be four motors of 100 h.p. on each car The full 
rated voltage of each motor is approximately 220 volts The 
motors are arranged in two pairs, each consisting of two arma- 
tures in series, and two fields in series, and the two pairs are 
connected in parallel The motors are connected permanently 
in this manner As voltage control is used, there is no necessity 
for series parallel operation, as with D. c. motors. To ensure 
equal voltage to the armatures in scries, a balancing or equalizing 
action is obtained by the use of a small auto-transformer con- 
nected permanently across the two armatures in series with its 
middle point connected between them. The fields are arranged 
m two pairs, with two fields in scries and two pairs in multiple 
This parallels the fields independently of the armatures, which 
was formerly the practice with D c motors. It was a defective 
arrangement with such motors, as equal currents in the field did 
not ensure equal field strengths in the motors, and the armatures 
connected in parallel would be operating m fields of unequal 
strength, with unequal armature currents as a direct result. 
With alternating currents in the fields, the case is different 
The voltage across the fields is dependent upon the field strengths, 
and the current supplied to the fields naturally divides itself for 
equal magnetic strengths The chief advantage m paralleling 
the fields and armatures independently is, that one reversing 
switch may serve for the four motors and one balancing trans- 
former may be used across the two pairs of armatures The 
usual D. c. arrangement of armatures m series with their own 
fields can be used, with a greater number of switches and con- 

The general arrangement of the auto-transformer, regulator, 
jnotors, etc , is shown in Fig. 1 


The induction regulator or controller, resembles an induction 
motor in general appearance and construction. The primary 
winding is placed on the rotor, and the secondary or low voltage 
winding on the stator. The rotor also has a second winding 
which is permanently short-circuited on itself. This function 
of this short-circuited winding is to neutralize the self-induction 
of the secondary winding as it passes from the magnetic influ- 
ence of the primary. The regulator is wound for two poles, and 
therefore is operated through 180 for producing the full range 
of variation of voltage for the motors. One end of the primary 
winding of the regulator is connected to the trolley, and the 
other to a point between the regulator and the motors. It thus 
receives a variable voltage as the controller is rotated. There 
are several advantages in this arrangement of the primary in 
this particular case. First, the regulator is worked at a higher 
induction at start, and at lower induction when running, the 
running position being used in these equipments for much longer 
periods than required for starting Second, when the motors 
are operating at full voltage the current in the primary of the 
regulator passes through the motors but not through the auto- 
transformer or the secondary of the regulator. This allows con- 
siderable reduction m the size of auto-transformer and regulator. 
The motors on the car are all of the straight series type. The 
armature and fields being connected in series, the entire current 
of the field passes through the armature as in ordinary series 
D c. motors. The motor has eight poles, and the speed is 
approximately 700 revolutions at 220 volts. The general con- 
struction is similar to that of a D. c. motor, but the field core is 
laminated throughout, this being necessary on account of the 
alternating magnetic field. There are eight field-coils wound 
with copper strap, and all connected permanently in parallel. 
The parallel arrangement of field-coils assists in the equalizing 
of the field strength in the different poles, due to the balancing 
action of alternating circuits in parallel. This arrangement is 
not really necessary, but it possesses some advantages and 
therefore has been used. With equal magnetic strength in the 
poles, the magnetic pull is equalized even with the armature out 
of center. The armature is similar in general construction to 
that of a D. c. motor. The fundamental difficulty in the opera- 
tion of a commutator type of motor, on single-phase alternating 
current lies in the sparking at the brushes. The working current 
passing through the motor should be practically no more difficult 


to commutate than an equal direct current, and it is not this cur- 
rent which gives trouble. The real source of trouble is found in 
a local or secondary current set up in any coil, the two ends of 
which are momentarily short-circuited by a brush. This coil 
encloses the alternating magnetic field, and thus becomes a 
secondary circuit of which the field-coil forms the primary. In 

40 ,10 00 TO 80 UO 300 110 130 130 140 

PIG 2 'Westinghouse Alternating Current Railway Motor No, 91, Single-Phase 220 


the motors of the Washington, Baltimore and Annapolis Rail- 
way, this commutation difficulty has been overcome by so con- 
structing the motor that the secondary -or short-circuit current in 
the armature coil is small, and the commutating conditions so 


perfect that the combined working and secondary currents can 
be commutated without sparking. This condition being ob- 
tained, the motor operates like a D. c, machine and will give no 
more trouble at the commutator than ordinary D. c. railway 
motors. Experience covering a considerable period in the opera- 
tion of motors of 100 h.p. capacity indicates that no trouble need 
be feared at the commutator. 

An extended series of tests were made at the Westinghouse 
shops at East Pittsburg, both in the testing room and under a 
car. Fig. 2 shows curves of the speed, torque, efficiency and 
power factor plotted from data from brake tests. 

It should be noted that the efficiency is good, being very 
nearly equal to that of high-class D. c. motors. The power 
factor, as shown in these curves, is highest at light loads and 
decreases with the load. This is due to the fact that the power 
developed increases approximately in proportion to the current, 
while the wattless component of the input increases practically 
as the square of the current. The curve indicates that th e 
average power factor should be very good. The calculations 
for the W. B. and A. Railway show that the average power factor 
of the motors will be approximately 96 per cent. 

The average efficiency of these equipments will be much 
higher during starting and acceleration than that of correspond- 
ing D. c. equipments, and rheostatic losses are avoided. When 
running at normal full speed, however, the efficiency will be 
slightly less than with D. c. This is due to the fact that the A. c. 
motor efficiency is slightly lower than the D. c., and in addition 
there are small losses in the transformer and the regulator. The 
A. c equipments are somewhat heavier than the D. c., thus re- 
quiring some extra power, both in accelerating and at full speed. 
Therefore, for infrequent stops the D. c, car equipment is more 
efficient than the A. c., but for frequent stops the A. c. shows the 
better efficiency. Tests on the East Pittsburg track verified 
this conclusion. But the better efficiency of the D. c. equipment 
with infrequent stops is offset with the A. c. by decreased loss in 
the trolley wire, by reason of the higher voltage used, and the 
elimination of the rotary converter losses.. The resultant effi- 
ciency for the system will therefore be equal to or better than 
that of the D. c. 

In the W. B. and A Railway contract the guarantee given by 
the Westinghouse Electric and Mfg. Co. states that the efficiency 
of the system shall be equal to that of the D. c. system with rotary 
converter substations 


There is one loss in the A. c. system which is relatively much 
higher than in the o c This is the loss in the rail return. Tests 
have shown that at 2,000 alternations this is three to four times 
as great as with an equal direct current This would be a 
Serious matter in cases where the D. c rail loss is high. But the 
higher A. c trolley voltage reduces the current so much, that 
the A. c. rail loss is practically the same as with direct current 
at usual voltages In many city railways the D. c. rail loss is 
made very low, not to lessen waste of power, but in order to 
reduce electrolysis. In such cases the A. c. rail loss could be 
higher than D. c , thus decreasing the cost of return conductors. 
More frequent transformer substations, with copper feeders 
connected to the rails at frequent intervals will enable the rail 
loss to be reduced to any extent desired. As a frequency of 
2,000 alternations per minute is used, the lighting of the cars and 
the substations was at first considered to be a serious difficulty, 
due to the very disagreeable winking of ordinary incandescent 
lamps at this frequency. Two methods of overcoming the 
winking were tried, both of which were successful. One method 
was by the use of split phase. A two-phase induction motor 
( was run on a single-phase 2,000 alternating circuit, and current 
was taken from the unconnected primary circuit of the motor, 
This current was, of course, at approximately 90 from the cur- 
rent of the supply circuit. A two-phase circuit was thus obtained 
on the car. Currents from the two phases wore put through 
ordinary incandescent lamps, placed close together. The 
resulting illumination a few feet distant from the lamps showed 
about the same winking as is noticed with 3,000 alts. With two 
filaments in one lamp the winking disappears entirely. A three- 
phase arrangement would work in the same way. 

A muoh simpler method was tried which worked equally well. 
This consisted in the use of very low-voltage lamps. I^ow volt- 
age at the lamp terminals allows the use of a thick filament with 
considerable heat inertia. Tests were made on lamps of this 
type at a frequency of 2,000 alts., and the light appeared to be as 
steady as that from the ordinary high-frequency incandescent 
lamp. The low voltage is not objectionable in this case, as a 
number of lamps can be run in a series, as in ordinary street 
railway practice, and any voltage desired can readily be obtained. 
as alternating current is used on the car. 

There will be an air compressor, driven by a series A. c. motor, 
on each car, for supplying air to the brakes and for operating 


the driving mechanism of the controller. The details of this 
mechanism are not near enough to completion to permit a de- 
scription of it The method used will be one which readily 
allows operation on the multiple-unit system. 

The generating station contains some interesting electrical 
features, but there is no great departure from usual A. c, prac- 
tice. There will be three 1,500 k w. single-phase alternators. 
These are 24-pole machines operating at 83 revolutions and 
wound for 15,000 volts at the terminals. They are o,f the 
rotating field type, with laminated magnetic circuits aftd field- 
coils of strap on edge. The field-coils are held on the pole-tips 
by copper supports, which serve also as dampers to assist in the 
parallel running The armatures are of the usual slotted type. 
The armature coils are placed in partially closed slots. There 
are four coils per pole The proportions of these machines are 
such that good inherent regulation is obtained without saturation 
of the magnetic circuit. The rise in potential with non-inductive 
load thrown off will be approximately 4 per cent. An alterna- 
tive estimate was furnished for the generators proposing 20,000 
volts instead of 15,000. The simplicity of the type of winding 
used, and the low frequency, are both favorable for the use of 
very high voltage on the generator. As 15,000 volts was con- 
sidered amply high for the service, the engineers for the railway 
considered it unadvisable to adopt a higher voltage. 

There are to be two exciters, each of 100 k w. capacity at 250 
revolutions The exciters are wound for 125 volts normal. The 
armature of each exciter has, in addition to the commutator, 
two collector rings, so that single-phase alternating current can 
be delivered It is the intention to use the exciters as alter- 
nators for .supplying current to the system for lighting when the 
large generators are shut down at night. The main station 
switchboard comprises three generator panels, one load panel, 
and three feeder panels. High-tension oil-break switches are to 
be provided, operated by means of controlling apparatus on the 
panels. The switches, bus-bars and all high-tension apparatus 
will be in brick compartments separate from the board. In 
each generator circuit there are two non-automatic oil-break 
switches in series; and on each feeder circuit there are two over- 
load time-limit oil-break switches in series. The two oil-break 
switches in series on the same circuit can be closed separately 
and then opened to test the switches without closing the circuit. 
With the switches in the closed position they are both operated" 


at the same time by the controller, to ensure opening of the cir- 
cuit, and to put less strain on the switches, although either one 
is capable of opening the load There will be nine transformer 
substations distributed along the railway line. Each station 
will contain two 250 k.w. oil-cooled lowering transformers, 
supplying approximately 1 ,000 volts to the trolley system. The 
transformers are used in each station so that in case of accident 
to one transformer the station will not be entirely crippled. It 
is the intention of the railway company to operate a n, c. road 
already equipped with the direct-current system The present 
D. c car equipments arc to be retained, but the current will be 
supplied from a rotary converter substation fed from the main 
system of the W B and A. Railway. As this system is single- 
phase, it is necessary that single-phase rotarics be used in the 
substations. There are to be two k.w. 550-volt rotary con- 
verters. These are 4-pole, 500-rcvolution machines. The 
general construction of these machines is very similar to that of 
the Westmghouse polyphase rotary converters. The armature 
resembles that of a polyphase rotary except in the number of 
collector rings, and in certain details of the proportions made 
necessary by reason of the use of single-phase. The commutat- 
ing proportions are so perfect that any reactions due to the use 
of single-phase will result in no injurious effect. The field con- 
struction is similar to that of a polyphase rotary. . The lamin- 
ated field-poles are provided with dampers of the " grid " or 
11 cage " type, a form used at present in the Westinghouse poly- 
phase rotary converters. This damper serves to prevent hunt- 
ing, as in the polyphase machines, and also to damp out pulsa- 
tions due to single-phase currents in the armature. The damper 
acts to a certain extent as a second phase. Each rotary con- 
verter is started and brought to synchronous speed by a small 
series A. c. motor on the end of the shaft. The voltage at the 
motor terminals can be adjusted either by loops from the lower- 
ing transformer or by resistance in series with the motor, so that 
true synchronous speed can be given to the rotary converter, 
before throwing it on the A. c. line. 

From the preceding description of this system and the appar- 
atus used on it, some conclusions may be drawn as to the various 
fields where it can be applied to advantage. It is evident that ?i 
good field for it will be on intcrurban long-distance lines such as 
the W B. and A, Railway. On such railways, high trolley 
voltage and the absence of converter substations are very 
important factors. 


For heavy railroading also, this system possesses many ideal 
features.. It allows efficient operation of large equipments at 
practically any speed and any torque, and also avoids the con- 
troller troubles- which are ever present with large direct current 
equipments. It. also permits the use of high trolley voltage, 
thus reducing the current to be collected Tn this class of serv- 
ice the .advantages of this A c system are so great that is it 
possible that heavy railroading will prove to be the special field 
for it. 

For general city work, this system may not find a field for some 
time to come, as the limitations m the present system are not so 
great that there, will be any great necessity for making a change. 
It is probable that at first this system will be applied to new 
railways, or in changing over steam roads rather than in replac- 
ing existing city equipments. One difficulty with which the 
new system will have to contend, is due to the fact that the 
A. c equipments cannot conveniently operate on existing city 
lines, as is the present practice where interurban lines run into 
the cities. It will be preferable for the A. c. system to have its 
own lines throughout, unless very considerable complication is 
permitted. When the A. c. system applied to interurban and 
steam railway systems finally becomes of predominant import- 
ance, it is probable that the existing D. c. railways will gradually 
be changed to A. c. as a matter of convenience in tying the vari- 
ous railway systems together 

As was stated above, A. c equipments cannot conveniently be 
operated on direct current lines. It does not follow that the 
motor will not operate on direct current. On the contrary, the 
motor is a first-class direct current machine, and if supplied with 
suitable control apparatus and proper voltage it will operate 
very well on the D. c. lines. This would require that the motors 
be connected formally in series, as the voltage per motor is low. 
A complete set of D. c. control apparatus would be needed when 
the A. c. equipment is to be run on direct current, and con- 
siderable switching apparatus would be necessary for disconnect- 
ing all the A. c. control system and connecting in the D. c. The 
complication of such a system may be sufficient to prevent its 
use, at least for some time to come 

In some cities, very strict laws are in force in regard to the 
voltage variations in various parts of the track system. The 
permissible variations are so small in some cases, that an enor- 
mous amount of copper is used for return conductors; and in 


some cases special boosters are used in the return circuits to 
avoid large differences of potential between the various parts 
of the track system. The object in limiting the conditions in 
this manner is to avoid troubles from electrolysis. The A. c. 
system will, of course, remedy this. 

For city work, it is probable that voltages of 500 or 600 would 
be employed instead of 1,000 or higher. The transformers and 
controllers can be designed to be readily changed from full to 
half voltage, so that low voltage can be used on one part of the 
line and high voltage on another As the car equipments of 
such railways arc usually of small capacity, it is probable that 
speed control will be obtained by means of a transformer with a 
large number of leads carried out to a control drum, rather than 
by means of the induction regulator, as the latter device is much 
more expensive in small units. This is chiefly a question of cost, 
and if the advantages of the induction regulator are found to 
over-weigh the objection of high first cost, then it will be used 
even on small equipments. 

In the W. B. and A. Railway, the generators are wound for 
single-phase, In the case of large power-stations with many 
feeders, the generators may be wound for three-phase, with 
single-phase circuits carried out to the transformer substation, 
or three-phase transmission may be used, with the transformers 
connected in such a -manner as will give a fairly well-balanced 
three-phase load. 

There are many arrangements and combinations of apparatus 
made possible by the use of alternating current in the car equip- 
ments, which have not been mentioned, as it is impracticable 
to give a full description of all that can be done. But enough 
has been presented to outline the apparatus and to indicate the 
possibilities of this new system which is soon to see the test -of 
commercial service. 


FOREWORD In 1890, the author discovered, during certain ex- 
periments in the Westinghouse testing room, that a synchronous 
motor could affect the power factor of the supply system, by 
variations in^its field strength. Later, he proposed the use of 
such a machine for regulating the pressure of a supply system 
and for changing the relation between e.m.f. and current in 
alternating-current systems. However, even as late as 1904, 
when the paper was presented, the value of this method of 
operation was but little appreciated. 

This paper should be read from the viewpoint of the time 
when it was written. Hand regulation was the ordinary prac- 
tice. Consequently, alternators with inherently good regula- 
tion, that is, which would give three to four times full load 
current on sustained short circuit, were preferred, in order that 
much hand regulation would not be needed. This paper was pre- 
sented at a meeting of the American Institute of Electrical 
Engineers in June, 1904. (ED.) 

IT is well known that the synchronous motor, running on an 
alternating-current circuit, can have its armature current 
varied by varying its field strength. A certain adjustment of field 
strength will give a minimum armature current. Either stronger 
or weaker fields will give increased current. These increased cur- 
rents are to a great extent wattless. If the field is weaker than 
the normal (or field for minimum armature current), the increased 
armature current is leading with respect to the e.m.f. waves in the 
motor and lagging with respect to the line e.m.f. For stronger than 
the normal field, the current is to a great extent lagging and 
tends to lessen the flux in the motor and the current is leading 
with respect to the line e m f A synchronous motor therefore 
has an inherent tendency to correct conditions set up by im- 
proper adjustment of its field strength The correcting current 
in the motor being drawn from the supply system has a correct- 
ing effect on sucli system, tending to produce equalization between 
generated pressures in the motor and the supply pressure. This 
characteristic of the synchronous motor can readily be utilized 
for two purposes; namely, for varying the amount of leading 
or lagging current in a system for producing changes in the 
power-factor of the system (including transmission line, trans- 
formers, and generators), or a synchronous motor can be utilized 
for pressure regulation in a system. 



As the synchronous motor can be made to impress a leading 
current upon the system, and as the amount of this leading 
current will depend upon the field adjustment of the synchronous 
motor, it is evident that this property can be used for neutral- 
izing the effects of lagging current due to other apparatus on 
the system. The resultant leading or lagging current can be 
varied and the power-factor controlled over a fairly wide range 
depending upon the location of the synchronous motor or motors, 
and upon the current capacity of the motor, etc. 

As the wattless current in the motor is primarily a corrective 
current, it is evident that for most effective purposes for ad- 
justing power-factor on the system the corrective action of 
this current on the motor should not be too great. When 
used for such purpose the synchronous motor should therefore 
be one which would give a comparatively large current if short- 
circuited as a generator Also the motor should preferably be 
one in which the magnetic circuit is not highly saturated, for 
in the saturated machine the limits of adjustment in the field 
strength are rather narrow. 

As has been noted above, if the field strength bf the motor 
be varied, a leading or lagging current can be made to flow in 
its armature circuit, this current being one which tends to 
adjust the pressure of the armature and that of the supply 
system. It is evident that if the armature pressure is held con- 
stant and the supply pressure varied, a leading or lagging 
current would also flow. If for instance the line pressure were 
dropped below that of the motor, then a lagging current would 
flow in the motor tending to weaken its field, and a leading 
current would flow in. the line, tending to raise the pressure, on 
the line. 1 If the line pressure should be higher than that of 
the synchronous motor, then the current in the motor woftld 
be leading, tending to raise its pressure; while it would be lag- 
ging with respect to the line, tending to lower it pressure. 
The resultant effect would be to equalize the pressures of the 
line and motor, and there would thus be a teadency to regulate 
the line pressure to a nioro nearly constant value, It is evident 
that the less the synchronous motor is affectecL by the <correc-' 
tive current and the more sensitive the line is Jo such corrective 
action, the greater the tendency will be toward constant pres- 
sure on the line It is therefore evident that the synchronous 
motor Which gives the largest current on short circuit as n gen- 
erator would be the one which gives the greatest corrective 
action as regards pressure regulation of the system. 


For such regulation, the synchronous motor which gives a com- 
paratively large leading or lagging current with small change to 
the pressure of the system is the most suitable one. Or, the 
motor which gives the greatest change in the leading or lagging 
current is the one which gives best regulation. It is the change 
in the amount of wattless current which produces the regulation. 
This current could vary from zero to 100 leading, for example, 
or could change from 50 leading to 50 lagging, or could change 
from 100 lagging to zero lagging. Any of these conditions 
could produce the desired regulating tendency, but all would 1 
not be equally good as regards the synchronous motor capacity. 
If in addition to the regulating tendency it is desired to correct 
for lower power-factor due to other apparatus on the circuit, 
it would probably be advisable to run a comparatively large 
leading current on the line due to the synchronous motor, and 
the regulating tendency would be in the variations in the 
amount of leading current, and not from leading to lagging, 
or vice versa. A larger synchronous motor for the same regu- 
lating range would be required than if the motor were used 
for pressure regulation alone. It is evident that the current 
capacity of a -motor regulating from 50 leading to 50 lagging 
need be much less than for current regulating from 100 leading 
to zero. It is evident therefore that if there is to be compensa- 
tion for power-factor as well as regulation of pressure, that 
'additional normal current capacity is required. 

In case such synchronous motors are required for regulation 
purely, it may be suggested that such machines be operated 
at very high speeds compared with ordinary practice. At first 
glance it would appear that such a synchronous motor could 
be operated at the highest speed that mechanical conditions 
would allow, but there are other conditions than mechanical 
ones which enter into this problem. For instance, it is now 
possible to build machines of relatively large capacity for two- 
poles for 60-cycle circuits, and for very large capacities say 
1500 kilowatts having four poles. Therefore mechanical con- 
ditions permit the high speeds, and the electrical conditions 
should be looked into carefully to see whether they are suitable 
for such service. As such synchronous motors should give rela- 
tively large currents on short circuit the effect of high speeds 
and a small number of poles on short-circuit current should be 

In order to give full-load current on short circuit, the field 


ampere-rums of such a machine should be practically equal to 
the armature ampere-turns, taking the distribution of windings, 
etc., into account By armature turns in this case is not meant 
the ampere wires on the armature, but the magnetizing effect 
due to these wires Therefore to give, for instance, five or six 
times full-load current on short circuit, the field ampere-turns 
should be relatively high compared with the armature 
This means that the field ampere-turns per pole should be very 
high, or the armature ampere-turns per pole very low Ex- 
perience shows that for very high speed machines, such as used 
for turbo-generators, there is considerable difficulty m finding 
room for a large number of field ampere-turns, and therefore 
in such machines it is necessary to reduce the armature ampere- 
turns very considerably for good inherent regulating charac- 
teristics This in turn means rather massive construction, as 
the magnetic circuit m both the armature and field must have 
comparatively large section and the inductions must be rather 
high. This in turn means high iron losses in a relatively small 
amount of material compared with an ordinary low-speed ma- 
chine, and abnormal designs are required for ventilation, etc , 
and for mechanical strength 

An increase in the number of poles usually allows increased 
number of field ampere-turns without a proportionate increase 
in the number of armature ampere-turns This condition is 
true until a large number of poles is obtained when the leakage 
between poles may become so high that the effective induction 
per pole is decreased so that there is no further gain by increas- 
ing the number of poles, unless the machine is made of abnormal 
dimensions as regards diameter, etc. Experience has indi- 
cated that in the case of very high-speed and very low-speed 
alternators, it is more difficult to obtain a large current on short 
circuit than with machines with an intermediate number of 
poles For example, it is rather difficult to make a 600 kilovolt- 
ampere, 3600-rev. per min., 2-pole machine which will give 
three times full-load current on short circuit A 4-pole, 1800 
rev. per min. machine can more easily be made to give three 
times full-load current on short circuit and with comparatively 
small additional weight of material The material in the ro- 
tating part of the four-pole machine, while of greater weight, 
may be of considerably lower cost per pound. The stationary 
part of the four-pole machine may have a sorhewhat larger in- 
ternal diameter, but the radial depth of sheet -steel will be less 


than in a two-pole machine. The total weight of material in 
the armature of a four-pole machine may be practically no 
greater than in a two-pole machine. Therefore a two-pole 
machine of this capacity should cost more than a four-pole 
machine, if designed to give the same current on short circuit. 
A six-pole machine would show possibly a slight gain over the 
one with four poles, but not nearly as much as the four-pole 
machine would over the one with two poles The real gain of 
the six-pole over the four-pole construction would be in ob- 
taining a machine which would give more than three times 
full-load current on short circuit. It would possibly be as 
easy to obtain four, times full load current on short circuit 
with a six-pole machine as to obtain three times full load cur- 
rent on four 7 pole machine. An eight-pole machine would be 
in the same way somewhat better than the six-pole machine 
Therefore if a 600 kilovolt-ampere machine giving six times 
full-load current on short circuit is desired, it would be advan- 
tageous to make the machine with possibly eight to twelve 
poles. The question of which would be the cheaper would de- 
pend upon a number of features in design. 

If very large short-circuit currents are desired, then, as in- 
dicated above, the number of poles for a given capacity should 
be increased, or the normal rating of the high-speed machine 
should be decreased. If, for example, the 600 kilovolt-ampere, 
3600 rev. per min. machine, mentioned above, should be rated 
at 200 kilovolt-amperes, then it could give nine times full-load 
current on short circuit ; but such a method of rating is merely 
dodging the question. 

In general, the following approximate Umits for speeds and 
short circuit currents for 40-cycle apparatus can be given. 
These limits are necessarily arbitrary, and are intended to rep- 
resent machines which could probably be made without using 
too abnormal dimensions, 

600 kilovolt-amperes, 3600 rev. per min., two to three times 
full-load current on short circuit. 

1000 kilovolt-amperes, 1800 rev. per min., three to four 
times full-load current on short circuit. 

1500 kilovolt-amperes, 1200 rev. per min., four to five times 
full-load current on short circuit. 

2500 kilovolt-amperes, 900 rev. per min , four to five times 
full-load current on short circuit. 

For 25 cycles it is more difficult to give limiting conditions, 


as the choice of speeds is very narrow. If, for example, a 1500 
kilovolt-ampere, 2-pole, 1500 rev per min. machine can be 
made to give three times full-load current on short circuit, 
then as machines of smaller rating cannot run at higher speed, 
the limiting condition of such machines must be the amount 
of current which they will give on short circuit In the same 
way a 4-pole machine running at 750 rev per min. may be 
made for 5000 kilovolt-amperes for three times full-load current 
as the limiting rating, and there is no choice of speeds for 
ratings between 1500 kilovolt-amperes and 5000 kilovolt-amperes. 

It should be noted that the above speeds are very high com- 
pared with ordinary alternator practice and are up to high- 
speed turbo -generator practice, but machines with the above 
short-circuit ratings and speeds are probably more costly to 
build than machines of corresponding ratings at somewhat 
lower speeds It will probably be found therefore that for 
the above maximum current on short circuit the cheapest 
synchronous motors for the given ratings will have somewhat 
lower speeds than those indicated above It is certain that 
tae lower-speed machines will be easier to design and will be 
slightly quieter in operation Probably best all-round condi- 
tions will be found at about half the above speeds. 

The above limiting conditions are given as only approxi- 
mate and are based upon machines having ventiliation as is 
usually found on rotating field generators for high speed. Arti- 
ficial cooling, such as obtained with an air-blast or blowers 
could modify the above figures somewhat; but in general it 
has been found that high-speed alternators can be worked up 
to the limit imposed by saturation before the limit imposed 
by temperature is attained. Therefore if higher saturation is 
not permissible, then there may be relatively small gain by 
using artificial cooling. 

One of the principal applications of such regulating syn- 
chronous motors would be for controlling or regulating the 
pressure at the end of a long transmission line for maintaining- 
constant pressure at the end of the line, independent of fluc- 
tuations of load or change of power-factor. In this case, in- 
creased output of the transmission line may more than con- 
pensate for the cost of the regulating synchronous motor. In 
such a case the synchronous motor not only acts as a regulator 
on the system but costs nothing in the end. In general, the 
more current that such a synchronous motor will give on short- 


circuit, the better suited it will be for its purpose at the end 
of a long transmission line 

Where a number of such synchronous motors are installed 
in the same station, the field adjustment must be rather care- 
fully made, to avoid cross-currents between machines, and the 
saturation characteristics of the various machines should be 
very similar. The better such machines are for regulating 
purposes, the poorer they are for equalizing each other by 
means of cross-currents 

As to the use of dampers with such synchronous motors, 
it is difficult to say just what is required A synchronous 
motor on a line with considerable ohmic drop is liable to hunt 
to some extent, especially if the prime mover driving the gen- 
erator has periodic variations in speed If the synchronous 
motor gives very large current on short circuit, then its syn- 
chronizing power is high , this will tend to steady the operation 
of the motor and decrease the hunting. The writer believes 
that such motors in practice will be found to operate better 
and have better regulating power for constant pressure if pro- 
vided with rather heavy copper dampers effectively placed on 
the field poles. With such heavy dampers reaction of the 
armature on the field is retarded, and therefore the armature 
may give a larger momentary current than would flow it there 
were no damping effect; in other words, the motor is more 
sluggish than one without dampers Therefore the addition 
of heavy dampers on such a machine may produce the same 
regulating effect which would be obtained by a machine without 
dampers which gives a larger current on short circuit. Also 
a machine with heavy dampers will usually be the one with 
the least hunting tendency and therefore will have the least 
effect on the transmission line due to hunting currents. 

In the above, the synchronous motor has been considered 
only as a regulator and not as a motor. It may be worth 
considering what would be the effect if the synchronous motor 
can do useful work at the same time that it regulates the 
system. In this case, with a given rated output, one com- 
ponent of the input will be wattless, and the other part will 
be energy. The ratio of these two components could be varied 
as desired. For example, considering the input as 100, the 
wattless component could be 60 when the energy component 
is 80; or the synchronous motor could carry a load of 80% 
of its rated capacity, this load including its own losses, and could 


have a regulating component of 60% of its rated capacity. 
If the motor is used as a regulating machine only, then its 
wattless component caa be practically 100 It appears there- 
fore" that the machine could be used more economically as both 
motor and regulator than as a regulator alone, but in such case 
it would probably be advisable to run the motor at somewhat 
lower speed than if operated entirely as a regulator. This 
reduction in speed may practically offset the gain in apparent 
capacity by using the machine for a double purpose, Also 
there is comparatively limited use for large synchronous motors 
for power purposes, as better results are usually obtained by 
subdividing the units and locating each unit nearest to its- 
load. If a load could be provided which would permit very 
high-speed driving, then it would probably be of advantage 
to utilize the synchronous motor for driving 

As the synchronous converter is one form of synchronous 
motor, the question of utilizing such machines for regulators- 
should be mentioned* Upon looking into the question of dis- 
tribution of losses in the converter, it will be noted that the 
losses in the armature winding are not uniform Investigations 
show that at 100% power-factor, the lowest heating in copper 
is obtained, and that any departure from this power-factor 
shows considerably increased loss in the copper, such loss being 
very high in certain portions of the winding Next to the 
taps which lead* to the collector there are strips of winding 
which at times are worked at a very high loss. Experience 
shows that it is not advantageous to operate converters at a 
low power-factor, and that if so operated continuously, or 
for any considerable periods > the winding should he made much 
heavier than for higher power-factors, Also in the usual de- 
sign of converters the field is not made as strong compared 
with the armature as in alternator practice, and therefore the 
regulating tendency of the converter compared with a generator 
or ordinary synchronous motor, is low. Synchronous con- 
verters can and do act as regulators of pressure for sudden 
changes of the supply pressure, but such correcting or regu- 
lating action should not be continual; that is, the pressure 
supplied to a converter from a line should nominally be that 
required by the converter for best operation as a synchronous 
converter. Unless designed for the purpose, a synchronous 
converter should not be used to correct low powgr-factors. 
due to other apparatus on the circuit. 


In the above considerations only general reference has been 
made to the cost of synchronous motors for regulating pres- 
sure and power-factors It is difficult to give even approxi- 
mate figures for relative costs of such apparatus As inti- 
mated before, there is some mean speed or number of poles 
which will be the most suitable for giving a certain maximum 
current on short circuit. For speeds slightly above or below 
such mean speed, the cost of the synchronous motor should 
vary almost in proportion to the speed, provided the maximum 
short-circuit current can be diminished somewhat at the same 
time If the speed is further increased or further decreased, 
the cost will tend to approach a constant figure As the ex- 
treme conditions are approached, the cost will begin to rise. 
The above assumptions are on the basis of continuous opera- 
tion at a given current capacity, this being the same in all cases. 
The above assumption is on the basis of decrease in the max- 
imum short-circuit current, as the machine departs from the 
mean, or best speed. If the same maximum current is re- 
quired, then the lowest cost should be at the mean or best 
speed, while at either side the cost should rise. 

It is evident that it would be difficult to give any figures on 
relative costs of such apparatus. The machine for the best 
or mean condition, should cost practically the same as an alter- 
nating-current generator of the same speed, output, and short- 
circuit characteristics. As this speed would probably be some- 
what higher than usual generator speeds, the cost of such 
machine would therefore be somewhat lower. This cost would 
be to a considerable extent, a function of the current on short 
circuit for a given rated capacity of machine. As mentioned 
before, in giving a table of limiting speeds and short circuits, 
it is probable that one-half this limiting speed would be near 
the best condition. Such machines would probably cost from 
60% to 80% as much as similar machines for usual commercial 
high-speed conditions, neglecting turbo-generator practice. The 
frequency has considerable effect on this, as, for example, there 
is small choice of speed as regards high-speed 25-cycle machines. 
Taking very general figures only, it is probable that in the 
case of a given capacity of machine for say three or four times 
full-load current on short circuit the cost cannot be expected 
to be lower than one-half that of machines of similar rating at 
ordinary commercial speeds, turbo-generator practice being ex- 
cluded. The costs in general should approximate more nearly 


those of turbo -generators, but again, an exact comparison 
cannot be made because in usual practice the turbo-generators 
do not give three to four times full-load current on short circuit. 

There are a number of other conditions in this general problem, 
such as advantage or disadvantage of placing synchronous 
motors in the main power-house, or distnbuting them m a 
number of sub-stations Also there is the question of the eSect 
of the cost on the generating plant when used with such regu- 
lating synchronous motors. If higher power-factors are main- 
tained on the transmission system and generator, a cheaper 
form of generator can probably be used. The high power- 
factor permits a larger output from the transmission system 
and thus represents a gam. If the synchronous motor can be 
operated at its best speed and also do work, then there is a 
further gam If the synchronous motor should be located * at 
the center of power distribution, and the power is" distributed 
through induction motors, then there is a possibility of re- 
ducing the cost of such motors by lowering the power-factor, 
this being compensated for by the synchronous motor deliver- 
ing leading currents As the cost per horse power of small 
motors will be much greater than the cost per horse power of 
a large regulating motor, there is a possibility of gain from this 
source. If the induction motors are distributed over wide 
territory, this gain would "be lessened and might disappear. 

It should be mentioned that the powers-factor of a system 
as influenced by difference in wave form has not been con- 
sidered in the preceding discussion. It is obviously impossible 
to neutralize by a synchronous motor the effect of currents 
in a system due to difference in wave form. Such currents will 
in general be of higher frequency than the fundamental wave 
of the system, and the synchronous motor obviously could not 
correct for them, unless it impressed upon the system opposite 
waves of the same frequency. This would mean a synchronous 
motor with a different wave form from that of the system. 

The power-factor of a system will also be affected by any 
hunting of the apparatus on the system. It is evident that 
the synchronous motor could not correct or neutralize such 
effects, except through exerting a damping effect on the system 
And other apparatus on the system. A synchronous motor 
with heavy dampers can reduce the hunting in a system, but 
such hunting can also be damped by induction motors with 
low -resistance secondaries, especially if of 'the cage type. This 


correcting effect should therefore be credited to the damper 
rather than to synchronous -motor action. There are a number 
of other questions which arise in connection with this regu- 
lating feature of the synchronous motor, but the subject is too 
broad to permit even mention of them. 

The substance of the preceding statements can be summarized 
as follows: 

1. A synchronous motor can be-used to establish leading or 
lagging currents in its supply system by suitable field adjust- 
ment, and can thus affect or control power-factor or phase 
relations of the current in the alternating current system 

2. A synchronous motor will set up leading or lagging cur- 
rents in its supply system if its field strength is held constant, 
and the pressure of the supply system is varied above or below 
that generated by the synchronous motor Such leading or 
lagging currents in the supply system will tend to vary the 
pressure of the system A synchronous motor can thus act 
as a regulator of the pressure of its supply system. 

3. This regulating action is greatest with synchronous motors 
which have the closest true inherent regulation (as indicated by 
high field magnetomotive force compared with the armature 
magnetomotive force) in distinction from machines which have 
close apparent regulation obtained by saturation of the mag- 
netic circuit. 

4. If the synchronous motor is used both for regulating the 
power-factor for neutralizing the effect of other apparatus on 
the circuit, and for regulating or steadying the pressure of 
the supply system, its normal capacity for regulating will be 

5. The most suitable speeds for best electrical conditions 
will in general be considerably below highest possible speeds 
as limited by mechanical conditions. 

6. Heavy dampers will increase the effectiveness of the reg- 
ulating tendency. 

7. If the synchronous motor can be used for power purposes 
as well as for regulation, its apparent capacity is increased. 
This is due to the fact that the regulation is obtained by means 
of a wattless component and the power from the energy com- 
ponent, and the algebraic sum of these two is greater than their 
resultant which fixes the current capacity of the machine. 

8. Synchronous converters in general are not suited for reg- 
ulating the pressure or controlling the power-factor of an alter- 

t svstem. 


9. The costs of synchronous motors for regulating purposes 
will in general be lower than for alternating-current motors 
or generators of customary speeds, and will approach more 
nearly -to turbo-generator practice 


FOREWORD In 1902, the author undertook the construction of 
10 000 cycle per second alternator. This problem was a very 
new and radical one at that time and it was considered worth 
while to put the record of results m permanent form. There- 
fore, this paper was prepared on the subject and presented before 
the American Institute of Electrical Engineers in May, 1904. 
This is interesting merely as a record of a relatively early 
construction. (ED ) 

IN the early part of 1902, M. Leblanc, the eminent French 
engineer, was in this country, and spent considerable time at 
the Westinghouse Electric & Manufacturing Company's works at 
East Pittsburg. M. Leblanc was very much interested in cer- 
tain special telephone work, and in connection with such work 
he desired for experimentation a current of very high frequency. 
He took up with the writer the question of building a successful 
alternator for generating current at frequencies between 5000 
and 10 000 cycles per second. He was informed that the ma- 
chine would necessarily be of very special construction, but that 
it was not an impossible machine. Later he took up the 
matter with Mr. Westinghouse, who, upon receiving satisfactory 
assurance that such a machine was possible, advised that the 
generator be built. A preliminary description of the general 
design was given M. Leblanc before he returned to Paris. He 
was somewhat surprised at certain of the features proposed, 
especially at the fact that an iron-cored armature was consid- 
ered feasible for a frequency of 10 000 cycles per second. 

The machine was designed and built on practically the lines 
of the preliminary description furnished M. Leblanc. The fre- 
quency being so abnormal, the writer believes that many features 
in the machine, with the results obtained, will be of scientific 
interest, and therefore the data of the machine, and the tests 
obtained are presented herewith 

The starting point in this machine was the sheet-steel to be 
used in the armature. No direct data were at hand showing 
losses in sheet-steel at such high frequencies, nor was there 'at 



hand any suitable apparatus for determining such losses. As 
preliminary data, tests at frequencies up to about 140 cycles 
per second were used and results plotted in the form of curves; 
these results were plotted for different thicknesses of sheet-steel. 
Also, tests were obtained showing the relative losses due to 
eddy currents and hysteresis, and these were plotted, taking 
into account the thickness of the sheets. These data were not 
consistent throughout; but the general shape of the curves was 
indicated, and in this way the probable loss at the frequency 
of 10 000 cycles per second was estimated for the thinnest 
sheet-steel which could be obtained. The steel finally obtained 
for this machine was in the form of a ribbon about 2 in. wide, 
and about 0.003 in. thick, which was very much thinner than 
any steel used in commercial dynamos or transformers, which 

varies from 0,125 to 0.0280 inch. Therefore the machine had to 
be designed with the intention of using this narrow ribbon of 
steel for the armature segments. 

A second consideration of great importance in the construc- 
tion of such a machine is the number of poles permissible for 
good mechanical construction. For instance, at 3000 revolu- 
tions which was adopted as normal speed the number of 
poles required is 400 for 10 000 cycles per second. The fre- 
quency, expressed in terms of alternations per minute, multi- 
plied by the pole-pitch in inches, gives the peripheral speed in 
inches. At 1 200 000 alternations per minute (or 10 000 cycles 
per second) and a pole pitch of 0.25 in., for example, the peri- 
pheral speed of the field will be 25 000 feet per minute. It 
was therefore evident that either a pole construction should be 


adopted which would stand this high peripheral speed, or the 
pole-pitch should be less than 0.25 in. It was finally decided 
that an inductor type of alternator would be the most convenient 
construction for this high frequency; with the inductor type 
alternate poles could be omitted, thus allowing 200 pole projec- 
tions, instead of 400. The field winding could also be made 
stationary instead of rotating, which is important for such a 
high speed. This construction required a somewhat larger ma- 
chine for a givan output than if the usual rotating type of 
machine were adopted; but in a machine of this type where 
everything was special, the weight of material was of compara- 
tively little importance, and no attempts were made to cut the 
weight or cost of the machine down to the lowest possible limits. 

The following covers a general description of the electrical 
and magnetic features of the machine. 

Armature. The armature was built up in two laminated 
rings dovetailed into a cast-iron yoke, as indicated in Fig. 1. 

^" Ffe-B 

The laminations were made in the form of segments dovetailed 
to the cast-iron yoke (Fig. 2). Special care was taken that the 
laminations made good contact with the cast-iron yoke, as the 
magnetic circuit is completed through the yoke. 

The armature sheet-steel consisted of plates of 0.003 in. 
thickness. The sheet-steel was not annealed after being re- 
ceived from the manufacturer; it was so thin that to attempt 
annealing was considered inadvisable. To avoid eddy currents 
between plates each segment was coated with a thin paint of 
good insulating quality. This painting was a feature requiring 
considerable care and investigation, as it was necessary to obtain 
a paint or varnish which was very thin, and which would adhere 
properly to the unannealed laminations. These laminations 
had a bright polished appearance quite different from that of 
ordinary steel. They were so thin that the ordinary paint or 
varnish used on sheet-steel made a relatively thick coating, 
possibly almost as thick as the plates themselves. A very thin 
varnish was finally obtained which gave a much thinner coating 
than the plate itself, so that a relatively small part of the arma- 
ture space was taken up by the insulation between plates. 


Each armature ring or crown has 400 slots. Each slot is 
circular and 0625 inch diameter "(Fig 3) There is 0.03125 
inch opening at the top of the slot into the air-gap, and the 
thickness of the overhanging tip at the thinnest point is 03125 

g. 4 

The armature winding consists of No 22 wire, B. & S gauge, 
and there is one wire per slot. The entire winding is con- 
nected in series (Fig 4) The measured resistance of the wind- 
ing is 1.84 ohms at 25 cent. 

After the sheet-steel was built up in the frame, it was ground 
out carefully. The laminations were then removed, all burred 
edges taken off and the laminations again built up in the frame. 
The object of this was to remove all chances of eddy currents 


between the plates due to any filing or grinding. The finished 
bore of the armature is 25.0625 inch. 

Field or Inductor. This was made of a forged-steel disc 
25 in. diameter turned into the proper shape, and the poles 
were formed on the outside by slotting the periphery of the ring. 
The general construction is indicated in Figs, 1 and 5. The 
poles were 125 in wide and about 75 in. long radially and 
were round at the pole-face. Fig. 6 shows the general dimen- 
sions of a pole. 

The field winding consisted of No 21 wire, B. & S. gauge. 
There were 600 turns total arranged in 30 layers of 20 turns per 
layer. The field coil after being wound was attached to a light 
brass supporting ring. The general arrangement of the field or 
inductor, armature yoke, and bearings, is as indicated in Fig 1. 
The measured resistance of the field winding is 53 8 ohms at 
25 cent. 

Tests. The machine was designed primarily for only a small 
output, but was operated on temporary test up to 2 kw. A 
series of curves were taken at 500, 1000, 1500, 2000, 2500, and 
3000 revolutions, giving frequencies from 1667 to 10 000 per 
second. At each of the above speeds, saturation curves, iron 
losses, and short-circuit tests were made. Friction and wind- 
age were also measured at each speed. 

On account of the high frequency, the machine was worked 
at a very low induction; consequently there is an extremely 
wide range in pressure, the normal operating pressure being 
taken at approximately 150 volts. 

On curve sheet No. 1, the saturation curves for the various 
speeds are given. These curves check fairly well, the pressure 
being practically proportional to the speed with a given field 
charge. This is to be expected at the lower speeds, but it was 
considered possible that at 3000 revolutions the air-gap might 
be slightly lessened, due to the expansion of the rotor under 
centrifugal action; and it was also thought that eddy-current 
loss due to the high frequency might affect the distribution of 
magnetism at the armature face, but the armature iron losses 
were comparatively small, and there appeared to be no such 
effect. Also there appears to be no effect due to expansion at 
high speed. The air-gap specified for this machine is 03125 in/ 
on each side or 0.0625 in total gap. A very small varia- 
tion in the diameter of the inductor or the bore of the armature 



would make a relatively large per cent, in the effective air-gap, 
Therefore no reliable calculations can be made on the saturation 
curves of this machine based upon the specified air-gap. 

Curve sheet No. 2 shows the iron losses at various speeds 
from 500 to 3000 rev. per min. 1667 to 10 000 cycles per second. 
These losses are plotted in terms of watts for a given exciting 



LO 1.2 1.4 
Field Amperes 

current. These curves show a rather unexpected condition as 
regards the losses. According to the original data showing the 
relative losses due to eddy currents and hysteresis, the eddy- 
current loss even with these thin plates should have been much 
higher than the hysteresis loss, but these iron-loss curves sho^v 



losses with a given field charge almost proportional to the fre- 
quency, which is the ratio that the hysteresis loss alone should 
show. As the eddy-current loss varies as the square of the 
frequency, the writer expected this to be a large element in 
the total iron loss, especially at the higher inductions. 
The six curves shown on this test-sheet are fairly consistent 




10000 Cycles per Second 

Iron-Loss Tests 

0.2 0.4 

0.8 LO lj> 1.4 
Field Amperes 


with each other, but it should be remembered that in making 
measurements of such abnormal apparatus little discrepancies 
in the curves could easily creep in. For instance, in the satu- 
ration curve a series of experiments were first made to find 
whether usual types of voltmeters were satisfactory, and a num- 



ber of different methods for checking these readings were used; 
In determining the iron losses in curve sheet No. 2, the machine 
was driven by a small motor and the losses measured with difn 
ferent field charges. Under most conditions of test the iron los$ 
was a small element of the total loss, and therefore slight varia-. 
tions in the friction loss would apparently show large variations 


10000 Cycles per Second 

Short-Circuit Tests 

Field Amperes 

in the iron losses. Also the fly-wheel capacity of the rotating 
part of the alternator was comparatively high. Therefore, if 
there are any variations in the circuits supplying the driving 
motor, there would tend to be considerable fluctuations in the 
power supplied. Considering all the conditions of test, the 
curves appear to be remarkably consistent. 



Curve sheet No. 3 shows the short-circuit curves at speeds 
of 1000, 2000, and 3000 rev. per min., or frequencies of 3333, 
6667, and 10 000 cycles per second, respectively. It should be 
noted that at a given frequency the short-circuit current is pro- 
portional to the field current over the entire range measured 


rev. permia. 




tut that the short-circuit current is not the same for the same 
field current at the various frequencies. According to these 
curves the current on short circuit increases somewhat with the 
given field charge as the frequency is increased. 

Curve sheet No. 4 shows the measured windage and friction 
losses plotted at speeds from 500*to 3000 rev. per min. This 


curve indicates clearly that the windage is the principal friction 
loss at the higher speeds. The writer has added two curves, 
one showing the estimated bearing friction loss, and the other 
the estimated windage, based upon the assumption that the 
bearing friction varies directly as the revolutions and the wind- 
age loss with the third power of the revolutions. The small 
circles lying close to the measured loss curve show the sum of 
these estimated losses, and the agreement with the measured 
loss is fairly close over the entire range. 


117 11*? 

10000 Cycles per Second 
Regulation Test artOOOO Cycles per Second 

No 2 


t eiiLf 







*~ *- 






&o 1 Duncan 

c Field 




OLD loo 











Ampei-es Load 

Curve sheet No. 5 shows regulation tests made at ISO volts. 
The power-factor of the load on this test was not determined, 
and it was extremely difficult to make accurate measurements. 
The load consisted of incandescent lamps and the wiring from 
the machine to the lamps was non-inductive for the usual fre- 
quencies; but at the abnormal frequency of 10 000 cycles per 
second it is more difficult to obtain a true non-inductive load 
with ordinary apparatus. The tested regulation indicates that 
the load was practically non-inductive. 

In first undertaking tests on this machine there was consid- 


erable difficulty in measuring the pressures. It was found that 
at a frequency of 10 000 cycles per second the Weston voltmeter 
did not work satisfactorily. Practically the same deflection 
was obtained on the high and low scales of a 60-120 volt Weston 
alternating-current direct-current voltmeter with the same 

Very good results were obtained by the use of a form of 
static voltmeter devised by Mr. Miles Walker. This voltmeter 
is of the same form as the static wattmeter described by Mr. 
GINEERS, May 1902,* Tests were also made with the Cardew 
hot-wire voltmeter with the high frequencies, and the results 
checked very satisfactorily with the static voltmeter. 

For measuring the current a current dynamometer was used 
which had wood upright supports and a celluloid dial. The 
only metal parts outside of the copper coils were brass screws. 
It was found that the current dynamometer is not affected by 
frequency, unless there are adjacent metal parts in which eddy 
currents can be generated which react upon the moving element. 
The dynamometer used had but a few turns in order to reduce 
the pressure drop across it. This dynamometer "was checked 
very carefully at different frequencies and apparently gave 
similar results for any frequency between 25 and 10 000 cycles. 

Several temperature tests were made on this machine. The 
heaviest load on any test was 13.3 amperes at 150 volts, or 
2-kw. output. This test was of two hours' duration, and at the 
end the armature iron showed a rise of 16 cent.; the armature 
copper 21 cent, by resistance, and the field copper 17.3 cent. 
Air temperature 19 cent. The machine showed a relatively 
small increase in temperature at this load over the temperature 
rise with one-third this load. This was probably due to the 
fact that the windage loss was so much higher than the other 
losses of the machine that the temperature was but little affected 
by the small additional loss with increase in load. 

Attempts were made to utilize the current from this machine 
for various experiments, but difficulty was at once found in 
transforming it. At this high frequency no suitable iron-cored 
transformer was available. Transformers with open magnetic 
circuits were tried and operated better than those with iron cores 
but were still rather unsatisfactory. It was decided that nothing 
could be done in this line without building special transformers. 
[TRANSACTIONS of the A. I. E. E^ Vol. xix. p. 1035J 


Among the few experiments made was that of forming an arc 
with current at this high frequency. This arc appeared to be 
like an ordinary arc so far as the light was concerned, but had 
a very high-pitched note corresponding to the high frequency. 
This note was very distressing to the ears. 

This machine is in reality of the nature of a piece of labora- 
tory apparatus; and at present it has no commercial value. It. 
was designed primarily for scientific investigation, and appears* 
to be a very good machine for that purpose. 



FOREWORD This paper was presented before the Philadelphia 
Section of the American Institute of Electrical Engineers in 
February, 1908. It describes, as simply as possible, the general 
construction and characteristics of compensated series single- 
phase motors (ED.) 

THE broad statement may be made that it is no more difficult 
to commutate an alternating current than an equal direct 
current Such a statement would appear to be entirely contrary 
to the usual experience, but a little study of the matter will show 
where the apparent discrepancy lies. In commutator type alter- 
nating-current motors, as usually built, a relatively large number 
of commutator bars pass off under the brush during one alternation 
of the supply current. While the current supplied is varying 
from zero to maximum value and back to zero, possibly 50 bars 
have been passed under the brush, and therefore 50 coils in 
the armature have been reversed or commutated. Some of 
these reversals occur at the top of the current wave which has 
a value of about 40% higher than the mean or effective value 
which is read by the ammeter. The motor is therefore at times 
commutating 40% higher current than that indicated by the 
instruments. It is thus evident that in comparing the com- 
mutation of 100 amperes direct-current with 100 amperes 
alternating-current we should actually compare the direct- 
current with 141 amperes alternating. In other words, for com- 
mutating equal currents alternating-current or direct-current, 
the alternating-current ammeter should register only 71% as 
much current as the direct-current. Another way of expressing 
it is that we have to commutate the top or maximum of the 
alternating-current wave, while our instruments only record 
the mean value. 

If the above represented the only difference between the 
alternating current and direct current the problem to be solved 
in commutation of alternating current would not be serious. 




However, the current to be commutated by an alternating- 
current motor is not merely the working current supplied tc 
the motor and measured by the ammeter, but there is, in addi- 
tion, a current which is generated m the motor itself, both a1 
standstill and during rotation, which has to be reversed or com- 
mutated along with the working current. It is this latter cur- 
rent, usually called the local or short-circuit current, which has 
been the source of greatest trouble in commutating alternating 
current; for this short-circuit current may have a value any- 
where from three to ten times the working current, depending 
on the design of the machine. Therefore in comparing the com- 
mutation of an alternating current, as indicated by an ammeter, 









1 * ! f ; 





FIG. 1 

with an equal direct current, we should, in reality, consider 
that the alternating-current motor is commutating a maximum 
current from five to ten times the value of the indicated current 
Furthermore, it would not do to reduce the ammeter current to 
one-fifth or one-tenth value in order to compare commutation 
with direct current, because by so doing we would simply be 
reducing the small applied component of the total current 
commutated by the brushes, the local or short-circuit current 
still retaining a rather high value. In order to compare with 
direct-current commutation, it would be necessary for the 
total maximum of the combined supply and the short-circuit 
current to be reduced to the same value asr direct current. 


It is the local current in the armature turn short-circuited 
by the brush which is the source of practically all the trouble in 
commutating alternating currents Fig 1 illustrates a portion 
of the field and armature structure of a commutator type 
alternating-current motor. It will be noted that the armature 
conductor, which is in the neutral position between poles, sur- 
rounds the magnetic flux from the field pole, just as the field 
turns themselves surround it. The field flux being alternating, 
this armature turn will have set up in it an electromotive force 
of the same value as one of the field turns. Short-circuiting 
the two ends of this armature turn should have the same effect 
as short-circuiting one of the field turns, which is the same 
thing as short-circuiting a turn on a transformer. Such a short- 
circuited turn, if of sufficiently low resistance, should have as 
many ampere-turns set up in it as there are field ampere-turns 
In single-phase motors of good design the field ampere-turns 
per pole are about twelve to fifteen times the normal ampere- 
turns in any one armature coil. Therefore, if the armature coil 
in the position shown in this Fig. 1 should have its ends closed 
on themselves the current in this coil would rise to a value of 
twelve to fifteen times normal In reality, it would not rise 
quite this much, because this armature turn is placed on a 
separate core from the field or magnetizing turns with an air- 
gap between, so that the magnetic leakage between the primary 
(or field winding) and this armature (or secondary winding) 
would tend to protect this coil somewhat, just as leakage between 
the primary and secondary windings of a transformer tends to 
reduce the secondary electromotive force and current. Also, 
this armature coil is embedded in slots, thus adding somewhat 
to its self-induction, and tending further to reduce the short- 
circuit current. In consequence, with its ends closed together 
the current in this armature coil would probably not rise more 
than ten to twelve* times above normal value under any con- 
dition. It is evident, therefore, that if the brush shown in 
Fig. 1 as bridging across two commutator bars to which the 
ends of this coil are connected is of copper or other low-resistance 
material, then there could be an enormous local current set up 
in the coil when thus short-circuited by the brush. This local 
current of about ten times the normal working current would 
have to be commutated as the brush moves from bar to bar, 
and therefore the operation of the machine would be similar to 
that of a direct-current motor if overloaded about ten times 
in^ current. In other words, there would be vicious sparking. 


Even if the low-resistance brush were replaced by one of 
ordinary carbon, the short-circuiting current would still be rela- 
tively high, due to the fact that it is not possible to make the 
brush contact of very high resistance by reducing the size or 
number of the brushes, because these -same brushes must carry 
the working current supplied to the motor, and there must be 
brush capacity sufficient to handle this current. This brush 
capacity will, in practice, be of such amount that the resistance 
in bridging from one bar to the next is still rather low, although 
much higher than if a copper brush were used. Experience 
shows that with not more than four or five volts generated in 
this short-circuited coil by the field flux, the resistance of the 
carbons at the contact with the commutator would be such that 
a short-circuit current of three to fouf times the normal working 
current in the coil can still flow. Therefore, if the motor were 
equipped with carbon brushes and had but four or five volts 
generated in the short-circuit coil, the motor would have to- 
comrnutate the main or working current and also a short-circuit 
current of possibly three times the amount. This short-circuit 
current would also have a maximum or top of its current wave. 
Assuming 100 amperes as the current supplied to the motor, 
the machine therefore actually commutates a supply current of 
141 amperes and an additional short-circuit current of possibly 
three times this value, or from 400 to 500 amperes; therefore, 
the motor actually comtoutates the equivalent of about 600' 
amperes direct current when the alternating-current ammeter 
is reading 100, It is evident from this that any one who tries 
to commutate alternating current with an ordinary type of 
commutating machine would at once draw the conclusion that 
alternating current in itself is very difficult to commutate, 
naturally overlooking the fact that it is the excessive "current 
handled by the brush that is back of the trouble, and not the 
current indicated by the ammeter. 

From what has been stated, it is evident that the excessive 
local current is back of the difficulty in commutating alternating 
current. All efforts of designers of alternating-current com- 
mutator motors have been in the direction of reducing or elimina- 
ting this local current. The present success of the motor, in 
the various forms brought out, is largely due to the fact that 
this current has been successfully reduced to so low a value that 
it does not materially add to the difficulties of commutating the 
main current. No successful method has yet been practically 


developed for entirely overcoming the effects of this short-circuit 
current under all conditions from standstill to highest speed. 
Some of the corrective methods developed almost eliminate this 
current at a certain speed or speeds, but have little or no cor- 
rective effect under other conditions; other methods do not effect 
a complete correction at any speed, but have a relatively good 
effect at all speeds and under all conditions. The former 
methods would appear to be applicable to motors which run at, 
or near, a certain speed for a large part of the time; the latter 
method would be more applicable to those cases where the motor 
is liable to be operated for considerable periods with practically 
any speed from standstill to the highest. While several methods 
have been brought forward for correcting local current when 
the motor has obtained speed, yet up to the present time but 
one successful method has been developed for materially re- 
ducing this current at standstill or very low speeds. It may 
be suggested that the short-circuit voltage per coil be reduced 
to so low a value, say four or five volts, that the local current 
is not excessive and does not produce undue sparking. This 
would certainly reduce the sparking difficulty, but is open to the 
very great objection that the capacity of the motor is directly 
affected by a reduction in the short-circuit voltage. This voltage 
per turn in the armature coil is a direct function of the value of 
the alternating field-flux and its frequency. Assuming a given 
frequency, then the short-circuit voltage is a direct function of 
the induction per field pole, and the lower the short-circuit volt- 
age the lower must be the field flux. But the output of the 
machine, or the torque with a given speed, is proportional to 
the product of the field flux per pole by the armature ampere- 
turns. In a given size of armature the maximum permissible 
number of ampere-turns is pretty well fixed by mechanical and 
heating considerations, and therefore with a given armature 
the torque of the motor is a direct function of the field flux. 
Using the maximum permissible armature ampere-turns, the 
output of a given motor would be very low if the field flux were 
so low that the short-circuit voltage would not be more than 
three or four volts. Increasing the field induction, and there- 
fore increasing the short-circuit voltage, increases the output. 
Experience shows that on large motors, such as required for 
railway work, the induction per pole must necessarily be so high 
that the electromotive force in the short-circuit coil must be 
about double the figure just given; therefore, with such heavy 



flux the short-circuited current will necessarily be excessive 
unless some corrective means is used for reducing it. 

I will consider the standstill or low-speed conditions first 
For this condition only one practical arrangement has so far 
been suggested for reducing the local current to a reasonably 
low value compared with the working current. This method 
involves the use of preventive leads, or, as they are sometimes 
called, resistance leads These consist of resistances connected 
between the commutator bars and the armature conductors. 
Fig. 2 illustrates the arrangement. The armature is wound 
like a direct-current machine, except that the end of one arma- 
ture coil is connected directly to the beginning of the next 




FIG 2 

without being placed m the commutator Between these con- 
nections separate leads are carried to the commutator bars, and 
in these leads sufficient resistance is placed to cut down the 
short-circuit current. The arrangement is very similar in effect 
to the preventive coils used in connection with step-by-step 
voltage regulators which have been in use for many years. In 
passing from one step to the next on such regulators, it is common 
practice to introduce a preventive coil or resistance in such a 
way that the two contact bars are bridged only through this 
preventive device. 

In. an armature winding arranged in this way, the working 
current is introduced through the brushes and the leads to the 
armature winding proper. After entering the winding, the 


current does not pass through the resistance leads because the 
connections between coils are made beyond these leads. In 
consequence, only a very small number of these leads are in 
circuit at any one time, when the armature is in motion all the 
leads carry current in turn so that the average loss in any one 
lead is very small. As the brush generally bridges across two 
or more commutator bars, there is usually more than one lead 
in circuit, but generally not more than three. When the brush 
is bridging across two bars, there is not only the working cur- 
rent passing into the two leads connected ta these two bars, 
but there is the local current, before" described, which passes in 
through one lead, through an armature turn, then back through 
the next lead to the brush. There are losses in these two leads 
due to these two currents. By increasing the resistance, the loss 
due to the working current is increased, but at the same time 
the short-circuit current is decreased. As the Toss due to this 
latter is equal to the square of the current multiplied by the 
resistance, it is evident that increasing this resistance will cut 
down the loss due to the local current in direct proportion as the 
resistance is increased. When the working current is much 
smaller in value than the short-circuit current, an increase in 
the resistance of the leads does not increase the loss due to the 
working current as much as it decreases the loss due to the 
short-circuit current. Both theory and practice show that 
when the resistance in the leads is so proportioned that the 
short-circuit current in the coil is equal to the normal working 
current, the total losses are u minimum. Calculation, as well 
as experience, indicates that a variation of 20% to 30% at either 
side of this theoretically best resistance gives but a very slight 
increase in loss, so there is considerable flexibility in the adjust- 
ment of this resistance. The resistance of the brush contacts 
and of the coil itself must be included with the resistance of the 
leads in determining the best value. In practice it is found that 
with ordinary medium-resistance brushes, the resistance in the 
leads themselves should be about four or five times as great as 
the resistance in the brush contact and the coil; that is, we 
usually calculate the total necessary resistance required and 
then place about 70% or 80% of it in the leads themselves. 
When leads of the proper proportion are added to the motor, it 
is found that practically twice as high field flux can be used as 
before with the same sparking and burning tendency as when 
the lower flux is used without such leads. But even with six 


to eight volts per commutator bar as a limit, we are greatly 
handicapped in the design of the motors, especially when the 
frequency is taken into account. This limited voltage between 
bars also indicates at once why single-phase railway motors are 
wound for such relatively low armature voltages. Direct- 
current railway motors commonly use from 12 to 20 volts per 
commutator bar, or from 2 to 2.5 times the usual practice on 
alternating-current motors. With this low voltage between 
bars in alternating-current machines, with the largest practic- 
able number of bars, the armature voltages become 200 to 250,. 
or about 40% of the usual direct voltages. The choice of low 
voltage should, therefore, not be considered as simply a whim 
of the designers; it is a necessity which they would gladly 
avoid if possible. 

Assuming preventive leads of the best proportions, let us 
again compare the current to be commutated in an alternating- 
current motor with that of the direct-current. Considering 
the ammeter reading as 100, the working alternating current 
has a maximum value of 140 and in addition there is a short- 
circuit current of same value. Even under this best condition, 
the alternating-current mo tor must commutate a current several 
times as large as in. the corresponding direct-current motor. 
The design of such a motor, therefore, is a rather difficult prob- 
lem, even under the best conditions. . 

While resistance leads theoretically appear to give the most 
satisfactory method for 'obtaining good starting and slow- 
speed running conditions, yet other methods have been pro- 
posed. The only one of any practical importance is that in 
which the short-circuit voltage is reduced at start and at slow 
speed by sufficiently reducing the field induction. As this 
reduced field induction would give a proportionately reduced 
torque, it is necessary at the same time to increase the armature 
ampere-turns a corresponding amount above normal. This is 
only a part solution of the problem, however, for the decrease 
in short-circuit current by this means is partly offset by the 
increase in the working current, so that the total current to be 
commutated is not reduced in proportion to the field flux. 
Where the period of starting and slow running is very short> 
this method is fairly successful in practice. However, with this 
arrangement it is rather dangerous to hold the motor at stand- 
still for any appreciable length of time, for in such a case the 
large short-circuit current is confined to a single coil and the 


effect is liable to be disastrous if continued for more than a very 
short period. With this method of starting, the total current 
handled by the brushes will usually be at least two to three 
times as great as when preventive leads are used. 

The preceding statements refer mainly to starting or slow- 
speed conditions. When it comes to full-speed conditions, 
however, there are various ways of taking care of the commuta- 
tion. One of these methods is based on the use of preventive 
leads, as described ; the other methods depend upon the use of 
commutating poles or commutating fields in one form or another. 

It is evident, from what has been said, that at start the pre- 
ventive leads which reduce the short-circuit current to low 
values will also be effective in a similar manner when running 
at normal speed Such a motor with proper proportion of 
leads will, in general, commutate very well at full speed when 
the starting conditions have been suitably taken care of Nothing 
further need be.said of this method except that the tests show 
that the short-circuit current has considerably less value at 
high speed than at start. 

The other methods of commutation at speed, involving corn- 
mutating poles and commutating fields, necessarily depend upon 
the armature rotation for setting up a suitable electromotive 
force in the short-circuit coil to oppose the flow of the short- 
circuit current. As the electromotive force in the short-cir- 
cuited coil is a direct function of the field flux, and is inde- 
pendent of speed, while the correcting electromotive force is a 
function of the armature speed, it is evident that either the 
commutating pole can produce the proper correction only at 
one particular speed, or the strength of this commutating pole 
must be varied as some function of the speed Usually the 
strength of these poles is made adjustable with a limited number 
of adjustments and approximate compensation only is obtained 
on the average. In the Siemens-Schuckert motor the corn- 
mutating poles are of small size and placed between the main 
poles. These are for the purpose of obtaining commutation 
when running. In addition the armature is provided with pre- 
ventive leads for improving the operation at start and at slow 
speed. In the Alexanderson motor, according to published 
description, no separate commutating poles are provided, but 
the edges of the main poles are used as commutating poles, 
the armature coil having its throw shortened until its two sides 
come under the edges of the main poles. In this motor the field 


'is weakened and the armature ampere -turns are increased 
while starting The commutating-pole scheme in this motor is, 
m some ways, not as economical as in the Siemens-Schuckert 
arrangement, as the motor requires a somewhat higher mag- 
netization with a consequent reduction in power-factor The 
Winter-Eichberg motor is quite different in arrangement from 
any of those which I have mentioned I will not attempt to 
describe this motor in full, but will say that it has two sets of 
brushes in the armature, one of which is short-circuited on 
itself, and carries the equivalent of the working current in the 
types I have described, while the other carries the magnetizing 
or exciting current which is supplied to the armature winding 
instead of the field. The arrangement is such as to give prac- 
tically the same effect as a commutating pole or commutating 
field. When starting, the field flux is decreased and the arma- 
ture ampere-turns increased 

All of the above motors are nominally of low armature voltage 
and all of them appear to comrnutate reasonably well at speed. 
Two of them use the full-speed induction at start, while the 
other two use reduced induction and increased armature ampere- 
turns at start. 

There has been considerable discussion during the last year 
or two regarding the most suitable frequency for single-phase 
commutator type motors It may therefore be of interest to 
consider what effect reduction in frequency would have on the 
commutation, output, and other characteristics of the motor. 

The short-circuit voltage, as I have stated before, is a function 
of the amount of field flux and of the frequency. For a given 
short-circuit voltage the induction per pole can be increased 
directly as the frequency is decreased If a certain maximum 
induction per pole is permissible at 25 cycles, then with 12.5 
cycles, for example, the induction per pole may be double, with 
the same short-circuit voltage, This would at once permit 
double output if the saturation of the magnetic circuit would 
permit the doubling of the induction. But on 25-cycle motors, 
as usually built, we work the magnetic flux up to a point just 
on the verge of saturation, so to speak, as indicated in Fig. 3. 
It is evident that double induction, under such conditions, 
would not be practicable unless the 25-cycle motor had been 
worked at an uneconomically low point. However, an increase 
of 30% to 40% in the induction would appear to be obtainable, 
but a large increase in excitation is required. With but 30% 



to 40% higher induction, and with the frequency halved, the 
short-circuit voltage would be but 65% to 70% of that with 
25 cycles or, in other words, the voltage per turn in the field 
coil is but 65% to 70%. As the higher induction raises the 
armature counter electromotive force the field electromotive 
force can be increased in proportion for the same power-factor, 
or can be 30% to 40% higher than with 25 cycles. As the total 
field voltage, therefore, can be 30% to 40% higher, and the 
voltage per field turn is but 65% to 70%, it is evident that the 
number of field turns can be doubled without changing the 






FIG. 3 

ratio of the field inductive volts to the armature electromotive 
force. In other words, the field turns can be doubled if the 
frequency is halved. With the double field turns the field 
excitation can therefore be doubled, which is the requirement 
for the increased induction shown in Pig. 3., It is thus evident 
that halving the frequency will permit higher pole inductions, 
and therefore higher torque and output, with lower short-circuit 
voltage and better commutatmg conditions throughout. Also, 
this higher field induction is not necessarily accompanied by an 
increased iron loss, for the lower frequency of the alternating 


flux compensates for this. On the above basis it may be asked 
why a reduction to 15 cycles is proposed instead of to 12 5, or 
even to 10 cycles. There are several reasons for the choice of 
15 cycles. 

1. The motor can be worked up to so high a saturation at 15 
cycles that there is relatively small gain with a reduction to 
12 5 cycles, which would be about the lowest frequency to con- 
sider when the transformers and other apparatus is taken into 

2 As the torque of the single-phase motor is pulsating in- 
stead of being constant, as in a direct- current machine, there 
is liability of vibration as the frequency of the pulsation is de- 
creased. This effect becomes more pronounced the larger the 
torque of the motor, and is, therefore, of most importance in the 
case of a large locomotive Expenence shows that this ten- 
dency to vibrate can be damped out effectively in very large 
motors with a frequency of 15 cycles, but becomes more difficult 
to suppress as the frequency is further reduced. This is, in 
reality, one of the fundamental reasons for keepmg up to 15 
cycles instead of reducing to 12 5 or lower. 

3. The lower the frequency the heavier the transforming 
apparatus on the car or locomotive. It is probable that with 
12^ cycles instead of 15 cycles, the increase in weight and cost 
of the transforming apparatus would about counter-balance 
the decrease in the same items in the motors themselves, al- 
though the efficiency and power factor of the equipment would 
be slightly better with the lower frequency. 

4. As synchronous converters will be used to some extent m 
connection with the generating plants for single-phase systems 
in order to feed existing direct current railways, the frequency 
of 15 cycles will be slightly more favorable than 12.5 as regards 
cost of the converters and the step-down transformers The 
same will be true if motor-generators are used for transforming 
to direct current, also for induction motors. 

Against the choice of 15 cycles may be cited the fact that 
there are other frequencies which represent a better ratio to 
25 cycles when frequency-changers are to be taken into account. 
A low-frequency railway generating plant may require to tie 
up with some existing 25-cycle or 60-cycle plant; this can be 
done by interposing frequency-changers. Or it may be desired 
to obtain a lower frequency with a single-phase current from 
some existing higher frequency^ polyphase plant, By inter- 


posing the frequency-changer the single-phase railway load will 
not exert any unbalancing effect on the polyphase supply 
circuit, and at the same time the railway circuit can be regulated 
up or down independently of the three-phase generator circuit. 
In case the three-phase plant is operated at 25 cycles, then a 
two-to-one ratio of frequencies; that is, 12.5 cycles on the rail- 
Way circuit, would give the best conditions as regards choice 
of poles and speeds in the frequency-changer sets A five-to- 
three relation is given by 15 cycles, which is not nearly as good 
as the two-to-one ratio. A frequency of 16| cycles would give 
a three-to-two ratio, which represents considerable improve- 
ment over the five-to-three ratio Therefore, this slightly 
higher frequency may prove of advantage in some cases. The 
choice of this frequency, however, does not mean a new line 
of apparatus; for a well designed line of 15-oycle motors tran- 
formers, etc, should operate very well on a 16-cycle circuit 
without any change whatever 

When transforming from 60 cycles, however, the 15 cycle 
gives a four-to-one ratio which is very good, and neither 12,5 
nor 16f cycles is very satisfactory. Therefore this 15-cycle 
frequency represents the best condition in transforming from 
60 cycles, and fairly good conditions for transforming from 25 
cycles; and by operation of 15-cycle apparatus at 16$ cycles a 
very good transformation ratio is obtained from 25 cycles. It 
may be of interest to recall that the old Washington, Baltimore 
and Annapolis Railway, which was the first road contracting 
for single-phase commutator motors, was laid out for 16-J 
cycles. There was considerable criticisms at that time of the 
use of this frequency, but the statement which I have just made 
shows one very good reason for this frequency A second rea- 
son is that 16f cycles per second is 2000 alternations per minute, 
which permits a steam turbine driving a two-pole generator to 
use a speed of 1000 rev. per min., which is a very good one for 
large turbo-generators. 

I have gone into the question of induction and frequency 
as affecting the commutation and torque. I will now take up 
the question of power-factor in the single-phase commutator 
motor In a direct-current motor we have two electromotive 
forces which add up equal to the applied electromotive force , 
namely, the counter electromotive force due to rotation of the 
armature winding in. the magnetic field, and the electromotive 
force absorbed in the resistance of the windings and rheostat. 


In the alternating-current motor there are these two electro- 
motive forces, and there is also another one not found in the 
direct-current machine, namely, the electromotive force of self- 
induction of the armature and field windings due to the alter- 
nating magnetic flux m the motor This inductive electro- 
motive force exerts a far greater influence than the ohmic 
electromotive force for it has much higher values 

The inductive electromotive force lies principally in the main 
field or exciting winding of the alternating- current motor 
There is a certain voltage per turn generated in the field coils, 
depending upon the amount of the field flux and its frequency, 
as stated before. This electromotive force per field turn is 
practically of the same value as the short-circuit electromotive 
force generated in the armature coil, as already referred to. I 
have stated that a short-circuit voltage of three or four volts 
per armature turn gave prohibitive designs and that it was 
necessary practically to double this This means that the field 
coils also have six to eight volts per turn generated in them 
The total number of field turns must, therefore, be very small 
in order to keep down the field electromotive force, for this 
represents simply a choke-coil in series with the armature. If 
the armature counter electromotive force should be 200 volts, 
for instance, which is rather high m practice with 25-cycle 
motors, then a field self-induction of half this value would allow 
about 14 turns total in the field winding. Compare this with 
direct-current motors with 150 to 200 field turns for 550 volts, 
or 60 to 80 turns for 220 volts The alternating-current 25-cycle 
motor, therefore, can have only about 20% to 25% as many 
field turns as the ordinary direct-current motor. This fact 
makes it particularly hard to design large motors where there 
must be many poles. In the single-phase motor the induction 
per pole being limited by the permissible short-circuit voltage, 
it is necessary to use a large number of poles for heavy torques; 
but the total number of field turns must remain practically 
constant on account of the self induction, while in reality the 
number of turns should be increased as the number of poles is 
increased. With a given number of poles we may have just 
sufficient field turns to magnetize the motor up to the required 
point; but if a large number of poles should be required, then we 
at once lack field turns and must either reduce the field induc- 
tion, and thus reduce the output, or must add more field turns 
and thus get a higher self-induction or choking action in the 


field, with a consequent reduction in power- factor Here is 
where a lower frequency comes in to advantage, fdr, as I showed 
before, with the same relative inductive effect, the field turns 
can be increased directly as the frequency is decreased The 
use of 15 cycles thus permits 67% more field turns than 25 
cycles and raises our permissible magnetizing limits enormously 
This problem is encountered particularly in gearless locomotive 
motors of large capacity. For increased capacity the driving 
wheels are made larger, thus permitting a larger diameter of 
motor, the length, axlewise, being fixed. But with increased 
diameter of drivers, the number of revolutions is decreased for a 
given number of miles per hour. With 25-cycle motors we 
soon encounter the above mentioned limiting condition in field 
turns; beyond this point the characteristics of the motor must 
be sacrificed, and even doing this we soon reach prohibitive 
limits By dropping the frequency to 15 cycles, for instance, 
we change the whole situation. The induction per pole can be 
increased and the number of poles, if desired, can also be in- 
creased. The practical result is that, in the case of a high-speed 
passenger locomotive with gearless motors, a 700-h p. 15-cycle 
motor can be got in on the same diameter of drivers as required 
for a 500-h.p. 25-cycle motor. Also a 500-h.p. 15-cycle motor 
goes in on the same drivers a.s a 360-h.p., 25-cycle motor. At 
the same time these 15-cycle motors have better all round 
characteristics than the 25-cycle machines as regards efficiency, 
power-factor, starting, over-load commutation, etc. 

Returning to the design of the motor, there is one other 
electromotive force of self induction which may be considered; 
namely, that generated in the armature winding and in the 
opposing winding in the pole face, usually called the neutralizing 
or compensating winding 

Fig. 4 shows a section of the field and armature corresponding 
to the usual direct-current motor, or an alternating-current 
motor without compensating winding. In the direct-current 
motor the armature ampere-turns lying under the pole face 
tend to set up a local field around themselves, producing what 
is known as cross-induction. This produces no harmful effect 
except in crowding the field induction to one edge of the pole, 
thus shifting the magnetic field slightly and possibly affecting 
the commutation in a small degree. But if the armature is 
carrying alternating current this cross flux will generate an 
electromotive force in the armature winding, and this will be 



added to the field self-induction, thus increasing the self-induc- 
tion or choking action of the machine. As the armature turns 
on such motors are much greater, in proportion, than the field 
turns, it is evident that the ampere-turns under the pole face 
can exert a relatively great cross-magnetizing effect. This high 
cross-magnetization generates a high armature self-induction 
which may be almost as much as the field self-induction. Further, 
this great cross-induction would tend to shift the magnetic 
field quite appreciably, thus affecting the commutation to some 

To overcome this serious objection, the neutralizing winding 
is added. This is a winding embedded in the pole face and so 

o \o\o\o\o\o \o V p IQI ooj o IO/QI o 


FIG. 4 

arranged that it opposes the armature cross-magnetizing action. 
The arrangement is shown in Fig. 5. As it opposes and thus 
neutralizes the cross-induction set up by the armature winding, 
it eliminates the self-induction due to the cross-magnetization. 
It also prevents shifting of the magnetic field and thus eliminates 
its injurious effect on commutation. As the cross-flux is 
practically cut out the armature winding becomes relatively 
non-inductive. There is, however, a small self-induction in 
the armature and neutralizing windings, due to the small flux 
which can be set up in the space between the two windings, 
they being on separate cores with an air-gap between. 

I have stated that the field turns of the alternating-current 
motor can be only 20% to 25% as many as in ordinary direct- 



current practice. It may be questioned how the field caa be 
magnetized with so few field turns. This has been one of the 
most difficult problems in the motor. Obviously, one solution 
would be the use of a very small air-gap, but in railway practice 
there are objections to making the air-gap unduly small. Furth- 
ermore, if the armature has large open slots, as shown in Fig. 6, 
experience shows that a reduction in the clearance between the 
armature and field iron does not represent a corresponding de- 
crease in the effective length of the air-gap, due to the fact that 
the fringing of the magnetic flux from the tooth tip of the pole 
face changes as the air-gap is varied. The most effective con- 
struction yet used consists in making the armature slots of the 
partially closed type as in the secondary of an induction motor. 
This is shown in Fig. 7. 




^"-v V 






^>x X ; 





o /o fo /o /b't o\ o\ o\ o^ 


[ 1 

I 1 I 1 i . \ 

\0 \0 \ \ \0 \0 \pl O/' O y ' O/ O/O / O/ 

\ \ \ \ \ ^_^ y v X / / / 

FIG. 5 

With this construction practically the whole armature surface 
under the pole becomes effective, and the true length of air-gap 
is practically the same as the distance from iron to iron. With 
the increased effective surface, due to this construction, the 
length of air-gap need npt be unduly decreased, which is of con- 
siderable importance in railway work. 

A further assistance in reducing the required field turns is 
the field construction used in the single-phase motor. The 
magnetic circuit consists of laminations of high permeability 
and usually without joints across the magnetic path. The iron 
is also worked either below the bend in the saturation curve or, 
at most, only slightly up on the bend, except in the case of very 
low frequency motors where more field turns are permissible. 


Taking the whole magnetic circuit into account, on 25-cycle 
motors about 80% of the whole field excitation is expended in 
the air-gap, while in direct-current motors, even with a much 
larger air-gap, as much as 40% to 50% of the magnetization 
may be expended in the iron and In the joints, 

This armature construction with the partly closed slots has 
been found very effective in large, slow-speed, single-phase 
motors in which a relatively large number of poles is required. 
This construction is used on the New Haven 250-h p., 25-cycle 

/> / > vj 


motors; also on the 500 h.p., 15-cycle motor on the Pennsylvania 
locomotive exhibited at Altantic City at the Street Railway 
convention, last October. Geared motors for interurban service 
can be constructed with ordinary open slots with bands, and 
many have been built that way. The semi-closed slot, however, 
allows more economical field excitation. 

It may be asked what the objection is to low power-factors 
on single-phase railway motors, aside from the increased watt- 
less load on the generating station and transmission circuits. 
There is an objection to the low power-factor in such motors, 



a very serious one. This lies in the greatly reduced margin for 
overload torque in case the supply voltage is lowered. In 
railway work it is generally the requirement of abnormal loads 
or torques which causes a reduction in the line voltage ; that is, 
the overload pulls down the trolley voltage just when a good 
voltage condition is most necessary. This is true of direct cur- 
rent as well as alternating current. In the direct current 
motor, however, such reduction in voltage simply means reduced 
speed but in the alternating current motor the effect may be 
more serious. 

To illustrate, assume a motor with a power-factor of 90% at 
full load. The energy component of the input being 90%, 

FIG. 7 

the inductive component is about 44% or, putting it in terms of 
electromotive force the inductive volts of the motor are 44% 
of the terminal voltage, Neglecting the resistance of the motor, 
a supplied electromotive force of 44% of the rated voltage 
would just drive full-load current through it and develop full- 
load torque. With full voltage applied the motor could develop 
from five to six times full-lgad torque. Under abnormal con- 
ditions a drop of 30% in the line voltage would still give suffi- 
cient voltage at the motor terminals to develop two and one half 
to three times full-load torque. Let us next take a motor of 
80% power-factor at full load. The inductive voltage would 
then become 60% of the terminal voltage, and therefore 60% 
Of the rated voltage must be applied to send full-load current 


through the motor. This neglects the resistance of the motor, 
which, if included, means that- slightly more than 60% of the 
voltage is required. With full voltage applied, this motor 
would develop about three or four times the rated torque. 
With 30% drop in the line voltage the motor could develop 
from one and one half to two times rated torque, which is hardly 
enough for an emergency condition 

Taking, next, a motor with 70% power-factor at full load it 
would require 70% of.the rated voltage to send full-load current 
through the motor; with 30% drop in line voltage the motor 
could just develop full-load torque, and even with 15% drop it 
would develop only about one and one half times torque. As 
15% drop is liable to occur on any ordinary system", this latter 
motor would be a very unsafe one. 

It is evident from the above that it would be bad practice in 
railway work to install motors with very low full-load power- 
factors. In general, the higher the power-factor the more 
satisfactory will be the service, other things being equal. 

I have endeavored to explain some of the problems which have 
been encountered in the design of single-phase commutator 
railway motors of sizes suitable for all classes of railway service. 
Here is a type of machine which has been known for a great 
many years, but which, until the last few years, has been con- 
sidered utterly bad. In a comparatively short time it has been 
changed from what was considered an unworkable machine to 
a highly satisfactory one and this has been accomplished, not 
by any radically new discoveries, but by the common-sense 
application of well known principles to overcome the apparently 
inherent defects of the type. As an indication that the motor 
is making progress in the railway field, I will mention that the 
first commercial single-phase railway motors have not been in 
use more than four or five years, and yet at the present time 
there have been sold by the various manufacturers in this 
country and Europe, a total capacity of approximately 200,000 
to 250,000 h.p., a very considerable part of which has been put 
in operation. Considering that the motor was a newcomer in a 
well established field, the above record is astonishing. How- 
ever, it may be safely predicted that what has been done in the 
last five years will hardly make a showing compared with what 
will be done during the next five years, for the real field for such 
motors, namely, heavy railway work, has hardly been touched. 


FOREWORD This is part of a discussion by the author, of papers by 
Dr. Steinmetz and Mr. Slichter before the American Institute 
of Electrical Engineers, January 2d, 1904. The major part of 
the author's discussion covered the comparison between the 
series and repulsion type motors, in which, he showed that the 
repulsion type was simply a series type motor with a trans- 
former added. The several references to Mr. Shchter's paper 
given in the discussion have but little bearing on the technical 
matter contained but could not be eliminated without consider- 
able remodeling of the paper. (ED.) 


IN the paper presented before the American Institute of Elec- 
trical Engineers, in September, 1902, the speaker called atten- 
tion to the fact that there were but two types of single-phase al- 
ternating-current motors having suitable characteristics for rail- 
way service; viz., that called the "Series Type/* and the "Repul- 
sion Type." Attention was called to the fact that both 
motors have suitable characteristics for railway service, as 
both automatically give variable-speed characteristics with 
changes in load. That paper primarily described a single- 
phase railway system, and the motor formed but an ele- 
ment in the general system. It was a very general opinion 
at that time that, the success of the commutator type of motor for 
large sizes was doubtful, and the sparking feature was considered 
a fundamental source of trouble. It was generally conceded that 
if a motor with series characteristics could be made to operate suc- 
cessfully, it would be a great step in advance in the railway field. 

Since that time single-phase railway systems have been more 
fully developed. Practically, no departures from the general 
system then indicated have been furnished, and the types of 
motors developed have been along the lines of the two motors 
indicated in that paper. 

Up to the present time the only suitable motors suggested 
for this work have been of the commutator type, and have been 
those having series characteristics. The speaker has suggested 
that all these motors can be considered broadly under the one 
class of series motors, as they all have the series characteristics 



of the direct-current series motor. The speaker further sug- 
gested that they be sub-divided into the "Straight-Series" type 
and the "Transformer-Series" type. The transformer-series 
could also be arranged in two classes; viz., one in which the 
armature or field is supplied by an external transformer, and 
one in which the transformer is placed in the motor itself, this 
latter is the repulsion type of motor. 

Figs 1, 2, and 3 illustrate the three classes. Fig 1 being the 
straight-series, Fig 2 the transformer-series and Fig. 3 the repul- 
sion motor Fig 2 would be considered as a true series motor, 
although the armature and field are not directly in series, yet most 
of the characteristics described as belonging in the repulsion 

FIG 1. 

PIG 2 

FIG 3 

motor apply directly to the transformer motor shown in the 
figure. Comparing the relations of these motors, viz., the 
straight-series and the repulsion motor, we will first take up the 

In this motor, if properly designed, two pressures can be con- 
sidered; viz,, that across the field circuit, and that across the 
armature circuit. The armature pressure can be made practi- 
cally non-inductive so that the input of the armature will repre- 
sent practically true energy. The pressure across the field 
is practically at right angles to the armature pressure, and repre- 
sents very closely the wattless component supplied by the motor. 
The resultant of these two pressures will then be the line pressure. 
The power-factor of the motor when running is represented prac- 


ticaUy by the pressure across the armature winding, increased 
slightly by the losses in the field-core and winding. Therefore, 
for high power-factors it is important that the pressure across 
the armature circuit be made as high as possible, relatively to 
the applied pressure, and that across the field as low as possible. 

There are three ways in which to increase the pressure across 
the armature; viz., by increase in speed, by increase in the 
number of wires in series on the armature, and by increase in flux 
through the armature. 

By increase in speed and increase of the wires in series, the 
armature pressure will be increased without affecting the field 
pressure, and therefore the ratio of the armature pressure to the 
line pressure is increased. Increasing the flux in the armature also 
increases the flux in the magnetizing-coil in the field, and the 
pressures of both are increased. Therefore this increase does not 
improve the power-factor of the machine. 

Instead of increasing the armature pressure, the pressure 
across the field winding may be decreased ; this can be done in two 
ways; viz., by reducing the turns in the field coil, or by reducing 
flux through the coil. Reducing the flux through the field re- 
duces the flux in the armature winding also, and therefore repre- 
sents no gain; reduction in field-turns, therefore, is the feasible 
means of reducing the field pressure. Reduction in field-turns 
can be accomplished in two ways; viz., by decreasing the effec- 
tive length of air-gap in the motor, and by increasing the cross- 
section of gap. By making the gap very small the pressure 
across the field could be made very small compared with the 
line pressure, and extremely high power-factors could be ob- 
tained, whether the motor is of the straight-series or the repul- 
sion type. Also by increasing the section of the air-gap the 
turns of the field can be decreased with a given total flux through 
the coil, and the power-factor can thus be very considerably 
increased. The first method, viz., decrease in gap, is limited by 
practical conditions which have been determined from long ex- 
perience with direct-current work. It should be borne in mind 
when published descriptions of such motors are given, that the 
results, as regards power-factor, generally depend upon data 
which are not given in the description; such as the magnetic 
dimensions of the armature and field, the length of gap, etc. 
Therefore, a machine may be described as showing an extremely 
high power-factor, which may in practice not be a commercial 
machine, from the standpoint of American railway experience. 


Increasing the section of air-gap without decreasing the 
length of gap also improves the power-factor, but makes a larger 
and heavier machine, as a rule. 

Both these modifications reduce the ampere-turns in the field. 
The direction of the improvement in the armature was shown 
to be in increased armature ampere-turns with a given speed. 
It therefore follows that almost any result desired can be ob- 
tained as regards power-factor by increasing the armature am- 
pere-turns and decreasing the field, or exciting ampere-turns. 
Reference will be made to this point in considering the repulsion 

It should be noted that in all these motors there should be but 
little saturation in the magnetic circuit and but few ampere- 
turns expended in saturation of the iron under normal conditions. 
This consequent low saturation in such motors leads to certain 
characteristics in the torque curves which have been cited this 
evening as an indication of superiority of alternating-current 
motors over direct-current motors; namely, a torque increasing 
approximately as the square of the current. In fact, this superi- 
ority of torque should be charged to the low flux-density of the 
motor rather than to the alternating current. If direct-current 
motors were worked normally at as low density as the alternating- 
current motor, then the direct-current motor would show better 
torque characteristics, and would be comparable with the alter- 
nating-current motor. This claim for a better torque in the 
alternating-current motor compared with the direct-current 
motor seems to be making a virtue of a necessity. 

It is evident from what has been said that the power-factor of 
the straight-series motor can be made anything desired, it being 
a question of proportion between armature and field, length of 
air-gap, amount of material used, etc. In practice a compromise 
would naturally be made among the various characteristics, and 
a slight reduction in power-factor is probably of less importance 
than a corresponding reduction in size and weight. Also large 
clearance is probably of more importance than an extremely high 
power-factor at normal load. In practice it will be found that 
the armatures of such motors have a large number of ampere- 
turns compared with the fields, in order to obtain comparatively 
high power-factors with large air-gaps. The number of poles 
need not be made such that the product of the poles by the normal 
speed represents the frequency of the supply circuit; good series 


motors can be made, and have been made, in which the number 
of poles were very much larger or much smaller than represented 
by this relation. 

Taking up next the transformer type of motors Fig. 2; the 
field is in series with the primary of the transformer, the second- 
ary of which is connected to the terminals of the motor. I would 
call this a true series motor, although it is not a straight-series 
motor. In this motor the pressure across the armature can be 
made practically non-inductive and the pressure across the 
primary of the transformer will be practically non-inductive. 
The voltage across the field winding will have practically 90 
phase relation to that across the primary of the transformers, and 
the magnetic field, set up by the field winding, will have a 90 
relation in time to the magnetic field in the transformer, as in 
the repulsion motor. In this motor the voltage across the trans- 
former will be highest at light loads and will decrease with load 
until zero speed is reached. At start there is lowest flux in the 
transformer and highest flux in the field winding. Such a motor 
will have speed-torque characteristics very similar to those of a 
straight-series motor, except as affected by the actions taking 
place in the transformer itself. If the transformer possesses no 
reactance, then at start the current in the armature should be 
the same as if connected as straight-series motor, and the condi- 
tions of torque at start should be the same. If the transformer 
has reactance, then at start the current in the armature will not be 
quite equal to the current which the armature will receive if 
coupled as a straight-series motor, assuming the transformer to 
have a 1 to 1 ratio. Neither will the armature current be ex- 
actly in phase with the field current; therefore the starting 
torque of a motor connected in this way will be slightly less than 
the torque of the same motor if connected in straight-series. 
This is on the assumption that the transformer is one propor- 
tioned for small reactance; but if the primary and secondary 
windings of the transformer should be on separate cores with 
air-gap between, then the reactances of the windings are con- 
siderably greater than in the above case. Therefore, we should 
expect a motor with such a transformer to give still lower torque 
than the straight-series with the same current supplied from 
the line. 

In a repulsion motor the transformer is combined with the 
motor itself and the primary and secondary windings are upon 



different cores with an air-gap between. The starting condi- 
tions of such a motor as indicated above should be poorer than 
the straight-series motor, or for the same starting torque some- 
what greater apparent energy should be required. It stands to 
reason that applying the current directly to the armature wind- 
ing should give greater ampere-turns and better phase relations 
than generating this current in a secondary circuit, and not 
under ideal transformer conditions. The tests which have been 
made, as well as the results shown in the curves of the papers 
given tonight indicate this. It is to be noted that the torque 
curve is not the same shape near the zero speed point as the 
torque curve of the series motor. 

PIG 4. 

FIG 5. 

Series motors and repulsion motors may be indicated in the 
simple form shown in Figs. 4 and 5. In the diagrams of the 
repulsion motor (Fig. 5), two field-poles FF, are shown, and two 
transformer-poles, TT. To obtain high power-factors on such a 
motoi the ampere-turns in T must be very much greater than in 
F, which means that the ampere-turns in the secondary or arma- 
ture are much greater than in the exciting field, as in the series 
motor. The high power-factor obtained with these motors is 
therefore due principally to the small ampere-turns in the field 
and the small pressure across the field. 

For instance, with brushes set at an angle of 16, from the 
primary or resultant field, the ratio of armature to exciting field- 
turns would be almost 5 to 1, a ratio which will also permit of 
extremely high power-factors in well-designed straight-series 


motors over wide ranges of speed. To this feature should be 
credited the good power-factors claimed for the repulsion motor. 
In either the series or repulsion type of motors, high power 
factors, especially at low speeds, are directly dependent upon this 
fact of high ratio of armature to field, and with a high ratio, high 
power-factors should be obtained without crediting the result to 
leading currents in the armature. In the diagram of the repul- 
sion motor, the line current indicated flows through both the 
field winding and the transformer winding. The primary cur- 
rent sets up a magnetic field in the exciting windings in phase 
with the line current. If it also set up a field in the transformer 
in phase with the line current, then the electromotive force gen- 
erated in the armature winding due to rotation would have a 90 
relation to the electromotive force set up by the transformer, and 
a correcting or magnetizing current would flow. This flow is 
in such direction that it corrects the relation between the two 
pressures in the armature by shifting the transformer magnetism 
one-quarter phase later than the exciting field magnetism. This 
armature corrective current may thus be considered as mag- 
netizing the transformer, making the primary input to the trans- 
former practically non-inductive; but this magnetizing or cor- 
recting current may be considered as flowing in a circuit at right 
angles to the field magnetic circuit, and having practically no 
effect on the field circuit. Therefore as a rough approximation, 
the exciting field may be considered to represent the wattless 
component of the input, and the transformer field the energy 
component, as in the series motor. As to the statement that the 
magnetizing current in the armature reduces the wattless com- 
ponent of the exciting field, the speaker does not accept it broadly. 
If this component is reduced, then another component of practi- 
cally equal value is introduced somewhere else, for the power- 
factors obtained with such motors can be accounted for by the high 
ampere-turns in the armature winding, compared with the field 
or exciting ampere-turns. If the armature current improves the 
power-factor by diminishing the magnetizing or exciting field, 
then the curves in Figs. 1 and 4 of Mr. Slichter's paper should 
show it. The speaker has gone over both sets of curves calculat- 
ing the wattless components from the power-factors. From this 
and other data in these curves, he finds that beginning near 
synchronous speed the wattless component in the motor goes tip 
.slightly faster than would be represented by the field excitation, 


assuming it to be entirely wattless. Therefore, according to 
these curves, the power-factors at lower speeds are not quite as 
good as would be obtained by a field entirely inductive and the 
armature entirely non-inductive, in a straight series motor. 
These calculations are rather approximate, as the curves do not 
check at all well with each other. For instance, the output of 
the motor as represented by the input multiplied by the power- 
factor and by the efficiency, does not check with the output as 
represented by the product of speed by torque, in either set of 
curves, the discrepancies being as high as 10 percent. In Fig. 4, 
for instance, either the torque or the speed is too high for the 
lower speeds. Checking back on this curve, using either the 
speed and torque or the power-factor and efficiency for deter- 
mining the output, the speaker finds that the wattless component 
in the motor at 190 revolutions is approximately 20 percent 
higher than it would be if the field excitation alone were wattless, 
assuming at 440 revolutions the wattless component is represented 
purely by field excitation; that is, from 440 to 190 revolutions the 
wattless component is increased 20 percent over that which 
would be represented by field excitation alone. This indicates 
that not only should the field excitation be considered as practi- 
cally wattless, but that in addition there is a wattless component 
due to reactances in the armature windings. 

The armature current can be split into two components, one 
of which is partly magnetizing and represents no torque. The 
other component is in phase with the field magnetism and there- 
fore represents torque. The magnetizing or wattless element 
may be comparatively small, as the number of turns in the arma- 
ture is relatively large, but the armature thus carries at times a 
slightly larger current than the straight-series motor. 

A further inspection of the diagram (Fig. 5) indicates how the 
power-factor of the motor can be made very high at synchronous 
speed. At all speeds the pressure generated in the armature due 
to rotation in the field of F, is practically equal to the pressure 
generated by the transformer T, thus making zero pressure 
across the terminals. But also at synchronous speed the pressure 
generated by the exciting field acting as a transformer, between 
the points a b, will be practically equal to the pressure generated 
in the winding by rotation of the winding in the transformer 
field. Therefore across a b the pressure is practically zero with 
these conditions, but the frequency remains the same as that in 


the field. If now the magnetizing current be supplied across the 
points a 6, then the required ampere-turns for magnetizing the 
motor can be supplied at practically zero pressure, and the turns 
of the external magnetizing field can be omitted. Therefore, 
under this condition the wattless component is practically zero 
and the power-factor becomes practically 100 per cent. This 
is the method of excitation used on certain European single-phase 
motors in which high power-factors are claimed for full-load 
running. But this method of excitation does not improve con- 
ditions at start, as the same excitation will be required at stand- 
still, whether the excitation be supplied to the armature or to the 
field. Therefore this method of excitation does not help the 
motor at that condition of load which is the severest on the 
generating and transmission system. It has the advantage of 
omitting the field exciting winding, but has the great disadvan- 
tage of requiring a double set of brushes on the commutator, 
with but half the distance between the brushes found in the 
straight-series or the ordinary repulsion motor. I do not believe 
that such methods of compensation are of sufficient advantage to 
overcome the complications attendant upon them. 

At zero speed, both the straight-series and the repulsion 
motors have low power-factors and with equal losses in the 
motors, the repulsion should have slightly lower power-factor 
than the series. This question of power-factor at start is largely 
a question of internal losses in the motor at rest, ajid the repul- 
sion motor in individual cases may show higher than the series 
motor, because it may be designed with higher internal losses. 
The real measure of effectiveness is not the power-factor at start, 
but the apparent input or kilovolt-amperes at start required for a 
given starting torque. With equally good designs of motors, the 
speaker's experience is that the kilovolt-amperes will be found to 
be considerably less with the straight-series than with the repul- 
sion motor, due to the fact that the current is fed directly into 
the armature and not by transformer action, and therefore the 
conditions of phase-relation and amount of current in the arma- 
ture windings are more favorable. Therefore it follows that in 
order to have the same kUovolt-ampere input for the same start- 
ing torque, the repulsion motor should have a smaller length 
of air-gap tira-n the corresponding straight-series motor, or should 
have a greater section of air-gap, which means greater weight 
of motor. This is one of the conditions which has led the speaker 



to the advocacy of the series motor rather than the repulsion 
motor, as he has considered this condition of starting of more 
importance than running; although he is satisfied that many of 
the running conditions of a well-designed series motor will be 
found in practice to be superior to those of an equally well-de- 
signed repulsion motor. 

Referring again to Fig. 5, it will be noted that two fields are 
set up in such a motor, and that at synchronous speed these two 
fields are equal. In the straight-series motor there is but one 
field set up, the other being omitted. It is evident that the 
straight-series motor with the current supplied directly to the 
brushes can have a smaller section in certain parts of the mag- 
netic circuit than is required for the repulsion motor, and that 
therefore the weight of material would be less, and the external 

PIG. 1. 

dimensions can be less. In Pig. 7 the heavy line represents 
outlines of series and the dotted line those of repulsion motor; 
therefore, it follows that for equally good designs and same fre- 
quency, the straight-series motor should be more compact, and 
should weigh less than the repulsion motor. It is reasonable 
to expect this, as the repulsion motor contains a transformer in 
addition to the other parts found in the straight-series motor. 
Futherrnore, the transformer found in such a motor is one with 
an air-gap, and with the windings on two separate elements, and 
therefore cannot be so well proportioned as a separate trans- 
former could be. Also, there is a transformer for each motor, 
and in a 4-motor railway equipment, for instance, there would be 
four transformers of smaller size against one largfer transformer 
used with the series motor, this larger transformer having a 



closed magnetic circuit, and of a highly efficient design com- 
pared with the transformers in the motors themselves. 

A further point should be taken up in the comparison of these 
motors ; viz , the current in the coil short-circuited by the brushes. 
This coil is a secondary to the field and the current in it is neces- 
sarily greatest at the period of strongest field. Therefore, this 
current will be greatest at the time of starting. If the repulsion 
motor and the straight series motor have the same field strength 
at start, then the short-circuited current should be the same in 
each But as the current is fed into the armature in the repulsion 
motor through transformer action, it will as a rule be found that 




the starting field strength of such a motor is slightly greater 
and the starting armature strength slightly less for a given torque 
than is found in the straight-series motor having same ratio of 
armature to field windings. Therefore the short-circuit current 
at start will be somewhat larger for the repulsion motor than for 
the corresponding straight-series motor. This short-circuit cur- 
rent may be somewhat less near full speed than in the straight- 
series motor, but it is not the full-speed condition which is the 
serious one. The short-circuit current at start is one of the 
most serious conditions which confronts us in alterrating-cur- 
rent motors, and is also of great importance where there is any 
considerable operation on low speeds. The speaker advocates 


a type wHch he considers gives the easiest condition in this regard. 
This short-circuiting cannot be entirely avoided in any of the 
motors brought out without adopting abnormal and question- 
able constructions, although devices like narrow brushes, sand- 
wich windings, etc., have been proposed. In certain foreign 
motors the brushes used are so narrow that they cover practically 
the width of one commutator-bar. As such motors are gener- 
ally built with a veiy large number of bars, the brushes used are 
extremely narrow, being approximately 0.2 inch thick at the tip. 
This will undoubtedly lessen the short-circuiting, but simply 
transfers trouble to another point; a brush 0.2 inch thick is not 
practicable for commercial railway service; at high speeds, 
with only a moderately rough commutator, such brushes will 
be liable to chip and break; further, the brush on a street-car 
motor should bridge at least two bars to give good, smooth, 
brush operation; in practice, a 5 inch brush on motors of 100 h.p. 
should be used. 

The sandwich winding, which consists of two or more wind* 
ings side by side 7 will prevent short-circuiting at the brushes, but 
is only another way of transferring trouble to another point; it 
has been found in practice that it is difficult to run a sandwich 
winding without trouble at the commutator with direct current, 
without a tendency to blackening and pitting the commutator, 
and with alternating current this tendency to pitting and burn- 
ing of the bars would be equally great. 

As a rule, there is little difference between the operation of 
repulsion and straight-series motors as regards sparking, except 
that the repulsion motors generally have greater current in the 
short-circuited coil near zero speed, and therefore show greater 
tendency to heat and spark. At or near synchronous speed, 
there appears to be very Httle difference in the commutation, 
although the speaker has never given the repulsion motor the 
same test of long-continued setvicfe as lie has in the case of the 
series motor. These series motors have never shown any tend- 
ency to give trouble on the commutator, and on an exhibition 
car equipped with four 100-h.p. motors, the commutator nave 
never been sandpapered since the equipment has been put into 
service. This exhibition car is used principally for showing the 
accelerating properties of the motors; therefore, the speaker 
does not hesitate to say that the commutation of the straight- 
series motor will prove to be equal to that of the direct-current 


motor. Wide brushes axe used with it, such as have been used 
in street-railway motors. 

It is well known that with large direct-current motors, espe- 
cially when operated at very high speeds, there is a tendency to 
flash across the commutator, or to the frame of the motor, if the 
field circuit be opened for a period long enough to allow magnet- 
ism to drop to zero, and then the field be closed again. In this 
case there is a rush of current before the field has had time to 
build up, and this rush of current, together with field distortion, 
may cause serious flashing. In the alternating-current motor, 
whether of the straight-series or the repulsion type, this ten- 
dency should be entirely absent. In the straight-series motor the 
magnetism falls to zero once in each alternation, and therefore 
if this tendency existed, flashing would occur continuously. 
Furthermore, a properly designed straight-series motor can be 
short-circuited across the brushes without injury to the motor, 
and can continue to operate in this way; therefore, if the ma- 
chine can be short-circuited in this way, there is evidently no 
tendency to maintain an arc. 

Returning to the subject of power-factors it should be noted 
that high-power-factors are very frequently found in motors of 
low or only moderately good efficiency. This low efficiency to a 
slight extent explains the high power-factor in some motors, both 
polyphase and single-phase. Low efficiency means higher true 
energy expended, and with a given wattless component it means 
higher power-factor. It is the old problem of increasing the 
power-factor by wasting energy in a circuit instead of reducing 
the wattless component. The power-factor of any alternating- 
current motor can be very considerably increased by putting 
resistance in series with it. Instead of this resistance the internal 
losses of the motor may be made higher, which will accomplish 
the same results. The motor will therefore appear to have a 
higher power-factor than it really deserves, if efficiency of the 
motor is taken into account. If, for instance, the efficiency at 
300 rev. shown in Mr. Slichter's Fig. 4 would be made as high as 
on direct-current motors, then the power-factors with the same 
magnetizing and other conditions, would have been approxi- 
mately four percent lower. This lower power-factor would not 
have made any harder condition on the supply circuit, but actually 
would have made a somewhat easier condition, as the supply sys- 
tem would have furnished about eight percent less Mlovolt- 


amperes. For lower speeds this difference in power-factor will 
be greater, and less for higher speeds A high power-factor at 
start, obtained by the use of resistance in series with the motor 
by high internal losses which do not represent torque, is there- 
fore a detrimental condition rather than a good one, as it means 
increased kilo volt-ampere expenditure for a given torque. This 
is merely given as an illustration showing that power-factor in 
itself is not a true indication of conditions, but must be accom- 
panied by other data; this is not a criticism of these motors, 
but is a general condition, found to a greater or less extent in all 
alternating-current motors. 


FOREWORD This article was prepared many years ago for the 
information of the younger technical men of the Westinghouse 
Company. It was considered of sufficient value by the Com- 
pany to publish in pamphlet form, of which there were several 
editions from time to time. It should be considered as purely 
educational. Only the types of windings which were in use up 
to the time the paper was prepared, are included. (ED.) 

THERE are a number of popular misconceptions regarding the 
relative polyphase and single-phase capacities which can be 
obtained from a given winding. For instance, there appears to be 
a half -formed opinion that a given winding connected for two- 
phase will give a slightly less output than when connected for three- 
phase; but, on the other hand, it seems to be generally assumed 
that the various three-phase windings all give the same rating. 

Also, it is a widespread idea that when any polyphase machine 
carries a single-phase load the permissible rating, with the same 
temperature, is approximately 71 percent of the polyphase rating. 
While there are a few cases where this may be true, yet, in general, 
it is far from being the fact, as will be explained below. 

This fallacy regarding single-phase ratings arose partly from 
early practice with polyphase machines, which were ofttimes 
designed with a view to carrying single-phase load almost exclu- 
sively. In consequence, the type of armature winding chosen was, 
in many cases, that which gave a high output on the single-phase, 
with some sacrifice in the polyphase rating, and the single-phase 
rating in many cases was a relatively large percent of the polyphase 
rating simply because the polyphase rating was less than could 
have been obtained with a different type of winding. 

The 71 percent (or 70.71 percent) ratio of single-phase to poly- 
phase ratings in a given armature arose partly from tho fact that 
at these relative loads the total armature losses were practically 
equal. On old designs of machines, in many cases it could be as- 
sumed safely that with equal armature losses the temperature of 
the armature parts would be practically equal. This assumption 




does not hold, in general, on modern designs of machines in which 
each individual part is proportioned for a specified result. The 
distribution of the armature losses is just as important as the total 
losses. If the temperature drop between the inside of the arma- 
ture coil and the armature core is small compared with the tem- 
perature drop from the core to the air, then the temperature of the 
armature, or its rating, will depend largely upon its total losses, 
equal ventilation being assumed in all such comparisons If, how- 
ever,the temperature drop from the coil to the iron, or from the in- 
side of the coil to the outside, is relatively high, then the temper- 
ature limit may be fixed by the loss in an individual coil rather 
than by the total loss. This is particularly the case in high voltage 

machines where there is a considerable amount of insulation over 
the individual armature coils. Also, in many of the later designs 
of machines (especially turbo-generators) each armature coil is 
practically separated from all other coils, so that one coil can have 
but little direct influence on the temperature of its neighboring coils. 
In such an armature it is possible to completely roast out an indi- 
vidual coil or group of coils without seriously heating any other 
coils or groups of coils. It is obvious that in such a machine the 
loss in the individual coils is what fixes the rating of the machine, 
and not the armature loss as a whole. It is evident, therefore, that 
when a polyphase machine, with such a winding, is loaded single- 
phase, the maximum current which can be carried in any single 
coil must be the same for either polyphase or single-phase rating. 
As this type of winding is used in the majority of large capacity 
machines of the present day, the following comparison will show 



the relative rating of such machines on polyphase and single- 
phase loading. Three-phase ratings will be considered first, 
because the great majority of modern machines are wound for 


All the various types of commercial three-phase windings 
with their current and voltage relations can be derived in a very 
simple manner from the consideration of a ring armature with 
its windings arranged in six symmetrical groups, each covering 
60 degrees of the ring, which may all be closed together to form 
the ordinary closed winding, or which may be separated into either 
three or six groups and connected to form various delta and star 
types of windings. 

Let Fig. 1 represent such a ring armature closed on itself and 
with six taps brought out, these being designated as A, a t B, b, C, c. 

By connecting together the points Aa, aB, Bb, etc., as shown 
in Fig. 2, a six-sided figure is obtained, which represents the 
various voltage and phase relations which can be obtained with 
all commercial three-phase (and six-phase) windings. It will be 
noted that Aa and bC are of equal length and are parallel in di- 
rection. The length represents e. m. f . and the direction re- 
presents phase relation. Therefore, these two groups or legs are of 
equal e. m f.'s and of the same phase. The same holds true of aB 
and Cc t and of Bb and cA. Beginning at Aa, these groups have 
also been numbered consecutively from 1 to 6, so that in the 
following diagrams a given leg or group can be identified by 



Fig. 3 is the same as Pig. 2 with three leads carried out from 
A, B and C to form the three terminals of a three-phase winding 
The dotted lines from A to B, B to C, and C to A represent the 
voltage and phase relations obtained from this combination 
This is known as a closed coil type of winding and is the standard 
arrangement of winding on a three-phase rotary converter. The 
comparative e m. f . values for this and other combinations will be 
given later 

By opening the closed arrangement of Fig 2 at the points A, B 
and C, as shown in Fig. 4, then an open coil arrangement is obtained 
and the three parts resulting can be recombined in several ways, 
keeping the same voltage and phase relations of the individual 

Fig. 6 

However, only one of these combinations, that shown in Fig. 
5 has been used to any extent. This is one form of star winding 
which is sometimes used to give certain voltage combinations, as 
will be explained later. 

By splitting Fig. 2 at six points instead of three, as shown in 
Fig. 6, various other open coil combinations of windings can be 
obtained while keeping each group or leg in its proper phase and 
voltage relations. 

One of these combinations is shown in Fig. 7, in which the 
groups which are similar in e. m. f . and phase are connected in 
parallel and the three resulting combinations are connected to form 
a delta winding. 

In Fig. 8 the two groups of similar phase are shown in series 
instead of parallel and connected to form a delta. Obviously 



Pig. 7 and 8 are equivalent, except that the terminal e. m. f of one 
is double that of the other 

By reconnecting the three components of Fig 7 in the manner 

shown in Fig. 9, a parallel star winding is obtained. Two arrang- 
ments are shown, one with all the legs connected together at the 
middle point, and the other with the two stars not connected at the 

Fig 10 is equivalent to Fig. 9 except that the two e. m. f.'s of 



equal phase are in series instead of in parallel, thus giving just twice 
the voltage of Fig. 9. 


The foregoing covers all of the usual combinations for three- 
phase windings, open and closed coil types. The same general 
scheme may be used to illustrate the usual six-phase combinations 
of windings which are frequently used in connection with rotary 



In Fig. 3, a three-phase winding is shown with terminals at 
A, B and C. If three other terminals be formed by a, b and c, 
then a second three-phase winding is obtained. The dotted lines 
in Pig. 11 illustrate the voltage and phase relation of these two 
windiags. This is the so-called double delta arrangement some- 
times used with six-phase rotaries, the dotted lines representing 
the voltage and phase relations of the transformers which supply 
the rotary converters. 


-/- y--? 

/ ^ ' 

Fig. II 

It is evident that the voltage represented by ac is equal in 
value and phase to that represented by BC. Therefore, one 
transformer with, two secondaries of equal value could be tapped 
across these two circuits. Similar arrangements can be applied to 
the other two-phase relations in this diagram. 

In Pig. 2 it is evident that, if six terminals are used, a voltage 
can be obtained across Ab Similar voltages can be obtained 
across aC and Be. These three voltages axe equal in value but 
have the 60-degree relation to each other. It is evident therefore 
that three transformers connected to a three-phase circuit can have 
their secondaries connected to this winding across the indicated 


points. This arrangement is indicated in Fig. 12, and is the 
so-called diametral connection of six-phase rotaries. The middle 
points of the transformer winding from these three circuits can be 
connected together, if desired. 


From an inspection of the above diagrams, and the application 
of but very little mathematics, all the e. m. f. relations of these 
various combinations can be readily obtained. In the following 
comparisons the magnetic field is assumed to be of such distribu- 
tion that the e. m. f. waves will be of sine shape, as this greatly 
simplifies the various relations. 

Let E represent the effective e. m. f . of any one of the six legs 
or groups in Fig. 2. Then, combining the various groups geo- 
metrically, taking into account the angular relation between the 
legs in the diagram, the various e. m. f.'s can be readily derived. 
The results are as follows: 

In Fig. 3, the e. m. f . across AB, BC, etc., = VT x E = 1 .732 

In Fig. 5, the e. m.f . Ad is the same as AB in Fig. 3 and is 
therefore equal to -x/lT x E, but the e. m. f . AB in Fig. 5 is equal 
to VlfxAd. Therefore, the e.m.f. of AB VlTx VlTxS^ 

In Fig. 7, AB is evidently equal to one group or side of Fig. 2 
and therefore the e. m. f . of AB = E. 

In the same way the e. m. f . of Fig. 8 = 2E. 

In Fig. 9 the e. m. f . Ad is evidently equal to E and the e. m. f . 
AB= VTxe. m. f. of Ad. Therefore the e. m. f. of AB= i/lT* 
E = 1.732 E, or same as Fig. 3. 

In the same way the e. m. f. of ABinFig. 10= 2times V 3 
x = 3.464E. 

For the six-phase combinations the following e. m. f.'s are 

In Fig. 11 each of the deltas is the same as in Fig. 3 and 
therefore the e. m. f.'s are the same and are equal to 1.732 E. 

In Fig. 12, Ab is geometrically equal to twice aB and the 
e. m. f . Ab is therefore equal to 2E. 


It might be assumed from casual inspection that all of the 
different combinations of three-phase and six-phase would give 


the same capacities when cajrrying the same limiting current per 
armature coil, or per leg. This however, is not correct, as will be 
shown by the following: 

Let A equal the limiting current which can be carried by one 
coil or by one group of windings. This is not necessarily the 
current per terminal, but it is the current permissible in an indi- 
vidual coil without exceeding a certain prescribed temperature. 
Then the following ratings are obtainable with the above combina- 
tions of windings. 

In Pig. 3 the rating = 3A x y"~T x E =* 5.196 AE. 

In Pig. 5 the current per coil and per terminal = A. The 
e. m. f . becomes A x \/ir x 3E = 5.196 AE. Therefore the three- 
phase ratings of the windings in Pigs. 3 and 5 are equal. 

In Fig, 7 the current in each leg of the delta is 2A, as there 
are two groups in parallel, each carrying current A . As the e. m. f. 
across terminals is E, the rating becomes 3 x 2A x E *= 6 AE. 

In Fig. 8 the current per side or leg of the delta is equal to A 
and the e. m, f . is 2E. The capacity therefore becomes the same as 
for Fig. 7 or = 6 AE. 

In Fig. 9 the current per terminal is 2A as there are two 
groups in parallel for each terminal. The e. m. f . across the term- 
inals is V1T x E. The capacity is therefore 2A V~ x V1T E = 

The rating of Pig. 10 is also 6 AE, the same as Fig. 9. 

In Figs. 11 and 12 the ratings can be determined by direct 
inspection from the following method of considering the problem: 

In a closed coil, polyphase machine, for example, such as 
shown in Pig. 2, one circuit can be taken off from A and a, a 
second circuit from a and B, etc., and the total number of circuits 
which can be taken off corresponds to the number of armature taps. 
Each circuit can be considered as having its own rating. There- 
fore, the effective voltage of each of such circuits times the current 
per circuit, times the number of circuits, equals the rating. In 
Figs. 11 and 12 six circuits can be taJsen off, each with voltage 
E and carrying current A. The rating therefore becomes 6 AE. 
The same method could be applied to any other number of phases 
from closed coil windings, 

It is evident from the foregoing that the same rating can not be 
obtained from the armature winding with all methods of con- 
nection. In those three-phase arrangements in which two groups 
of similar phase relations are thrown in series or parallel, the high- 


est output is obtained. In those cases where two e m f.'s out of 
phase with each other are combined to form one leg of the three- 
phase circuit, it is evident that the resultant e. m. f. is at once 
reduced by such combinations and that the capacity of the ma- 
chine is therefore reduced, simply because the most effective use of 
the windings is not obtained. The three-phase closed coil winding 
is therefore not as effective as the true delta or star type of winding. 
For this reason the closed coil winding is used in only those cases 
where some condition other than the current capacity itself is of 
greater importance. Otherwise, delta and star windings are 
always used, the star being preferred as it gives a higher voltage 
with a given number of conductors, or a smaller number of conduc- 
tors for a given voltage, and is therefore somewhat more effective 
in the amount of copper which can be gotten into a given space 

SINGLE-PHASE RATING Any three-phase machine with one of 
the above windings can be used to carry single-phase load by using 
two of the three terminals. The single-phase e. m f.'s obtained will 
therefore be the same, in each case, as the three-phase. The 
current capacity per coil, or group, on single-phase can be no 
greater than on three-phase. On this basis, therefore, the following 
single-phase ratings are obtained with the above combinations: 

Fig. 3, calling A and B the single-phase terminals, then with 
the limiting current A per coil, the windings 1 and 2 in the diagram 
will carry current A, and 3, 4, 5 and 6 will carry %A. The total 
current at the terminals will therefore be 1}^A and the e. m. f . per 
terminal will be V~3~ E. The single-phase rating then becomes 1 .5 
A x V1T E = 2.598 AE. The corresponding three-phase rating is 
3A x V1TE = 5.196 AE. The single-phase rating is therefore just 
50 percent of the three-phase for this combination. 

In Fig. 5, the current per leg is A> while the e. m. f. is 3E. 
The single-phase rating therefore becomes 3 AE. The correspond- 
ing polyphase rating is .A x V1T x 3E = 5.196 AE. The single- 
phase rating is therefore 57.7 percent of the polyphase rating 

In Fig. 7 the total current in two legs is 2A t while in the 
other four legs of the delta the total current is A. The total 
current at the terminals therefore becomes 3 A. The e. rn. f . is E 
and therefore the single-phase rating becomes 3 AE. The corres- 
ponding three-phase rating is 6 AE. The single-phase rating is 
therefore 50 percent of the polyphase for a true delta winding. 

The same holds true for Fig. 8. 


In Fig. 9 the current per group Is A and with two groups in 
parallel the current per terminal is 2A. The e. m. f. across the 
terminals is VT E. The single-phase rating therefore becomes 
2A x VT E or 3 464 AE. The three-phase rating for the same 
combination is 6 AE. The single-phase rating therefore becomes 
57.7 percent of the three-phase when a true star winding is used. 

Pig. 10 gives the same results as Fig. 9. 

It may be noted that in the three-phase star arrangement 
two legs are carrying all of the current, while the third leg is idle 
and could be omitted. This means that the active winding covers 
two-thirds of the armature surf ace, while an idle space of one-third 
the surface lies in the middle of the winding. 

In the delta winding it may be noted that one leg, covering 
one-third the surface, is directly in phase with the single-phase 
e. m. f . and is therefore in its most effective position The other 
two legs carry current also, but are relatively ineffective as the 
e. m f.'s generated in these two legs are displaced 60 degrees in 
phase from the single-phase e. m. f . delivered. The delta arrange- 
ment therefore has two-thirds of its winding acting in a very 
ineffective manner. One-third of the winding is very effective. 
In the star arrangement, two-thirds of the winding is almost in 
phase with the terminal e. m. f . (being 86 6 percent effective), while 
one leg is entirely idle. The star arrangement is about 15 percent 
more effective than the delta arrangement. 

The single-phase rating which can be obtained from the two 
six-phase combinations shown in Figs. 1 1 and 12 should also be con- 
sidered. In either of these diagrams, if two opposite terminals, 
such as AB, be taken as the single-phase terminals, then the e. m f . 
will be 2E. As each half of the winding can carry the current A, 
the total which can be handled is 2A. The single-phase rating 
therefore becomes 4AE. The corresponding polyphase rating is 
6 AE. The single-phase rating is therefore 66.7 percent of the 
polyphase, or is higher than in any of the other three-phase com- 
binations shown. It should be noted, however, that in order to 
obtain three-phase from this combination, transformers are neces- 
sary in order to transform from six-phase at the winding to three- 
phase on the line. Therefore, while this combination gives the 
highest single-phase and polyphase ratings, yet if three-phase is 
used on the transmission circuit, transformers must be interposed. 
Therefore, the highest obtainable rating of single-phase and three- 
phase from the same winding implies the use of transformers. 


The high single-phase rating obtained in this case is due to the 
fact that the arrangement is equivalent to the star arrangement 
with the idle leg added, as illustrated in Fig. 13. The addition of 
this extra leg increases the terminal e. m. f . in the ratio of 100 : 86.6, 
while the current per terminal remains the same. This arrange- 
ment, when used for both single-phase and three-phase, implies 
the use of a closed coil type of winding which, as shown before, 
cannot give the maximum three-phase rating unless six terminals 
axe used. 

It should be noted that the three legs shown in Fig. 13 have 
the same phase relations as a delta winding when used on single- 
phase; that is, one of the three legs is in phase with the terminal 
voltage, while the other two legs have a 60-degree relation, How- 
ever, these two legs, with the 60-degree relation, carry the full cur- 
rent A; while in the delta arrangement they carry one-half current. 

Therefore, although the voltage relations are the same, the current 
relations are quite different; which accounts for the increased 
capacity with the groups connected as in Fig. 13 or Fig. 12. 

Fig. 13, like Fig. 12, is equivalent to covering the entire arma- 
ture surface with copper which is equally active in carrying current 
when the machine is operated single-phase. However, compared 
with the three-phase star arrangement where two legs only are 
active, it may be seen that the voltage and the output have been 
increased in the ratio of 100: 86.6, or about 15 percent, by the 
addition of 33 1-3 percent in copper, and 33 1-3 percent in total 
armature copper loss. It is evident, therefore, that the addition 
of a third leg when operating single-phase does not give results in 
proportion to the material used. 



All the foregoing comparisons have been on the basis of equal 
losses in a given coil or group; but it has been shown that with 
some of the windings, when operated on single-phase, the currents 
are not divided equally. In consequence, in such cases the total 
copper loss in the windings must be less than where the current is 
divided equally. In the following comparisons the total copper 
losses for three-phase and single-phase are given, and the possible 
increase in single-phase rating for the same total copper loss is 

Let r = the resistance of one group. 

Let A = the limiting current per group, which has been used 
in the above comparisons. 

Then in Fig 3, for three-phase, 6A z r = the armature copper 

loss. For single-phase { jx4r + 2A 2 r = 3AV = total armature 

copper loss. 

The three-phase loss is therefore twice the single-phase on the 
basis of equal limiting current For equal total loss the single- 
phase current could therefore be increased as the V~2^ as the loss 
varies as the square of the current. As the former single-phase 
output was 50 percent of the three-phase, then for equal losses the 
single-phase output "becomes 50 x V~T = 70.7 percent of the 
corresponding polyphase rating. 

In Fig. 5, the three-phase loss = 6AV. The single-phase loss 
with the same limiting current = 4AV, as there are but four legs in 
circuit instead of six, each leg carrying the same current as when 
operating three-phase. The three-phase loss is thus 6/4 single- 
phase, and for equal losses the single-phase current can be in- 
creased in the ratio of -v/6/4- The former single-phase rating was 
57.7 percent. This therefore can be increased to 57.7 x Vs/4 = 
70.7 percent of the corresponding three-phase rating. 

In Figs. 7 and 8, the three-phase loss = 6AV. The single-phase 
loss = 3A*r, as determined by direct inspection of currents and 
resistances, For equal losses, therefore, the single-phase current 
can be increased as the VT- The output then becomes 50 x \/~ir 
= 70.7 percent of the corresponding three-phase rating. 

In Figs. 9 and 10, the three-phase loss = 6A 2 r. Single-phase 
loss = 4A 2 r. The single-phase output - 57 7 percent and for 


equal loss this can be increased in the ratio of VoTI- The output 
then becomes 70.7 percent of the corresponding three-phase output. 

In Fig. 12, the six-phase loss = 6AV. The single-phase 
loss = 6A 2 r, as all the groups carry equal currents and all are in 
circuit. Therefore the single-phase current cannot be further 
increased and the single-phase output remains at 66.7 percent of 
the six-phase output (or three-phase beyond the transformers.) 

From this it would appear that most of the above wind- 
ings would give, for equal armature copper loss, 70.7 percent of the 
three-phase rating. However, it should be taken into account that 
the three-phase ratings are not all equal on the basis of equal 
copper loss. 

In Figs. 3 and 5, for instance, the three-phase ratings are 
equal to 5.196 AE. The three-phase ratings with the arrangement 
shown in Figs. 7, 8, 9 and 10, are equal to 4. Therefore Figs. 3 
and 5 have only 86.6 percent of the three-phase ratings of 7, 8, 9 
and 10. The single-phase ratings of Figs. 3 and 5 therefore are 
70.7 percent of 86.6 percent, or 61.2 percent of the best three-phase 
rating which can be obtained. Therefore, on the basis of 6AE 
being the best three-phase output, then with equal copper loss, the 
arrangements in Figs. 3 and 5 give 61.2 x 6AE = 3.792AE as the 
single-phase rating with equal copper loss, while Figs. 7, 8, 9 and 10 
give 4 243AE as the single-phase ratings with equal copper loss, 
and Fig. 12 gives 4AE as the single-phase rating with the same 
copper loss. Therefore, the arrangements in Figs. 7, 8, 9 and 10 are 
better than any of the others for single-phase rating, if total copper 
loss is the limit rather than the loss in an individual coil or group. 

However, if total copper loss is the limit, then there is still a 
difference between the true delta and star windings. With the 
delta winding the current A is increased 41 percent, which means 
that one of the groups will have double the copper loss which it has 
on three-phase, while with the star winding the current A will be 
increased slightly over 22 percent, which means that two groups of 
the winding will have their copper losses increased SO percent. The 
star arrangement, even with the same total copper loss, works the 
individual coils on single-phase easier than in the delta arrange- 

The following table summarizes the above relationships, 







^ ^ * 





ftj W 









M < 



w w 








The two-phase windings may be analyzed in a manner similar 
to the preceding. Starting with a closed-ring arrangement, just as 
in the three-phase, the various relations may be readily determined. 
Assuming a ring, as in Fig. 14, with four taps brought out at 90 
degrees apart and assuming that this winding is the same in every 
way as that in Fig. 1, then the following e. m. f.'s and capacities 
are obtained. 

Fig. 14 

Fig. 15 represents a closed coil two-phase winding correspond- 
ing to the three-phase winding in Fig. 3. Calling E the e. m. f. of 
the groups of legs, then the e. m. f.'s AC and BD = V1T x E. 

Opening Fig. 15 at two opposite points as in Fig. 16, the two 
parts may be rearranged to give Fig. 17. This is an interconnected 
open coil two-phase winding; that is, the central points are con- 
nected together so that there are fixed e. m. f . relations between all 
four terminals. The e. m. f . Ad is equal to E, and the e. m. f . across 
AB, BC, etc. = V!T x Ei, while the e. m. L across AC and BD = 

Splitting the winding of Fig. 16 at four points, then the ar- 
rangements shown in Figs. 18 and 19 are obtained. These two 
windings are equivalent, except that in Fig. 18 the two legs which 
are in phase are connected in parallel, while in Fig. 19 they are in 
series. If the middle points in. Fig. 19 are connected together tie 
arrangement becomes equivalent to Fig. 17. In Fig. 18, e. m. f.'s 
AC and BD are equal to EI, while there is no fixed e. m. f . relation 
between AB, BC, etc. 

In Fig. 19 the e. m, f.'s ACaad BD are equal to 2Ei and there 
is no fixed relation between AB, BC, eta, unless the middle points 



are interconnected, in which case the e.m.f.'s become the same as 
in Fig. 17. 

In Figs. 20 and 21 the usual two-phase, three-wire arrange- 
ment is shown. In Fig. 20, AB = EI and AC = \/~2~ x EI. In 
Fig. 21 AB = 2Ei and AC = 2 x V 2 EI. 

Fig .17 

rent per coil, this current being the same as for the three-phase 
winding. Then 

In Fig. 15 the capacity equals 4 

" " 
" " 


It is obvious therefore that the two-phase capacities are equal for 
all the various windings which have been commonly used. 


As the same winding has been assumed for both two-phase and 
three-phase, it is of interest to compare their ratings. Comparing 
E and EI in Fig. 22, it may be seen that EI -\f~2~ x E. Therefore 
the two-phase capacities given above, when put in terms of three- 
phase e.m.f.'s become, in all cases, 4A x V1T x E = 5.656 AE. 
The closed coil three-phase capacity = 5 196 AE. The closed coij 
six-phase capacity = 6 AE. The open coil (star or delta) three^ 
phase capacity = 6 AE. Therefore, the three-phase closed coij 
arrangement gives the least output, while the two-phase, (which i$ 



in reality, four-phase with a closed coil winding) gives somewhat 
better results and the six-phase closed coil gives still better results. 


Two of the terminals of the two-phase windings may be used 
for single-phase. Assuming the same current A per coil as in 

Fig, 19 

two-phase or three-phase, then the single-phase capacity 
In Fig. 15 = 2A x VT Ei = 2.828 AEi 
In Fig. 17 = A x 2E l = 2 
In Figs 18 and 19 = 2 AE l 
In Figs 20 and 21 = 2 


Fig. 2 i 

Fig. 22 

Comparing the best single-phase obtained from the two- 
phase with the best single-phase from the three-phase windings, EI 
being equal to V^IT E, the following is obtained 


Then 2,828 EiA = 4 AE, or same as obtained from the 
six-phase closed coil winding. 

Comparing the three-phase closed coil winding with the two- 
phase closed coil winding for both polyphase and single-phase 
ratings, the following is obtained on the basis of same loss per coil: 
The 3-phase closed coil winding gives 3-phase rating of 5 196 AE. 
" 3 " " " " " single" " " 2.596 

2 " " " " " 2 " " " 5. 656 AE 

2 " " " " " single" " " 4 AE. 

It is therefore apparent that with the closed coil winding the 
two-phase arrangement (or four-phase in reality) gives higher out- 
puts, for both polyphase and single-phase, than the three-phase 
closed coil arrangement will give. 

It may be of interest to note that in the earlier Westinghouse 
polyphase machines, when the single-phase rating of a polyphase 
generator was frequently of more importance than the polyphase 
rating, the closed coil two-phase winding shown above was gener- 
ally used. One reason for the selection of this type of winding was 
the high single-phase rating which could be obtained without un- 
due sacrifice in the polyphase rating. 


All of the preceding comparisons have had to do with sym- 
metrical arrangement of windings. However, by putting on one 
or more additional connections, which are used for single-phase 
operation purely, the windings can sometimes be made to give 
larger single-phase ratings than where the straight polyphase con- 
nections are used for single-phase operation. Two such arrange- 
ments will be shown below: 

It is shown in Pig. 12 that by taking off single-phase at -A 6, a 
high single-phase rating can be obtained. For supplying three- 
phase circuits, however, it was stated that transformers would have 
to be interposed to transform from six-phase to three-phase. 
However, by using A and b as the single-phase terminals and 
using A, B and C as the three-phase terminals, thus having four 
terminals total on the winding, as shown in Pig. 23, the machine 
can supply three-phase directly to the circuit and can also deliver 
single-phase with the best utilization of winding. In this case the 
three-phase rating equals 5.196 AE and the single-phase rating 
-equals 4 AE. The single-phase thus becomes approximately 
77 percent of the polyphase. This high relative rating, however, 



is due to the fact that the three-phase rating is only 86.6 percent 
of the maximum three-phase which could be obtained. 

In a similar way, with the delta winding shown in Fig. 8, an 
improved single-phase rating can be obtained by putting an 
additional terminal at the middle of one of the legs, as shown in Fig. 
24. The single-phase is then taken off at A and 6, while A, B and C 
are the three-phase terminals. In this case two of the delta legs are 
almost in phase with the single-phase, while the third leg is prac- 
tically idle as far as voltage is concerned, although it carries the 
full current. If the e. m. f . of AB is 2E then the e. m f . of AB is 
v"~3~ E. The total single-phase current is 2A, Therefore, the 

Fig. 24 

single-phase rating becomes 3.464 AE. The single-phase rating in 
this case is therefore 57.7 percent of the three-phase, instead of 50 
percent where the single-phase was taken off at the terminals AB. 
The above two arrangements are therefore more effective than 
the usual single-phase from the same types of windings. However, 
as will be shown later, the true delta and the closed coil three- 
phase windings are seldom used on alternating-current generators 
and therefore the above special arrangements are of no particular 
commercial advantage. 


If direct current be taken from the same winding as described, 
the limiting current per coil should be the same as the effective 
(or square root mean square) current when delivering alternating 


current. This is the value A used in the preceding comparisons. 
The direct-current e. m. f . is taken off from two opposite points of 
the armature, This e. m. f . therefore corresponds to the two op- 
posite terminals of either the two-phase closed coil or six-phase 
closed coil winding shown in the preceding diagrams. The direct- 
current e. m. f . will be equal to the maximum or peak value of the 
alternating-current e. m f . taken off from these two points This 
will be \J~~z times the effective value used in the preceding 

For the six-phase diametral arrangement, it was shown that 
the effective alternating-current e m. f . = IE. Therefore the peak 
value of direct-current e. m. f . will be equal to VIT x IE. As the 
limiting current is A, and as there are two direct-current branches, 
the total direct current will be 2. The direct-current output 
therefore becomes 4 x \/~2~ AE = 5.656 AE. 

The following interesting comparisons can therefore be made: 

Direct-current capacity = 5 656 AE 

1-Phase closed coil capacity =4 AE =70.7% of D C. 

3-Phase " " " =5.192 AE =91 8% " " 

2 " (4-phase) " " =5656AE =100% " " 
6 " " " " =6 AE =1061% u " 

3 " open " " =6 AE =1061%" " 

From the above it appears that the two-phase closed coil (and 
two-phase open coil) capacity is equal to the direct-current capacity 
from the same armature winding. The three-phase closed coil is 
less than the direct current, while the six-phase is greater than the 
direct current. The three-phase true star or delta winding and the 
six-phase closed coil winding are all slightly more effective than 
when the same winding is used for direct-current. 

The question may be raised whether still higher ratings could 
not be obtained from a given winding by taking off more phases. 
An examination will show that higher ratings can be obtained with 
the number of phases increased, with the dosed coil winding; but 
it can be shown that the possible increase over the six-phase 
arrangement is very small. 

An easy way of comparing the ratings of closed cofl windings, 
with different numbers of phases, is to compare the number of 
circuits which can be taken off between adjacent taps or terminals 
all around the winding, as referred to in first paragraph of page 120. 
This is equivalent to comparing the perimeters of the poly- 



gonal figures shown in the diagrams for the various closed coil 
combinations and is illustrated in Figs. 25, 26, 27 and 28. 
In Fig. 25, calling one side E, than the perimeter = 6E 
In Fig.^26, the perimeter = 5.656 E. In Fig 27 the perimeter 
= 3 V 3 E = 5,196 E. In Fig. 28, which represents single- 
phase, the two sides of the polygon coincide, making a straight line 
Therefore, double the length of this line should represent the 
perimeter, which = 4E. A comparison of these values shows that 
they are exactly in proportion to the alternating-current capacities 
given above 


Fig. 25 

Fig. 26 

It is evident that the greater the number of phases obtained 
from the closed coil winding, the more nearly the perimeter of the 
polygon approaches the circumference of the circle. With an infinite 
number of phases a true circle would be obtained and in this case 
the perimeter becomes 2irE = 6 283 E. Therefore, the maximum 

/\ 983 

possible polyphase rating is == 1.047, or 4.7 percent greater 


than the six-phase closed coil rating or the true star and delta rat- 
ing. Also, the greatest possible polyphase rating is greater than the 
direct-current rating in the proportion of 6.283: 5.656, or approx- 
imately 11 per cent. 


In the above comparisons of the relative ratings of the three- 
phase, two-phase and single-phase windings, only the armature 
copper losses have been taken into account: but if the problem is 
to be considered in its completeness, other armature conditions and 
the field conditions must also be taken into account. 



A comparison of the three-phase and two-phase ratings shows 
that they are usually so close^together that the field conditions 
would probably not exert a controlling influence on the relative 
capacities In general, it may be taken that those combinations 
of polyphase windings which give lower ratings at the same time 
give lower armature reactions. 

In comparing single-phase with polyphase ratings, however, 
the field conditions, both as regards the field winding and field core, 
must be taken into account The armature reaction of the single- 
phase winding is pulsating and tends to produce magnetic disturb- 





, 1 

\ 2 


Fig. 2 7 

Fig. 28 

ances in the field poles or core which may result in very considerable 
iron losses, both eddy and hysteretic. In general, these disturb- 
ances are relatively much greater on larger capacity machines, so 
that provision must be made on such machines for suppressing or 
avoiding the ill effects of the armature reaction. This can be 
accomplished to some extent, by completely laminating the field 
poles. Another method which has been used on very large ma- 
chines is the employment of heavy cage damper in the pole faces, 
similar to that of the secondary of an induction motor. This damper 
must have current capacity such that when developing ampere 
turns sufficient to completely neutralize the armature pulsations, 
the heating effect in the damper winding, due to the current in it, 
is relatively low. 

Field copper heating, in most cases, is not a controlling con- 
dition, owing to the fact that the single-phase rating, defined by the 
armature heating, as indicated above, is so much lower than the 
polyphase rating that the field copper is usually worked some- 
what easier than on the polyphase loading. This is particularly 


true when the rating is fixed by the heating of individual armature 
coils.' However, if the single-phase rating is determined by the 
total armature loss and not by the loss in individual coils, then the 
permissible armature capacity on single-phase may be such that in 
some instances the field copper is worked harder than on polyphase. 
In such cases, if the field copper is the limiting condition, then the 
single-phase rating cannot be as high as the armature would 
permit. It may be assumed, however, that in large machines the 
armature conditions, as fixed by the loss in individual coils, 
determine the safe single-phase rating; and under this assumption 
the field conditions, except in regard to the use of dampers or the 
elimination of the effects of armature reaction, need not be con- 



The three-phase true star type of winding is the one which, in 
general, lends itself to best advantage to the various types of 
alternating-current machinery. It may be a question then as to 
why any other types of windings are used. However, it was 
intimated , before that where other than the true star winding is 
used, there is usually some condition other than the output which 
is of first importance. In the following will be given some of the 
principal applications of the different types of windings: 


The closed coil type of winding is always used with rotary 
converters. The controlling feature in this case is that the rotary 
converter carries a commutator, which naturally requires a closed 
coil type of winding. Rotary converters are, in practice, wound 
for three-phase as in Fig. 3, four-phase (usually called two-phase) as 
in Fig. IS and six-phase as in Figs. 11 and 12; and the number of 
collector rings is 3, 4 and 6 respectively. The three-phase winding 
is generally used in small capacity rotaries. While the three-phase 
winding allows less output than the fofttr-phase or six-phase, on 
small rotaries the capacity is usually not limited by the armature 
copper loss, while the use of three rings somewhat simplifies the 

Four-phase rotaries are used to a very considerable extent in 
connection with two-phase circuits. However, where the supply 


circuit is three-phase it is rare that the transformation is from three- 
phase on the supply circuit to the two-phase on the rotary, as there 
are certain disadvantages in such transformation which more than 
offset the slight advantage of the four-phase rotary over the three- 
phase Moreover, where a higher number of phases is of advantage 
in a rotary converter, it is practicable to transform from the three- 
phase supply circuit to six-phase for the rotary Two arrange- 
ments of such six-phase transformation are in use, as illustrated in 
Figs 11 and 12 

One of these is the so-called " Double Delta" arrangement, in 
which each of the step-down transformer circuits is equipped with 
two secondaries, as indicated in Fig 11 These are connected to 
form two separate deltas, one being inverted with respect to the 

The other arrangement is the so-called "Diametral" arrange- 
ment, as shown in Fig 12 This has advantages over the double 
delta in that only one secondary circuit is required for each phase 
and the middle points of these secondary circuits may be connected 
together for a neutral or middle wire between the direct-current 
leads from the rotary converter. 

In a rotary converter the armature copper loss is generally so 
small, compared with that of the straight direct-current or straight 
alternating-current machine with the same winding, that all con- 
siderations of the comparative heating of three-phase, four-phase 
and six-phase windings, as on alternating-current generators, has 
practically no bearing on the rotary converter rating. In a rotary 
converter, an increase in the number of phases over six represents 
a considerable reduction in the armature copper loss, much more 
so than in the closed coil alternating-current generator This is 
due, in the rotary converter, to the fact that one armature winding 
carries both the direct and the alternating currents, which are to a 
certain extent, flowing oppositely. 

Closed coil windings are also occasionally used on the second- 
aries of induction motors in order to give a better choice in the 
number of slots than would be allowed otherwise Such windings 
when used on induction motors are usually of the two-circuit or 
series type, for the purpose of increasing the voltage as much as 
possible and at the same time keeping the number of conductors 
as small as possible, while retaining the closed coil arrangement. 
A two-circuit closed coil winding will close upon itself symmetrical- 
ly if the ntonber of turns or coils is one more or less than a multiple 


of the number of pairs of poles. This sometimes allows the use 
in the secondary of an induction motor, of a number of coils or slots 
which has no close numerical relation to the number of primary 
slots. For instance, if the primary of a four-pole induction motor 
has 48 slots with an open coil, star or delta winding, then with 
39 coils and slots in the secondary, a symmetrical closed coil three- 
phase winding could be obtained, while if an open coil secondary 
were used, the number of slots should preferably be 36 or 42, which 
might not be as desirable as 39 in some cases. This simply illus- 
trates an occasional use of the closed coil winding. 

Closed coil windings were at one time used very extensively on 
low voltage, rotating armature, two-phase generators. Such genera- 
tors were very satisfactory for delivering a relatively large percent- 
age of their rating as single-phase. Furthermore, with one conduc- 
tor per slot and with bolted-on end connectors, the potential bet- 
ween adjacent end connectors was at all points relatively low. The 
symmetrical arrangement of such windings also rendered them 
very suitable for use with supporting bands or end bells over the 
end winding. However, with the advent of the rotating field ma- 
chines, and particularly with the use of higher voltages, the open 
coil star winding has entirely superseded the closed coil type of 
generator winding. 


Two types of star windings have been shown, namely, those in 
Figs. 5 and 10. That of Fig. 5 gives less output than that of Fig. 10 
in the ratio of 86 6: 100. There would appear therefore to be no 
use for the Fig. 5 arrangement; but, in certain cases, in using a 
given winding it may be desired to reduce the voltage from 12 
percent to 15 percent while retaining normal conditions otherwise. 
In such a case the lower voltage could be obtained, if a new winding 
were used, by simply chording the winding one-third the pitch. 
On the other hand, if an existing winding is to be used, the same 
result could be obtained by coupling as in Fig. 5. 

In induction motors the arrangement shown in Fig. 5 may be 
used occasionally where the windings are arranged for coupling for 
two different speeds. In some cases this type of winding may give 
better average field distribution for the two numbers of poles 
than the one shown in Fig, 10. In this case therefore it is the dis- 
tribution of the magnetic field, and not the capacity of the winding, 
which is the important feature. 


The arrangement shown in Fig. 10 is the true star winding 
which is used almost universally on three-phase machines. For a 
given voltage it requires fewer conductors than any other type of 
winding, This is of very material advantage in allowing, with a 
given number of slots, a smaller number of conductors per slot, 
which, as a rule, allows a better utilization of the star space: 
That is, more copper can be gotten into a given slot. Furthermore, 
in relatively high voltage machines where the conductors may be 
very large in number and small in size, the star winding with its 
smaller number of conductors, each of much larger size, gives more 
substantial coils than any other arrangement. Another advantage 
of the three-phase winding is its fairly good utilization of copper 
when operated on single-phase. When operated on purely single- 
phase load, one leg of the star could, of course, be omitted, but if 
it is retained it becomes a reserve winding which may be used in 
case of an accident to one of the active legs of the winding By 
opening any defective coils in an active leg and connecting in the 
reserve leg in place of the defective one, the machine can still 
develop its specified rating on single-phase. 

Another advantage of the star type of winding is the readiness 
with which the central or neutral point can be grounded, which is 
a very considerable advantage in some high voltage systems. 


The true delta type winding, as illustrated in Figs. 7 and 8, 
is not used to any great extent in either alternating-current 
generators or induction motors. For a given voltage it requires 
73 percent more conductors, each of 58 percent of the capacity of 
those of the true star type of winding. As the terminals of all three 
legs are connected in a closed circuit it is necessary that the e. m. f .'s 
generated in the three legs should balance each other at all instants 
or there is liable to be circulating current around the windings. 
This means that the winding must be applied only where the 
conditions are favorable, or the conditions in the design must be 
made to suit the type of winding. This winding is occasionally 
used on low voltage turbo generators of fairly large capacity, due 
to the fact that the delta type winding requires more conductors 
than the star type. For example, in a large capacity two-pole 
turbo-generator, wound for relatively low voltage, the number^of 
conductors for the star winding may be so small that a satisfactory 
number of slots is not obtained, even with only one conductor per 


slot and even using the double-star winding, shown in Fig. 9. In 
such case a double delta winding will allow 73 percent more con- 
ductors and slots than the double star will give. Also, each 
conductor will be much smaller than in the star arrangement, 
which may be of considerable advantage in the case of low voltages 
and very heavy current per conductor. 

Delta windings are occasionally used on machines which are 
arranged for connection for two different voltages, such as 6600 
volts and 11,000 volts. If an armature is wound for star connection 
at 11,000 volts, then it can be coupled in delta for 6600 volts with 
practically the same inductions, losses, field currents, etc. The 
delta type of winding is also used occasionally in the primaries 
of induction motors for special purposes, such as multi-speed 
combinations where the winding is changed from one number of 
poles to another. In general, however, the star type winding is 
used on induction motors. 

The delta winding is not well adapted for single-phase opera- 
tion on account of its low capacity. Also, it does not admit of 
grounding of the neutral or central point of the system. Taking 
everything into account, the true delta winding is only used where 
some special condition is imposed upon the winding which puts 
the star arrangement at a disadvantage. 


All the foregoing comparisons have been made on the basis of 
the same armature winding being vised for three-phase, two-phase 
or single-phase The relations shown do not hold true in general for 
machines which are wound initially for single-phase service, such 
as for single-phase railway or electro-chemical or electro-fusion 
work. In such cases the amount of armature copper used and its 
distribution are such that the armature coils, either individually or 
as a whole, do not determine the true Emits of output; but the 
armature as a rule can carry anything that the field winding will 
stand, so that the field temperature becomes the true limit in such 
machines. Also, very massive, well distributed cage dampers are 
used with such machines when they are of large capacity and these, 
in turn, have a certain effect on the characteristics, such as the 
regulation, and thus have an indirect influence on the permissible 
capacity. It is well known that if the inherent regulation of an al- 
ternator is made poorer, the capacity can usually be increased with 
the same limiting field temperature. In large single-phase gener- 


ators, especially for railway service, the capacity is increased by 
sacrifice in the inherent regulation of the machine. However, the 
massive dampers greatly improve the regulation for quick changes 
in load; while the poorer inherent regulation only affects the 
regulation over considerable intervals of time, and automatic 
regulators, acting on the alternator excitation, readily take care of 
the slow fluctuations. In consequence, single-phase generators of 
large capacities may be built for ratings which bear no definite 
relation to any of those given above. 

The armature windings of single-phase generators, when ar- 
ranged for single-phase purely, are frequently distributed over 
only part of the surface. Usually they cover considerably more 
than half the surface, and in extreme cases they cover 80 percent 
or more. Of course, when spaced like a true three-phase winding 
they cover two-thirds the surface. This arrangement admits of an 
extra leg being added to the winding, which is normally idle, if the 
winding is connected in star, this leg being a reserve in case of 
accident, as mentioned before. However, when such a leg is not 
added, the winding generally covers more than two-thirds the 
surface, rather than less, but rarely covers the whole surface. 


FOREWORD This formed part of the discussion of a paper presented 
before the Institute of Electrical Engineers, December, 1908, 
by Mr. Murray, describing the operation of the New Haven 
single-phase railway. The effect of the addition of the massive 
dampers on the rotors of the New Haven generators was so 
pronounced, and the results were so beneficial, as a whole, that 
it was considered advisable to publish it as new and interesting 
material, in the form of a discussion of Mr. Murray's paper. 
Immediately after the publication of this, heavy dampers were 
adopted very generally by manufacturers of large single-phase 
generators, throughout the world, who had encountered more or 
less trouble of the same nature as found in the New Haven 
generators. Practically all large single-phase generators since 
then have been built with such dampers as part of their con- 
struction. (ED.) 

WHEN the New Haven single-phase generators were put on. 
load test, the first, and most pronounced, difficulty was in 
heating, not in the winding, but in the field or rotor structure, due 
to the pulsating reaction of the armature winding when carrying a 
heavy load in single-phase current. This reaction was known 
previously to building these machines, but on machines of smaller 
capacity it had not developed destructive tendencies. It was 
proved later that this was simply because it had not been tried out 
under the conditions which would develop its most harmful 
effects. This pulsating armature reaction may be analyzed in the 
following manner : 

Consider the armature winding as a magnetizing coil fixed 
in ( Space and carrying an alternating current. This coil may 
be considered as setting up an alternating field fixed in space. 
For analysis, this alternating-current field, fixed in space, may be 
considered as made up of two constant fields of half value, rota- 
ting in opposite directions at the synchronous speed of the machine. 
One of these fields therefore rotates at the same speed and in the 
same direction as the rotor. The other field is traveling round the 
rotor core in the direction opposite to its rotation. This field may 
therefore be considered as equivalent to one fixed in space with 
the rotor running in it at double speed. This core thus becomes 
an armature core subject to a heavy induction at a high frequency* 



When the first rotor was built, the structure was laminated as 
completely as mechanical conditions would permit. However, in 
the case of high-speed turbine-generators of very large capacity, it 
is almost impossible completely to laminate everything, due to the 
fact that the mechanical requirements call for rigidity in some of 
the structural features. Upon testing the first machine it was 
found that there was local heating, with heavy load, sufficient to 
create hot spots in the core; and in a comparatively short time in 
turn these hot spots damaged the insulation on the coils from the 
outside, thus causing grounds on the winding. As soon as this was 
noticed, an effort was made to eliminate these hot spots ; but it was 
found, after several attempts, that as soon as one was eliminated 
others would show up in some different place as soon as a higher 
load condition was reached. It was evident, after considerable 
work had been done, that the correct remedy was not being applied 
to this trouble It was then decided to take a bold step by at- 
tempting to eliminate all pulsating reactions from the armature by 
putting a short-circuited winding on the rotor, of such value that 
a very large current could flow in it with but very little loss. It 
was the idea to damp out the field in very much the same way that 
the armature of a polyphase alternator will demagnetize, or kill its 
magnetic field, if the armature terminals are all short-circuited 
together. It is known that under this condition the armature 
current will rise to such a value that the field flux is practically 
eliminated. In order to maintain this condition indefinitely 
without overheating, it is only necessary to put enough copper on 
the armature so that the PR losses in it under this condition are 
within the temperature capacity of the windings. Working on 
this theory, a complete cage winding was placed on one of the 
rotors of the New Haven generators. This rotor had not been 
designed originally for this purpose, and it was therefore difficult 
to adopt the most suitable proportions in this winding, but what 
was put on, immediately showed in practice that a practicable 
remedy had been applied for this trouble Meanwhile the new 
rotors designed for the application of heavy cage windings were 
under construction, and upon the installation of these, the field or 
rotor trouble all disappeared. It is interesting to note that the 
fourth machine installed, which has a 4260 kilovolt-ampere single- 
phase rating, has a solid steel core, in the surface of which the 
copper cage winding is embedded. As this winding completely 
eliminated the pulsating armature reaction, there was no further 


occasion for laminating the field as a protection from magnetic 

I might add that a number of the earlier tests, leading up to 
the design of the first New Haven rotors, were misleading, in the 
fact that turbine-generators were used for obtaining the prelimin- 
ary data for single-phase operation and, in all cases, the machines 
had solid steel cores. These cores acted as dampers to a certain 
extent, and this in itself eliminated part of the pulsation. It thus 
developed afterwards, that in the very act of lamination to avoid the 
trouble, we had gotten into it deeper. 

Practically all this work on the generators was done before 
the full electric service was established, and while only one or two 
generators were required to be operated at one time. With one 
generator running, there was apparently but little or no disturb- 
ance due to short-circuits on the system. As the service was in- 
creased and two generators put in operation, the effect of short- 
circuits became more pronounced. When, in June, 1908, the entire 
electric service was established, and three generators were con- 
nected to the system, it soon became evident that there was some 
serious condition existing in the system, as indicated by the ex- 
tremely violent shocks to everything in case of a short-circuit. 
This was particularly noticeable in the switching system, and, as 
Mr. Murray intimates, in the case of a short-circuit, all the 
switches in the system felt it their duty to jump in and open the 
circuit. This indicated an abnormal current condition. It was 
calculated that these machines would give possibly six or seven 
times full-load current on the first rush, in the case of a dead short- 
circuit, this excess current dying down to possibly two or three 
times normal full-load current. All indications were, however, 
that this current was being greatly exceeded, and therefore a 
series of oscillograph tests were made to determine the current 
rush when the lines were purposely short-circuited tinder various 
conditions. These tests indicated that under certain conditions 
each machine could give, at the moment of short-circuit, almost 
5000 amperes on one phase, the normal full-load current being 340. 
With three machines in parallel, this would therefore mean that 
approximately 15,000 amperes could be delivered momentarily. 
This enormous current rush was sufficient to explain many of the 
difficulties, but this was not all the explanation. The oscillo- 
graph tests also showed that this short-circuit current would be 
maintained at almost its maximum value for a very considerable 


period, due to the cage winding on the rotors of the generators. 
Apparently this current at the first rush, was not appreciably 
greater than that on the machine before the dampers were added, 
but without the dampers the field was killed more quickly by this 
enormous current, so quickly that apparently the breakers did not 
open until the current had fallen somewhat. However, with the 
heavy cage winding on the field structure, secondary currents were 
set up in this winding, tending to maintain the field strength, and 
thus the current rush was maintained at almost full value for 
possibly 20 to 30 alternations. These oscillograph tests indicated 
very clearly that the armatures of these generators did not have 
nearly so great internal self-induction as our calculations indicated. 

Meanwhile, the generators in the power house had been suf- 
fering from the tremendous shocks which accompanied short- 
circuits on the line. There is necessarily considerable local field 
around the end-windings of all these machines, and this stray 
field is especially large on machines with a small number of poles, 
and, in consequence, high ampere-turns per pole. These stray 
fields at the ends tend to exert a bending or distorting effect on the 
end-windings. In any given machine the distorting force varies 
as the square of the current carried by the coils. Our experience 
with the windings on these machines indicated that they were 
being subjected to enormous forces in the end-windings. The 
oscillograph tests gave an indication as to the amount of this 
force. As the machines could give about IS times full-load cur- 
rent momentarily on short-circuit, the force acting on these end- 
windings would be 225 times normal; in this case, therefore, these 
forces were so great that it became a serious problem to devise a 
type of bracing on the end-windings sufficient to withstand such a 
force. It should also be borne in mind that probably as many 
short-circuits came, in one day, on these generators, as the or- 
dinary high-voltage power-house generator is called upon to 
sustain in one year. While ninety-nine shocks out of a hundred 
might not be sufficient to do damage, yet if the shocks occur fre- 
quently enough, the hundredth one will soon be reached. In our 
endeavors to support these windings against movement, probably 
the most complete system of bracing ever applied to alternating- 
current generators was developed and used on these machines. 

But in spite of this there was evidence of movement at times,' 
It thus became evident that some method of limiting this short- 
circuit current to the value originally intended; namely, about 


six times full-load current, would have to be applied. This was 
done by placing an unsaturated choke-coil, or impedance coil, on 
the trolley side of each machine This coil takes up a comparatively 
small voltage tinder normal operation, but in case of a short- 
circuit, the electromotive force generated in it is sufficient to limit 
the current rush to less than half the value it would attain without 
this coil Thus as the shock on the end-windings of the generators 
varies as the square of the current, it is evident that cutting this 
current in half would cut the shock to one-quarter of its former 
value, which, with the method of bracing used on these machines, 
would mean the difference between good and bad. 

When these choke-coils were installed, the results on the 
power house were evident The shocks on the machines were very 
greatly reduced, so reduced that we do not fear future trouble 
from this source. It is interesting to note that No. 4 machine; 
that is, the 4260 kilovolt-ampere generator, referred to before, 
was put in service a considerable time before the choke-coils were 
installed, and it went through the most severe short-circuits ever 
encountered on this system. Its armature winding has never 
shown any distress. This is partly because, in the design of this 
machine, the difficulties to be overcome were known, and the 
remedies could be applied in the most suitable manner. 

An interesting point in connection with the use of the cage 
windings on these generators, is that the apparent regulation of the 
system has been improved. This was anticipated, but the actual 
result in practice was more pronounced than was expected. In 
installing new rotors for these machines with the heavier cage 
dampers, the inherent regulation of the generators was made 
somewhat poorer than before, partly in order to accommodate 
certain structural features in the rotor. It was anticipated that 
the cage winding with its damping effect would, to a certain 
extent, mask this poor regulation by making the machine sluggish 
as regards fluctuation in voltage with sudden variations in load. 
In practice it was found that, with the later rotors with their 
poorer inherent regulation, the average regulation of the system 
was considerably better than before, thus indicating that most of 
the disturbances in the voltage, when the old rotors were used, 
were due to sudden changes in load, while the slow variations 
were taken care of by the automatic regulators. With the new 
rotors the voltage changes are so slow, that the Tirrill regulator 


has plenty of time to act before any serious disturbance can take 

It must be borne in mind that in one way this New Haven 
power-house installation was more difficult than anything under- 
taken heretofore, and that is, in the use of 11,000 volt generators 
with one terminal connected directly to ground Taking this 
condition into account, together with the enormous current 
rushes with consequent shocks on the winding, and the single- 
phase operation of units of such large capacity, it may reasonably 
be claimed that this was the most difficult case of alternating- 
current generation ever undertaken. 


FOREWORD In 1906, the Westinghouse Company contracted to 
build a 2000 kw uni-polar type generator direct coupled to a 
1200, revolution steam turbine. Many difficulties were en- 
countered in shop tests on this machine, winch were apparently 
corrected, but upon installation and operation on the customer's 
premises, many new and totally unexpected difficulties arose. 

This paper illustrates how a responsible manufacturing com- 
pany will throw its whole engineering and manufacturing en- 
deavors into correcting serious difficulties. It also serves to give 
student engineers a good idea of the practical side of manufac- 
turing engineering. Fearing the results of the engineering 
efforts expended on this machine would eventually be lost, the 
author prepared them for presentation at the twenty-ninth 
annual convention of the American Institute of Electrical 
Engineers at Boston, June, 1912. (ED.) 

THIS paper is not intended to be a theoretical discussion of 
the principles of unipolar machines; neither is it a purely 
descriptive article. It is a record of engineering experiences 
obtained, and difficulties overcome, in the practical development 
of a large machine of the unipolar type. Some of the conditions 
of operation, with their attendant difficulties, proved to be so 
unusual that it is believed that a straightforward story of these 
troubles, and the methods for correcting them, will be of some 

Two theoretical questions of unipolar design have come up 
frequently; (1) whether the magnetic flux rotates or travels 
with respect to the rotor of the stator; and (2) whether it is 
possible to generate e.m.fs. in two or more conductors in series 
in such a way that they can be combined in one direction, with- 
out the aid of a corresponding number of pairs of collector rings, 
to give higher e.m.fs. than a single conductor. 

To the first question the answer may be made that in the 
machine in question, it makes no difference whether the flux 
rotates or is stationary; the result is the same on either assump- 
tion. To the second it may be said that when the theory of inter- 
linkages of the electric and magnetic circuits is properly con- 
sidered, it is obvious that the resultant e.m.f . is always equivalent 
to that of one effective conductor. It has been proposed in the 




past, by means of certain arrangements of liquid conductors in 
insulating tubes, to add the e m fs of several conductors in series 
but such a scheme does not appear to be a practical device There- 
fore, the theoretical considerations being largely eliminated, the 
author confines himself to the practical side only. 

In 1896 the writer designed a small unipolai generator of 
approximately three volts and 6000 amperes capacity at a 
speed of 1500 rev. per mm. This machine was built for meter test- 
ing and the occasion for its design lay in the continued trouble 
encountered with former machines of the commutator type 
designed for very heavy currents at low voltages. 

The general construction of this early machine is shown in 
Fig. 1. The rotating part of this machine consisted of a brass 
casting, cylindrical shaped, with a central web, very similar 
to a cast metal pulley. The two 
outer edges of this pulley or ring 
served as collector rings for col- 
lecting the current as indicated 
in the figure, while the body of the 
same ring served as the single 
conductor. The object of this 
construction of rotor was to obtain 
a form which could be very quickly 
renewed in case of rapid wear, as 
this arrangement would allow a 

small casting to he made and simply turned up to form a new- 
rotor. However, this renewal feature has not been of very 
great importance for the rotor of the first machine was replaced 
only after 12 years' service. This period of course did not 
represent continuous service, for this particular machine was 
used for meter testing purposes or where large currents were 
required only occasionally, 

A number of peculiar conditions were found in this machine. 
In the initial design the leads for carrying the current away 
from the brushes were purposely carried part way around the 
shaft in order to obtain the effect of a series winding by means 
of the leads themselves. In practice, they were found to act in. 
this manner and, in fact, they over-compounded the machine 
possibly 30 to 40 per cent. In consequence, it was necessary 
to shunt them by means of copper shunts around the shaft in 
ihe opposite direction. 

FIG. 1 


Shortly after this machine was put in operation there was con- 
siderable cutting of the brushes and rings, especially at very 
heavy currents. It was found that block graphite, used as a 
lubricant, gave satisfactory results. This machine was operated 
up to 10,000 to 12,000 amperes for short periods. 

The description of the above machine has been gone into rather 
fully, as it was a forerunner of the 2000-kw. machine which will 
be described in the following pages. The general principle of 
construction and the general arrangement of the two parts, or 
paths, of the magnetic circuit are practically the same in the two 
machines, as will be shown. 

In 1904, due to the rapidly increasing use of steam turbines, 
the question of building a turbo-generator of the unipolar type 
was brought up, and an investigation was made by the writer 
to determine the possibilities. This study indicated that a 
commercial machine for direct connection to a steam turbine 
could be constructed, provided a very high peripheral speed was 
allowable at the collector rings or current collecting surfaces. It 
appeared that the velocity at such collector surfaces would have 
to be at least 200 to 250 feet per second, in order to keep the 
machine down to permissible proportions of the magnetic 
circuit, and to allow a reasonably high turbine speed. Con- 
trary to the usual idea, the very high speeds obtainable with 
steam turbines are not advantageous for unipolar machines. 
For example, while maintaining a given peripheral speed at the 
current collecting surface, if the revolutions per minute of the 
rotor are doubled, then the diameter of the rotor collecting 
rings is halved, and the diameter of the magnetic core surrounded 
by the collector rings is more than halved, and the effective 
section of core is reduced to less than one-fourth. The e.m.f. 
generated per ring or conductor, therefore, on the basis of flux 
alone, would be reduced to less than one-fourth, but allowing 
for the doubled revolutions per minute, it becomes practically 

On the other hand, if the revolutions are reduced, while the 
speed of the collector ring is kept constant, then the e m.f . 
per ring can be increased, as the cross section of the magnetic 
circuit increases rapidly with reduction in the number of revo- 
lutions. But-at a materially reduced speed, the total material 
in the magnetic circuit becomes unduly heavy. In consequence, 
if the speed is reduced too much, then the machine becomes too 
large and expensive, while with too great an increase in speed, 



the e,m.f. per ring becomes low or the peripheral speed of the 
rings must be very high. It is desirable to keep the number 
of collector rings as small as possible, for each pair of rings handles 
the full current of the machine, and therefore any increase 
in the number of rings means that the full current must be col- 
lected a correspondingly large number of times. Therefore, 
it works out that the range of speeds, within which the unipolar 
machine becomes commercially practicable, is rather narrow. 
In 1906, an order was taken for a2000-kw. 1200-rev.permin., 
260-volt, 7700-arnpere unipolar generator to be installed in a 
Portland cement works near Easton, Pa. The fact that it is 
a cement works should be emphazised, as having a considerable 
bearing on the history of the operation of this machine, as will 
be shown later* 

jnnnnnn nn 

PIG. 2 

This 2000-kw. machine does not represent any theoretically 
radical features, being similar in type to the smaller machine 
already described, but modified Somewhat in arrangement to 
allow the use of a large number of current paths and collector 
rings. The general construction of this machine is indicated 
in Fig. 2. 

The stator core and the rotor body are made of solid steel, 
the stator being cast, while the rotor is a forging. There are 
eight collector rings at each end of the rotor, the corresponding 
rings of the two ends being connected together by solid round 
conductors, there being six conductors per ring, or 48 con- 
ductors total. In each conductor is generated a normal 
e.m.f. of 32.5 volts, and with all the rings connected in series, 
the total voltage is 260. 

The stator core, at what might be called the pole face, is built 



up of laminated iron, forming a ring around the rotor. This 
was laminated in order to furnish an easy method for obtaining 
the stator slots in which the conductors lie which connect to- 
gether the brushes or brush holders for throwing the pairs of 
rings in series. The slots in the stator laminations were made 
open, as-indicated in Fig. 3, in order to readily insert the stator 
conductors. There are 16 slots in this ring, and in each slot 
there is placed one large solid conductor. 

As first assembled, non-metallic wedges were used to close 
these slots, but later these were changed to cast iron for reasons 
which will be explained later. 

The rotor core consists of one large forging, as indicated in 
Fig. 2. Lengthwise of this rotor are 12 holes for ventilating 

FIG. 3 

FIG. 4 

purposes originally 2f in. diameter. Each of these holes con- 
nected to the external surface by means of nine If in. radial 
holes at each end of the rotor, these holes corresponding to mid- 
positions between the collector rings. It was intended to take 
air in at each end of the rotor and feed it out between the collec- 
tor rings for cooling. In addition, as originally constructed, 
there was a large enclosed fan at each end, as indicated in Fig. 
4. These fans took air in along the shaft and directed it over 
the collector rings parallel to the shaft. The object of this was 
to furnish an extra amount of air for cooling the surfaces of the 
rings, and the brushes and brush holders, as it was estimated 
that the brushes and brush holders themselves could conduct 
away a considerable amount of heat from the rings by direct 



contact, and that the cooling air from the fans, circulating among 
the brush holders, would carry away this heat. These fans 
were removed duriag the preliminary tests, for reasons which will 
be given later. 

The rotor collector rings consisted of eight large rings at each 
end, insulated from the core by sheet mica, and from each other 
by air spaces between them. Each ring has 48 holes parallel 
to the shaft. These holes are of slightly larger diameter than 

FIG. 5 

the rotor conductors outside their insulation. Six holes in 
each ring were threaded to contain the ends of six of the conduc- 
tors which were joined to each ring. The six conductors con- 
nected to each ring were spaced symmetrically around the core. 
Fig. 5 shows this construction. 

The rotor conductors, 48 in number, consist of one in. copper 
rods, outside of which is placed an insulating tube of hard ma- 
terial. Each conductor, in fact,- consists of two lengths arranged 
for joining in the middle. The outer end of each conductor 
is upset to give a diameter 
larger than the insulating tubes, , 
and a thread is cut on this ex- ~~ 
panded part. After th^ rings 
were installed on the core, the 
rods were inserted through the E 
holes to the threaded part of a 
ring and were then screwed p IG . 5 


At the middle part of the rotor core, a groove is cut as shown 
in Fig. 6. Into this groove the two halves of each conductor 
project. These two ends are then connected together by strap 
conductors in such a way as to giye flexibility in case of expan- 
sion of the conductors lengthwise. This arrangement is also 
shown in Fig. 6. 

With this arrangement there is no possibility whatever of 
the conductors turning after once being connected. There is 



a series of holes from the axial holes through the shaft to this 
central groove, for the purpose of allowing some ventilating air 
to flow over the central connections. 

As originally constructed, the conductors passed through com- 
pletely enclosed holes near the surface of the rotor core, as in- 
dicated in Fig. 7. This construction was afterwards modified 
to a certain extent. The face of the rotor at this point was also 
solid, as originally constructed. This was afterwards changed, 
as will be described later. 

The collector rings, as originally constructed, consisted of 
a base ring with a wearing ring on the outside, as shown in Fig. 
8. Both rings were made of a special bronze, with high elastic 
limit and ultimate strength. On the preliminary tests these 
rings showed certain difficulties and required very considerable 
modifications, and several different designs were developed 
during the preliminary operation, as will be described. 

FIG. 7 


FIG. 9 

FIG. 8 

The eight sets of brush holders at each end are carried by 
eight copper supporting rings. These supporting rings are 
insulated from the frame of the machine but are connected in 
series by means of the conductors through the stator slots. 
There are 16 brush holders studs per ring and two brush holders 
per stud, each capable of taking a copper leaf brush f in. wide 
by If in. thick. These. brush holders are spaced practically 
uniformly around the supporting rings. The supporting copper 
rings are continuous or complete circles, so that the current 
collected from the brushes are carried in both directions 
around the ring. There are two conductors carried from each 
ring through the stator slots to a ring on the opposite side of the 
machine, in order to connect the various brush holders in series. 
The arrangement is illustrated in Fig. 9, 

The above description represents the machine as originally 


constructed and put on shop test. From this point on, the 
real story begins. Various unexpected troubles developed, each 
of which required some minor modification in the construction of 
the machine and, moreover, these troubles occurred in series* 
that is, each trouble required a certain length of time to de- 
velop, and each one was serious enough to require an immediate 
modification in the machine. In consequence, the machine 
would be operated until a certain -difficulty would develop; that 
is, that trouble would appear which took the least time to de- 
velop. After it was remedied, a continuation pf the test would 
show a second trouble which required a remedy, and so on. 
Some of these troubles were of a more or less startling nature 
as will be described later. 

This machine, after being assembled according to its original 
design, was operated over a period of several weeks in the testing 
room of the manufacturing company. It was operated both 
at no load and at full load, and a careful study was made of all 
the phenomena which were in evidence during these tests. 

The machine was first run at no-load without field charge 
to note the ventilation, balance, and general running conditions 
of the machine. The ventilation seemed to be extremely good, 
especially that due to the fans on the ends o the shaft. The 
noise, however, was excessive so much so that anyone working 
around the machine had to keep his ars padded. At first it 
was difficult to locate the exact source of this noisfe, but it was 
determined that the end fans were responsible for a considerable 
part of it. 

On taking the saturation curve of the machine, it was found 
to be extremely sluggish in following any changes in the field 
current. The reason for this sluggishness is obvious from the 
construction of the machine, each magnetic circuit of the rotor 
core being surrounded by eight continuous collector rings of 
very heavy section, and also by eight brush holder supporting 
rings of copper of very low resistance. These rings, of course, 
formed heavy secondaries or dampers which opposed any change 
in the main flux. The total effective section of these rings was 
equivalent in resistance to a pure copper ring having a section 
of 49 sq. in. One can readily imagine that such a ring would be 
very effective in damping any sudden flux changes. This slug- 
gishness of the machine to changes in flux, however, was not 
an entirely unexpected result* 

The saturation curve showed that the machine could be carried 


considerably higher in voltage than originally contemplated, for 
apparently the magnetic properties of the heavy steel parts 
were very good, and it was possible to force the inductions in 
these parts to much higher density than was considered prac- 
ticable in working out the design. This gave considerable lee- 
way for changes which later were found to be necessary. 

In taking the saturation curve, the power for driving the 
machine was measured and it was found that there were prac- 
tically no iron losses in the machine; that is, at full voltage 
at no-load the total measured losses were practically the same 
as without field charge. This apparently eliminated one pos- 
sible source of loss which was anticipated, namely, that due to 
the large open slots in the stator pole face, these slots being very 
wide compared with the clearance between the stator and rotor 

After completion of this test the machine was then run on 
short circuit Apparently, as there was no iron loss shown in 
the no-load full voltage condition, the short circuit test with full 
load current should cover all the losses in the rotor which would 
be found with full" load current at full voltage. Experience 
afterward proved this assumption to be correct, for in its final 
form the machine would operate under practically the same 
condition as regards temperature, etc., at full voltage as it would 
show at short circuit, carrying the same current, the principal 
difference being the temperature of the field coil. 

It was in this short circuit temperature run that the real 
troubles with the machine began. The measured losses, when 
running on short circuit, were somewhat higher than indicated 
by the resistance between terminals times the square of the 
current. These extra losses were a function of the load and 
increased more rapidly with heavy currents. The measured 
power indicated that these excess losses were principally due to 
eddy currents. However, the total losses indicated in these 
preliminary tests, although somewhat higher than calculated, 
were still within allowable limits, as considerable margin had been 
allowed in the original proportions to take care of a certain 
amount of loss. It was therefore considered satisfactory to go 
ahead with the short circuit tests, and in making these it was 
the intention to operate long enough to determine the neces- 
sary running conditions as regards lubrication, heating, etc. 

As mentioned before, the original collector rings of the machine 
each consisted of a base ring upon which was mounted a second- 
ary or wearing ring, it being the intention to have this latter 


ring replaceable after it was down to the lowest permissible 
thickness, as it would be rather expensive and difficult to replace 
the base ring which carried the rotor conductors. As the inner 
*ing was shrunk on the core and the outer ring was shrunk on 
over the base ring, with a very small shrinkage allowance, it 
was considered that the outer ring was in no danger of loosening 
on the inner ring, especially as both rings, being of bronze, and 
in good contact, should heat each other at about the same rate. 
This assumption, however, was wrong. The machine was put 
on short circuit load of about 8000 amperes early one evening 
and an experienced engineer was left in charge of it until about 
midnight. Up to that time the machine was working perfectly, 
with no under heating in the rings and no brush trouble, although 
vaseline lubrication was used. About midnight the engineer 
left the machine in charge of a night operator, and at about three 
o'clock in the morning this operator saw the brushes beginning 
to spark and this very rapidly grew worse, so that in a very few 
minutes he found it necessary to shut 
the machine down. An examination 
then showed that several of the outer 
rings had shifted sideways on the base 
ring, as indicated in Fig. 10. One of 
these rings had even moved into contact 
with a neighboring ring so as to make FIG. 10 

a dead short circuit on the machine. It 

was also noted that all the rings which loosened were on one side 
of the machine, and that the surfaces of the rings exposed to 
the brushes were very badly blistered. The brushes also were 
in bad shape, indicating that there had been excessive burning 
for a short time. An investigation of the loose rings showed 
that they had loosened on their seats on the inner or base rings. 
Investigation then showed that a temperature rise of 70 to 80 
deg. cent., combined with the high centrifugal stresses, would 
allow the rings to loosen very materially. It was then assumed 
that as the ring had heated up, bad contact had resulted be- 
tween the inner and outer rings and this, in turn, had caused 
additional heating, so that the temperature rose rather suddenly 
after bad contact once formed. It developed later that this 
was probably not the true cause of the trouble, but at the time 
it was considered that the remedy for the trouble was in the 
tise of rings which could be shrunk on with a greater tension. 
It was then decided to try steel outer rings instead of bronze 


on the end where the bronze rings had loosened. However, upon 
loading the machine, after applying the steel rings, a new diffi- 
culty was encountered. It was found that the loss was very 
greatly increased over that with the bronze rings. This loss 
was so excessive as to be prohibitive, as far as efficiency was 
concerned, and also the tests showed excessive heating of the 
rings and of the machine as a whole. Also, there were continual 
small sparks from the tips of the brushes, these sparks being 
from the iron itself, as indicated by their color and appearance. 
However, during the time these rings were operated there did 
not seem to be any undue wear of either the brushes or the rings, 
but obviously there was continued burning, as indicated by the 
sparks. With thes,e steel rings it was found to be impossible 
to operate continuously at a current of 8000 amperes, due to 
the heating of the steel rings in particular and everything in 
general. At a load of 6000 amperes the loss was materially re- 
duced and it was possible to operate continuously but with very 
high temperatures. The tests showed that, 
with the steel rings, at full rated current, the 
loss was approximately 200 kw. greater than 
with the bronze rings, or about 10 per cent 
of the output. With both ends equipped 
_ with steel rings, this would have been prac- 
FIG 11 tically doubled. 

While this was recognized as an entirely 
unsatisfactory operating condition, yet it allowed the machine to 
be run for a long enough period to determine a number of other 
defects which did not develop in the former test. One of these 
defects was an undue heating of the rotor pole face. This was 
obviously not due directly to bunching of the flux in the air gap 
on account of the open stator slots, for this heating did not appear 
when running with normal voltage without load. Further investi- 
gation showed that this was apparently due to some flux dis- 
torting effect of the stationary conductors in the stator slots, 
which carried about 4000 amperes each at rated load. On 
account of ample margin in the magnetizing coils the air gap 
was then materially increased, with some benefit. A further 
improvement resulted in the use of magnetic wedges, made of 
cast iron, in place of the non-magnetic wedges used before. 
These wedges are illustrated in Fig. 1 1 . This produced a further 
beneficial effect, but there was still some extra heating in the 
pole face. Cylindrical grooves alternating $ in. and 1 in. deep 



and about J in. wide, with a J ia. web of steel between, were 
then turned in the pole face. The resultant pole face was there- 
fore crudely laminated, as shown in Fig. 12. Also, on account 
of an apparent local heating of the metal bridge over the rotor 
slots, a narrow groove was cut in the closed bridge above each 
rotor slot, thus changing it to a partially open slot, as shown in the 
figure. This effectively eliminated the excess loss in the rotor 
pole face. This, however, led to another unexpected difficulty, 
which will be described later. 

After this trouble was cured, the short circuit test was con- 
tinued with a current of about 6000 amperes. After a con- 
siderable period of operation, a very serious difficulty in the 
operation of the machine began to show up ? namely, trouble 
with lubrication. At first the lubrication was vaseline fed on 
to the rings by lubricating pads. This was apparently very 
effective for awhile, but eventually it was noted that slight 
sparking began, which, in some cases, would increase very 
rapidly and, in a comparatively 
short time, became so bad that 
the rings or brushes would be- 
come badly scored or blistered. 
Examination of the sparking 
brushes showed a coating of black 
" smudge " over the surface which 
seemed to have more or less in- 

sulating qualities, A series of tests then showed that when- 
ever sparking began, the contact drop between a brush and the 
collector ring was fairly high and this drop increased as the spark- 
ing increased. For instance, it was found that on good, clean 
surfaces, the voltage drop between the brushes and the ring 
might be 0.3 to 0.5 volt. As each brush carried about 250 
amperes at full load, this represented 75 to 125 watts per 
brush. When this contact resistance rose to about one volt, 
noticeable sparking would begin-, the watts being, of course, 
proportionally higher, afid when the contact drop became as high 
as two volts, representing about 500 watts per brush, very bad 
burning of the brushes and rings was liable to occur. A series 
of tests then showed that vaseline, or any other lubricating oil, 
would tend to form a coating over the brush contact and this 
coating would gradually burn, or be acted upon otherwise by 
the current, so that its resistance increased and the black 
smudge was formed which had more or less insulating qualities. 

FIG. 12 


A great number of tests were then carried out with various 
kinds of lubricants and it was found that anything of an dil 
or grease nature was troublesome sooner or later, as the smudge 
was formed on the brush contact. Then graphite, formed into 
cakes or brushes by means of high pessure, was tried on the rings 
and the results were very favorable compared with anything 
used before. In fact, the tests indicated that soft graphite 
blocks or brushes could furnish proper lubrication for the rings. 
The graphite is a conducting material, and a coating of it on 
the brush contact does not materially increase the resistance 
of the contact. This was supposed to have practically settled 
the question of lubrication and brush contact trouble, but ex- 
perience later gave an entirely new turn to this matter. 
While these tests were being carried on, a study of the ventila- 
tion of the machine was being made. Tha tests indicated that 
the end rings, that is, those next to the exciting coils, were con- 
siderably cooler than those near the center of the machine. 
However, as there were excessive losses and heating in the steel 
rings themselves, it was not possible to make any material im- 
provement until the rings were changed. 

The steel rings at one end of the rotor, and the bronze rings 
at the other end, were then removed and a second set of bronze 
rings was tried. These rings were specially treated in the manu- 
facture so that the elastic limit was very high, and they were 
put on much tighter than in the former case. The load tests 
were then continued and the excess losses were agaia measured 
at various loads. It was found that the losses were very small 
compared with those of the steel rings, thus verifying the former 
results. The temperatures of the rings were much lower than 
with the steel, but it was found that the heating of the rings war 
unequal. It was finally determined that this unequal heating was 
due to the large external blowers which were driving the air over 
the rings in such a way as to heat those next to the center of the 
rotor to a much higher temperature than those at the outer 
ends. It was assumed at first that the air entering the axial 
holes through the core and blowing out between the rings as 
shown in Fig. 4, was more effective on the outer rings, and that 
this possibly caused the difference in temperatures. However, 
the radial holes at the outer ends were dosed, and this made 
but little difference. The axial holes were then closed, and 
while the temperattires of the rings, as a whole, were increased, 
about the same difference as before was found between the end 
rings and the center ones. 


It was then decided to remove the two large bloxvers to de- 
termine whether some other method of ventilation would be 
more effective. \Vhen this change was made the windage of 
the machine was greatly reduced and there was greater uni- 
formity in the temperatures and the average temperature of 
the rings wasonly about 10 deg. higher than with the fans. More- 
over, the windage loss was only about one-seventh as great as 
before, although the average temperature rise was not much 
higher, which indicated that the ventilation through the rotor 
holes was much more effective than that due to the blowers. 
In consequence, it was decided to increase the size of the axial 
holes through the rotor core from 2f in. to 3f in. diameter, and 
to " bell-mouth " them at their openings at the ends, in order 
to give a freer admission of air to the holes. When this was done 
it was found that the temperatures of the rings were lower than 
in any of the preceding tests, and moreover, they were fairly 
uniform. Also after the removal of the blowers, the objection- 
able noise, already referred to, was largely eliminated, so that it 
was not disagreeable to work around the machine. The graphite 
lubrication was continued "with the bronze rings, on this test, 
and no difficulty was encountered, although the machine was 
operated for very considerable periods at approximately 8000 

On the basis of these tests, the machine was shipped to its 
destination and put in service. Then the real difficulties began 
difficulties which were not encountered in the shop tests, princi- 
pally because the conditions under which the machine operated 
in service were radically different from those at the shop, and 
also, because the shop test had not been continued long enough. 
This machine was operated in service, although not regularly, 
for a period of about two months, being shut down at times due 
to difficulties outside of the generating unit itself. However, 
this period of operation of the generator was suddenly ended 
by the stretching of one of the outer collector rings, which 
loosened it to such an extent that it ceased to rotate with the 
inner ring. This required the return of the rotor to the manu- 
facturer. , 

This two months' operation gave data of great practical 
value, and in consequence, a number of minor difficulties were 
eliminated in the repaired rotor. 

Upon the return of the rotor to the shop, an examination of the 
collector rings showed that the separate shrtmk-on type of ring 



was not practicable with any design of nng then at hand. There- 
fore, it \vas decided to make the collector rings in one solid piece 
with a very considerable wearing depth. This necessitated 
the removal of all the base rings and, in fact, it required a com- 
plete dismantling of the entire rotor winding. As the outer ring 
had loosened, there was a possibility of the base nngs loosening 
in the same way, and therefore it was considered necessary to 
apply some scheme for preventing this loosening in case of sudden 
heating and expansion of any of the collector rings. It was then 
decided to apply some form of spnng support underneath these 
nngs, which could follow up any expansion in such a way as to 
keep the rings tight under any temperature conditions liable to 
be met with in practice. The spnng support used consisted of 
a number of flat steel plates arranged around the rotor core, as 
indicated in Fig. 13. These plates were of such length and stiff- 
ness that a very high pressure was required to bend them down to 

conform with the rotor surface. 
These plates were arranged 
around the rotor core and drawn 
down with clamp rings until 
they fitted tightly against the 
mica. The collector ring was 
highly heated and slipped over 
the springs, the clamps being 
removed as the ring was slipped 
on. Tests were made to find 
at what temperature such a ring would loosen. While the best 
arrangement without springs would loosen at about 100 to 125 
deg. cent., it was found that a ring supported, in the above 
manner, was still fairly tight at 180 deg. cent., which was far 
above any temperature which the machine would attain under 
any condition. It may be said here that, after several years' 
operation, this construction still appears to be first class, and 
no loosening of any sort has occurred. 

In removing the winding from the rotor, it ws discovered 
that the insulating tubes over the rotor conductors had traveled 
back and forth along the rods a certain amount* This travel, 
if continued for a long enough period, would apparently have 
injured the insulation, although no trouble had yet developed. 
Apparently, during heating and cooling, the expansion and con- 
traction of the rods would carry the tubes with them lengthwise 
a very small amount. The tubes would then seat themselves in 

FIG. 13 


the supporting rings or core and would not return to their original 
positions. It was found that in the slotted pole face already 
described, the webs or laminations of metal overhanging the 
rotor slots would hold the tube when the rod was traveEng in 
one direction, but would sometimes allow the tube to move 
slightly when the rod traveled in the other direction, so that 
there was a sort of extremely slow ratchet action taking place. 
It was evidently necessary to have the tubes fit rather tightly 
in the retaining or supporting holes in the rings and the core, and 
to have the rods fit rather loosely in the tubes. Also, it ap- 
peared that shellac or other " gummy " material on the inner 
surface of the insulating tubes, was harmful, for wherever shellac 
was, present the insulating tube always stuck to the rod and 
would tear at either side of such place. In consequence, the 
new set of tubes was made with a dry, hard finish on both the 
outside and the inside, and the inside surface was also paraf- 


fined. This, when carried out properly, served to remedy this 

The reconstructed rotor, with the solid collector rings, was 
shipped to the customer and the service was continued. After 
operation for a considerable time, certain extremely serious dif- 
ficulties appeared. One of these was brush trouble, and another 
was undue wear of the rings. 

The brush trouble was a most discouraging one. The machine 
was located in an engine room adjacent to a rock-crushing build- 
ing. Fine dust was always floating around the machine and, 
this dust continuously passing through the machine tended to 
form a deposit immediately behind the brushes as shown in Fig. 
14. This dust packed in rather solidly behind the brush, due to 
the high speed of the rings, and eventually it tended to lift the 
brushes away from the rings. It also showed a tendency to get 
Tinder the brush contact, with consequent increased resistance 
of contact. Frequent removal and cleaning of the brushes 


impracticable, as they were not sufficiently accessible to do 
this readily. This rock dust, packed behind the brushes, 
also had a scouring or grinding action on the rings themselves. 
Accompanying this was an undue rate of wear of the rings. This, 
however, was not entirely mechanical wear, as it appeared also 
to be dependent upon the current carried and was, to some ex- 
tent, due to a burning action under the brush which tended to 
eat away the surface of the rings. However, while the undue 
wear was not altogether due to dust back of the brushes, yet 
this accumulation of dust appeared to have a very harmful 
action on the machine. Various methods were considered for 
overcoming this collection of dust, one of which consisted of 
enclosed air inlets to the machine, fitted with screens for sifting 
out the dust. This lessened the trouble to some extent, but 
it was evident that it would not cure it entirely, as the entire 
machine was so located that dust could come in around the brush 
holders without going through the ventilating channels. 

The method finally adopted for overcoming the difficulty of 
accumulation of dirt was rather startling. It was casually sug- 
gested that the copper leaf brushes be turned around so that 
the rings would run against the brushes, so that the dirt or dust 
over the rings would be " skimmed off " by the forward edge of 
the brushes. This obviously would prevent the collection of 
dirt y but the question of running thin leaf copper brushes on 
a collector ring operated at a speed of about 220 feet per second 
(or 13,200 feet per minute) looked like an absurdity to any one 
with experience in electrical machinery, so that we all hesitated 
at first to consider the possibility of it. However, as something 
had to be done, the writer suggested to the engineer in charge, 
that he change the brushes on one of the rings so that they would 
be inclined against the direction of rotation. This gave no 
trouble and the other brushes were then changed to the same 
direction and the operation ever since has been carried on with 
this arrangement. To the writer this has always seemed an 
almost unbelievable condition of operation, but as there has 
not been a single case of trouble from this arrangement during 
several years of operation, one is forced to believe that it is all 
right. This change entirely overcame the trouble from accumu- 
lation of dirt. However, it did not entirely cure the burning of 
the brushes and rings above described, but rendered the matter 
of lubrication somewhat easier than at first. 

As to the other serious trouble, it was mentioned that there 


was a burning action under the brushes which tended to " eat " 
or " wear " away the surface of the rings. This also tended to 
burn away the brush surfaces, the amount of burning in either 
case depending, to a considerable extent , upon the direction 
of the current. At one side of the machine the brushes 
would wear more rapidly, while at the other side the rings 
would wear faster. The polarity of the current was influential 
in this action. Particles of the metal appeared to travel in 
the direction of the current; that is, where the current was from 
the ring to the brushes, the ring would wear more rapidly, 
while the brush would show but little wear, while at the other 
end of the machine, the opposite effect would be found. How- 
ever, the particles of metal taken from the ring did not deposit, 
or " build up," on the brushes. 

During all this operation, graphite had been used for lubri- 
cation. In the earlier stages, powdered graphite compressed 
into blocks, had been used. Later it was found that very soft 
graphite brushes in insulated holders would give ample lubri- 
cation for the rings. However, even with this lubrication and 
the removal of the dirt trouble, there was still an appreciable 
burning of the brushes and rings as indicated by the more rapid 
wear of the rings at one end of the rotor, and of the brushes at 
the other end. Extended tests showed that this burning was 
a function of the contact drop between the brushes and the rings. 
Neither the nngs nor the brushes would burn appreciably if 
the contact drop between the brushes and the ring could be kept 
very low. When this drop became relatively high (about one 
volt), the rings or brushes would show an undue rate of wear. It 
was found also that, after a considerable period of operation, 
it was very difficult to obtain a low brush contact drop, as the 
brush wearing surface became coated with a sort of " smudge," 
which seemed to have resisting qualities. An analysis of this 
coating showed a very considerable amount of zinc in it, and 
it was determined that the zinc in the collector rings was burning 
out and forming an insulating coating on the brush contacts. 
The remedy for this condition was the application of some clean- 
ing agent which would chemically act on the smudge and dis- 
solve it or destroy its insulating qualities. The right material for 
this purpose was found to be a weak solution of muriatic acid 
about 4 per cent in water. When this was applied to the rings by 
means of a " wiper," at intervals, the brush contact drop could be 
reduced to a very low figure frequently to 0.1 or 0.2 of a volt, 


and the rings would take on a very bright polish. Also, while 
this low contact drop was maintained it was found that the rings 
showed an almost inappreciable rate of wear. However, one 
set of rings continued to wear somewhat faster than the other. 
This difficulty of unequal wear of the two sets of rings was over- 
come by arranging a switch so that the polarity of the two ends 
of the machine could be changed occasionally. 

The temperature of the machine was reduced by the above 
treatment of the rings. Obviously, part of the heat was due to 
the loss at the brush contacts, which, of course, was reduced 
directly as the contact drop was reduced. 

The machine was now running quite decently with compara 
tively heavy loads, from 7000 to 10,000 amperes, and the only 
trouble was in several minor difficulties which were then taken 
up, one at a time, in order to ascertain a suitable remedy. 
These difficulties, however, were not interfering with the regular 
operation of the machine. 

One of the difficulties which finally developed was due to 
stray magnetic fluxes through the bearings. These fluxes, pass- 
ing out through the shaft to the shell of the bearing, consti- 
tuted, in themselves, the elements of a small unipolar machine, 
of which the bearing metal served as collecting brushes. The 
e.m.f . generated in the shaft was a maximum across the two ends 
of the bearing. Consequently the current collected from the 
shaft by the bearing metal -should have been greatest near the 
ends of the bearing, and least at the center. This was the case 
as. indicated by the appearance of the bearing itself, which 
showed evidence of pitting near the ends but none at the center. 

To remedy this trouble, a small demagnetizing coil was placed 
outside the stator frame, at each end of the rotor, between the 
rotor core and the bearings. These coils were excited by direct 
current which was adjusted in value until practically zero e.m.f . 
was indicated on the shaft at the two ends of each bearing. This 
indicated that the unipolar action was practically eliminated. 
This arrangement has been in use ever since it was installed, and 
no more trouble of any sort has been encountered from local 
currents in the bearings or elsewhere. 

Some of the brushes did not show as good wearing qualities 
as desired and various experiments were made with different 
combinations of materials and various thicknesses and arrange- 
ment of the brush laminae. Brass leaf brushes were tried; also, 
mixtures of copper, brass, aluminum and various other leaf metals 


in combination. None of these showed any better than the thin 
copper leaf brush. The tests finally showed that such a brush, 
very soft and flexible, with a suitable spring tension, would 
give very satisfactory results. Also instead, of two brushes 
side by side, a. single brush, covering the full width of a ring, was 
found to be more satisfactory. Some tests were also made with 
carbon brushes, consisting of a combination of carbon or graphite 
combined with some metal, such as copper, in a finely divided 
state. These brushes were claimed to have a very high carrying 
capacity and also to have a certain amount of self -lubrication. 
A set of these brushes was tried on one of the rings, but lasted 
only for a very short time. The apparent wear was rapid, but 
it is not known whether this was due to the very high speed of the 
collector rings, or rapid burning away of the brush or4;he inability 
of this type of brush to quickly follow any inequalities of the 
collector rings. This test was abandoned in a comparatively 
short time. 

After getting rid of the old troubles, a new and unexpected one 
had to appear. For some unknown reason, the insulating tubes 
on the rotor conductors began to break down j also grounds oc- 
curred between the collector rings and the core. 

On account of the delay required in making any changes in 
the rings or rotor winding, the customer arranged with the 
manufacturer to have a new rotor built as a reserve, as it was 
obvious that sooner or later there would have to be considerable 
reconstruction of the insulation on the first rotor due to unex- 
plained short circuits and grounds. A new rotor was at once 
constructed, embodying all the good features of the first rotor, 
with some supposedly minor improvements. The old rotor'was 
then removed for investigation and repairs. The cause of the 
breakdowns of the insulation on the tubes was then discovered. 
The air entering through the axial rotor holes and passing out 
through the radial holes between the rings, carried fine particles 
of cement or Crushed stone dust and this had " sand-blasted M 
the under side of the tubes. When the rotor had been operated 
during the preliminary two months* period, previously described, 
before the replacement of the rings, no evidence of thfe sand- 
blasting had been visible. Investigation showed that the in- 
sulating tubes in the former winding had been made with a fuller- 
board base, which is rather soft and fibrous in its construction. 
The tubes on the second winding had been made with " fish " 
paper instead of fullerboard, in order to give a hard finish on the 



inside and outside. It was due to this hard material that the 
troubles from sand blasting occurred. However, fish paper 
tubes were superior to the fullerboard in strength and other 
qualities, and as they were inferior only, in this one character- 
istic, they were used again in rewinding the rotor, but where- 
ever the tubes were exposed in passing from one ring to the next, 
they were taped over with several layers of soft tape whch was 
also sewed. This gave a soft finish which would resist sand- 
blasting, and no trouble from this source has occurred for several 

From the breakdowns to ground, it was evident that an entire 
replacement of the rings was necessary in order to repair the 
mica bush or sleeve lying beneath the rings. This required 
the removal of the entire rotor winding and rings. It was found 
that cement dust coming up through the radial holes had sifted 

in through various crevices- or openings 
around the holes and that, finally, con- 
ducting surfaces and paths were formed 
which allowed the current to leak to 
ground sufficient - to eventually burn the 
insulation. Therefore, when replacing 
the mica sleeve over the rotor, extra care 
was taken to fit insulating bushings at 
the top of the radial holes in such a 
way as to seal or -close all joints, thus 
allowing no leakage paths between 
collector rings and the body of the cote. This is shown in Fig. 15. 
No further trouble has occurred at this point. 

In removing the collector rings for these repairs, it was found 
that the flat spring supports shown in Fig. 13 had been entirely 
effective and there was no evidence whatever of any disturbance 
of the rings on the core, and there was no injury to the mica, 
such as would be shown by any slight movement. The rings 
were also very tight so that it took a very considerable temper- 
ature to loosen them sufficiently for removal. 

In view of the delay and expense of repairing one of these 
rotors when the collector rings had to be removed, with the pos- 
sibility of damaging the insulating tubes over the conductors, 
and the insulating bush over the core, it was then decided that 
a movable wearing ring was practically necessary in order to 
make this mgdhine a permanent success. Therefore, the 
problem of a separate outside wearing ring, as originally con-' 

FIG. 15 



templated, was again taken up. The difficulty, already de- 
scribed, of the zinc burning from the rings and forming a coating 
on the brushes, indicated that some other material, without 
such a large percentage of zinc, should give better results. The 
difficulty was to obtain such a material, with suitable charac- 
teristics otherwise. All data at hand showed that rings, with 
desirable characteristics electrically, did not have the proper 
elastic limits, or proper expansion properties when heated. In 
other words, when such rings were shrunk on the base or sup- 
porting ring they would stretch to such an extent, when cooled, 
that they would become loose again with very moderate in- 
crease in temperature. The solution of this problem of a separate 
ring construction was found in the use of some spring arrange- 
ment -underneath the outer ring which would still keep it tight 
on the inner ring even when hot. The spring arrangement 

PIG. 16 

used under the inner rings, as shown in Fig. 13 was then applied 
with certain modifications. In order to get good contact be- 
tween the inner and outer rings for carrying the current, each 
of these steel springs or plates was covered by a thin sheet of 
copper as shown in Pig, 16. While each copper shest was of 
comparatively small section, the large number of springs used 
gave sufficient total copper to carry the current from the outer 
to the inner or base ring without any danger of current passing 
through the spring plates themselves. This arrangement was 
used in reconstructing this rotor and has proven entirely 

In order to determine the effects of various materials without 
zinc, or with but a small quantity of it, a number of rings were 
fitted up on a test rig and were operated for long periods with 
currents, up to 12,000 amperes in some cases. In these tests, 


four different kinds of material were used, all of them representing 
different mixtures of copper with a small percentage of other 
materials but with little zinc in any of them. It was feared 
that copper brushes on the copper rings would not work satisfac- 
torily, but while there was apparently some difference between 
the action of the different rings, it was found that copper brushes 
running on copper were, in general, satisfactory. The brushes 
were in-clined against the rings, as in the actual machine, 
during this series of tests. 

These tests were carried through with various numbers of 
brushes, etc. It was found that the number of brushes could 
be reduced to about one-third the full number, and still collect 
the total rated current, but that any great reduction from the full 
number of brushes made the operation of the rings and brushes 
more sensitive, and more attention was required to keep them 
in perfect condition. It was also found that any hardness or 
undue "springiness" in the brushes, or brush material, would 
tend to give increased wear. Brushes of very thin leaf copper, 
eventually gave best results. It was also shown by these tests 
that if a very good polish could be maintained on the rings, 
the rate of wear from day to day was practically unmeasurable 
on account of its smallness. 

As a result of these ring tests, the rotor undergoing repair 
was equipped with outside copper wearing rings, spring sup- 
ported. The material in the rings was about 92 per cent pure 
copper, 2 per cent zinc and 6 per cent tin. 

The rotor was then installed in service and has been operating 
steadily for several years, with entire success. The other rotor, 
which had been operating while this rotor was being repaired, 
was then thoroughly examined after removal, to determine any 
possible defects. It was noted that the insulating tubes over 
the rotor conductors were badly cracked or buckled in a number 
of places. Upon removal of the rods or conductors it was found 
that the insulating tubes were stuck so tightly to the copper 
rods that they would be torn in pieces in trying to remove them. 
As it had been intended that these tubes should move freely 
on the rods or conductors, as previously described, it was evident 
that there was something radically wrong. The true cause 
of the "trouble was then discovered. In first fitting this set of 
tubes over the rods, they had been too tight, and, in 
order to make them fit easily, the men who assembled 
the machine had reamed them* on the inside to enlarge 


them, and, in doing so, had cut away the inner hard sheet of 
fish paper which had formed the lining, thus exposing a shellaced 
surface. As soon as heated, this shellac stuck the tube to the 
rod so that there could be no possible movement between the 
two. In consequence, when the rods expanded or contracted, 
the tubes moved backward and forward in the supporting holes, 
and wherever they stuck fast in the outer holes, something had 
to give, so that eventually the tubes buckled or cracked or pulled 
open. This was readily remedied by putting on new tubes prop- 
erly constructed. As the rings on this rotor were in very good 
condition with but little worn away, the removable type of 
ring was not added, as this would require turning off a large 
amount of effective material on the existing rings and replacing 
it with new outer rings. It was decided that as there was several 
years' wear in the old rings, it would be of no material advantage 
to throw this away when it could be worn away in service, 
just as well as it could be turned off in a lathe. After the rings 
in this machine are worn down the permissible depth, they will 
be refilled by the addition of the removable type. 

This unipolar generator has now been in service for quits 
a long period, with no difficulty whatever, and with an average 
ring wear of less than 0.001 in. per day, or less than f in. per 
year. This may seem like an undue rate of wear; but in reality 
it is an extremely low rate, if the high peripheral speed, and the 
number of brushes, are considered. This machine operates 
day and night, seven days in the week, and practically contin- 
uously during the entire year. Taking the peripheral speed 
into account, the above rate of Wear represents a total travel 
of each, ring of about 3.6 million miles for each inch depth of wear, 
or about 150 times around the earth along a great circle. Con- 
sidering that there are brushes bearing on each ring at intervals 
of about eight in., a wear of one in., for every 3.6 million miles 
traveled, does not seem unduly large. If, at the same time, it 
is considered that the brushes are collecting from 7500 to 10,000 
amperes from each ring on a total ring surface of about 3 in. 
wide by 42 in. diameter, it is not surprising that there should 
be more or less " wear " due to the collection of this current. 
In fact, the current collected averages from 16 to 20 amperes 
per square inch of the total ring wearing surface. This may be 
compared with standard practice with large d-c. commutators, 
in which H to 2 amperes per square inch of commutator face 
is usual and 3 amperes is extreme. 


On account of the final success of this machine, the story of 
its development is a more pleasant one to tell than is the case 
in some instances where entirely new types of apparatus are 
undertaken. It might be said, after reviewing the foregoing 
description, that many of the troubles encountered with this 
machine could have been foreseen; but such a statement would 
be open to question, for the engineers of the manufacturing 
company were in frequent session on all the various phases and 
difficulties which developed. The writer knows that in many 
cases, after any individual trouble was known, suggestion for 
remedies were not readily forthcoming. The writer does not 
know of any individual machine where more engineering and 
manufacturing skill was expended in endeavoring to bring about 
success, than was the case with this machine. As an example 
of engineering pertinacity, this machine is possibly without a 
rival. A mere telling of the story cannot give more than a 
slight idea of the actual fight to overcome the various difficulties 
encountered in the development of this machine. 

The results obtained were valuable in many ways. Many 
data were obtained which have since been of great use, both from 
a theoretical as well as a practical standpoint, in other classes 
of apparatus. Certain fundamental conditions encountered in 
this machine have led to the study of other allied principles 
which point toward possibilities in other lines of endeavor. 
Therefore this machine, which was very costly in its develop- 
ment, may eventually pay for itself through improvements and 
developments in other lines of design. 

The writer wishes to say a good word for the purchaser of this 
new apparatus. He was long-suffering, and was undoubtedly 
put to more or less trouble and inconvenience, but nevertheless 
he gave opportunity to correct difficulties. He recognized that 
the engineers were confronted with a new problem in this ma- 
:hine and he gave them an opportunity to cany it through to 
success. Apparatus of this type could only be developed to 
full success in commerical operation, as all the difficulties en- 
countered would never have been found on shop test. There- 
fore, the attitude of the customer was of prime importance in 
the development of such a machine. 


FOREWORD About 1909, the use of commutating poles in syn- 
chronous converters was being studied. Suggestions were made 
from time to time that our usual slow speed rotary converters 
shoiild have interpoles. The author, therefore, prepared a short 
article, explaining wherein commutating poles would be of less 
value to rotary converters, of the then usual speeds and con- 
structions, than they would be on direct-current generators. 

11 Late in September, 1910, the Chairman of the Papers Com- 
mittee of the American Institute of Electrical Engineers asked 
the author whether he had any material which could be pre- 
pared for the Institute on very short notice, A rough draft 
of this article was submitted and was at once accepted, with 
instructions to complete it for the following November meeting. 
The author called to his aid, Mr. F. D. Newbury, who added 
about one-half more, covering principally material on existing 
types of rotaries. Most of this latter part has been omitted from 
this reprint, but the author's discussion at the Institute meeting 
has been incorporated as it forms a technical continuation of the 
first part of the paper and brings out that the real need for 
commutating poles in rotary converters would come with higher 

This paper states that the short-circuiting effect of the 
dampers surrounding the commutating poles is considered 
harmful. However later experience has shown that the in- 
creased damping effect of this arrangement more than compen- 
sates for the harmful effects. 

As this paper was written before the term "commutating 
pole" was adopted as standard, the term "interpole" has been 
used throughout. (ED.) 

SYNCHRONOUS converters with interpoles have been used 
^ but little in this country up to the present time (1910). Con- 
sidering that interpole generators and motors have come into 
extensive use in this country, the question will naturally be raised 
why interpole converters have not come into similarly extensive 
use. The reply might be that the introduction of any new type of 
apparatus is a relatively slow process; but, on the other hand, 
interpoles on direct current generators and motors came into 
general use in a relatively short time, especially so in railway 
motors. This indicates that there has been a more or less pressing 
need for interpoles in certain classes of apparatus and the greater 
the need for the change the quicker was the change made. 

Any important change in design or type must be justified 




by engineering and commercial reasons, such as improved per- 
formance greater economy, or lower cost. In the railway motor, 
placed under the car, and more or less inaccessible, improved 
operation at the brushes and commutator, when equipped with 
interpoles, represented a pressing reason for the change in type, 
although the cost and efficiency were not appreciably changed. 
In the direct-current generator with the modern tendency toward 
higher speeds with lower cost, the interpoles represented a 
practical necessity. This has been recognized for several years 
and the change to the interpole type has been made as rapidly 
as circumstances will permit. Also, in variable-speed direct- 
curretit motors interpoles have been in general use for a number 
of years, simply because the interpoles represent a very definite 
improvement in a number of ways. 

New types of apparatus should only be introduced where they 
represent some distinct improvement or advance over existing 
types. Where a new type does not represent such improvement 
and is simply introduced to gratify a personal whim of the 
purchaser, or desire on the part of a manuf acturing company to 
produce something different from other companies, the new 
apparatus, as a rule, will not advance quickly into public favor 
since there is no real necessity for it. 

It is therefore a question 
whether the slowness in the 
introduction of interpoles in 
synchronous converters is due 
to lack of sufficient advan- 
tages, or American engineers 
do not sufficiently appreciate 
their advantages. There ap- 
pears to be room for wide 
differences in opinion on this 
subject. The synchronous 
converter and the direct- 
current generator are two 
qttite different machines, in 
their characteristics, and no 
one can say off hand, that interpoles will give the same results 
in both. In the following is given a partial analysis of the condi- 
tions occurring in the two classes of machines, which will indicate 
wherein interpoles are of greater advantage on direct-current 
generators than on converters. 

FIG. 1 


Taking up first, the direct current generator, it may be 
considered as containing two sets of magnetizing coils, namely, 
the armature and the field windings. Considering the armatiire 
winding alone, the magnetomotive force of the armature winding 
has zero values at poirxts midway between two adjacent brush 
arms or points of collection of current and rises at a uniform rate 
to the point of the winding which is in contact with the brushes. 
This is illustrated in Fig. 1. Therefore the armature winding 
has Its maximum magnetizing affect or magnetomotive foi<~- 
at that part of the core surface where the winding is directly 
in contact with the brushes. However, the magnetic flux- set 
up by the armature winding will not necessarily be a maximum 
at this point, as this depends upon the arrangement of the mag- 
netic or other material surrounding the armature. If this point 
occurs midway between two field poles, then, while the mag- 
netizing effect is greatest at this point, the presence of a large 
air-gap at this same point may mean a relatively small magnetic 
flux, while a much higher flux may be set up by the armature 
winding at the edges of the adjacent field poles. In the usual 
direct-current generator construction without interpoles, the 
position of commutation is almost midway between two adjacent 
poles and therefore the point .of maximum magnetomotive 
force of the armature is also practically midway between poles. 
The absence of good magnetic material over the armature at 
this point serves to lessen the magnetic flux due to the armature 
magnetizing effect, but even with the best possible proportions 
there will necessarily be a slight magnetic flux set up at this 
point. While this field is usually of small value, yet unfor- 
tunately it is of such a polarity as to have a harmful effect on the 
commutation of the machine. During the operation of comma* 
tation, the coil which is being commutated has its two terminals 
short-circuited by the brushes. If this short circuited coil at 
this moment is moving across a magnetic flux or field, it will 
have an e.m.f . set up in it which will tend to cause a loal or 
short circuit current to flow. Such a current is set up by the 
flux due to the armature magnetomotive force described above 
and unfortunately this current flows in such a way as to give the 
same effect as an increased external or working current to be re- 
versed as the coil passes from under the brush. In other words, 
the e.m.f. set up in the short circuited coil by the above field 
adds to the e.m.f. of self induction in the coil due to the reversal 
of the working current. 


Another cause of difficulty in the commutation of a direct 
current machine is the self induction of the armature coils as they 
individually have the current reversed in them in passing from 
one side of the brush to the other. Each coil has a local magnetic 
field around itself, set up by current in itself and its c-sighboring 
coils. The value of this local magnetic field depends upon the 
arrangement of the winding, the disposition of the magnetic 
structure around the coil, the ampere turns, etc. During the 
act of commutation, that part of the local field due to the coil 
which is being commutated must be reversed in direction. It is 
therefore desirable to make the local field due to any individual 
coil as small as possible. This means that the number of tisms 
per coil should be as low as possible, the amperes per coil aiso 
should be as small as possible, while the magnetic conditions sur- 
rounding the coil should be such as to give the highest reluctance. 
By the proper arrangement of the various parts, it is usually 
found that the e.m.f- of self induction, due to the reversal of the 
coil passing under the brush, can be made of comparatively small 
value so that, if no other conditions interfere, good commutation 
could be obtained under practically all commercial operating 
conditions. t However, the magnetic field between the poles set 
up by the armature magnetomotive force as a whole, as described 
above, adds very greatly to the difficulties of commutation. If 
the armature magnetomotive force, or the field due to it, could 
be suppressed, then one of the principal limitations in the design 
and operation of direct-current generators would be removed, 
and the commutation limits would be greatly extended. Or, 
better still, if a magnetic flux in the reverse direction were estab- 
lished at the point of commutation, then the e.m.f. set up by this 
would be in opposition to the e-.m.f. of self induction of the 
commutated coil and would actually assist in the commutation! 

This latter is what is accomplished by interpoles. When these 
are used the brushes on the commutator are so placed that the 
short circuited or commutated coils are directly under the inter- 
pole. Consequently, the maximum magnetomotive force of the 
armature is in exact opposition to that of the interpoles. There- 
fore, the total ampere turns on the interpoles should be equal to 
the total ampere turns on the armature in order to produce zero 
magnetic flux under the interpole or at the point of commutation 
But, for best conditions there should not be zero field, but a 
slight, field in the opposite direction from that which the arma- 
ture winding alone would produce. Therefore, the magneto* 


motive force of the interpole must be greater than that of the 
armature by an amount sufficient to set up a local field under 
the interpole which will establish an e.m.f . in the short circuited 
coils opposite to that set up by the commutated coils themselves 
and practically "equal to it. The excess ampere turns required 
on the commutated poles is therefore for magnetizing purposes 
only and the amount of extra ampere turns will depend upon the 
value of the commutating field required, depth of air-gap under 
the commutating pole, etc. The commutating field required is 
obviously a function of the self induction of the commutated coil 
and evidently the lower the self induction the less commutating 
field will be required. It is evident therefore that the commu- 
tating field under the commutating pole bears no fixed relation 
to the armature ampere turns or to the main field ampere turns, 
but is, to a certain extent, dependent upon the proportions of 
each individual machine. 

It is evident that the magnetomotive force of a given arma- 
ture varies directly with the current delivered, regardless of the 
voltage. Therefore, that part of the interpole magnetomotive 
force which neutralizes that of the armature should also vary 
directly in proportion to the armature current. Also, the self 
induction of the commutated coils will vary in proportion to the 
armature current carried, and therefore the magnetic field under 
the interpole for neutralizing this self induction should also 
vary in proportion to the armature current. It is therefore 
obvious that if the main armature current be put through the 
interpole winding, the magnetomotive force of this winding will 
vary in the proper proportion to give correct commutating con- 
ditions as the armature current varies, regardless of the voltage 
of the machine. This is on the assumption that the entire 
magnetomotive force of the interpole winding is effective at the 
air gap and armature, which implies an absence of saturation in 
the interpole- magnetic circuit. In the usual construction, the 
interpole winding always carries the main armature current 
as indicated above. 

One consequence of the use of the interpole is that somewhat 
less regard need be paid to keeping the self induction of the 
commutating coil at its lowest value. In consequence, there is 
somewhat more freedom in proportioning the armature wading, 
slots, etc., than in a non-interpole machine, and advantage can be 
taken of this in bettering the proportions for other characteristics. 



The conditions of design are therefore not as rigid in the interpole 
as in the non-interpole type. 

The above description of the interpole generator has been gone 
into rather fully, as many of the points mentioned ^vill be re- 
ferred to again in connection with interpoles on synchronous 

The synchronous converter differs from the direct current 
generator in one very important particular, namely, it may be 
considered as motor and generator combined. It receives cur- 
rent from a supply system the same as a motor and it delivers 
current to another system like a direct-current generator. The 
magnetomotive force of the armature winding as a motor acts 
in one direction, while the magnetomotive force of the armature 
winding as a generator acts in the opposite direction. As the 
input is practically equal to the output, it is evident that these 
two armature magnetomotive forces should practically neu- 
tralize each other, on the assumption that the armature mag- 
netomotive force, due to the polyphase current supplied has 
practically the same distribution as -that of the corresponding 
direct-current winding. Assuming that the two practically 
balance each other, then it is evident that one of the principal 
sources of commutation difficulty in direct current generators 

-45-75 -I5 

FIG, 2 

FIG. 3 

is absent in the converter and therefore the limits in commuta- 
tion should be much higher than those of direct-current ma- 



The following diagrams show the distribution of the alternating- 
current and direct-current magnetomotive forces on a six-phase 
rotary converter. The magnetomotive force distribution for 
the alternating-current input is plotted for several different 
positions of the armature. Three different positions are shown 
with the armatures displaced successively 15 electrical degrees. 
The general forms of these distributions repeat themselves for 
further similar displacements. 

These distributions are illustrated in Figs. 2, 3 and 4. It is 
evident from these three figures that the peak value of the mag- 
netomotive force the armature varies as the armature is rotated, 
as indicated by the heights of the center line in the three figures. 

In Fig. 5, the magnetomotive force distribution of Fig. 2 and 
the corresponding direct-current distribution of Fig. 1 are both 
shown, but in opposition to each other. In this figure both are 
shown in proper proportion to each other, taking into account 
the alternating current amperes and the direct-current amperes 
output. The resultant of these two distributions is also indi- 
cated in these figures. 

In Fig. 6 the distributions correspond to Figs. 3 and 1 combined 
and the resultant is also shown. 

Fig. 7 combines Figs. 4 and 1. 

FIG. 4 


It is the resultant magnetomotive force in these three figures 
which is important, as this is the effective magnetomotive force 
which tends to produce a flux or field over the commutated 



coil. It is evident from these figures, which are drawn to scale, 
that this resultant varies in height as the armature is rotated, but 
the maximum is only a relatively small per cent of the direct- 
current magnetomotive force. Therefore, it is obvious that 
one of the principal sources of difficulty in the commutation of 
the direct-current generator is practically absent in the converter, 
and it is also evident from this that the commutating conditions 
in the latter should be materially easier than in the former. 
This has proved to be true by wide experience in the construction 
and operation of converters. 

In the above figures the magnetomotive forces have been 
plotted to scale on the following basis: 

The six-phase converter winding is connected to three trans- 
formers with the so-called diametral arrangement; each of the 
three secondaries is connected across the diameter, or across 
180 deg. points on the winding, the three diameters being dis- 
placed 60 deg. with respect to each other. Assuming the direct 
current in the winding as A, then the maximum value of the 
alternating current in- any one phase of the alternating-current 
end will be equal to f A, or 0.667 A, assuming 100 per cent 
efficiency. However, as the alternating-current input must be 
somewhat greater than the direct-current output, due to certain 

losses in the machine, it is evident that the maximum alter- 
nating current in any one phase must be somewhat greater than 
0.667 A. The field copper losses may be considered as part of 



the output of the rotary. The armature copper Ipsess maybe 
considered as due to an ohmic drop between the counter e.m.f. 
of the armature and the transformer e.m.f., and simply a higher 
transformer e.m.f. must be supplied to overcome this drop and 
therefore it does not effect the true current input of the rotary. 
However, the losses due to rotation, such as iron loss and. the 
friction and windage are excess losses which represent extra 
current which must be supplied to the alternating-current end 
of the rotary. These rotational losses will usually be relatively 
small in a 25-cyde converter, being possibly 4 per cent or 5 per 
cent in a small machine and li per cent to 3 per cent in a large 
machine. In the 60-cyde converters, where the iron losses are 
relatively higher and the speeds are somewhat higher, giving 
greater friction and windage, the rotation losses may be con- 
siderably greater than on 25-cycle machines. Assuming these 
rotation losses will be 3 per cent, then the maximum alternating 

rt ftftrr A 
current per phase = ' y = 0.687 A. The foregoing Figs. 5, 

6 and 7 are worked out on this assumption of 97 per cent rota- 
tional efficiency and on this basis of -mini-mum value of the 
resultant magnetomotive force of the armature at the direct- 
current brush is about 7 per cent of the direct-current magneto^ 

motive force of the same word- 
ing, while the maximum value is 
about 20 per cent. The lower the 
rotational efficiency the smaller 
would be these values, and with 
a rotational efficiency of about 
) per cent, the mfmrmini result- 
ant would fall to zero, while the 
maximum value would be about 
13 per cent. 

The resultant magnetomotive 
force of a synchronous converter 
might be compared with that of 
a direct-current generator with 
compensating windings in the 
pole faces. It is generally known 
that such direct-current generators have much better com- 
mutatiag conditions than ordinary uncompensated machines. 
If such compensating winding on the field of a direct-current 
machine covered symmetrically the whole armature surface. 

p IG 


then the armature reaction could be completely annulled, which 
is not the case in the converter. But with compensating wind- 
ings located only in the pole faces, then the armature magneto- 
motive force midway between the poles could not be completely 
annulled, unless over-compensation is used, and the resultant 
would be as shown in Fig. 8, which is not quite as good as the 
average resultant in the converter. The commutating con- 
ditions in the converter can therefore be considered as at least 

In the application of interpoles to the synchronous converter 
the same principles should hold as in a direct-current generator, 
namely, the interpole magnetomotive force should be sufficient 
to neutralize that of the armature winding and, in addition, should 
set up a small magnetic flux sufficient to overcome the self in- 
duction of the commutated coil. As the magnetomotive force 
the armature varies between 7 per cent and 20 per cent shown in 
the above figures, it is evident that perfect compensation of this 
cannot be obtained and that therefore only some average value 
can be applied. Assuming that 15 per cent will be required on 
the average to compensate for this, then in addition the inter- 
pole winding must carry ampere turns sufficient to set up the 
small magnetic field for commutation. Thus the total ampere 
turns on the interpole will be equal to 15 per cent of the armature 
direct-current ampere turns plus a small addition for setting up 
the useful or commutating field. In the direct-current gen- 
erator, the ampere turns on the interpoles must equal the total 
armature ampere turns plus a corresponding addition for the 
commutating field. It is therefore evident that an interpole 
winding on a converter will naturally be very much smaller than 
on a direct-current generator, and in general it is between 25 per 
cent and 40 per cent of the direct-current. 

In the pulsating resultant magnetomotive force in the con- 
verter there lies one possible source of trouble with interpoles 
Assume, for example, the total ampere turns on the interpoles 
are equal to 30 per cent of the direct-current ampere turns on the 
rotary and that 15 per cent of this is for overcoming the average 
value of the resultant magnetomotive force, then an average 
of 15 per cent will be available for setting up a commutating 
field; but, according to the above diagrams, the resultant mag- 
netomotive force of the armature varies from 7 per cent to 20 per 
cent. With a total interpole winding representing 30 per cent, 


then the effective or magnetizing part will vary from 30-7 
to 30-20; that is, from 23 per cent to 10 per cent. The effective 
magnetomotive force therefore tends to vary over quite a wide 
range so that the commutating field would also tend to vary up 
or down over a velry considerable range, which is an undesirable 
thing for commutation. However, as this pulsation is at a 
fairly high frequency it tends to damp itself out by setting up 
eddy currents in the structure of the magnetic circuit. If a 
good conducting damper or closed circuit were placed around the 
interpole, it is probable that this pulsation would be almost 
completely eliminated, but such a damper possesses certain dis- 
advantages, as will be shown later. 

In practice this pulsation of the armature reaction under 
the interpoles is apparently not noticeably harmful in most 
cases, as evidenced by the fact that well-proportioned interpole 
converters in commercial service show no undue trouble at the 
commutator or brushes, 

Due to the relatively small number of ampere turns required 
on the interpole of a converter compared with those required on 
a direct-current generator, the design of the interpoles in the two 
cases presents quite different problems. In the direct-current 
generator the interpoles carry ampere turns, which in all cases 
are greater than the armature ampere turns, as explained before. 
As the field ampere turns on the main poles are, not infrequently, 
but little greater than the armature ampere turns, it is evident 
that the interpole winding may, in some cases, carry as many 
ampere turns as the main field windings. While but a small 
per cent of these interpole windings is effective in producing 
flux under the pole tip, yet they are all effective in producing 
leakage from the sides of the poles. As the interpoles are gen- 
erally small in section compared with the main poles, and as 
they may carry ampere turns equal to the main poles, it is 
evident that the effect of leakage may be relatively great on the 

For instance, if the leakage on the main poles is 15 per cent of 
the useful flux, then, with the same total leakage on the inter- 
poles, this may represent a very high 'value compared with the 
useful flux, due to the small section of the interpole and the 
relatively low useful interpole flux. In consequence, it is con- 
siderable of a problem to proportion the interpoles of a direct 
current generator so that the leakage flux will aot saturate the 
interpoles at some part of the circuit. If they saturate, then 


part of the ampere turns on the interpole are expended in such 
saturation and the part thus expended must be counted off from 
the extra or excess interpole ampere turns. If, for example, 
the interpole winding requires 100 per cent for overcoming the 
armature and there is 20 per cent extra ampere turns for setting 
up a useful flux, then any saturation in the interpole circuit must 
represent additional ampere turns on the field, as the above 
120 per cent is necessary for useful flux and for neutralizing the 
armature. With reduced current, and consequent lower satura- 
tion, these additional interpole turns become effective in mag- 
netizing the gap and thus the commutating flux is too strong. 
At greatly increased load, more ampere turns are required for 
saturation, and the commutating flux is altogether too weak. 
It is thus evident that a machine with highly saturated inter- 
poles will not comtnutate equally well for all loads. Herein 
lies a problem in the design of interpole generators, as it is 
difficult to maintain a relatively low saturation in the interpoles 
due to their small section and high ampere turns which cause 
leakage. It is well known that in the main poles of the generator, 
a leakage flux which is higher than the useful flux is objection- 
able, from the designer's standpoint; and yet in the use of inter- 
poles this is a normal condition rather than an exception. 

In the synchronous converter the conditions are somewhat 
different due to the fact that the interpole ampere turns are 
usually only 25 per cent to 40 per cent as great as on a correspond* 
ing direct-current generator. The leakage at the -sides of the 
poles becomes relatively much less, while the usefrjl induction 
remains about the same as on the direct-current generator. In 
consequence, saturation of the poles is not so difficult to avoid. 
Irl some cases, due to the smaller ampere turns on the interpole 
winding, the interpole coils can be located nearer the pole tip 
and thus the leakage can be further reduced. However, the 
placing of the interpole coil over the whole length of the pole 
is not as objectionable in the converter interpole as -it is on the 
direct-current generator as the ampere turns are less. It is those 
ampere turns which are located close to the yoke, or furthest 
away from the pole tip, which produce the highest leakage, while 
those close to the pole tip usually produce much less leakage, 
but in interpole generators with their high number of ampere 
turns on the interpoles it is often difficult to find space for the 
iaterpole winding, even if distributed over the whole pole length. 
In some cases s a direct-cuirent machine may be larger than 


would otherwise be zequired, simply to obtain space for the inter- 
pole winding. This is not true' to the same extent in the appli- 
cation of interpoles to converters. 

In the above the leakage is referred to as a function of the 
interpole winding as if the main winding had little or nothing 
to do with it. The reason for this nxay be given as follows: 

Fig. 9 represents two main poles and an interpole of a direct- 
current generator or converter, with their windings in place. 
The direction of current or polarity of each side of each coil is 
also indicated by + or . It is evident that between the inter- 
pole and one main pole, the interpole winding and the main 
field winding are of the same polarity, while on the opposite side 
of the interpole, these two windings are in opposition. Let A 
equal the ampere turns of the interpole and B the ampere turns 
in the main coil. Then, A+B will represent the leakage ampere 
turns at one side of the interpole and A - B will represent the 
leakage ampere turns at the other side. Therefore, the leakage 
at the two sides of the poles is represented by (A +B) -f (A - B) 
= 2A; that is, the leakage could be considered as due to the 
interpole winding entirely and may also be considered as due to 
double the interpole turns acting as one side of the interpole 
only. Another way of looking at this is to consider that the 
windings on the main pole produce leakage in' the interpoles, 
but the leakage due to one main pole acts radially in, one direc- 
tion in the interpole, while that due to the other main pole is 
in the opposite direction. 

Considering therefore the interpole leakage as being due to 
the interpole ampere turns only, it is evident that the syn- 

chronous converter will not be 
troubled with saturation of the in- 
terpoles to the same extent as a 
direct-current generator. With the 
same size of interpole it is evident 
that tk e converter should be able 
to carry heavier overloads than the 

direct-current generator before saturation of the interpoles is 

It was mentioned before that a closed conducting circuit 
around the interpoles would be objectionable. This has be$n 
proved by experience with irtterpole generators. It is evident 
from the preceding analysis that the ampere turns on the inter- 
pole of a direct-current generator should always rise or fall in 


proportion to the armature ampere turns in order to give best 
commutation, assuming, of course, no saturation of the poles. 
If the interpole turns are directly in series with the armature 
winding, with no shunt across the interpole winding, it is evident 
that the interpole ampere turns must vary in direct proportion 
to the armature ampere turns. However, if a non-inductive 
shunt, for instance, were connected across the interpole winding 
in order to shunt part of the current, then in the event of a sudden 
change in load, the interpole winding being inductive due to its 
iron core and the shunt being non-inductive, the momentary 
division of current during a change in load would not be the same 
as under steady conditions. In other words, if the armature 
and interpole current were suddenly increased, then a large part 
of the increase would momentarily pass through the non-induc- 
tive interpole shunt until steady conditions were again attained. 
In consequence, the interpole ampere turns would not increase 
in proportion to the armature ampere turns just at the critical 
time when the proper commutating field should be obtained. 

The same condition is approximated when a separate con- 
ducting circuit is closed around the interpole. A sudden change 
in the current in the interpole winding, causes a change in the 
flux, and secondary currents are set up in the closed circuit, 
which always act in such a way as to oppose any change in the 
flux, whereas, the flux in reality should change directly with the 
current. The above described non-inductive shunt across the 
interpole winding might be considered also as completing a 
closed circuit with the interpole winding, and therefore retarding 
secondary currents would be set up in this closed circuit with any 
change in the flux in the interpole. 

In some cases it may be impracticable to get exactly the right 
number of turns on the interpole winding to give the correct 
interpole magnetomotive force. For example, on a heavy 
current machine, 1.8 turns carrying full current might be re- 
quired on each interpole. If two turns were used, with the 
extra current shunted, the right interpole strength would be 
obtained. A non-inductive shunt, however, is bad, as shown 
above. However, if an inductive shunt is used, instead of non- 
inductive, and the reactance in this shunt circuit is properly 
adjusted, then it is possible to get the right interpole strength for 
normal conditions and still obtain satisfactory conditions with 
sudden changes in load. Also, by arranging the interpole 
winding so that a very considerable percentage of the current 


is shunted normally by an inductive shunt having a relatively 
high reactance compared with the interpole, it should be possible 
to force an excess current through the interpole winding in case 
of a sudden increase in load, in case a stronger commutating 
field were needed at this instant. 

On the interpole synchronous converter a non-inductive shunt 
across the interpole winding should act very much as on an inter- 
pole generator and therefore non-inductive shunts are inad- 
visable. If any shunting is required it should be by means of an 
inductive shunt in those cases where the current from the con- 
verter is liable to sudden fluctuations, as in railway service. 
"Where the service is practically steady, a non-inductive shunt 
should prove satisfactory for the interpoles of converters or 
direct-current generators. 

Under extreme conditions of overload current, that is, in 
case of a short circuit across the terminals, it is questionable to 
what extent interpoles are effective. It is practicable to design 
interpoles on direcUcurrent generators which will not unduly 
saturate up to possibly three or four times normal load. How- 
ever, in case of a sudden short circuit the current delivered by 
the machine is liable momentarily to rise to a value anywhere 
from 15 to 30 times full load current. With this excessive cur- 
rent the interpoles of the direct-current generator must -neces- 
sarily be more or less ineffective. On account of saturation, the 
commutating flux under the interpole cannot rise in proportion 
to the current. However, there should still be some commutating 
field present, which condition is probably Considerably better 
than no field at all, or a strong field in the opposite direction 
as would be found without commutating poles. Therefore, in 
direct-current generators with well-proportioned interpoles, 
the conditions on short circuit are generally less severe than in 
non-interpole machines. 

If the pole is highly saturated by the heavy current rush on 
short circuit, then it is evident that a highly inductive shunt, 
as described above, which would increase the interpole current 
in a greater proportion than the armature current, would simply 
mean higher saturation with little or no increase in the useful 
flux tinder the interpole. 

In the synchronous converter at short circuit the conditions 
may be somewhat different. When the converter is short cir- 
cuited it can also give extremely high currents, possibly touch 
greater than the corresponding direct-current generator can give. 


Both, the armature winding tied to an alternating-current supply 
system, and the presence of the low resistance dampers on the 
field magnetic circuit, tend to make the short circuit conditions 
more severe in the converter. The worst condition, however, 
would appear to be in the relation of the interpole ampere turns 
to the armature ampere turns on short circuit. As shown before, 
the normal ampere turns on the interpole winding will be only 
25 per cent to 40 per cent of the direct-current ampere turns on 
the armature. In the case of a sudden short circuit the armature 
momentarily may deliver a very considerable current as a direct- 
current generator, and the armature reaction, or the resultant 
magnetomotive force, may approach that of a direct-current 
generator. In such case the ampere turns on the interpole will 
be very much smaller than the* armature resultant magneto- 
motive force at this instant and thus there will be no commu- 
tating flux under the interpole, but, on the contrary, the arma- 
ture being stronger, there will be a reverse flux which may be 
considerably higher than if no interpole were present, as the iron 
of the interpole represents an improved magnetic path for such 
flux. While the converter armature will probably never deliver 
all its energy as a direct-current generator at the instant of short 
circuit, yet it may be assumed that it will deliver some of its 
load thus, and it does not require a very large per cent to be 
generator action in order to neutralize, or even reverse the effect 
of -the interpoles. In consequence, on a short circuit the con- 
verter may have a reverse field under the commutating pole, 
while the direct-current generator under the same condition 
will have a field of the proper direction but of insufficient strength 
which, however, is a much better condition than a field of the 
wrong polarity. 

The inductive shunt mentioned before, which normally shunts 
a considerable portion of the interpole current, might be more 
effective in a converter than in a direct-current generator in the 
case of a short circuit. In a direct-current generator, the inter- 
poles would be so highly saturated, as described before, that the 
increase in current in the interpole winding due to the inductive 
shunt would be relatively ineffective. In the converter, how- 
ever, the saturation of the interpole can normally be very much 
lower than in the direct-current generator and it might be 
practicable to so proportion these interpoles that they do not 
saturate highly, even on short circuit. In consequence, a 
strong inductive shunt might force up the interpole ampere 


turns so that the negative field under the interpole would be 
much decreased, or might even be changed to a positive field 
and thus become useful in commutation. This would be helpful 
only during the short circuit. However, converters not infre- 
quently flash over or " buck " when the circuit breaker is 
opened on a very heavy overload or a short circuit and not 
when the first rush of current occurs. If the flash tends to 
occur at the opening of the circuit, then the above mentioned 
inductive shunt might have just the opposite effect from what is 
desired, for it would tend to develop or maintain a stronger 

field under the interpole after the armature reaction is removed. 
In consequence, the heavy inductive shunt might prove harmful 
in such a case. 

Another condition exists in a converter which does not exist 
in a generator* When a short circuit occurs on a direct-current 
generator, the armature reaction tends to distort the main field 
very greatly so much so that the field of the machine is very 
greatly weakened. This decreases the terminal voltage and 
the resultant decrease in the shunt excitation will still further 
tend to weaken the field. In consequence, the machine tends 
to " kill " its magnetic field and the voltage tends to drop to a 
low value. Therefore, when the breaker opens on a short circuit 
the direct-current voltage may be falling rapidly. When the 
armature current is removed from the machine the voltage may 
rise slowly, depending upon the natural rate of building up the 
field. Consequently, after the breaker opens there is little or no 
tendency to flash, and practically all difficulties occur dttting the 
current rush, before the breaker oftens. In a converter, how- 
ever, the conditions are different. The armature of the con- 
verter is tied to an alternating-current supply system which 
tends to maintain the voltage on the converter. The machine 
cannot " kill " its field in the same way as the direct-current 
generator, for the alternating-current system tends to maintain 
the field by corrective currents which act in such a way as to tend 
to hold up the voltage. An enormous current may be drawn from 
the alternating-current system momentarily in case of a short 
circuit on the direct-current side of the converter. This heavy 
alternating current may cause a drop in the alternating-current 
lines, step-down transformers, etc., so that the supply voltage 
does fall very considerably and the direct-current voltage does 
drop materially in case of a short circuit. However, the instant 
the short circuit is removed by opening the breaker, then the 


converter at once tends to attain full voltage as the alternating- 
current supply system tends to bring the armature up quickly 
to normal voltage conditions. In consequence there may be a 
relatively heavy current flow in the alternating-current side of 
the machine, while there is no direct-current flow in the armature 
Part of this alternating-current flow represents energy in bringing 
the machine back to a normal condition, and part is purely mag- 
netizing or wattless current. The energy component tends to 
produce an armature magnetomotive ibrce giving an active field 
at the point of commutation. This energy component alter- 
nating-current flow, however, cannot be corrected by inter- 
poles, as there is no direct current flowing. 

A further difference between the synchronous converter and 
the direct-current generator, in case of a short circuit, lies in the 
results of field distortion. The enormous short circuit current 
from the converter with the armature acting partly as a direct- 
current generator, may very greatly shift or distort the field 
flux. The dampers on the field poles tend to delay this distor- 
tion. Also, after distortion has occurred they tend to maintain 
the distorted or shifted field so that momentarily after the circuit 
breaker opens the converter may be operating without direct- 
current load but with a very badly distorted or shifted field. 
This also tends to produce sparking or flashing after the direct- 
current breaker has opened. 

Another condition which may affect the action of interpoles 
on converters, but which does not occur in direct-current gen- 
erators, is hunting. When Jaunting occurs in a converter the 
energy current delivered to the alternating-current side of the 
converter pulsates, or varies up and down over a certain range, 
which may be either large or small. At the same time the direct- 
current flow is apparently varied but little. In consequence, 
the resultant magnetizing effects of the alternating current and 
direct current do not nearly neutralize each other at all times. 
When the alternating-current energy input is least the converter 
delivers part of its direct-current load as a generator, the stored 
energy in the rotating armature being partly given up to supply- 
ing the direct-current power. In this case the resultant magneto- 
motive force may be a very considerable per cent, of the maxi- 
mum direct-current magnetomotive force of the armature wind- 
ing. Also, the magnetic field under the main poles is distorted 
or shifted toward one pole edge. The armature necessarily 
slows down during this operation, the field polarity of all 


the poles being shifted toward one pole edge. The position 
'of maximum e.m.f . of the alternating-current end and also the 
position of maximum alternating-current flow may be shifted 
to a certain extent also. In consequence, the magnetomotive 
force due to the alternating-current flow will be shifted cir- 
cumferentially a certain amount, while the direct-current mag- 
netomotive force cannot be shifted, being fixed in position by 
the brushes. In consequence, the alternating-current magneto- 
motive force may not be in direct opposition to the direct-current 
at this instant, and the resultant magnetomotive force may be 
much higher than at normal condition. A moment later the 
swing may be in the opposite direction; that is, the alternating 
armature current may be greater than direct current and the 
energy being received from the alternating-current system is 
considerably greater than is given out by the direct current* 
Again, the two magnetomotive-forces will not nearly neutralize 
each other and there also will be field distortion, but in the 
opposite direction, and again, the two magnetomotive forces 
will not be in direct opposition to each other circumferentially. 
If hunting is very severe, the resultant magnetomotive force of 
the armature due to the inequality of the input and output, and 
to the circumferential shifting of the magnetomotive forces 
with respect to each other, may vary enormously and may pass 
from positive to negative values periodically. It is evident that 
under such condition the presence of an interpole may give much 
worse results than if no interpole were present; for, as mentioned 
before, if there is a magnetomotive force in the wrong direction 
at the interpole, the interpole magnetic circuit apparently makes 
conditions worse. In consequence, an interpole synchronous 
converter should be especially well designed to avoid hunting* 

All of the above considerations have taken into account only 
the energy currents delivered to the alternating-current side of 
the converter. Some consideration should be given to the effect 
of wattless currents in connection with interpoles. 

As is well known, when a synchronous converter has its field 
strength improperly adjusted for the required alternating-current 
counter e.m.f., alternating currents will flow in the armature in 
such a way as to correct the effect of the improper field strength; 
that is, if the field is too weak wattless currents will flow in the 
armature which tend to magnetize the field of the converter. 
These currents will be leading in the armature, but will be lagging 
with respect to the Hue. On the other hand, if the converter 


field is too strong, these wattless or corrective currents will tend 
to weaken the field and will lag with respect to the armature, but 
will lead with respect to the line. These corrective currents 
will have a lead or lag of 90 deg. with respect to the energy 
currents. Their magnetomotive forces also will have a lead or 
lag of 90 deg. from the magnetomotive force of the energy com- 
ponent of alternating-current input. As this latter practically 
coincides with the direct-current magnetomotive force, which is 
midway between the main poles, the corrective armatur.e cur- 
rents will have a maximum magnetomotive force practically 
under the middle of the main poles and therefore become purely 
magnetizing or demagnetizing due to such position* Also, 
being at right angles to the energy component, the magneto- 
motive forces of the corrective currents will have zero value 
where the energy component has maximum, and therefore 
should have no direct effect upon the resultant magnetomotive 
force midway between the main poles, or under the interpoles 
if such are used. It might be assumed therefore that the usual 
wattless or corrective currents, which the converter may carry 
on account of improper field strength, will have no direct harmful 
effect on the commutation- However, there ar6 apparently 
some indirect effects due to this corrective current, for when a 
converter is operated at a bad power-factor, either leading or 
lagging, there is generally more trouble at the commutator and 
brushes than when a high power-factor is maintained. 

It has been shown that the maximum possible benefit to be 
derived from interpoles in neutralizing armature reaction is much 
less in synchronous converters than in direct-current generators. 
In direct-current generators and motors interpoles have also been 
oi great advantage, due to variable speed and variable voltage 
requirements- In synchronous converters, however, the re- 
quirement of variable speed is obviously absent and that of 
variable voltage very limited. The converter has constant 
voltage characteristics and variable voltage can only be obtained 
through the agency of such relatively expensive devices as in- 
duction regulators, synchronous boosters or split-pole construc- 
tions. The advantages of interpoles in synchronous converters 
are then to be looked for only in the direction of increased -outputs 
and higher speeds. 



Some question has been raised this evening regarding the 
statements in the paper that in case of sudden overload or short 
circuit the alternating-current and direct-current magneto- 
motive forces will not balance each other and that the machine 
will operate momentarily as a direct-current generator, with a 
correspondingly high armature reaction. The basis of the 
criticism is that the converter, being a synchronous machine, 
cannot change its speed except for a very short period, namely, 
that occurring within a small fraction of one cycle, otherwise 
the machine would fall out of step. For such a small change in 
speed, it Was argued, very little energy could be given up as a 
direct-current generator, as there is not enough stored energy in 
the converter armature to giv up much energy as a direct-cur- 
rent generator without falling OUT; of step. 

At first thought, such an argument seemed reasonable, but 
one answer to it is found in the operation of a synchronous con- 
verter on a single-phase circuit. In such operation the energy 
supplied to the alternating current end falls to zero twice in each 
cycle, while the direct-current output remains practically constant. 
The alternating-current input must therefore vary from zero 
to far above the direct-current output of the machine. The 
converter must therefore act as a direct-current generator, for a 
brief period, twice during each cycle. When it is considered 
that such a converter can operate with more or less spkrking up 
to three or four times full-load current, or even much more, 
depending upon the design of the machine, it is obvious that the 
converter can deliver very heavy outputs momentarily as a 
direct-current machine without falling out of step. 

Also, a little calculation will show that with an ordinary design 
of synchronous converter the stored energy in the armature is 
such that, in dropping back as much as 45 electrical degrees in 
position, the armature could give up an enormous energy com- 
pared with its normal rated capacity. If it were not for this it 
would not be possible to run the machine on heavy load on a 
single-phase circuit. 

That is all I will say in regard to the points brought up in the 
discussion. However, there are several points I want to bring-, 
out in connection with the paper itself. In the first part of the 
paper it is stated that the ampere turns on the interpoles of a 
direct-current generator are always greater than the ampere 


turns of the armature winding. This statement is not correct in 
all cases but in those arrangements which depart from this rule, 
direct-current generators and converters would be affected in 
the same way so that for comparative purposes the statement 
in the paper may be considered as correct. 

When there are as many interpoles as there are main poles it is 
correct to say that the ampere turns on the interpoles should 
always be greater than on the armature. However, in some 
cases, especially on small machines, the number of interpoles on 
direct-current machines is made only half as great as the number 
of main poles. There are several advantages in this arrange- 
ment and they apply equally well to generators and synchronous 
converters. Obviously where only half as many interpoles are 
used the commutating flux or field of each interpole must be at 
least twice as strong as when the full number of interpoles is used, 
as the opposing e.m f . set up by the interpoles must be sufficient 
to overcome the e m.f . of self-induction, regardless of the number 
of interpoles. This opposing e.m.f. need not be distributed 
over the whole armature coil, but could be located over either 
side of the commutated coils or even along a short portion of 
its length. It is only necessary that this opposing e.m f . should 
have the proper value, while the distribution of it seems to be of 
relatively less importance. It should be understood, however, 
that the use of half the interpoles is permissible only with drum- 
wound armature windings, where each armature coil spans 
approximately one pole pitch. Ring-wound armatures require 
the full number of interpoles. 

Experience shows that when but half the number of interpoles 
is used the demagnetizing ampere turns, or those which directly 
oppose the armature magnetomotive force, should have about 
the same value per interpole as when the full number is used. 
However, the effective ampere turns which set up the commu- 
tating flux must be doubled in value, as just stated. Therefore 
the total ampere turns per interpole would be greater than when 
the full number of interpoles is used, but the total number of 
ampere turns on all the interpoles is much less than with the 
full number of interpoles. In consequence, there is a very con- 
siderable saving in the amount of copper required. 

On account of the increased number of ampere turns per 
interpole when half the number of poles is used, the interpole 
leakage will be increased in proportion. This is particularly 


objectionable on large machines where the design of the inter- 
pole becomes difficult on account of magnetic leakage. There- 
fore this arrangement is usually confined to small machines. 

A very considerable advantage in this arrangement is that 
the ventilating conditions are improved due to the fact that the 
interpoles and main poles do not so completely enclose the arma- 
ture, for, with alternate interpoles omitted, the circulation of air 
between the armature and the field poles can be materially im- 

With interpole converters, with their smaller ampere turns 
per interpole, the omission of alternate interpoles will not have 
as much influence on the general design as in the case of direct- 
current generators. As the interpole ampere turns are only 
about 35 per cent, as great as on a direct-current machine, and 
as about half is useful and half demagnetizing, it is evident that 
the useful component would readily be doubled, thus doubling 
the useful flux, while the total leakage would still be far less on a 
direct-current machine. Therefore the smaller number of inter- 
poles is much better adapted to the synchronous converter than 
to the direct-current generator. 

In the converter the use of the small number of interpoles 
also possesses a further advantage. In the case of a short cir- 
cuit, and assuming a negative field to be set up by the armature 
reaction, as described in the paper, the use of half the number of 
interpoles would cut this reverse field to half value. In conse- 
quence, any flashing tendency would be proportionately re- 
duced. Half the neutral spaces being without interpoles, and 
the other half having interpoles, it is evident that such an 
arrangement should be practically midway between a non-inter- 
pole and a full interpole converter as regards any flashing 

It is also evident that with half the number of interpoles the 
ventilating conditions will be improved just as on the direct- 
current generator. 

The lower leakage in the interpoles of the converter allows 
another material difference between the design of the converter 
interpoles and those of tie direct-current generator. In ordinary 
direct-current generators, especially those of large capacity, 
the interpoles, as a rule, are made almost the full width of the 
armature core, principally in order to maintain a lower satura- 
tion of the interpole core. As the width u i the interpoles is 


varied the leakage flux varies practically in proportion to the 
-width, but the total useful flux remains practically constant. 
Therefore, with wider interpoles the flux density due to the com- 
bined leakage and useful fluxes will be lower than if a narrower 
pole were used, and the saturation will be correspondingly re- 
duced. In the interpole converter, however, the leakage flux 
being so much lower than in a direct-current generator, it is 
evident that the useful flux could be correspondingly increased 
while maintaining no higher saturation than on a direct-current 
machine. This, therefore, permits a much narrower interpole 
on the converter than on a direct-current machine. As the 
interpole becomes narrower than the armature the reverse 
field which may be set up on short circuit also should be pro- 
portionately reduced, so that with interpoles of practically 
half the width of the armature, the conditions should be practi- 
cally equivalent to those where half the number of poles is used, 
as far as flashing conditions are concerned The use of narrow 
interpoles should also allow better ventilation than when the 
full width is used. Narrower interpoles, of course, allow con- 
siderably less copper for the same total number of ampere turns 
However, unless the interpoles can be made less than half the 
width of the armature, the amount of copper required for this 
arrangement would be still greater than would be required with 
only half the number of interpoles, each of full width of the 

There are many other points in connection with the use of 
interpoles on converters which were not mentioned in this 
paper. I will describe briefly a few interesting features which 
are encountered in the design of such machines, but which are 
not found in direct-current machines. 

One of these concerns the application of dampers to interpole 
converters. It is found that the usual distributed cage type of 
damper supplied with self-starting converters is not directly 
applicable to the interpole converter Dampers are supplied 
to synchronous converters for two purposes, namely, to prevent 
hunting and to obtain good self-starting conditions. To prevent 
hunting the damper should be thoroughly distributed through 
and around the pole face in the form of numerous low resistance 
bars or rods which are joined together at each end by low re- 
sistance connectors. There may or may not be any connection 
between the dampers on adjacent poles. In practice, with well 


proportioned dampers, such connection between the poles may 
be of some benefit, but this is difficult to determine as far as 
hunting is concerned. Those conductors embedded in the pole 
and immediately surrounding it appear to give all the damping 
action which is necessary if the damper is well proportioned. 

However, when it comes to self-starting converters, that is, 
those which are started and brought up to speed by direct 
application of alternating-current to the collector rings, it is 
claimed by some designers that the interconnection between 
the adjacent dampers is of benefit at the moment of starting, 
by reducing the tendency toward dead points or points of very 
low starting torque. When started in this manner the armature 
of the converter becomes the primary of an induction motor, 
while the cage damper in the field poles becomes the equivalent 
of a cage winding on the secondary of an induction motor. It 
is claimed that the interconnection between the dampers to 
form a complete cage allows better polyphase action in the 
secondary winding. Any beneficial result of this should show 
in more uniform torque at start, but not to any pronounced 
extent in the apparent input required to start the converter 
and bring it up to speed. 

When hunting occurs the magnetic field in the main poles is 
alternately shifted or crowded toward one pole edge or the other 
and the parts of the damper embedded in and immediately 
surrounding the pole face aare particularly effective in preventing 
such shifting. Also, the lower the resistance and the better 
distributed this damper, the more effective it appears to be in 
general as regards damping. 

On the other hand, for self starting, the damper, acting as a 
cage secondary of an induction motor, will have the character- 
istics of such secondary and therefore for best and most uniform 
starting torque conditions, a relatively high resistance is de- 
sirable and a continuous cage is usually preferred. In conse- 
quence, the two conditions of best damping and best starting 
are, to a certain extent, opposed to each other. 

In the use of a continuous cage damper is found a difficulty 
in the application of interpoles to the synchronous converter. 
If adjacent dampers are connected together, as shown in Fig. 1 
then the interpole between the two main poles is actually sur- 
rounded by the low resistance damper circuit, a condition which 
is very objectionable, as explained in the paper. Conse- 



quently, tlie usual arrangement of the cage damper for 
self starting is not advisable on an interpole con- 
verter which is subject to sudden fluctuations in load. In 
other words, the continuous cage damper should not be used, 
or its design should be modified very considerably, in the case of 
self-starting converters, which are subject to considerable 
fluctuations in load in service. If the continuous cage construc- 
tion is desired, the individual dampers might be connected 
together by high resistance connectors. 

A second interesting point in the design of interpole converters, 
but not found in direct-current generators, comes up in connec- 
tion with the copper loss in the tap coils, that is, those armature 
coils which are tapped directly to the collector rings. As is well 
known to those familiar with synchronous converter design, 

the copper loss in the tap coils 
of a rotary is relatively high 
compared with the average loss 
in all the coils, the loss per coil 
falling ofi to a minimum value 
between the taps. The real 
limit in carrying capacity of the 
armature is fixed by the heating 
of the tap coils and not by the armature copper as a whole. 
It is possible to overload an armature so that the tap coils will 
roast out while the remaining coils will show very much less 
signs of heating. The heating in these tap coils also increases 
rapidly as the power-factor of the alternating-current input is de- 
creased, the output remaining constant. Therefore by reducing the 
power-factor of a converter while keeping the direct-current 
output constant it is possible to roast out the tap coils. The 
true limit of heating in a converter armature therefore is found 
in these coils. Herein is found a difference between the inter- 
pole and the usual non-interpole converter. In the non-inter- 
pole type, as usually constructed, the armature coils are of the 
fractional pitch or "chorded" type in which the "throw" or 
"span" of a coil is one or more slots less than the pole pitch. 
The primary object of this is to improve commutation. In the 
ordinary direct-current winding there are two coils in each slot 
one above the other. With a full pitch winding, when the 
upper coil is being commutated or reversed the lower coil in 
the same slot is also being reversed so that the e.m.f. of self- 

FiG. 1 


and mutual-induction of the commutated coils is due to the 
reversal of the local field of both upper and lower commutated 
coils in the slot. With a fractional pitch winding, the upper 
coil which is being commutated lies in a different slot from the 
lower one which is being commutated at the same instant. 

This same arrangement of fractional pitch winding puts the 
upper tap coil in a different slot from the lower one so that the 
maximum heating does not occur in the upper and lower coils 
in the same slot, as would be the case if a full pitch winding were 
used. Therefore, with a fractional pitch winding the heating 
is somewhat better distributed than in the full pitch winding. 
However, with interpoles, a full pitch winding would naturally 
be used, as a fractional pitch winding would mean a relatively 
wide interpole with a corresponding increase in distance between 
the main poles. Therefore with the full pitch winding used with 
interpole converters the heating due to the tap coils will be more 
concentrated than in the non-interpole type. In other words, 
the machine will have less maximum capacity unless more copper 
is used in the armature coils, or an inferior type of interpole 
construction is used in order to allow a fractional pitch winding. 
This looks like a minor point, but when it is borne in mind that 
in modern converter designs the starting point in the design of 
the armature winding is the permissible copper loss in the tap 
coils, and not the armature copper loss as a whole, the import- 
ance of this point may be seen. 

A third point, not mentioned in the paper but which concerns 
design as well as operation, is found in self-starting converters. 
In such machines the alternating current is applied directly to 
the alternating-current end of the converter and a rotating mag- 
netic field is set up, just as in the primary of an induction motor. 
This field travels around the armature at a speed corresponding 
to the frequency of the supply circuit and the number of field 
poles and all the armature coils in turn are cut by this traveling- 
field. Those coils which are short circuited at the commutator 
by the brushes form closed secondary circuits and secondary 
currents are set up by the alternating field just as in commu- 
tating type alternating-current motors at start. As soon as the 
converter gets in motion the short circuit is transferred from coil 
to coil but the short circuit current must be broken as each coil 
passes out from tinder the brushes and this results in more or 
less sparking, depe&ding on the size and general proportions of 


the machine. It is a question to what extent this sparking is 
dependent upon the normal commutating characteristics of 
the armature winding. Other things being equal, presumably 
the better these characteristics the less should be the sparking 
and burning at the brushes when the converter is self started 
from the alternating-current end. On this basis then, a con- 
verter armature designed with poor commutating characteristics 
and in which the commutation at synchronous speed is ac- 
complished by interpoles, should spark considerably more when 
starting than a converter which has inherently very much 
better commutating characteristics. The presence of com- 
mutatiag poles should in no way help commutation at start as 
there is no current in the interpole winding. However, as con- 
verters are started very infrequently, such increased sparking 
at start would probably do but little real injury. This is simply 
mentioned as one of the points in which the designer is concerned. 

Some reference has been made this evening to the split pole 
converter in connection with interpoles. Some distinction 
should be made between the true interpole or commutating pole 
arrangement referred to in this paper and what is sometimes 
referred to as the interpole in the so-called "split-pole" con- 
verter. In the split pole converter, as usually built, there is a 
series of wide poles alternating with narrow poles, the field 
construction therefore resembling somewhat the ordinary inter- 
pole machine. In the split pole converter, however, the small 
pole is used primarily for the purpose of obtaining variations 
in the direct-current voltage and not for the purpose of ob- 
taining a true commutating field. The winding on this small 
pole on the split pole machine is usually in shunt with the 
armature instead of in series, and its circuit is so arranged that 
the polarity can be varied from maximum down to zero and to 
maximum in the opposite direction regardless of the armature 
current carried. In certain combinations this arrangement 
can be made to have the effect of commutating poles, but under 
other conditions it may have just the opposite effect. 

The small pole is usually placed dose to one of the main 
poles, thus allowing a fairly wide interpolar space between itself 
and one of the adjacent large poles and a very narrow space to 
the other large pole. Commutation occurs usually in the wider 
interpolar space and not under the small pole itself as is the case 
in the true interpole machine. The direct-current e,m.f. is 


generated by the resultant field due to one large pole and the 
small pole which is closest to it. When these two have the 
same polarity the direct-current e m.f . is highest and when they 
are of opposite polarity it is lowest. However, the alternating- 
current e.m f . is due to the flux of two adjacent poles, a large 
and a small one of like polarity. It is evident therefore that 
the maximum alternating-current e.m.f . will coincide in position 
with the direct-current only at the highest direct-current e.m.f.; 
that is, when both fluxes included in one direct-current circuit 
are of the same polarity. At lowest direct-current e.m.f . when 
one direct-current circuit includes two fluxes of opposite polarity, 
it is obvious that the maximum alternating-current e.m.f. 
must be shifted tircumferentially with respect to the direct- 
current. The alternating-current magnetomotive force will 
also be shifted in like manner with respect to the direct-current 
and the resultant of the two will vary both in height and position 
with variations in the strength and direction of the flux of the 
small pole. 

At highest direct-current e.m.f. a coil which is being corn- 
mutated lies midway between poles of opposite polarity and the 
conditions resemble those in an ordinary converter as regards 
commutation. At the lowest direct-current e.m.f. the commu- 
tated coil lies midway between two poles of like polarity and 
there will be a field flux in the interpolar space in which the 
armature coil must commutate. The direction of this field 
may be such that it will assist in commutation; that is, it will 
tend to overcome the higher magnetomotive force of the arma- 
ture currents resulting from the alternating-current and direct- 
current magnetomotive forces being shifted with respect to 
each other, as just mentioned. Therefore this interpolar field 
flux may act in a very beneficial manner tinder certain condi- 
tions. However, if this flux is in the right direction for assisting 
commutation when transforming from alternating-current to 
direct-current, it will evidently be in the wrong direction when 
operating from direct-current to alternating-current. Also, 
this field flux in the interpolar space will vary with any variations 
in the strength of the small pole; that is, with any change in the 
direct-current voltage, although the currents in the armature 
may be unchanged. Also, this interpolar field may remain of 
constant strength, while wide changes may occur in the armature 
currents, and thus in their resultant magnetomotive forces. It 


is obvious therefore that this interpolar flux can be equivalent 
to a true interpole of proper strength and polarity, only under a 
very limited range of operation. 

In conclusion I may say that, as brought out in the paper, 
the real field for interpoles in synchronous converters is found 
in connection with higher speeds and large outputs per pole. 
I am an advocate of the highest speeds which the public will 
stand, up to the point where no further real gain in cost and 
performance is obtained. If this highest speed in converters 
is such that interpoles are of material benefit, then in such ma- 
chines we may look forward to the use of interpoles. However, 
for the relatively low speeds represented by much of our present 
practice the use of interpoles can be considered as only a rela- 
tively small improvement, concerning which there may be honest 
differences of opinion regarding the commercial value. 


FOREWORD This paper was the result of many years of the author's 
work on the subject of commutation. In its presentation, it 
embodies a new method of looking at the problem. In this 
method, the armature winding, as a whole, is considered as 
setting up a magnetic field in the so-called neutral zone; and 
it is, primarily, the e.m.fs. set up by the armature conductors 
cutting this field, which are dealt with in this theory of commuta- 
tion. Before the completion of the paper, the commutating con- 
stants of many hundreds of direct-current machines of various 
kinds were checked to determine the correctness of the method. 
The paper was presented at a meeting of the American Institute 
of Electrical Engineers, October, 1911. 

Throughout this paper, it will be noted that the term "inter- 
pole" was used in place of the present accepted term ' ' commutat- 
ing pole" which came later. (ED.) 

IN the usual theory of commutation it is considered that , 
when the current in a coil is commutated or reversed, the local 
magnetic flux due to the current reverses also, and in so doing 
sets up an e.m.f. in the coil which opposes the reversal. This is 
the so-called reactance voltage referred to in commutation prob- 
lems. The fact that two or more coils may be undergoing 
commutation at the same time involves consideration of mutual 
as well as self-induction. The relation of the mutual to the self- 
induction, the probable value of each, etc,, lead to such mathe- 
matical complication in the analysis of the problem, that em- 
pirical methods have become the usual practice in dealing with 
commutation. The usual analytical methods do not permit a 
ready or easy physical conception of what actually takes place. 

According to the usual theory, during the commutation of the 
coil the local magnetic flux due to the coil is assumed to be 
reversed. However, in the zone in which the commutation 
occurs, certain of the magnetic fluxes may remain practically 
constant in value and direction during the entire period of com- 

The fact of part of the flux in tbe zone of commutation 
remaining practically constant in value and direction, led the 
author to a method of dealing with the problem of commutation 
which is based upon consideration of the armature flux as a 



whole, as set up by the armature ampere turns. The results 
obtained by the method were very satisfactory, and it was ap- 
parent that a much better conception could be obtained of some 
of the phenomena of commutation than was possible with former 

In the following pages the method is indicated in general, 
and its application to interpole machines is then worked out in 
greater detail. In non-interpole machines the problem is greatly 
complicated by the presence of local currents under the brushes 
which modify the distribution of certain of the armature magnetic 
fluxes, as will be shown. 

This theory of commutation, with the method of calculation, 
is based upon the broad principle of the armature conductors 
cutting across the magnetic field %et up by the armature "winding and 
thereby generating an e.mj. in the short circuited coils which w 
proportional to the product of the revolutions, the flux which is cut 
and the number of turns in series. The usual " reactance " 
voltage due to reversal of the local flux of an individual coil is 
not considered, although its equivalent appears under another 

The method in general is therefore the same as that used for 
determination of the main armature e.rn.f., except that the 
magnetic fluxes cut by the armature conductors are those due to 
the armature magnetomotive force instead of those due to the 

When the armature winding is carrying current its magneto- 
motive force tends to set up certain magnetic fields or fluxes, 
which have a definite relation to the position of the brushes. 
Considered broadly, the current after entering the commutator 
or armature winding, at any brush arm, divides into two paths- 
of opposite direction. As the winding on each of these paths is 
arranged in exactly the same way, and as the currents flow in 
opposite directions, the armature windings in these two paths 
have magnetomotive forces which are in opposite directions. 
The resultant armature magnetomotive force rises to a maximum 
at points corresponding to the brush positions. Midway between 
these points the magnetomotive force is zero. Magnetic fluxes 
are set up by these magnetomotive forces, which are a function 
o the force producing them, and the proportions, dimensions, 
and arrangement of the magnetic paths; and these magnetic 
fluxes will be practically fixed in position corresponding to the 
brush setting. 


The armature conductors cutting across these fluxes set up by 
the armature magnetomotive forces, will have e.m.fs. generated 
in them. In those conductors which have their terminals short 
circuited by the brushes, these e.m.fs. may be called the short 
circuit e.m.fs. 

There are three principal armature fluxes which are cut by the 
short circuited armature coils. In the order of their usual 
importance these are, 

1. That which crosses from slot to slot. It may be called the 
slot flux. 

2. The interpoiar flux which passes from the armature surface 
to the neighboring poles or yoke surface. It may be called the 
interpoiar flux, as distinguished from interpole flux, ,which term 
will be used later. 

3. That flux set up in the armature end-winding in the zone of 
the short circuited coil/ due to the magnetomotive force of the 
end windings as a whole. It may be called the end flux. 

The short circuited armature coils cutting across these three 
fluxes generate the short circuit e.m.fs. The whole problem of 
commutation may be considered as depending upon the prede- 
termination of these three fluxes. 

Consider, first, an armature conductor approaching the poini 
of current reversal or commutation. Under this condition the 
current carried by the coil always flows in the same direction 
as the e.m.f. generated by the conductor cutting across the magnetic 
field or flux set up by the armature winding is induced. When the- 
terminals of an armature coil pass under the brush and are short 
circuited, it is obvious that -the e.m.f. set up in the coil by the 
armature flux is unchanged in direction for the coil is still cutting 
a field of the same polarity. This e.m.f. tends to maintain the cur- 
rent in the short circuited armature coil in the same direction as 
before but the value the current attains will be dependent upon, 
the short circuit e.m.f . and largely upon the resistance in the 
circuit, which will usually consist of the resistance of the coil 
itself and of the brush contact. As the coil passes out of short 
circuit, that is, as it leaves the brush, the current must flow in 
the opposite direction, but the e.m.f . set up by the armature 
flux is still in the same direction as before. Therefore, after 
commutation, the armature current in the coil is flowing in. 
opposition to the e.m.f . set up in the coil by the armature flux. 

The following is a method for calculating approximately the 
three fluxes before described and the e^m-fs., generated by the 



armature conductors cutting them. The interpolar fluxes will 
be considered first, the end fluxes second, and the slot fluxes 
last, as these latter are greatly complicated "by the problem of 
local currents produced largely by the interpolar and end fluxes. 


By- this is meant the flux in the interpolar space between the 
armature core and the field poles and yoke, due to the magneto- 
motive force of the armature winding, as shown in Fig. 1. This 
magnetomotive force has its highest value at those parts of the 
armature winding corresponding to the brush contacts on the 
commutator and is zero midw;ay between such points. If the 
brushes are set -with zero lead then the maximum magneto- 
motive force of the armature lies midway between adjacent field 
poles and will taper off in value from this midpoint toward the 


FIG. 1 

FIG 2 

adjacent edges of the poles. The flux density between the arma- 
ture surface and the sides of the poles should therefore tend to 
taper off as the armature magnetomotive force is reduced but, in 
most types of field construction, it tends to increase in value due 
to the relatively shorter magnetic path as- the edges of the poles 
are approached. Usually this increase very considerably over- 
balances the decrease due to the lower magnetomotive forces 
and in consequence the interpolar flux density due to the arma- 
ture generally has a mini-cntrm value midway between the poles 
and rises toward tte edges of the poles. This is illustrated by 
Fig. 2. 

The density of this fltut in the interpolar space is dependent 
-upon many conditions such as the ampere turns of the armature 
winding per pole r distance between poles, conformation of the 
poles, yoke, etc. In Fig. 2 the ordinates of the dotted lines 
represent the flux densities at the armature interpolar surface 


due to each of the two adjacent poles. The resultant of these 
two is the full line a c b which represents the distribution of the 
armature inter polar flux. This interpolar flux might be con- 
sidered as a true magnetic field fixed in space with respect to the 
position of the brushes. This field being fixed and the, armature 
conductors rotating it is obvious that any conductor moving 
across this magnetic field must have e,m,f. generated in it, the 
value of which depends upon the flux which is cut at any instant. 
Therefore, the e.m.f. due to this interpolar field can be de- 
termined directly, if the intensity of the field itself can be cal- 

During the period of commutation the armature ooil is short 
circuited and has the current reversed in it under certain por- 
tions of this field. The problem is to determine the strength of 
the field corresponding to this point of com- 
mutation and then by direct calculation the 
corresponding e.m f. can be determined. 
In the following analysis two cases will be 
considered, namely, pitch windings, and 
" chorded " or " fractional pitch " windings. 
Pitch Windings. When commutating or 
reversing a coil with a pitch winding it is 
evident that if there were no lead at the 
brushes such a coil would commutate, on 
the average, at the midpoint between two 
poles. The e.m.f. generated in the coil by cutting the interpolar 
field would therefore be proportional to the strength of the inter- 
polar flux at the midpoint. This flux can be determined approx- 
imately in a fairly simple manner in the ordinary types of machines 
in which the poles are relatively long compared with the distance 
between adjacent pole tips and where the distance from the arma- 
ture surface to the yoke is relatively great. The following is 
a method which appears to give reasonably close results: 
Let W t = total number of wires on the armature. 
!<, =the current per conductor. 
p = number of .poles. 

Then, the armature ampere turns per pole = C * r 


neglectifig any change in ampere turns due to the short circuit- 
ing action of the brushes. 

In Fig. 3 let b represent the length of the mean flux path 
corresponding to the mid-interpolar position. This is assumed 


to" be a part of a circle which is poetically at right angles to the 
armature surface and the side of the field pole, as indicated in 
Pig. 3. 

Let P = widt& of body of pole. 

Let .Bt the flux density at the midpoint between the poles. 

2X3 197.X TV* 

Then E^-- 


But 6 = 27ra Q Z. . approximately, as angle (90+0) is 


only approximate. 

Also, a = ( ) approximately. 

\ & p i I 





(0.25+0.5) (IT D-Pp)X2 p 

The above gives the approximate flux density at the midpoint 
tetween poles. The flux densities at points at each side of the 
midpoint can be determined in a similar manner, taking into 
account the lower armature magnetomotive force as the mid- 
point is departed from. As the edge of the pole is approached 
the effect of pole horns may complicate the flux distribution so 
that the above method of calculating interpolar flux density will 
not apply for points close to the pole. 
E,mj. Due to Interpolar Flux. 

Let E c - The e.m.f. due to cutting the armature flux. 
D = diameter of armature. 
L = length of core including ventilating spaces. 
Tc = turns per individual armature coil. 
R s revolutions per second. 


Then, the e,m f . induced in a coil cutting the field at c (Fig. 2) 
can be represented by the formula, 

T c XRs 

c 10 


27rDLT c R s 

' (0.25^+0.5) (irD-Pp) ~ 10 8 


I c WtT c R s / ZpXirDL 

L \ 

D-Pp) ) 

10 8 \ (0.25p+0.5) (irD-Pp) 

Incidentally, with this method of dealing with the problem 
the effect of the addition of an interpole can at once be seen. 
The magnetomotive force of the interpole is superimposed on 
that of the armature and the resultant flux is then considered. 
The armature conductors cut this flux and thereby generate 
e.m.f. If the interpole magnetomotive force is stronger than 
that of the armature, then the flux established will be in the 
opposite direction in that part of the armature face which lies 
under the interpole. Therefore, the flux or field over the com- 
mutated coil in the non-commutating pole machine is replaced 
by flux in the opposite direction. The presence of the interpole 
does not increase the reactance of the armature coil as sometimes 
considered, but, on the contrary, the harmful flux is replaced by 
one which is of direct assistance in commutation. 

Effect of Brush Width. In cutting across the interpolar flux 
it is obvious that the e.m.f. set up in the short circuited coil is 
not a function of the length of time the coil is short circuited, 
for this interpolar flux is set up by the armature winding as a 
whole and not by individual coils. If two or more armature 
coils in series are short circuited by the brush, then their e.m.fs. 
will be in series while the total resistance in the path will be very 
little higher than in the case of a single coil short circuited, for 
the principal part of the resistance lies in the brush contact. It 
is evident therefore that considerably higher short circuit cur- 
rents can be set tip by the interpolar field when more commutator 
bars, and more turns, are short circuited. It can therefore be 
assumed that, as far as the interpolar field is concerned, the more 


commutator bars the brush covers the greater will be the short 
circuit current and the greater will be the difficulty in commuta- 
tion, assuming there is no external field assisting commutation 
Chord Winding. With a pitch winding, with no lead at the 
brushes, the commutation of a coil will occur in the lowest part 
of the armature interpolar flux, as a a in Fig. 4. With a chorded 
winding, as indicated at b b, the commutation will occur under 
somewhat higher flux than with a pitch winding. Therefore in 
considering the interpolar flux a full pitch winding commutates 
linder better conditions than a chorded winding. 


The armature winding as a whole sets up certain fluxes in the 
end windings. These fluxes are fixed in position with respect 
to the brushes, and the armature coils, in cutting across, them, 
generate e.m.fs. The only part of these end fluxes concerned in 

FIG. 4 FIG. 5 

the present problem is that which, the commutating coils cu 
during the operation of commutation. 

Fig. 5 illustrates an armature winding na which the heavy 
lines represent two coils in contact with the brushes and there- 
fore at the position of commutation. It is only the end flux 
density along the shaded portion or zone of this diagram which 
need be considered. If the various densities for this zone can 
be determined, then the e.m.f. in the commutated coil can be 
calculated. 0nly the usual cylindrical type of end windings 
will he considered, as practically all direct current machines at 
the present time use this type. Such windings are usually ar- 
ranged iii two layers, the coils of which extend straight out from 
the armature core for a short distance, usually | in. to If in., 
depending upon size and voltage of the machine, and then extend 
at an angle to the core of 30 deg. to 45 deg. The conductors of 
the upper and lower layers therefore usually He almost at right 
angles to each other 



Pitch Winding Let Fig. 6 represent a single coil of the end 
winding located in the commutating zone. Both theory and 
test show that the maximum flux density in this zone is at a 
and tapers off slightly to 6, then tapers off more rapidly from 
b until it reaches practically zero value at c. It may be assumed 
with but little error that the decrease from b to c is at a practi- 
cally uniform rate. The flux density along the commutating 
zone of the end winding may therefore be represented by Fig. 7, 
in which the ordinates represent flux density. On the above 
assumption the total flux in the commutating zone of the end 
winding can be determined with sufficient accuracy if the density 
at b, for instance, can be determined and the distances a b and 
c d in Figs. 6 and 7 are known. These latter can be determined 
directly from the winding dimensions. 

FIG. 6 

FIG. 7 

The following is an approximate formula for the flux density 
at b, including allowance for proximity of iron end plate r core, etc. 

N number of slots per pole. 
I c = current per conductor. 
W t total armature wires. 
D = diameter of armature. 
Let a b = h, and c d = m. 

Then the flux cut by one conductor at one end is 

2.15-1, WtXlog 2N 

Therefore the e.m.f. per single turn of the armature winding, 



due to the end flux, considering the end fluxes for both ends of 
the core, becomes 

m_ \ 2.15 l c W t Xlog 2 A T ir D R S X2 T c 

TT D sin 6 





This formula is on the basis of non-magnetic paths around the 
end windings, that is, with no bands of magnetic material and no 
magnetic supports under the coils, The effect of bands over the 
end winding is approximately equivalent to cutting the flux 
path to half length for those parts of the 
end winding covered by the bands. There- 
fore, with bands, the diagram representing 
flux density in the commutating zone of 
the end winding would be as indicated in 
Fig. 8. In this case the total flux corresponds ]\ 

to the total area of the curve including the 
dotted portion Of course the actual flux 
distribution would not be exactly as shown ^ 
in this diagram for there would be sonie 
fringing in the neighborhood of the bands FIG. 8 

The diagram simply serves to illustrate the 
general effect of magnetic bands and an approximate method of 
taking it into account. 

The effect of a magnetic coil support will be very similar to 
that of a steel band in reducing the length of path and therefore 
increasing the flux in the neighborhood of the cpil support. 
However, in case of magnetic bands over the winding and coil 
supports under it the limit lies in saturation of the bands them- 
selves. This usually represents a comparatively small total flux. 
The coil support, however, would probably not saturate in any 

The above formula for end flux can therefore be corrected 
for magnetic bands and coil supports by multiplying by a suit- 
able constant to cover the increased flux". 

It is obvious that the determination of the end flux is, to 
certain extent, a question of judgment and experience. No 


iixed method or formula can be specified for all types of machines, 
for this flux would be influenced very greatly by the bands, if of 
magnetic material, and by the material, size and location of the 
coil supports and their relation to the bands. Also, eddy cur- 
rents may be set up in the coil supports which will influence the 
distribution of the end flux in the zone of the commutated coil. 
However, in each individual case an approximation can be made 
whioh will, in general, be much closer than would be obtained 
from any empirical rule or by neglecting the effect of the end 
flux altogether. 

Chord Winding. The effect of chording the armature winding 
is to slightly diminish the flux density in the commutating zone 
which results in a slight reduction in the e.m.f . of the commu- 
tating coil. But a relatively much greater gain is obtained 
by the consequent shortening of the distance c d in Fig. 8 and 
the corresponding reduction of the total end flux. Due to the 
chording itself the flux density at b is reduced practically in the 

ratio of i o AT ' where NI number of slots spanned by the 

coil. For example, if the full pitch is 20 slots and the coil 
spans 18 slots, then the density at b will be reduced in the 

t O/* 

ratio of . s n =0.971 due to the chording itself; and the flux 

along cd, Fig. 7, will be further reduced in the ratio of ^ 


due to the shorter end extension. The average flux along c d 
therefore will be reduced to 0.9X0.971 = 0.874, or about 87 per 
cent of that of a pitch winding. 

Effect of Brush Width. As in the case of the interpolar flux the 
width of the brush, or the number of armature coils short cir- 
cuited by the brush, has practically no influence on the e.m.f. 
generated per turn. However, the total effective armature 
iampere turns will be reduced slightly, if the average current in 
the short circuited turns is less than the normal current. This 
will have a very slight effect on the e.m.f. 


By this is meant the magnetic flux across and over the arma- 
ture slots which does not extend to the yoke or field poles. 

Two general cases will be considered; first, that in which no 
local currents are present, which is the case in well designed 
interpole machines; and second, that in which there are local 



currents set up in the short circuited coils, which is almost 
invariably the case in machines without inter poles or some other 
form of compensation. Also, pitch and chorded windings will 
be considered. 

Pitch Winding. Let Fig. 9 represent an upper and a lower 
coil in the same slot, with, equal turns and currents. Then if 
there is no saturation in the adjacent teeth the flux density across 
the slot will be zero at the bottom of the lower coil and will 
rise to a maximum value at the top of the upper coil. There 
will also be a flux across the slot above the upper coil and also 
from the top of the tooth as indicated in Fig. 9. The total slot 
flux entering at the bottom of the teeth is therefore equal to the 
total flux which crosses the two adjacent slots, plus the flux 
crossing at the top of the slots. The interpolar fiux which ex- 





FIG 10 

tends from the armature surface to the poles or yoke is not in- 
cluded in this. 

As this slot flux is practically fixed in position the armature 
conductor in 'slot A, in passing from a to b must cut this flux. 
It is obvious that the flux which crosses above the uppermost 
conductor in the slot is cut equally by all the conductors in the 
slot, as the coil passes from position a to position b; but the flux 
crossing the slot below the uppermost conductor does not affect 
all the conductors equally, and therefore, for simplicity of cal- 
culation, an equivalent flux of lower value can be used which 
may be considered as cutting -all the conductors equally. 

Let d Fig. 10, represent the depth of the conductors of one 

complete coil. 
/ represent the distance between the upper and lower 


a represent the distance from the upper conductor to 
the core surface. 


s represent the width of the slot, assuming parallel sides. 
n represent the ratio of width of armature tooth to the 

width of the armature slot, at the surface of the core. 
T c represent turns per single coil, or per commutator bar, 
C s represent the number of individual coils, or commu- 

tator bars, per complete coil. 
L represent the width of armature core, including 

ventilating spaces. 

I c represent the current per armature conductor. 
Then, ampere turns per upper or lower coil = / c T c G. 

~ . i fl . 3.19 I c TcXCs-L (2 d+f) 

Total flux across coil space = - - - -- *- 

1*1. -1 

Flux across slot above coil = 

Flux from tooth top across the slot is approximately, 
3.19 I c Tc C,iX2X0.54 ^n 

Total flux above-upper coil=3:19 I c T c C S L 

The sum of the two fluxes represents the total flux across one 
slot which enters at the bottom of one tooth. As a similar flux 
passes across the slot at the other side of the tooth the total 
flux entering the tooth will be double the above and becomes 

Total slot flux=2X3.19 I, T f C. L < 2 *+*+* ' +L06 ' V "> 


This total flux cannot t>e used directly in the calculations as 
it does not affect all the conductors equally. It is therefore 
necessary to determine equivalent fluxes for the upper and 
lower coils which can be used instead of the above value. 

For the lower coil the following value has been calculated: 

JVM 1 J A 2X3.19 J.C TC CrS L ^., oOO J I *\ 

The equivalent flux= - (LSSSa+f) 
And for the upper coil, 

r> - * ^ a 2X3.19Xi<T TC (*S-Lr ~ OO Q i 

Equivalent flux = --- O< 0.833d 

To these equivalent fluxes shotild be added the total flux 
above the upper coil. This gives the total effective flux for the 
upper and lower Coils. Then, for the lower coil, 


Total effective flux 

-2X3I9A7- C,L 

And for the upper coil, 
Total effective flux 

= 2X3197, T. C5Lx t0.833d+2a+108sV B) 

The average value of the effective flux for the upper and 
lower coils then becomes, 

3i9/.r c c.L (2 - 67rf+f+4a+2 - 16aVir) 


(This average effective value is approximately 80 per cent of the 
total slot flux.) 

On the basis of a pitch winding and the assumption that only 
one armature coil is short circuited, that is, with the brush 
covering the width of only one commutator bar, then the above 
slot flux is cut by all the coils in the slot in passing through one 
slot pitch. From this the e.m.f in the commutating coil due to 
the slot flux can be calculated directly and may be expressed as 

& = TTtr-^-X number of slots 

But CsX number of slots = No of commutator bars 

_ total number of conductors W t 


Therefore the above expression for e.m.f may be changed to the 

If it is desired to compare this expression with a certain well 
known fonmda which has been much used heretofore* thea let 


the quantity in the parenthesis in the above expression be repre- 
sented by c x . The formula can then be changed to,, 

r? /ON/Q IQVX \v^ Ic T<?^< number commutator 
Xi - = (<& X o . lv X c x ) X 

It contains the same terms (except in the value of the constant) 
for the expression of the e.m.f. which has been used heretofore 
in determining the reactance of the commutated coil. 

Effect of Brush Width or Number of Commutator Bars Covered 
by Britsh. The above formulas are on the basis of the brush 
covering only the width of one commutator bar. In this case 
all the conductors of one slot cut across the entire slot flux in 
passing, through one tooth pitch. However, if the brush covers 
more than one commutator bar, then the full slot flux is not cut 
in passing through one tooth pitch, and a movement greater than 
one tooth pitch is required for full cutting. For example, if 
there is one commutator bar per armature slot and the brush 
covers a width equal to two commutator bars, then the total 
cutting of the slot flux will take place in two tooth pitches. 
Again, if there are three commutator bars per armature slot and 
the brush covers the width of one commutator bar, then the total 
cutting of the total slot flux would occur in one tooth pitch, while 
if the brush covered two bars, the total cutting would occur in 
1J tooth pitches; and if it covered three bars If tooth pitches 
are required. In other words, the total cutting will occur in a 
period corresponding to the number of commutator bars per 
slot plus one less than the number of commutator bars covered 
by the brush. 

On this basis the correction factor for the slot e.m.f. should be 

C * 

expressed by the term ^ , * f where C s = number of com- 


mutator bars per slot, and B = number of commutator bars 
spanned by the brush. However, \*ith several coils per slot, 
and with the brush spanning several bars, the rate of cutting of 
the tooth flux for the entire period is not quite the same as the 
rate for one tooth pitch. Taking this into account the correc- 


tion factor should not be equal to ~ . P J =-, but is slightly 

- I 

greater. Up to four commutator bars per slot, and three bars 

period of coeqsmtatlon is obtaitwd 

factorv^ , ^ _^ t the overage slot e, m. 1 for the 


spanned by the brush the correction factor can be expressed by 

1 1 * 

the term 1-i = ^- "7^~- 

JJi txj L,s 

Taking the lengthened period of reversal into account, it 
would appear that a wide brush covering a large number of 
commutator bars should be beneficial in reducing the e.m.f 
generated by the slot flux. This is true where the local currents 
ate very small, or are absent, as is the case in a properly designed 
interpole machine. In a non-interpole machine where the local 
currents in the short circuited coils may be relatively high, this 
condition does not hold, as will be explained later. 

The above formula for e.m.f. due to the slot flux should there- 
fore be modified by multiplying by a factor which takes into 
account the period of reversal as affected by brush width. 

Chord Winding. The armature winding may be chorded one 
or more slots and, in some instances, where there are several 
coils side by side there has been chord- 
ing of part of the conductors in the 
slot. In Fig. 11 is illustrated the 
conditions with one-slot chording 
The total slot flux now occupies two 
teeth instead of one. Therefore the 
e.m.f. set up by cutting across this 
slot flux will be approximately one- 
half that which is obtained with a 
full pitch winding, on the basis of the brush covering the 
width of one ba/ only, for the e.m.f generated by cutting this 
flux will be reduced in proportion as the period of cutting is 
increased There is one slight difference from the flux distribu- 
tion with a pitch winding, namely, that at the top of the teeth. 
With a chorded winding this flux will be slightly greater than 
with a pitch winding, but the total effect *of this difference should 
be relatively so small that ordinarily the value need not be 
changed. Therefore equivalent fluxes used with chord windings 
can be taken the same as for pitch windings. In consequence, 
the e.m.f , due -to the slot flux, with one-tooth chording, may be 
taken as one-half that for a pitch winding, with the brush cover- 
ing one commutator bar in both cases. 

For two-slot chording the slot flux may be considered as oc- 
cupying the space of two teeth only, while there will be a mag- 
netically idle tooth at the center. The e.m.f . per coil actually 
generated by cutting the slot flux will be, for part of the period 

^General use of this factor, 1 + B^Cs C^ should, be avoided. It is not 

applicable to all tyres or ccirbinaticns of vir dings Use instead the factor Divine averaee 


the same as for one-slot chording, but there will be an inter- 
mediate period where the slot e.m.f. is practically zero, which 
does not occur with a one-slot chording or with a pitch winding. 
The average results, however, should be practically the same as 
if the total slot flux were actually distributed over three teeth 
instead of two. 

Effect of Brush Width with Chord Winding. In the chord 
winding, when the brush covers two or more commutator bars, 
the period of cutting the slot flux will be lengthened just as with 
a pitch winding on the assumption of no local currents For 
example, if there are three commutator bars per armature slot 
and the winding is chorded one slot, then with the' brush covering 
one .commutator bar, complete cutting of the slot flux will occur 
in the space of six commutator bars. If the brush covers three 
commutator bars instead of one, then complete cutting will occur 
in the space of eight commutator bars, while in a corresponding 
full pitch winding it would occur in ths space of five bars. There- 
fore, the wide brush represents an improvement with the chorded 
winding, but not to the same extent, relatively, as with the pitch 
winding. This is on the assumption of absence of local currents 
in the short circuited coils. * 

Bands on A rmature Core. By the preceding method of analysis 
the effect of bands of magnetic material on the armature core can 
be readily taken into account. This effect represents simply an 
addition to the total flux which can pass up the tooth and across 
the top of the slots. From the ampere turns per slot, the clear- 
ance between the bands and the iron core, the total section o the 
band, etc., the flux due to the band can be calculated This flux 
can either be combined directly with the slot flu* already de- 
scribed and the resultant e.m.f. can then be calculated; or, the 
e.m.f. can be calculated independently for the baud flux alone 
Magnetic bands on the 'armature introduce a complication into 
the general e.m.f . formula due to the fact that in many cases the 
flux into the bands Is such as to highly saturate the band material 
at relatively low armature currents. This flux therefore is 
usually not proportional to the armature ampere turns. If the 
e.m.f. due to the band Sux is to be calculated separately, the 
following formula can be used: 

4& represents the total magnetic flux in the hand from the 
armature core considering both directions from the tooth, then 

2 <t> b N p Tc R s . 

^Considering chortling, the correction factor ^ -:--= - = becomes 


-= - = - 

& J, *- j -f *>t A -f 

wheie K slots chorded X C*. For tie general case, where there aie P f circuits and F 
poles, the correction factor becomes - *+$ - 


This formula holds true for the band flux which passes throug 

the one tooth in the pitch winding. Proper allowance must be 

made for the effect of chord windings and brush width, which 

can be done by the methods already described. 


PitchWinding. In the prsceding analysis local currents have 
not been included, as the method would be greatly complicated 
by taking such currents into account. In the general method, 
given below the effect of local currents in the short circuited coils 
can be most easily shown. 

As already explained, an armature coil, as it approaches the 
short-circuit condition, has an e,m.f. generated in it by the inter- 
polar and the end fluxes. After the coil is short circuited this 
e.m.f , is still generated by the coil and naturally a local or short 
circuit current tends to flow through the coil, brush contact 

and brush. In addition, the work, or supply, current is being 
furnished to the armature winding through the brushes. These 
two currents are superimposed in the short circuited winding 
in such a way as, to have a very pfonounced influence in the 
distribution of the slot fluxes. This effect can be bast seen by 
first determining the distribution of the work current in the 
various parts of the short circuited winding on the assumption 
of n& local current and second, determining the distribution 
of the local currents on the assumption of no work current, but 
with the same armature magnetomotive force as in the first assump- 
tion. The two distributions can then be combined and the re- 
sultant currents in the various parts of the short circuited coils 
can be obtained. 

Let Fig. 12 represent the first assumption in which no local 
currents are present. In order to illustrate conditions to better 
advantage, four commutator bars are assumed to be covered by 
the brush. Uniform distribution of current over the brush con- 
tact can be assumed in this case, as there are no local currents* 



Tracing out the current in each short circuited coil in Fig. 12, 
it will be seen that the current decreases at a uniform rate and 
then rises in the opposite direction at the same rate until the 
short circuit is removed The period of commutation is the 
longest possible with this number of commutator bars short 
circuited, and the brush conditions are ideal, as the current 
density at the brush contact is uniform at all parts The above 
are the conditions which the designer endeavors to obtain in the 
construction of good Interpole machines, as will be shown later. 

In Fig 13 the same arrangement of winding and brushes is 
chosen as in Fig. 12 except that only the local currents are shown 
and the values of these are assumed as proportional to the 
e.m.fs. in the short circuited coils and the resistance in circuit. In 
this diagram the current is a minimum 
in the coils at the moment that short 
circuit occurs, and rises to a maximum 
value and then diminishes to zero value 
again at the end of the short circuit 

FIG 14 

FIG 15 

In Fig 14 the currents of Fig. 12 and 13 are supenmposed 
The resultant currents in the various parts of the short circuited 
winding are seen to rise after short circuit until a maximum value 
is reached and then decrease rapidly and reverse to normal 
value m the opposite direction. Therefore, the period from 
normal value of the current to normal in the opposite direction 
is very much shorter than when no local currents are present 
It may therefore be considered that the period of reversal is 
much reduced by the presence of the local currents, so that the 
e rruf m the short circuited armature conductors generated by 
the slot flux is proportionately increased, compared with the 
value it would have in case the local currents were absent 
These conditions can be shown possibly in a somewhat better 
manner by curves a, b and c in Fig. 15 The curve a shows the 
distribution of current in the short circuited coils without any 
local currents Curve b shows the distribution of local currents 



while curve c shows the resultant of the two. The distance be- 
tween d and/ on curve c gives the period of reversal from normal 
current in one direction to normal current in the opposite direc- 
tion. This period is much shorter than the full period repre- 
sented by g f which would be obtained without local currents. 
The period df, however, may not differ much from the period 
of commutation with the brush covering the width of only one 
bar, when the local current is high compared with the work cur- 
rent. In such case the gain in the period of commutation which 
should be obtained by means of the wider brush may be practi- 
cally offset by the effect of the local currents which also increase 
with the wider brush, so that over a considerable rnge the 
resultant of the two effects may be practically constant. This 
is one indication why, in non-interpole machines, the brush 
width may be varied over quite a range with relatively small 
noticeable difference in the commutation. This may be il- 

FIG. 16 

FIG. 17 

lustrated by Pig. 16, in which is shown the current conditions 
with two to five bars spanned. In this figure a b, b c, c d, etc., 
each represent the width of one commutator bar. Therefore, 
curve A, extending over the width ac, represents two bars 
spanned. The period of reversal of the current from normal 
value in one direction to normal in the opposite direction is 
represented by g c for curve A , h d f or curve B, i e for C and 
kf for D. A comparison of these values is interesting. Calling 
a b the period of reversal with the brush covering one bar only, 
then g c with two bars covered, is greater than a b. kdis also 
greater than a &, but less than g c, while i e is slightly less than 
ab, and kf is considerably less. However, the variation be- 
tween g c and kf is much less than between a c and af which 
would be the corresponding periods with no local currents. 

It should be borne in mind that the above curves are only 
relative, depending upon the comparative values of the local 
and work currents and assuming a constant brush resistance' 


which is not correct, but they serve. Do illustrate the general 
principle This method of presentation is simply a skeleton of 
the problem of commutation when local currents are present 
in ths short circuited coils and it would be beyond the scope of 
this paper to attempt a full solution. 

Effects of Field Distortion. One of the " bugaboos " of the 
designer of commutating machines has been the question of field 
distortion. It has usually been considered that when the ma- 
chine is loaded the magnetic field is more or less distorted or 
shifted from its nprmal no-load position and that commutation 
is affected by this distorted field. 

To state the case plainly, the field distortion has practically 
nothing to do with the problem. The distorted field magnetism 
is simply a resultant of the no-load main field flux combined 
with that due to the armature winding. Therefore, the two 
components of the distorted full-load field are the no-load main 
field, which is fixed in space and is usually practically constant, 
and the armature field, which is also fixed in space but varies 
with the load. If the brushes are set in a certain position with 
respect to the no-load field, then, as this component of the re- 
sultant full load field is practically fixed in space and in value, 
It has no variable influence on the commutating conditions. 
The true variable element which does affect the commutation 
is the armature field, or flux, and it is in this very flux which is the 
basis of the preceding theory of commutation. Therefore, the 
distorted resultant field of a loaded machine does not present 
any new condition in the problem of commutation. Ose ex- 
ception, however, can be made to the above, namely, where 
there is any considerable saturation in the armature teeth or 
in the main field pole corners. The effect of the armature mag- 
netomotive force is to strengthen one corner or edge of the field 
pole and to weaken the other edge, but when saturation is pro- 
nounced the strengthening action is much less than the weaken- 
ing action. The resultant of these actions is a decrease in the 
total value of the main field flux. If, now, this main field flux 
be brought back to its normal total value, or higher, a very con- 
siderable addition to the main field magnetomotive force will be 
necessary, which wilt be effective in increasing the field flux at 
the weaker pole corner to a much greater extent than at the 
highly saturated pole corner. In consequence, with load, the 
main field distribution, or field form, may be considered as being 
changed from its no-load form A, to the fonn B, as indicated 
in Fig. 17. It is, in reality, strengthened at a point b, for 


example. In such case the main field will have a variable in- 
fluence on the commutation, if the brush is set with a lead, as 
at &, and, to a slight extent, the effect of an interpole is thus 


Effect of Brush Lead. Before taking up the problem of inter- 
poles on direct current machines it might be well to consider the 
effect of brush lead, as this gives a result intermediate between 
true intsrpole and non-interpole commutating conditions. 

The precedmg formulas apply to non-commutating pole 
machines without brush lead. However, except in case of re- 
versing machines, such as street railway motors, or hoist motors, 
etc., it is usual practice to give a forward lead to the brushes 
of direct current generators or a slight backward lead to direct 
current motors. The effect of giving a lead at the brushes of a 
non-interpole direct current machine may be considered as 
being equivalent to the effect of an interpole with the exception 

FIG. 18 FIG. 19 

that correct flux conditions and proper commutation, with any 
given brush setting, are obtained only for one given load. 

As described before, with a non-interpole machine the arma- 
ture winding sets up a flux in the interpolar space. With no- 
lead at the brushes this flux is usually a minimum midway be- 
tween the poles and ris,es toward the polar edges. The flux 
from the adjacent main poles has a zero value midway between 
the poles and rises toward the polar edges, but has opposite 
polarities at the two sides of the midpoint. This is illustrated in. 
Fig. 18. The resultant of the armature and field fluxes is indi- 
cated by the dotted line A. This resultant falls to zero at one 
side of the midpoint and then rises in the direction opposite to 
that of the flux due to the armature ampere turns. At the other 
side of the midpoint the two fluxes add, giving an increased re- 
sultant flux m the same direction as the interpolar flux due to the 
armature. Prom this figure it is evident that if the point of 


commutation is shifted from a to the point of zero interpolar 
flux 6, then commutation will occur without any interpolar flux 
to be taken into account, that is, the e.m.f. generated by the 
short circuited armature conductors may be due to the slot and 
end winding fluxes only. If the brushes are shifted still further 
in the same direction to c, then, not only will the interpolar arma- 
ture flux be annulled but a flux in the opposite direction would be 
cut by the short circuited armature conductor, which will gen- 
erate an e.m.f. in opposition to that due to the armature fluxes 
in the slots and end windings. Consequently, the commutation 
can be materially assisted by such lead at the brushes. 

The difficulty in the use of this method of commutation lies 
In the fact that the commutating or reversing flux at c is the 
resultant of the main field flux and tha armature interpolar flux 
at this point, and the latter flux varies with the load, while the 
former remains practically constant. Therefore the zero point 
of the resultant field shifts backwards or forwards with change in 
load and the density of the commutating field beyond the zero 
point will therefore change with the armature current. In con- 
sequence, if the brushes are shifted into a suitable resultant field 
c at a given current, then with a different load the intensity of 
this field at c will be changed, and unfortunately the change will 
be in the opposite direction from that desired. In other words, 
the density of this resultant field will decrease with increase in 
load, whsreas just ths opposite effect is desired for good commu- 
tation over a wide rangs in load. 

In practice, however, an average condition is found which, in 
many cases, will give reasonably good commutation over a rela- 
tively wide range in load. The brushes may be shifted at no- 
load into an active field in such a way as to generate an e.m.f. 
in the armature coils of a comparatively high value. This 
e.m.f. will circulate considerable local current through the brush 
contacts and the amount of lead which can be given is dependent, 
to a certain extent, upon the amount of local current which can 
thus be handled without undue sparking 

As the load is increased the strength of the resultant field, 
corresponding to this brush position, will be decreased, and with 
some value of the current this field will be reversed in direction. 
At this point the e.m.f due to this field is added to the e.m.f. 
due to the slot and end winding fluxes. Obviously the limiting 
condition of commutation will be reached at a much higher cur- 
rent than, would be the case if ao load at all had been given. This 
condition is represented in Fig. 19, in which curves 1, 1, 2, 2, 3, 3, 


etc., represent the armature and resultant flux distributions with 
various loads. In this figure the brushes are given a lead so that 
commutation occurs at a point corresponding to b. 

It is obvious that at heavy load a still greater lead at the 
brushes might give improved commutating conditions. How- 
ever, if the load were suddenly removed without moving the 
brushes toward a, then the short circuited coils would be cutting 
the main field at such density that serious sparking or flashing 
might occur. 

One serious objection to this method of commutation is that 
the distribution of the resultant field is practically such that 
equally good commutation cannot be obtained for all the coils 
in one slot when there are several coils or commutator bars per 
slot. All the coils of one slot must pass under a given position 
or value of the interpolar magnetic field at the same instant, 
while the commutator bars to which these coils are connected 
must pass under the brush Consecut-vely. If the field intensity 
is just right for good commutation a$ the first coil per slot 
passes under the brush, then it may be entirely too great by the 
time the last coil is commutated. For good commutation with 
a number of coils in one slot, the resultant interpolar flux should 
have practically constant- value over the whole range repre- 
sented by the period of commutation of all the coils in one slot. 
This condition, however, is extremely difficult, or is frequently 
impracticable, to obtain with the ordinary non-interpole ma- 

The above treatment of the problem of the effect of the brush 
lead has been based upon the armature interpolar magnetic 
field being located in the same position with lead as when there 
is no lead at the brushes. It has been assumed heretofore that 
the non-interpolar flux due to the armature winding has a mini- 
mum value midway between the main poles and rises uniformly 
toward two adjacent pole corners. This, however, is only true 
when the point of commutation, or brush setting, is midway be- 
tween the poles. When the brushes are shifted toward either 
pole the point of maximum armature magnetic potential is 
shifted in the same way. This means that the distribution of 
the armature interpolar flux will be modified directly by the 
position of the brushes. Instead of rising uniformly toward th3 
two pole corners, with a minimum value midway between, it 
will have a minimum at one side of the midpoint, this being at 
the opposite side from the point of brush contact, and will have 


an increased value on the side toward which the brushes are 
shifted. This is illustrated in Fig. 20 in -which A represents the 
armature .interpolar flux distribution with the brush at a, while 
B represents it with the brush at b. 

This increased armature interpolar flux due to the brush 
shifting means that the resultant interpolar flux due to both the 
armature and main field fluxes will cross the zero line at a point 
further removed from the midpoint than in the case of no lead 
at the brushes. Consequently, in order to obtain a given useful 
commutating field the brushes must be given a greater amount of 
lead and this in turn shifts the zero point still further Thus, 
the act itself of shifting the brushes makes the commutating con- 
ditions more difficult. 

The calculation of the comrnutating conditions with any given 
lead therefore resolves itself into a determination of the re- 

FIG 20 

sultant fluxes in which the coil is short circuited or commutated 
and the e.m fs generated by such fluxes. For the slot and end 
winding fluxes the calculation will be the same as for no-lead at 
the brushes The resultant flux in the interpolar space is the 
only condition which will introduce any variation from the pre- 
ceding formulae and methods of calculation This part of the 
problem resolves itself simply into the determination of the 
resultant interpolar flux at the point of commutation for any 
given load The corresponding e.m.f can then be calculated. 
This, combined with the e.m.fs. due to the slot and end windings, 
gives the total short-circuit e.m.f The method is, in principle, 
exactly the same as given before, except that the determination 
of the interpolar flux will be modified. 

Summation of Formula In order to obtain the total voltage in 
the short circuited coil a summation should be made of the four 
separate voltages which have been derived for the interpolar, 


end, slot and band fluxes. In reality it is the resultant fluxes 
which should be combined, but as the voltages to be derived 
from these fluxes represent somewhat different terms, a better 
procedure appears to be the summation of the voltages. Also, 
in practice it is the e.m.fs. generated by the different fluxes, 
rather than the fluxes themselves, which are desired. 
The e.m.f. derived from the interpolar flux is 

WiTcRs 2pirDL 

10 s (0.25 +0.5) (jrD-Pp) 

where Ci is a correcting factor for chord winding, etc. 
The formula for the, end flux voltage is, 

_ ^Ic W t T e R>^-4.3(2h+m) 
*~ - CaX HP x ihT0 

where d represents the correcting factor for chord windings, etc. 
The formula for the slot flux voltage is, 

3.19 I c W t T C R S L (2.666 d+4a+t+2.16s Vn) 
&c-c 3 X 1Q8 - 

where c 3 is the correcting factor for the brush width, chord 
winding, etc., and, 
For the bands, 

where c 4 is the correcting factor for chord winding, brush width, 



<-*-i-** '* . _ 7T 
total- ^ L' ((0.25/^+0.5) 

2<t>NpT c R s 


It is evident from this last equation that when there are no 
bands over the core the total e.m.f. in the short circuited coil 
is directly proportional to the current per armature coil or con- 
ductor. If the bands saturate, as would usually be the case 
with any considerable load, then the e.m.f. is no longer directly 
proportional to the current. Attention is called to this point 
as it has some bearing in the design of interpole machines. 

Condensed Approximate Formula. The above formula can 
be simplified very considerably by certain approximations which 
introduce but little error within the range of ordinary design 

First, the expression, (025/)+05 f (irD _ Pp} does not seem 

to be capable of any general simplification. In fact, as shown 
from its derivation, it is not a general term, but applies only to 
certain constructions and may appear in a quite different form 
for other constructions. Therefore this expression must be 
used with judgment in any case. Moreover, this term appears 
only in non-interpole machines or in interpole machines only 
when the interpoles are narrower than the armature core or the 
number of interpoles is less than that of the main poles. There- 
fore this term may be neglected in many cases where interpoles 
are used. 

/O 7L I ft-\ 

Second, the expression 4.3 -W-: ^p- log 2 N can be changed 
as follows: 

4-3 / m = - * with reasonable accuracy within the 
(sin u) p 

ordinary limit of design, 

And log 2 N= 0.9+0.035 N, with an error of about 4 per cent 
within the range of 6 to 24 slots. 

Therefore 4.3 -" log 2N=--- (0.9+0.035 N} r 

(sin u) p 

approximately . 

This is simpler to handle, in practice, than the original term. 

Third, the expression, *-wr'T*-"* v " can ^ sim , 

plified very materially. 

Let the total depth of slot be represented by d*, which is equal 
to 2 d+a+1.5 t, approximately. 

4d s 8 a 3 / 

Then, the term, 2.666 d+ a +t can be changed tc ^- ^ ^ 


Assuming a 0.25 and J0.15, then 

Q __ O j 

5 =0.52 approximately. 


_. , 2.6Qd+4a+t id, . 52 
Therefore , _ -37+ 

This is a very close approximation within the ordinary work- 
ing range of slot dimensions. Therefore, the above expression 

becomes, * H ' 1-2. 16 V#, which is much simpler to use in 

3 s $ 


Fourth, in the simplified equation TT appears in the first 
and second terms, and 3.19 appears in the third term. These 
are so nearly equal that TT may be used as a common factor for 
the three terms. 

The combined formulas for the total voltage per armature 
coil thus becomes, in approximate form, 


" 10* (0.25+0.5)(7rZ>-P/>) 

4 D 

+C* ^f- (0.9+0.035 N) + 


(1.33 &+0.52+2.16* Vn)1 , 2 <fr> N}> T c R, 
- - - - 


This appears to be about as simple a form as the equation can 
be put into when all the factors are to be included- It will be 
shortened for machines without magnetic bands on the core 
and in many interpole machines the term derived from the inter- 
polar flux may be omitted. For a given line of machines which 
are all of similar design, etc., it is probable that the terms can 
be further combined and simplified. 


In the interpole machine a small pole is placed between two 
adjacent main poles for the purpose of setting up a local magnetic 
flux under which the armature coil is commutated. This local 

*For ordinary working range of slot dimensions, 216 JV* 1.07 X tooth pitch at ar- 
mature surface. This formula may be further simplified by substituting 1,07 Pt for 2.16 
s v/n. Pt being the tooth pitch at the armature surface. 


flux, in order to assist commutation, must be opposite in direction 
to the interpolar flux set up by the armature winding itself. To 
set up this flux in the opposite direction the magnetomotive 
force of the interpole winding obviously must be greater than 
that of the armature winding in the commutating zone. 

An armature coil, cutting across this interpole flux, generates 
an e.m.f. proportional to the flux, the speed and the number of 
conductors in series. This e.m.f. is in opposition to the e.m.f. 
in the short circuited coils, generated by the slot and end winding 
fluxes. For ideal commutation these e.m.fs. are not only in 
opposition, but they should also be of practically equal value. 
For perfect commutation the current in a short circuited coil 
should die down to zero value at about a uniform rate and 
should then rise to normal value in the opposite direction by the 
time the coil passes out from under the brush, as was illustrated 
in Fig. 12. This is the condition when no local currents are de- 
veloped in the short circuited coils and this can only be obtained 
when the interpole e.m.f, at all times, balances the armature 
e.m.fs. in the short circuited coils. 

Looking at the problem broadly, the resultant magnetic 
fluxes and e.m.fs. may be assumed as made up of two com- 
ponents which can be considered singly. One of these com- 
ponents is that which would be obtained with the armature 
magnetomotive forte alone acting through the various flux paths, 
including the interpole. The other would be that which would be 
obtained with the full interpole magnetomotive force alone, the 
armature magnetomotive force being absent. Saturation is not 
considered in. either case. 

Considering the first component, due to the armature mag- 
netomotive force alone, there would be the slot and the end 
fluxes with their short circuit e.m.fs., as already described, 
and in addition, there would be a relatively high flux, and short- 
circuit e.m.f. due to the good magnetic path furnished by the 
interpole core. In case the interpole does not cover the full 
width of the armature, or the number of interpoles is less than 
the main poles, there will also be some interpolar flux and e.m.f. , 
as already described. 

Considering the second component, the entire interpole mag- 
netomotive force would set up a relatively high flux through the 
interpole magnetic circuit and a. correspondingly high e.m.f. 
would be generated in a sfoort drcuited armature coil cutting 

fhift flux. 


When these two components are superimposed, it is seen that 
the interpole flux due to the armature magnetomotive force is 
in direct opposition to that due to the interpole magneto- 
motive force and therefore only the e.m.f. due to their dif- 
ference need be considered. As the interpole winding has the 
higher magnetomotive force, the resultant interpole e.m.f. is 
in opposite direction to the armature e.m.fs., and should be 
sufficient to neutralize them. This way of considering the prob- 
lem avoids a number of confusing elements which would com* 
plicate the explanation if given in detail. 

In practice it is difficult to obtain exact equality between the 
interpole and armature e.m.fs. That due to the armature 
fluxes is generated in all parts of the coil including the end 
winding, while the e.m.f. due to the interpole flux is generated 
only in that part of the coil which lies in the. armature, slots, 
However, it makes no difference in what part of the coil the e.m.f. 
due to the interpole is generated provided it is of such value that 
it properly opposes and neutralizes the various e.m.f s., due to 
the armature fluxes. Therefore, in practice the interpoles need 
not have the same width as the armature core and, where space 
and magnetic conditions will permit, the number of interpoles 
can be made half that of the main poles. 

According to the method outlined, the whole problem of the 
design of the interpole depends, first, upon the determination 
of the a.m.fs. due to the armature fluxes, and, second, upon the 
determination of such interpole flux as will generate an e.m.f, 
in the short circuited armature coils which will equal, or slightly 
exceed, the armature e.m.fs. 

Interpole Calculations. Assuming that all the armature 
fluxes, except the interpolar, are unaffected by the presence of 
interpoles, the armature e.m.f. to be balanced by the interpole 
would be represented by the formula 


_ _ 
- (0,25 +0.5) 

(0.9+0.035 AO + 
1.88 A +0.52+2.16 * Vn 


However, the flux above the slot, from the tooth top, is very 
considerably modified by the interpolar flux. In fact most of 
this should be omitted It may be assumed that the flux across 
the slot, above the upper coil, simply " bulges " up slightly 
into the air gap, and the remainder of the usual tooth top flux 
is absent, except when the interpole does not cover the full 
armature width. Therefore, in the above formula, the term 
LX2 16 V^ should be changed to (L-Li)X2 16 Vn and the 

] .33 ck + 0.52 . , u 1.33 & +0.7 

term ! replaced by jL_L___ 

s s 

Then, the corrected resultant of all the armature e.m.fs. 

/T T \ % D P 

d (LL^ 


4. n 
+cjX-~ ^ (0.9+0.035 JV) 

In this formula 

L represents the width of the armature core. 

Li represents the effective width of interpole at the gap 
on the basis of the full number of interpoles. 

LLi is the difference between the width of the armature 
core and the interpole face. This term enters when the interpoie 
is narrower than the armature core. When alternate interpoles 
are omitted and the remaining interpoles are of the same width, 
as the armature core the conditions are practically the same as 
when the full number of interpoles are used but with their width 
equal to half the core width. Other combinations should be 
treated in the same way so that the above formula can be taken 
to represent the general conditions. 

In practice it is desired that the resultant interpole,, 
and therefore the interpole flux, vary in- proportion to the arma- 
ture short-circuit e.m.f. which is to be neutralized. As shown 
fcy the last ^equation, this e.m.f . is proportional to the armature 


current, except where there is saturation in the armature flux 
path, as in the case of magnetic bands over the core* Therefore 
the interpole magnetomotive force should vary in proportion 
to the armature current, neglecting core bands. In consequence, 
in practice the interpole winding is always connected in series 
with the armature winding. 

The interpole magnetomotive force can be considered as made 
up of two components, one of which neutralizes the armature 
magnetomotive force, and the other component represents the 
ampere turns which set up the actual interpole flux. The first 
component will be referred to as the neutralizing ampere turns 
or neutralizing turns , and the other as the magnetizing ampere 
turns or magnetizing turns. 

Let T represent the total interpole turns for one interpole, 

T v represent the total magnetizing interpole turns for one 

T a represent the total effective " armature turns per 

total eff . ampere turns of armature , 
~~ number poles X total current ' 

/ represent the amperes per interpole coil. 
Then IT t = IT-I T a , or, r= 

Let g = effective air gap per interpole. 

2J= flux density under the interpole, and 
JS, = e.m.f. in an armature coil of turns T f due to the 
interpole flux 

The e.mX due to one interpole is equal to 

Or, for two interpoles, 

*If average slot e. rru f. is used in calculating' EC (See Note on page 215), this expression 
should be multiplied by a factor Cp to obtain, average value of Ei. Cp is the ratio of the 
average flux density in the commutating zone, to the marfmijTn density, J5. 


This e.m.f. should be equal to the e.m.f. generated in the same 
coils by the armature flux, or Ei=E c . Therefore, 

3.19 I Tt TT D LiX2 T c R s -I e W,.T f R s ir 

, r _ r 

gXlO 8 10* 

2D p 


(L-i02.16 V^ T < 

In the second term of this equation 7 t W ( = IX T a 'X2p, where 
jT ' = total armature turns per pole, as distinguished from effec- 

tive turns per pole T a , and T a ' =- ? , where & = ^'T' as will 

1 op Wt 

be shown later under the subject of " Effective Armature Am- 
pere Turns." Therefore, neglecting magnetic bands on the core, 
the above expression becomes, 

r r a j>g r 

1 *~3.19 DZ-! (l-&^>) L 

(0.25 ^>+0.5) (irD-Pp) 

(0.9+0.035 (1 ' 33 

+c (L-iO 2.16 



(0.9+0.025 . 
+c (L LI) 2.16 


If the full number of interpoles is used, and each covers the 
full width of the armature, then Z~Z,i = 0, and 


Therefore the total interpole turns for one pole are equal to 
the effective armature turns, per pole multiplied by a constant 
which is a function of the' proportions of the machine. How- 
ever, this holds true only for the condition of no saturated path 
for the armature flux, such as magnetic bands. 

The above formula gives the interpole turns for two inter- 
poles acting on each armature coil. With but one interpole 
per coil the number of conductors per armature coil generating 
the interpole e.m.f. is halved so that the flux density must be at 
least doubled, and the effect of the armature flux in the inter- 
polar space over the other half of the armature coil must also be 
taken into account. This can be done in the preceding formula 
by using the equivalent value of LI. 

With half the number of interpoles the effective gap length, g r 
will not be the same as with the full number of interpoles with 
the sameinechanical gap, for the flux from the interpole maybe 
considered as returning across the gap of the two adjacent main 
poles and the value of g must be increased to represent the total 
resultant gap. 

Let ge represent the effective resultant gap, 

g m represent thfc effective gap under the main poles, 
A i represent the area of the interpole gap, and 
A m represent the area of one main pole gap. 

These areas can be derived from the field distribution or " field 
form " of the main and the interpoles. 

Then, the effective resultant gap g, = g-f p g mj and this 

" ./LIT* 

should be used instead of g 

With half the number of interpoles and on the basis of the 
interpole flux returning through the two adjacent main poles, 
it may be assumed that this flux weakens the total flux in one 
pole and strengthens that of the other pole a Ek^ amount. If 


there is no saturation in the main poles or armature teeth under 
them, then no additional ampere turns, other than for the 
increased gap, will be required on account of the main poles 
carrying the interpolar fluxes. However, where there is much 
saturation of the main poles or teeth, then additional interpole 
ampere turns will be required, as will be described later in con- 
nection with effects of saturation. 

Chord Windings with Interpoles. Chorded armature windings 
can be used with interpoles with satisfactory results provided 
the interpoles are suitably proportioned. There are apparently 
some advantages with such an arrangement, but there are also 
disadvantages of such a nature that it is questionable whether 
it is advisable to use chord windings with such machines, except 
possibly in special cases. When chorded windings are used with 
interpoles, the e.m.f. due to the armature flux is usually much 
smaller than with a pitch winding and thus fewer interpole 
magnetizing turns are required. Also, the effective armature 
turns which must be neutralized by the interpole are reduced 
somewhat, which also means a slight reduction in interpole 
tuftis. Against these advantages must be charged the disad- 
vantage of a wider interpole face. This in itself would not be 
objectionable where there is space for such wider pole face, but 
if the space between the main poles must be increased it may lead 
to sacrifice in the proportions of the main poles or changes in 
the general dimensions, such that the result as a whole is less 
economical than with a pitch winding. 

Effective Armature Ampere Turns. The term T a representing 
the effective artnature ampere turns should be considered, as the 
value of this term is influenced by a number of conditions, such 
as the number of bars covered by the brush, the amount of 
chording in the armature winding, etc. With a full pitch wind- 
ing and neglecting the reduction in current in the short circuited 
coils, the magnetomotive force of the total armature winding 

is represented by the expression, ^ L , and per pole it is, 

- . However, when the brush spans several coils, so that 


a number of armature coils are short circuited at the same time, 
the average current in these short circuit coils should be con- 
siderably less than the normal value so that the effective ampere 
turns per pole is correspondingly reduced. Allowance must be 


made for this reduction as it has considerable influence in de- 
termining the correct number of interpole turns. 

On the basis of no local currents, the average value of the 
current in the short circuited coils is just half that of the work 
currents per conductor. 

Let B represent the total number of commutator bars, 

BI represent the number of bars spanned by the brush, 
pi represent number of current paths, and 
p number of poles. 

n <TT 

Then, . ," = total number of armature turns per pole, and 

7? 7" 1 

* c = number of turns by which the total armature turns per 

& pi 

pole must be reduced to obtain the effective turns per pole, or, 

^.g-, ..,.___ 

Let BI T e be represented as a percentage of W h or BI T c = b Wt 
Then, r o = * (1 

W t IW t 


Therefore, T a / = - rr ' 

Chorded windings also have an influence on the effective arma- 
ture ampere turns per pole- When the winding is chorded one 
slot, for example, then, in one slot per pole, the upper and lower 
coils will be carrying current in opposite directions and tEeir 
magnetizing effects will be neutralized. In consequence, the 
total effective armature ampere turns are correspondingly re- 
duced and this must be allowed for in determining the interpole 



Conditions Affecting Interpole Proportions. The foregoing 
formulae have been based upon the use of interpoles of such pro- 
portions that the interpole flux varies directly as the magnetizing 
current and its distribution over the cojnmutating zone is such 
as will give the proper opposing e.m.f. at all times. 

However, the proportionality of flux to current can only be 
true as long as there is no saturation, in the interpole magnetic 
circuit. Such saturation is liable to be found in practice and not 
infrequently it is quite a problem df design to avoid it within 
the working range of the machine. 

Also, another difficult problem lies in so designing the interpole 
face that the flux distribution in the commutating zone is such 
that its e.m.f. will properly balance the armature e.m.fs. in the 
short circuited coils, especially as the latter are generated by 
cutting fluxes which may be distributed in a quite different man- 
ner from the interpole flux. 

FIG. 21 

Shape and Proportion of Interpole Face. As already shown, 
the effective interpole flux under the pole face is the "resultant 
of the total interpole magnetomotive force and the opposing 
armature magnetomotive force. As the armature windiag is 
distributed over a surface and the interpole winding is of the 
polar or concentrated type, the resultant magnetomotive force 
would normally be such as would not tend to give a uniform flux 
distribution under the interpole unless the interpole face is 
properly shaped or proportioned for such distribution. The 
conditions may be illustrated in Pig. 21. In this figure the lines 
A A represent the armature magnetomotive force, with a full 
pitch winding, and the brash covering one commutator" bar. 
The heavier p&rt of the lines a b c at the peak of the magneto- 
motive force diagram, represents the armature magnetomotive 
fdrce which would be obtained tinder the interpole face, a#d also 
the fttcc distribttfioa wMdi would be obtained, with no interpole 


magnetomotive force, and with, uniform gap under the pole 
faces. In opposition is shown the interpole magnetomotive 
force and flux distributions d ef for corresponding conditions. 
The resultant magnetomotive force is represented by g h i, and 
with a uniform gap under the pole, the resultant interpole flux 
would have a similar distribution. Instead of this, either a 
flat or, in some cases, the reverse distribution is required, that 
is, with a slight " hump " in the middle instead of a depression- 
By prbperly shaping the pole face so as to give an increased air 
-gap toward the edges, the flux distribution can be made practi- 
cally anything desired. In some cases a relatively narrow 
pole tip with a very large air gap will give a close approximation 
to the desired flux distribution. 

However, in practice the above distribution of the armature 
magnetomotive force is rarely found. The use of brushes 
which cover more than one commutator bar serves to cut off 
or flatten out the pointed top of the armature magnetomotive 
force diagram, as shown by the dotted line B, in Fig. 21, and thus 
lessen the depression at the center of the resultant magneto- 
motive force distribution. 

As intimated before, this problem of proportioning the inter- 
pole face turns upon the determination of the armature e.m.fs. 
in the short circuited coils which have to be balanced by the 
interpole. If the different armature e.m.fs. are determined for 
the whole period of commutation and then superimposed, the 
resultant e.m.f . indicates the flux distribution required under the 
interpole. Usually the e.m.fs. due to the end winding, and to 
the interpolar flux, if any, will be practically constant during the 
whole period of commutation. If no local currents are present 
the e.m.f. due to the slot flux will also be practically constant, 
although it may be slightly reduced near the beginning and end 
of the commutation period. The sum of these e.m.fs. should 
therefore be practically Constant over the whole commutation 
period and therefore, in a well designed machine, the interpole 
flux, density should be practically constant over the whole 
commutation zone. As explained before, special shaping of the 
poles and pole face will be necessary, in most cases, to obtain 
exactly this proper flux distribution. Large interpole air gaps 
are obviously advantageous in obtaining such distribution. In 
fact* a very small interpole gap makes the determination of the 
properjnterpole face dimensions very difficult in, many cases. 
On accocrat of the interpole usually covering less than two 


armature teeth, the ordinarily accepted methods of determining 
the effective length of air gap under a pole will not apply, iifc 
many cases, which may lead to a slight error in the results. 
Practically the effective gap tinder the narrow interpole will 
usually be longer than determined by the ordinary methods. 
This partly explains the fact that, in some cases, an increase 
in mechanical clearance between the interpole face and the 
armature core does not require anything like a corresponding 
increase in the interpole magnetizing ampere turns. The effec- 
tive interpole air gap increases, but at a much less rate than the 
mechanical gap. 

The brush setting in relation to the interpole is of great im- 
portance. The point of maximum armature magnetomotive 
force is definitely fixed by the brush setting. With the interpole 
fixed in position, any shifting of the biuishes backward or forward 
will obviously change the shape of the- resultant magnetomotive 
fores distribution under the interpole face and in consequence 
the flux distribution will be changed. With but one armature 
coil per slot and the brush covering but one commutator bar, 
good commutating conditions might be found over a considerable 
range of brush adjustment, by suitably varying the interpole 
ampere turns. However, with two or more coils per slot and 
with the brush short circuiting several bars, any marked change 
in the resultant interpole magnetomotive force and flux distribu- 
tion will mean improper commutation for some of the coils. 
Proper brush setting is therefore of first importance. 

It has been ''assumed in the foregoing treatment, that an exact 
balance between the interpole and armature e.m.f s. will give the 
best conditions. From certain standpoints, this is true, but in 
practice usually a slight excess in the interpole strength, or 
" over-compensation " of the interpole, as it is frequently called, 
is advantageous. Reference to Fig. 14 shows that in a machine 
without interpoles, and therefore without compensation, the 
current flowing between the brush contact and the commutator 
is crowded toward one brush edge, this being the edge at which 
the commutation of a coil is completed, that is, at the so-called 
forward brush edge. With over-compensation the opposite 
effect occurs that is, the brush current density is below the 
average at the forward edge. This is, to a certain extent, a de- 
sirable condition. Also, if there is any saturation of the inter- 
pole circuit at overloads, the over excitation of the interpole 
winding can take care of the saturation ampere turns, so that 


normal compensation can be obtained at considerably higher load 
than in a machine with no over compensation. Furthermore, 
over compensation is desirable on account of the effect of the 
resistance of the coils undergoing commutation, which heretofore 
has been neglected as being of minor importance. Such re- 
sistance tends to lower the current density at the middle of the 
brush contact, and increase it toward the brush edges. Over 
compensation will oppose this at the forward edge, but increase it 
at the back edge, which is less objectionable. Also, as shown in 
Fig. 21, there is liable to be a depression at the center of the 
interpole flux distribution, if the pole face is not properly 
shaped. This depression tends to cause higher current densities 
at the brush edges. Over compensation again tends to reduce 
this density at the forward brush edge. Thus there are several 
good reasons for slight over compensation, and practical opera- 
tion bears this out, especially on high voltage machines, where 
the short circuit e.m.fs. average higher than in other machines. 

Balanced Circuits. It has been assumed that the armature 
ampere turns per pole have been the same for all poles. This 
will be true for the usual two-circuit or series type of winding, 
or its allied combinations, but is not necessarily true of the 
parallel type of armature winding. Ill such a winding a number 
of circuits are connected in parallel at the brushes, and, unless 
ample provision be made for equalizing the different circuits, 
they may not carry equal currents at all times. As the re- 
sultant interpole flux and e.m.f. is directly dependent upon the 
opposing armature ampere turns, it is obvious that any ine- 
qualities in the armature currents would lead at once to incorrect 
interpole conditions. A poorly equalized parallel-wound arma- 
ture might furnish conditions such that the interpoles cannot be 
adjusted for satisfactory operation. Also paralleling of the 
interpole windings, unless care be taken to insure equal cutrent 
division among the circuits, is liable to lead to trouble, 

Saturation of the Inter.pole Circuit. Heretofore the interpole 
turns T, as determined, have been only those required for forcing 
the resultant interpole flux across the effective interpole air gap, 
and nothing has been allowed for any turns required for magnet- 
izing the parts of the interpole circuit other than the gap. 
Where such additional turns are required they must be added to 
the turns T, already determined. 

Saturation in the interpole magnetic path is the principle 
cause for such additional turns, but saturation in the various 


flux paths may occur in such a way as to be either harmful or 
beneficial, depending upon where it is located. Beneficial 
saturation may be assumed to be such as will reduce the arma- 
ture short circuit e.m.fs,, while harmful saturation tends to re- 
duce the interpole e.m.f. 

While the useful interpole flux passing into the armature may 
be relatively low say one-fifth that required for saturation of 
the interpole material the leakage flux between the interpole 
and the two adjacent main poles is often very much greater 
than the useful flux so that the interpole at the part where it 
carries the highest total flux may be worked up to possibly half 
saturation, or higher, with normal load on the machine. The 
interpole leakage flux is due to the total ampere turns on the inter- 
pole, while the useful interpole flux is due only to the magne- 
tizing component of the interpole ampere turns, which may be 
as low as 15 per cent to 25 per cent of the total interpole ampere 
turns. Ths leakage flux is thus liable to be a high percentage 
of the total interpole flux t 

"While the ampere turns on the interpole will rise in direct 
proportion to the current, the effective magnetizing component 
will rise in direct proportion only below saturation of the inter- 
pole circuit. Any ampere turns required for saturating this 
circuit will be taken from the magnetizing component of the 
interpole winding. Therefore, when any appreciable saturation 
occurs, the effective magnetizing component will not vary in 
proportion to the current, and the interpole e.m.f. will not vaiy 
in proportion to the armature e.m.f s. As the magnetizing com- 
ponent of the interpole winding usually represents a relatively 
small number of ampere turns per pole a comparatively slight 
saturation in the interpole circuit may have an appreciable effect * 
It is therefore advisable to work at as low a saturation as possible 
in the interpole circuit so that practically no saturation occurs 
within the ordinary working range of the machine. 

Where saturation occurs in any of the armature flux paths, 
as, for instance, with saturated bands over the armature core, 
the result of such saturation will serve to neutralize the effect 
of saturation in the interpole magnetic circuit. In other words, 
the armature e.m.f. will not rise in proportion to the current 
and thei*efore the opposing interpole e.m.f. does not need to 
increase in proportion either. 

The principal source of saturation in the interpole circuit lies 
in the magnetic leakage from the interpole to the adjacent main 


poles. Serious trouble has often been encountered by not 
making due allowance for such leakage. However, there may 
be other causes for saturation. When the full number of inter- 
poles is used the interpole magnetic path or circuit is independent 
of the main pole magnetic circuit, except in the yoke and in 
the armature core below the slots, as indicated in Fig. 22. 
In the yoke it may be seen that the interpole flux is in the same 
direction as the main flux at one side of the main pole and is in 
opposition to the main flux at the other side. The same is true 
in the armature core. Therefore the interpolar flux tends to 
reduce the flux in one part of the yoke and tends to increase it in 
the other part. If the saturation in these parts is relatively low, 
then the magnetomotive force required for forcing the low and 
the high fluxes through the yoke will be but little greater than if 
these fluxes were equal. However, if the yoke is highly saturated 
the increase in ampere turns required for the high part much 
more than offset the decrease in ampere turns for the low part, 
so that, as a result, additional am- 
pere turns are required for sending 
the interpole flux through this path. 
The interpole ampere turns therefore 
must be increased on this account, } 

^riien the saturation is high. The same F 22 

condition holds for the armature core. 

A similar condition occurs -where half the number of interpoles 
is used and when there is much saturation of the main pole and 
the armature teeth under it, as already referred to. This con- 
dition requires additional interpole ampere turns. 

In practice, with the ordinary compact designs of direct cur- 
rent machines, it is usually difficult to keep the total interpole 
flux as low as one-third that which gives any material saturation 
and, not infrequently, it is much higher than this. Therefore, 
by direct proportion it might be assumed that such machines 
could carry only double to treble load without sparking badly. , 
However, the resistance of the brushes, etc., will be of such as- 
sistance that relatively higher loads may be commutated rea- 
sonably well. For instance, with the interpole worked at about 
half saturation at normal load, the machine may be able to 
commutate considerably more than double load without undue 
sparking. It is also of material assistance, where heavy over- 
loads are to be carried, to over-excite the interpolevwinding, 
that is, to make the magnetizing component somewhat greater 


than required at normal load, as described before. In this case, 
at light loads, the interpole e.m.f. exceeds the armature e.m.f. 
a certain amount which is taken care of by the brush resistance 
as local currents will be less harmful when the work current is 
low. As partial saturation is obtained at overload, the two 
e.m.fs. become equal but at a higher load than would be the case 
without ovef -excitation of the interpole. 

Commutating Conditions on Short Circuit. When a direct 
current generator is short circuited across its terminals, either 
through a low external resistance or without such resistance, a 
current rush will occur which will rise to a value represented 
approximately by the generated e.m.f. divided by the resistance 
in circuit. This current rush is only of short duration as the 
excessive armature current will react to demagnetize or " kill " 
the field. If the short circuit is without external resistance the 
current rush may reach an enormous value as the internal re- 
sistance on large machines is usually very low. This means that 
currents from 25 to 40 times full load may be obtained on " dead " 
short circuit. Experience shows that under such current rushes, 
any kind of direct current machine will tend to flash viciously at 
the brushes. 

By the preceding theory and analysis a rough approximation 
to the commutating conditions on short circuit can readily be 
obtained. Assuming an interpole machine, the following con- 
ditions will be found: 

1. The interpole will be highly saturated so that it is of little 
or no direct benefit. 

2. The slot flux will rise to such a value that the armature 
teeth in the commutating zone are practically saturated. 

3. There may be some interpolar flux from the armature, as the 
high interpole saturation may allow this. 

4. The armature end flux, with the exception of that part due 
to magnetic bands, will rise practically in proportion to the 

The following short circuit e.m.f. conditions will be obtained: 

1. There will be possibly a slight e.m.f. due to the armature 
interpolar flux. 

2. There will be an e.m.f. due to the tooth flux which is almost 
as high, per conductor, as could be obtained by a conductor 
cutting the flux under the main field at no load, for saturation of 
the armature teeth may be assumed to be the limit in both cases. 

3. There will be an e.m.f. due to the end flux which may be 
10 to 20 times larger than at normal full load. 


Therefore, the total e.m.. in the short circuited coil due to 
cutting the armature flux on dead short circuit may be higher 
than would be obtained if the brushes were shifted at no load until 
the commutated coil lies under the strongest part of the main field. 

As very few machines of large capacity would stand this 
latter condition without flashing, it may be assumed that they 
would be no more able to stand a dead short circuit without 
flashing. In fact, 8 to 10 times full load current will make an 
interpolar machine of normally good design flash badly, as it is 
impracticable to make an interpole of the usual type which will 
not saturate highly at 8 to 10 times normal current. 

If, however, the interpole is combined with compensating 
windings in the main poles, the interpole leakage may be made 
so small that comparatively low saturation is obtained normally 
in the interpole circuit. In such case the interpole may be effec- 
tive with heavier currents and the flashing load may be very 
much higher than with the usual type of interpole machines. 


The foregoing is a general presentation of the problem of 
commutation, which is admittedly crude and incomplete in some 
points. In particular may be mentioned the part describing, 
the action of local currents. Also, the method of considering 
the resultant action in interpole machines as the superposition 
of two components does not tell the whole story, but the actual 
analysis, in detail, of a number of these phenomena would be so 
confusing and complicated, that a general physical conception 
of what takes place during commutation would be lost. In the 
ultimate analysis it will be found that a number of the methods 
described are, In reality, simply illustrations of the conditions 
of commutation rather than an analysis of the conditions them- 
selves. However, the method as given throws light on many 
things which take placie during commutation. It also includes 
a number of conditions which are not -covered in the usual 
methods of dealing with this problem. For example, the number 
of commutator bars spanned by the brush is an important ele- 
ment in this method of handling the problem, whereas, in many 
former methods, this point was either omitted, or treated in an 
empirical manner. In this method the results obtained would be 
very greatly in error if the brush span were not included. 

Any theory qr method of calculation is open to question until 
ti has stood the proof of actual test. In consequence, the above 


"jnethod has been tried on a very large number of direct current 
machines, including high speed direct current generators, 
direct current turbo-generators, direct current railway motors of 
all sizes, moderate- and low-speed generators of all capacities, 
industrial motors of various designs including adjustable speed 
motors and machines with half the number of interpoles. In 
those cases where tfce actual test data of the machines was very 
accurately obtained, the agreement between the tests and the 
calculated results by the above method was found to be close. 
In fact, the method in som cas6s indicated errors or inaccuracies 
in the test results. In a number of cases of early interpole 
machines there was considerable disagreement between the 
results of the calculation and the actual test, but, in many of 
these cases, later experience showed definitely that the proper 
interpole field strength or proportions had not been obtained 
in the actual test or that the proper brush setting had not been 
used. These cases were thus, to a certain extent, a verification 
of the method, for in general the greatest discrepancies between 
the calculated and the test results corresponded to the ma- 
chines which eventually proved to have the poorest proportions 
or adjustment. 

This theory of commutation looks complicated and cumber- 
some in its practical application, but it should be understood 
that it is, in reality, an exposition of a general method from 
which special and simpler methods may be derived -for different 
types and designs of machines. It indicates plainly that the 
problem is so complicated that no simple formulae or methods of 
calculation can be devised which will cover more than individual 
cases, and that such formulae, if applied generally, will lead 
to errtfr sooner or later. If, however, tt^e general derivation 
of such simplified formulae is well understood, then they may 
be used with proper judgment and with much less danger of 
error in the results. It is evident, from the general analysis, that 
the whole problem must be handled with judgment, for new or 
different conditions are encountered in almost every type of 

A great many problems, closely allied to that of commutation 
in interpole machines, have not been considered, because some 
of them represent special cases of the general theory, while 
others axe somewhat outside the subject of this paper. Of the 
former class may be mentioned, commutation of synchronous 
converters, niachiiies with distributed or true compensating 


windings, the so-called " split-pole " converter, and the commu- 
tator type alternating current motors, etc. In the latter class 
may be included such problems as the effect on commutation of 
closed circuits around the interpoles, losses due to commutation, 
current distribution at the brush contact, etc. Some of these 
subjects were included in this paper as originally prepared, but 
on account of its undue length they had to be omitted. 


FOREWORD This paper was presented before the American Insti- 
tute of Electrical Engineers in San Francisco, September 16, 
1915, at the Electrical Congress at the Panama-Pacific Inter* 
national Exposition. It gives the results of the author's work 
on determination of commutating limits covering the period of 
many years of work. As regards commutation, it is, in reality, 
a supplement to the paper, "Theory of Commutation and Its 
Application to Commutating Pole Machines." On the subject 
of flashing, it covers some very interesting limitations, based 
upon experience and special tests. 

This paper is, in reality, more or less of a general summary 
of the author's experience in direct-current machinery. Al- 
though usually looked upon as an "alternating-current man," 
he has probably spent as much total time on direct-current 
work as on alternating. Many of the limiting conditions in 
direct-current machinery, as described in this paper, were de- 
termined by the author himself. Many of the early more or 
less radical developments and improvements in direct-current 
machinery resulted directly from his work. A description of 
some of these is covered in his historical papers, entitled, "The 
Development of the Direct-Current Generator in America," 
and "The Development of the Street Railway Motor in Amer- 
ica" which appear in the latter part of this volume. (ED.) 

IN DIRECT-current commutating machinery there are 
many limitations in practical design which cannot be 
exceeded without undue risk in operating characteristics. 

Some of these limitations are actually physical ones, and, there- 
fore, cannot be avoided or over-stepped without very considerable 
departures from our present methods of construction and opera- 
tion; others are not wholly physical, but are fixed largely by 
practical experience, and are, in consequence, subject to modifica- 
tion, as our experience is increased. Seme of them are quite defin- 
ite in nature, while others axe indefinite. Some are measurable, 
in a quantitative sense, while others may be considered as quali- 
tative. Noise, for instance, is a distinct limitation, in many cases, 
but it is difficult to fix any definite value where it is prohibitive. 



Many of these limits are not sharply defined in practise, due, in 
many cases, to the impossibility of taking advantage of all" the 
helpful conditions and of avoiding the objectionable ones. There 
are many minor conditions which affect the permissible limits 
of operation, which are practically beyond the scope of reliable 
calculation. Usually, such conditions are recognized, and al- 
lowance is made for them. It is the purpose of this paper to 
treat of some of the major, as well as minor, conditions which 
must be taken into account in advanced direct-current design. 
These are so numerous, and are so interwoven, that it is difficult 
to present them in any consecutive order. 

Probably the most serious limitation encountered in direct- 
current electric machinery is that of commutation. This is an 
electrical problem primarily, but in carrying any design of direct- 
current machine to the utmost, certain limitations are found 
which are, to a certain extent, dependent upon the physical 
characteristics of materials, constructions, etc 

A second limitation which is usually considered as primarily 
an electrical one, namely, flashing, (and bucking) is in reality 
fixed as much by physical as by purely electrical conditions. 

A third limitation is found in blackening and burning of com- 
mutators, burning and honeycombing of brushes, etc. These 
actions are, to a certain extent, electrical, but are partly physical 
and "mechanical, as distinguished from purely electrical. 

There are many other limiting conditions dependent upon 
speed, voltage, output per pole, quality or kind of materials 
used, etc. As indicated before, these cannot all be treated 
separately and individually, as they are too closely related to 
other characteristics and limitations. 

In dealing with the limits of commutation, it is unnecessary 
to go into the theory of commutation, except to indicate the 
general idea upon which the following treatment is based. This 
has been given more fully elsewhere,* and therefore the following 
brief treatment will probably be sufficient for all that is required 
in this paper. 

In this -theory it is considered that the armature winding as a 
whole tends to set up a magnetic field when carrying current, 
and that the armature conductors cutting this magnetic field 

*Theory of Commutation and Its Application to Comxautatxng Pole Machines, Page 201. 


generate e.m.fs. just as when cutting any magnetic field. 
From consideration of the armature magnetomotive force alone, 
the flux or field set up by this winding would have a maximum 
value over those armature conductors which are connected to 
the brushes. If the magnetic conditions or paths surrounding 
the armature were equally good at all points, this would be true. 
However, with the usual interpolar spaces in direct-current 
machines, the magnetic paths above the commutated coils are 
usually of higher reluctance than elsewhere. However, what- 
ever the magnetic conditions, the tendency of the armature 
magnetomotive force is to establish magnetic fluxes, and, if 
any field is established in the commutating zone by the armature 
winding, then those armature coils cutting this field will have 
e.m.fs. generated in them proportional to the field which is cut. 
As part of this armature flux is across the armature slots them- 
selves, and part is around the end windings, both of which are 
practically unaffected by the magnetic path in the interpolar 
space above referred to, obviously, then no matter how poor the 
magnetic paths in the interpolar space above the core may be 
made, there will always be e.m.fs. generated on account of that 
part of the armature flux which is not affected by those paths. 
In the coils short circuited by the brushes, these e.m.fs. will 
naturally tend to set up local of short circuit currents during the 
interval of short circuit. 

In good commutation, as the commutator bars connected to 
the two ends of an armature coil which is carrying current in a 
.given direction, pass under the brush, the current in the coil 
itself should die down at practically a uniform rate, to zero value 
at a point corresponding to the middle of the brush, and it should 
then increase at a uniform fate to its normal value in the opposite 
direction by the time that the short circuit is opened as the coil 
passes from under the brush. This may be considered as the 
ideal or straight line reversal or commutation which, however, 
is only approximated in actual practice. This gives uniform 
current distribution over the brush face. 

If no corrective actions are present, then the coil while under 
the brush tends to carry current in the same direction as before 
its terminals were short circuited. In addition, the short circuit 
current in the coil, due to cutting the armature flux, tends to add 
to the normal or work current before reversal occurs. The 
resultant current in the coil is thus not only continued in the 
same direction as before, but tends to have an increased value. 


Thus the conditions at the moment that the coil passes out from 
tinder the short circuiting brush are much worse than if no short 
circuit current were generated. The reversal of the current 
would thus be almost instantaneous instead of being gradual as 
called for by the ideal commutation, and the resultant current 
reversed much greater than the work current alone. However, 
the introduction of resistance into the local circuit will greatly 
assist in the reversal as will be illustrated later. The ideal condi- 
tion however, is obtained by the introduction of an opposing 
e.mi . into the local short circuited path, thus neutralizing the 
tendency of the work current to continue in its former direction. 

As this opposing e.m.f. must be in the reverse direction to 
the short circuit e.m.f. which would set up by cutting the arma- 
ture magnetic field, it follows that where commutation is accomp- 
lished by means of such an e.m.f. it is necessary to provide a 
magnetic field opposite in direction to the armature field for 
setting up the commutating current. This may be obtained in 
various ways, such as shifting the brushes forward (or backward) 
until the commutated coil comes under an external field of the 
right direction and value, which is the usual practise in non- 
commutating pole machines; or a special commutating field of 
the right direction and value may be provided, this being the 
practise in commutating pole and in some types of compensated 
field machines. When the commutating emf. is obtained by 
shifting the commutated coil under the main field, only average 
conditions may be obtained for different loads; whereas, with 
suitable commutating poles or compensating windings, suffi- 
ciently correct commutating e.m.fs. can be obtained over a 
very wide range of operation. 

In practise, it is difficult to obtain magnetic conditions such 
that an ideal neutralizing e.mJ. is generated. However, the use 
of a relatively high resistance in the short circuited path of the 
commutated coil very greatly simplifies the problem. If the 
resistance of the coil itself were the only limit, then a relatively 
low magnetic field cut by the short circuited coil would generate 
-sufficient e.m.f. to circulate an excessively large local current. 
Since such current might be from 10 to 50 times as great as the 
normal work current, depending upon the size of machine,, it 
would necessarily add enormously to the difficulties of commuta- 
tion whether it is in the same direction as the work current or is 
in opposition. To illustrate the effect of resistance, assume, for 
example, a short circuit e.m.f. in the commutated coil of two 


volts, and also assume that a copper brush of negligible resist- 
ance short circuits the coil, so that the resistance of the short 
circuited coil itself -practically limits the current to a value 20 
times as large as the work current. Now replace this copper 
brush with one giving about 20 times as large a resistance (some 
form of graphite or carbon brush) then the total resistance in 
circuit is such that the short circuit current is cut down to a 
value about equal to that of the work current. This at once 
gives a much easier condition of commutation, even without any 
reversing field; while with such field, it is evident that extreme 
accuracy in proportioning is not necessary. Thus a relatively 
high resistance brush or brush contact, rather is of very great 
help in commutation; especially so in large capacity machines 
where the coil resistance is necessarily very low. In very small 
machines, the resistance of the individual armature coils has 
quite an influence in litrilting the short circuit current. 

It is in its high contact resistance that the carbon brush is 
such an important factor in the commutating machine. Usually, 
it is the resistance of the brush that is referred to as an important 
factor in assisting commutation. In reality, it is the resistance 
of the Contact between the brush and commutator face which 
must be considered, and not that of the brush itself, which usually 
is of very much lower resistance, relatively. As this contact re- 
sistance or drop will be referred to very frequently in the fol- 
lowing, and as the brush resistance itself will be considered 
in but a few instances, the terms " brush resistance rt and 
" brush drop " will mean contact resistance and contact drop 
respectively, unless otherwise specified. 

Short Circuit Volts per Commutator Bar. As stated before 
the armature short circuit e,m.f. per coil, or per commutator 
bar, is due to cutting a number of different magnetic fluxes, such 
as those of the end windings-, those of the armature slots, and 
those over the armature core adjacent to the commutating zone. 
Each of these fluxes represent different conditions and distri- 
butions, and therefore the individual e.m.fs. generated by them 
may not be coincident in time phase. Therefore, the resultant 
e,m.f . usually may not be represented by any simple graphical 
or mathematical expression. 

When an external flux or field is superimposed on the armature 
in the commutating zone, it may be considered as setting up an 
additional e.m.f. which may be added to, or- subtracted from, 
the resultant short circuit e.nuf. due to the armature fluxes. 


These component e.m.fs. are not really generated separately 
in the armature coils, for the external flux combines with part 
of the armature flux, so that the armature coil simply generates 
an e.m.f. due to the resultant flux. However, as part of the 
armature short circuit e.m.f is generated by fluxes which do not 
combine with any external flux, as in the end winding, for in- 
stance, it follows that, to a certain extent, separate e m.fs are 
actually generated in the armature winding in different parts of 
the coil. For purposes of analysis, there are advantages in 
considering that all the e m.fs in the short circuited armature 
coil are generated separately by the various fluxes. A better 
quantitative idea of the actions which are taking place is thus 
obtained, and the permissible limitations are more easily seen. 
In the following treatment, these component e.m.fs will be 
considered separately. As that component, due to cutting the 
various armature fluxes, will be referred to very frequently 
hereafter, it- will be called the " apparent " armature short 
circuit e.m f . per coil, or in abbreviated form, " the apparent 
short circuit em.f." In practise, on account of the complexity 
of the separate elements which make up the apparent short 
circuit e m.f., it is very difficult, or in many cases, impossible, 
to entirely neutralize or balance it at all instants by means of an 
e.m.f. generated by an extraneous field or flux of a definite distri- 
bution-. Therefore, it should be borne in mind that, in practise, 
only an approximate or average balance between the two com- 
ponent e.m.fs. is possible. With such average balance there are 
liable to be all sorts of minor pulsations in e.m.f. which tend to 
produce local currents and which must be taken care of by means- 
of the brush resistance. Pulsations or variations in either of the 
component e.m.fs. are due to various minor causes, such as the 
varying magnetic conditions which result from a rotating open 
slot armature, from cross jnagnetizing and other distorting effects 
under the commutating poles, variations in air-gap reluctance 
under the commutating pol^s, pulsations in the main field reluc- 
tance causing development of- secondary e.m.fs. in the short 
circuited coils, etc. Some of these conditions are liable to be 
present in every machine; some which would otherwise tend to 
give favorable conditions as regards commutation, are partic- 
ularly liable to -set up minor pulsations in the short circuit e.m.f. 
Therefore, brushes of high enough resistance to take care of the 
short circuit e m.f . pulsations are a requisite of the present types 
of d-c. machines, and it may be assumed that there is but little 


prospect of so improving the conditions in general that relatively 
high resistance brushes, or their equivalent, may be discarded. 
It is only on very special types of machines that low resistance 
brushes can be used. 

With ideal or perfect commutation, the two component e.m.fs 
in the short circuited coil should balance each other at all times 
However, as stated before, this condition is never actually ob- 
tained, and the brush resistance must do the rest. With ideal 
commutation, the current distribution over the brush contact 
face should be practically uniform, and a series of voltage read- 
ings between the brush tip and commutator face should show 
uniform drops over the whole brush face In most cases in 
practise however, such voltage readings will be only averages 
For example, instead of a contact drop of one volt at a given 
point, the actual voltage may be varying from zero to two volts, 
or possibly from minus one volt to plus three volts These 
pulsating e.m.fs. will result in high frequency local currents, 
which have only a harmful influence on the commutation and 
commatator and brushes These pulsations may be assumed 
to be roughly related in value to the apparent short circuit 
volts generated by the. armature conductor. In other words, 
the higher the apparent short circuit volts per conductor, the 
larger these pulsations are liable to be As the currents set up 
by these pulsations must be limited largely by the brush contact 
resistance, it is obvious that there is a limit to the pulsation* 
in voltage, beyond which the current set up by them may be 
harmful. A very crude practise, and yet possibly, the only 
fairly safe one, has been to set an upper limit to the apparent 
short circuit volts per bar, this limit varying to some extent with 
the conditions of service, such as high peak loads of short dura- 
tion, overloads of considerable period, continuous operation, etc. 
Experience has shown that in commutating pole machines, the 
apparent short circuit voltages per turn may be as high as four 
to four-and-one-half volts, with usually but small evidence of 
local high frequency currents, as indicated by the condition of 
the brush face. If this polishes brightly, and the commutator 
face does not tend to " smut," then apparently the local currents 
are not excessive. However, in individual cases, the above 
limits have been very considerably exceeded in continuoxis opera- 
tion, while, in exceptional cases, even with apparently well 
proportioned comxnutating poles, there has been evidence of 
considerable local current at less than fo'ur volts per bar. 


The contact droi> between brush and commutator with the 
usual brushes is about 1 to 1.25 volts As is well known, this 
drop is not directly proportional to the current, but increases 
only slowly with very considerable increases in current density 
at the brush contact For instance, with 20 amperes per sq in. 
in a given brush the contact drop may be one volt ; at 40 amperes 
per square inch, it may be 1 25 volts, while at 100 amperes per 
square inch, it may be 1 4 volts, and, with materially higher 
currents, it may increase but little further. This peculiar prop- 
erty of the brush contact is, in some ways, very much of a dis- 
advantage For instance, if the local currents are to be limited 
to a comparatively low density, then necessarily the voltages 
generating such currents must be kept comparatively low With 
the above brush contact characteristics, two volts would allow 
a local current of 20 amperes per square inch to flow, (there being 
one volt drop from brush to commutator and one volt back to 
the brush) If, however, the local voltage is three volts instead 
of two, or only 50 per cent higher, then a local current of possibly 
150 to 200 amperes per square inch may flow, and this excessive 
current density may destroy the brush contact, as will be de- 
scribed later 

It may be assumed in general that the lower the apparent short 
circuit voltage per armature conductor, the lower the pulsations 
in this voltage are liable to be. Assuming therefore, as a rough 
approximation a 50 per cent pulsation as liable to occur, then, 
from the standpoint of brush contact drop, the total apparent 
voltage of the commutated coil in continuous service machines 
should not be more than 4 to 4J volts, which accords pretty well 
with practise. For intermittent services, such as railway, 
materially higher voltages are not unusual 

As the main advantage pf the carbon brush is that it determines 
or limits the amount of short circuit current, it might be ques- 
tioned whether such advantage might not be carried much fur- 
ther by using higher short circuit voltages and proportionately 
greater resistance. However, there are reasons why this cannot 
be done The carbon brush is a resistance in the path of the 
local current, but it is also in the path of the work current As 
the brush resistance is increased, the greater is the short circuit 
voltage which can be taken care of with a given limit in short circuit 
current, but at the same time, the loss due to the work current is 
increased. Decreasing the resistance of the brush contact in- 
creases the loss due to the short circuit current, but decreases 


that due to the work current. Thus in each individual case, 
there is some particular brush resistance which gives minimum 
loss However, this may not always be the resistance desired 
for best commutation, from the operating standpoint, but these 
two conditions of resistance appear to lie fairly close together 
Practise is a continual compromise on this question of brush con- 
tact resistance In some machines, a low resistance brush is 
practicable, with consequent low loss due to work current. In 
other cases, which, to the layman, would appear to be exactly 
similar, higher resistance brushes give better average results 
Thus one grade of carbon brush is not the most suitable for dif- 
ferent machines unless they have similar commutating condi- 
tions However, it is impractica*ble to design all machines of 
different speeds, types, or capacities so that they will have equal 
commutating -characteristics In non-commutating pole ma- 
chines where only average commutating fluxes are obtainable, 
the resistance of the brush is usually of more importance than 
m the commutating pole type, for. in the latter, a means is pro- 
vided for controlling the value of the short circuit current How- 
ever, advantage has been taken of this latter fact to such an 
extent in modern commutating pole machines, that the critical 
or best brush resistance has again become a very important 
condition of design and operation 

"Apparent" Short Circuit em f per Brush. The preceding 
considerations lead up to another limitation, namely, the total 
e m f short circuited by the brush. This again may be considered 
as being made up of two components, the apparent short 
circuit e.m f per bar times the average number of bars covered 
by the brush, hereafter called " The apparent short circuit 
e m.f per brush *'; and the e.m.f. per bar generated by the com- 
mutating field, times the average number of bars covered by 
the brush 

As has been shown, ordinary carbon brushes can short circuit 
2 to 2^ volts without excessive local current. Obviously , if the 
resultant e.m.f. generated in all the coils short circuited by the 
brush, that is, the resultant of the short circuit e.m.fs , due to 
both the armature and the commutating field is much larger 
than 2J volts, large local currents will flow. Therefore, in a 
commutating pole machine, for instance, the strength of the 
commutating pole field should always be such that it also 
neutralizes the total short circuit e.m.f- across the brush within 
a limit represented by the brush contact drop, in order to keep 


within the limits of permissible local currents. With very 
low resistance brushes, the proportioning of the commutating 
field for neutralization of the apparent brush e.m.f. would 
have to be much closer than with higher resistance brushes. 
Moreover, not only should this e.m.f. generated by the com- 
mutating flux balance the total short circuit voltage across the 
brush within these prescribed limits, but these limits should 
not be exceeded anywhere under the brush. 

It might be assumed that if there is a pulsation of t\vo volts 
per coil, for instance, then the total pulsation would be equal 
to this value times the average number of coils short circuited. 
However, this in general is not correct, as the e.m.f. pulsations 
for the different coils are not in phase, and their resultant may 
be but little larger than for a single coil 

Based upon the foregoing considerations, the limiting values 
of the apparent brush e.m.f. may be approximated as follows: 
Assume ordinary carbon brushes with 1 to li volts drop with 
permissible current densities that is, with 2 to 2| volts opposr 
ing action as regards local currents. Also, assume, for example, 
an apparent brush short circuit e.m f . of 5 volts, with brush 
resistance sufficient to take care of 2 volts. Then the total 
e.m f. due to the commutating flux need not be closer than 50 
per cent of the theoretically correct value, with permissible 
local currents. This is a comparatively easy condition, for it 
is a relatively poor design of machine in which the commutating 
pole strength cannot be brought within 50 per cent of the right 
value. Assuming next, an apparent brush e.m.f. of 10 volts, 
then the commutating pole must be proportioned within 25 
per cent of the right value. In practise, this also appears to 
be feasible, without undue care and refinement in proportion- 
ing the commutating field. If this machine never carried any 
overload, this 25 per cent approximation would represent a 
relatively easy condition, for experience has shown that pro- 
portioning within 10 per cent is obtainable in some cases, which 
should allow an apparent brush e.m.f. of 25 volts as a limit. 
However, experience al<o shows that this latter is a compara- 
tively sensitive condition, which, while permissible on short 
peak loads, is not satisfactory for normal conditions. Where 
such close adjustment is necessary to keep within the brush 
correcting limits, any rapid changes in load are liable to result 
in sensitive commutating conditions, for the commutating pole 
flux does not always rise and fall exactly in time with the arma- 


ture flux, and ihus momentary unbalanced conditions of pos- 
sibly as high as 10 or 12 volts might occur with an apparent 
brush e.m.f. of 25 volts Also, very slight saturation in the 
commutating pole magnetic circuit may have an unduly large 
influence on unbalancing the e m.f . conditions. In other words, 
the apparent brush short circuit and neutralizing e.m.fs. must 
not be unduly high compared with the permissible corrective 
drop of the brushes Experience shows that an apparent e rn.f 
of 10 volts across the brush in well designed commutating pole 
machines is usually very satisfactory, while, in occasional cases, 
12 to 13 volts allow fair results on large machines, and, in rare 
cases, as high as io to 18 volts has been allowed on small ma- 
chines at normal rating. However, overloads, in some cases, 
limit this permissible apparent brush voltage. As a rule, 30 
volts across the brush on extreme overload is permissible, 
but, usually this is accompanied by some sparking, usually 
not of a very harmful nature if not of too long duration. Under 
such overload conditions, doubtless unbalancing of three volts 
or more may tye permissible, and thus, with 30 volts to be 
neutralized, this means about 90 per cent theoretically correct 
proportioning of the commutating pole flux. Cases have been 
noted where as high as 35 to 40 apparent brush volts have been 
corrected by the commutating pole on heavy overloads with 
practically no sparking. This, however, is an abnormally good 
result, and is not often possible of attainment. Obviously, with 
such high voltages to be corrected, any little discrepancies in 
the balancing action between 'the various* e.m.fs. are liable to 
cause excessive local current flow.- 

Incidentally, the above indicates pretty ' clearly why d-c 
generators are liable to -flash viciously when dead short cir- 
cuited. The ordinary large capacity machine can give 20 to 
30 times rated full load current'on short circuit. If this large 
current flows, then, neglecting saturation, the armature short 
circuit e.m.f. across the brush will be excessive. Assuming, 
for instance, a 10-volt limit for normal rating, then with only 
ten times full load current, the apparent short circuit e.m f 
would be 100 volts. The commutating pole, in the normal con- 
struction, does not have flux margin of 10 times before high 
saturation is reached, and in consequence, it may neutralize 
only 50 to 60 volts of the 100. Therefore a resultant actual 
e.m.f of possibly 40 volts must be taken care of by the brushes 
This means an enormous short circuit current in addition to 


the 10 times work current. Vaporization of the copper and 
brushes occurs and flashing results, as will be described more 
fully in the treatment of flashing limits. 

Brush contact drops of 1 to 1.5 volts have been assumed 
in the preceding, and certain limits in the apparent short cir- 
cuit e.m.f. based on these drops, have been discussed. How- 
ever, the conditions may be modifi ed to a considerable extent 
by effects of temperature upon the brush contact resistance. 
Usually it has been assumed that the well known decrease in 
contact resistance of carbon and graphite brushes with increase 
in temperature, is in some ways related to the negative tem- 
perature coefficient of carbon and graphite The writer has 
been among those who advanced this idea, but later experience, 
based upon tests, has shown that the reduced drop with increase 
in temperature does not necessarily hold any relation to the 
negative temperature coefficient of the carbon brush itself, for 
similar changes in the contact drop have been found with ma- 
terials, other than carbon, which actually had, in themselves, 
positive temperature coefficients. Moreover, in some tests, 
the changes in contact resistance with increase in temperature 
have proved to be much greater in proportion than occurs 
in the carbons themselves. In some cases, the measured drops 
with temperature increases of less than 100 deg. cent decreased 
to one-half or one-third of the drops measured cold 

Obviously, these decreased contact resistances or drops may 
have a very considerable effect on the amount of local current 
which can flo\v and, therefore, in such case the foregoing general 
deductions, should be modified accordingly However, the 
results are so affected by the oxidation of the copper commutator 
face, and other conditions also more or less dependent upon 
temperature, that, as yet. no" definite statement can be made 
regarding the practical effects of increase in temperature except 
the general one that the resistance is usually lowered to a con- 
siderable extent Apparently, oxidation of the copper face 
tends toward higher contact resistance Ofttimes, "sanding 
off " the glaze tends to give poorer commutation The above 
points to one explanation of this 

Assuming any desired limits for the apparent e m fs , such as 
4 to 4| volts per commutator bar, it is possible to approximate 
by calculation the limiting capacities of generators or motors in 
terms of speed, etc Appendix I shows one method of doing 
this In the writer's experience, a number of machines have been 



earned up to about the limits derived in the appendix, and the 
practical results were in fair accord with the calculations. In 
general, it may be said that in large machines, the upper limits 
of capacity in terms of speed, etc. are so high that they do not 
indicate any great handicap on future practise. 

In the foregoing, the limits for the apparent short circuit e m.f . 
per bar and per brush have been based upon the brush contact 
resistance However, it may be suggested that something other 
than the brush contact resistance might be used for limiting the 
local current, and thus the commutating limits might be raised. 
For instance, an armature winding could be completely closed on 
itself, with high resistance leads carried from the winding to the 

commutator bars. Each of such leads 
would be in circuit only where the 
brushes touched the commutator 
bars. Thus there could be very con- 
siderable resistance in each lead with- 
out greatly increasing the total losses; 
and, unlike the brushes, each lead 
would be in circuit only for a very 
small proportion of the time. 

About 10 years ago, the writer de- 
signed a non-commutating pole d-c. 
turbo-generator with such resistance 
leads connected between the winding 
and the commutator. The leads were 
placed in the armature slots below the 
main armature winding. The idea was 
to have enough resistance in circuit 


FIG. 1 

with the short circuited coils that the brushes at no load could be 
thrown well forward into a field flux sufficient to produce good 
commutation at heavy load, even if very low resistance brushes, 
were used. Tests of this machine showed that the non-sparking 
range, with the brushes shifted either forward or back of the 
neutral point was very much greater than in an ordinary machine. 
In this case, it developed that the leads were of too high resistance 
for practical purposes, as the armature ran too hot, the heat-dis- 
sipating conditions in a small d-c. turbo-armature not being any 
too good at best. These tests however, indicate one possibility 
in the way of increasing the present limits of voltage per bar and 
volts across the brush. Moreover, such resistances can have a 
positive temperature coefficient of resistance, instead of the 


negative one of the carbon brushes and contacts. Also, the 
corrective action in limiting local currents would vary directly 
with the current over any range, and not reach a limit, as in car- 
bon brushes 

Considerable experience with resistance leads in d-c operation 
has also been, obtained in large a-c, commutator type railway 
motors, designed for operation on both a-c and d-c. circuits. 
Apparently these leads have a very appreciable balancing action 
as regards division of current between brush arms in parallel. 
With but few brushes per arm, it appears that very high current 
densities in the brushes can be used without undue glowing or 
honeycombing. Presumably the reduction in short circuit 
current, when operating on d-c , also has much to do with this. 
Some special tests were made along this line, and it was found that 
a, very low resistance in the leads, compared with that which was 
best for a-c. operation, was sufficient to exert quite a decided 
balancing between the brush arms 

With properly proportioned resistance leads it should be pos- 
sible to use very low resistance brushes, and relatively high 
current densities. Advantage of this might be taken in various 
ways. There may prove to be serious mechanical objections to 
such arrangements However, if the objections are not too 
serious, the use of resistance leads in this manner may be prac- 
tised at some future time as we approach more extreme flesigns 


One of the limits in commtltating machinery is flashing. This 
tnay be of several kinds. There may be a large arc or flah 
from the front edge of the brush, which may increase in volume 
until it becomes a flash-over to some other part of the machine. 
Again, a flash may originate between two adjacent bars at some 
point between the brush arms, and may not extend further, or 
it may grow into a general flashover. Different kinds of flashes 
:may arise from radically different causes, some of which may be 
-normally present in the machine, while others may be of an 
.accidental nature. 

Whatever the initial cause, the flash itself means vaporized 
conducting material. If the heat developed by or in this vapor 
arc is sufficient to vaporize more conducting material that is, 
generate more 'conduct ing vapor then the arc or flash will grow 
or continue. Thus, true flashing should be associated with 
vaporization, ,and, in sttany cases, in order to get at the initial 


cause of flashing, it is only necessary to find the initial cause of 

Arcs Between Adjacent Cowrmit-ator Bars. This being one of 
the easiest conditions to analyze, it will be treated first, especially 
as certain flashing conditions are dependent upon this. 

A not uncommon condition on commutators in operation is 
a belt of incandescent material around the commutator, usually 
known as "ring fire" This is really incandescent material 
between adjacent bars, such as carbon or graphite, scraped off 
the brush faces usually by the mica between bars. As the mica 
tends to stand slightly above the copper, due to less rapid "wear," 
its natural action is to scrape carbon particles off the brush. 
These particles are conducting and if there is sufficient voltage,* 
and current to bring them up to incandescence, this shows as a 
streak of fire around the commutator In many cases, by its 
different intensities around the commutator, trlis ring fire shows 
plainly the density of the field flux, or e m.f . distribution around 
the machine It is practically zero in the commutating or 
neutral zone, and shows plainly under the main field. In loaded 
machines, this often indicates roughly the flux distortion. In 
machines which act alternately as motors and generators, as in 
reversing mill work, the point of highest incandescence shifts 
forward or backward over the "commutator, depending upon the 
direction of field distortion. 

In undercut commutators (those with mica cut below the cop- 
per surface) this ring fire is also observable at times, due to con- 
ducting particles in the slots between bars. Usually such 
particles consist of carbon or graphite, as already stated, but 
particles of copper may also be present. Also, oil or grease, mixed 
with carbon, will carbonize under incandescence, and will thus 
add to the ring fire. Often when a commutator is rubbed with 
an oiled cloth or wiper, ring fire will show very plainly, and then 
gradually die down. The burning oil exaggerates the action,, 
and also, the oil itself may enable a conducting coating to adhere- 
to the mica edges, thus starting the action, which disappears* 
when the oil film is 'burned away. However, when the oil can- 
penetrate the mica, the incandescence may continue in spots and 
at intervals, the mica being calcined or burned away so that it 
gradually disppears in spots. This is the action usually called. 
"' pitting ", which experience has shown to be almost invariably 
caused by conducting material in the mica, such as carbonized 
oil, carbonized binding material, copper and carbon particles 
been carried in with the oil, etc. 



This ring fire is not always a direct function of the voltage 
between bars, although, under exactly equivalent conditions of 
speed, grade of brushes, etc., it is closely allied with voltage condi- 
tions. In high voltage machines, usually hard high-resistance 
brushes are used, which tend to give off the least carbon in the 
form of particles; while in low voltage machines, soft, low-re- 
sistance brushes, with a good percentage of graphite in them, are 
common, and these naturally tend to coat the mica to a greater 

Under extreme conditions, this ring fire may become so intense 
locally that there is an actual arc formed between two adjacent 
bars, due to vaporization of the copper. This may show in the 
form of minute copper beads at the edge of the bar, or minute 
"pits" or "pockets" may be burned in the copper next to the 
mica. In extreme cases, where the voltage between bars is 
sufficient to maintain an arc, conical shaped 
cavities or holes may be burned in the 
copper. In such cases, the arc is usually 
explosive, resembling somewhat a small , 
"buck-over." An examination of the com- 
mutator will show melted-out places, as in 
Pig. 2. Part of the missing copper has 
been vaporized by the arc, while part may 
have become so softened or fused that it is 
thrown off by centrifugal force. Exper- ' 
ience shows that sometimes these explosive 


FIG. 2 

arcs grow into general flashes, while at other times, they are 
purely local. 

An extended study was made of such arcs to determine the 
conditions which produced them. Also, numerous tests were 
made, the results of which are given below, 
i It was determined first, that these explosive arcs between 
adjacent bars were dependent, in practically all cases, upon a 
fairly high voltage between bars. This was reasonable to expect, 
but it was found that the voltage between bars which would 
produce arcs in one case, would not do so in another. Apparently 
there were other limiting or controlling conditions. It developed 
that the resistance of the armature winding between two adja- 
cent bars has much to do with the arc. Apparently an excessive 
current is necessary to melt a small chunk out of a mass of good 
heat-conducting material like a large copper commutator; * and 
also, a certain amount of time is required to bring it up to the 


melting point Therefore, both time and current are involved, 
as \\ ell as voltage. A series of tests was made to determine some 
of the limiting conditions. 

The commutator of a small machine (about 20 kw., high speed) 
was sprinkled with iron filings, fme dust, etc, during several 
days' operation under various conditions of load, field distortion, 
etc. Such dust, whether conducting or not; apparently would not 
cause arcing between bars. Graphite was finally applied with a 
special "\\iper," and with this, small arcs or flashes could be 
produced at 50 to 60 volts maximum between commutator bars. 
It soon became evident that this was too small a machine from 
which to draw conclusions. Then numerous other much larger gen- 
erators were tested A slow-speed engine type generator of 200-kw. 
capacity at 250 volts, was speeded up to about double speed, 
in order to obtain sufficiently high e.m.f. between commutator 
bars. With a clean commutator nothing was obtained at 40 
volts maximum per bar. The commutator was then wiped with 
a piece of oily waste which had been used to wipe off other com- 
mutators. Arcs then occurred repeatedly between commutator 
bars, although all such arcs were confined to adjacent bars and 
there were no actual flashovers from brush holder to brush holder. 
Moreover, the arcs always appeared to start about midway 
between brush arms or neutral points, and lasted only until the 
next neutral point was reached. Quite large pits or cavities 
were burned in the bars next to the mica, as shown in Fig. 2. 
some of these being possibly J inch in width, and 1/16 inch deep 
or more at the center This indicated excessively large currents. 
These arcs would develop at about 32 to 34 volts between bars, 
and they were very vicious (explosive) above 35 volts 

Still larger machines were tested with various speeds, voltage 
between bars, etc It was found that, as a rule, the larger the 
machine or rather, the lower the resistance of the armature 
winding per bar the lower would be the voltage at which serious 
arcing would develop In these tests, it was found that graphite 
mixed with grease gave the most sensitive arcing conditions. 

In these various tests, no arcing between bars was developed 
in any case at less than 28 volts maximum, while 30 volts was 
approximately the limit on many machines. However, the 
results varied with the speed Apparently it took a certain time 
to raise the incandescent material to the arcing point and to build 
up a big arc. Therefore, the duration of the possible arcing 
period appeared to be involved. If this were so, then a higher 


voltage limit for a shorter time should be possible with the same 
arcing tendency. Also, if this were the case, then with 30 volts 
maximum, for instance, between commutator bars with an un~ 
distorted field flux, the arcing should be the same as with a some- 
what higher voltage with a highly distorted narrow peaked field. 
In other words, the limiting voltage between bars on a loaded 
machine might be somewhat higher than on an unloaded machine. 
This was actually found to be the case, the, difference being from 
10 per cent to 15 per cent in several instances. This, however, 
depended upon various limiting conditions such as the actual 
period within which the arc could build up to a destructive 
point, etc. 

One very interesting case developed which apparently illus- 
trated very beautifully the effects of lengthening or shorten- 
ing the period during which the arc could occur. A high-speed, 
600-volt generator of a motor-generator set was speeded up 
about 60 per cent above normal. Even at normal speed this 
was a rather high-frequency machine, so that the period of 
time for a commutator bar to pass from neutral point to neutral 
point was very short. At the highest speed the graphite-grease 
was used liberally on the commutator, but without causing arc- 
ing, even when the voltage was raised considerably higher 
than usually required for producing arcs between bars in other 
machines of similar size. Neither was there much ring-fire 
at the highest speed with normal voltage. Finally, after an 
application of graphite, without forming arcs or unusual ring- 
fire, the speed was reduced gradually with normal voltage 
maintained. The ring-fire increased with decrease in speed, until 
at about normal speed, it was so excessive that the on-lookers 
expected an explosion of some sort. However, the voltage 
was now below the normal arcing point and nothing happened. 
At still lower speed, but with reduced voltage on account of 
saturation, the ring-fire gradually decreased. Apparently at 
the very high speeds, the tirrie was too short for the ring-fire to 
reach its maximum; while with reduction in speed, even with 
somewhat reduced voltage, the, ring-fire increased to a maxi- 
mum and then decreased. This test was continued sufficiently 
to be sure that it was not an accidental case. Only a certain, 
combination of speed, frequency, voltage, etc. could develop 
this peculiar condition, and it was purely by accident that 
this combination was obtained, for the result was not foreseen 
in selecting the particular machine used. 


A summation of these and other tests led to th conclusion 
that there were pretty definite limits to the maximum volts 
per bar, beyond which it was not safe to go. These limits 
however, involved such a number of conditions that no fixed 
rule could be established, and apparently, the designer has 
to use his judgment and experience to a certain extent, if he 
works very close to the limits. The grades and materials of 
the brushes, the thickness of the mica, flux distortion from over- 
loads, etc. must be taken into account. For instance, the above 
tests were made on machines with 1/32-inch mica between bars. 
This thickness is fixed, to a great extent, in non-undercut 
commutators, by conditions of mica wear, as will be referred 
to later. But with undercut commutators, thicker mica can 
be used, and, while the gain in permissible safe voltage between 
bars is not at all in proportion to the mica thickness, yet it is 
enough to deserve consideration. 

The general conclusions were that with 1/32-inch mica, 
large current machines would very rarely flash with 28 volts 
maximum between bars; while with moderate capacities, 30 
volts is about the lower limit; and with still smaller machines, 
100 kw. for example, this might be as high as 33 to 35 volts, 
the limit rising to 50 or 60 volts with very 1 small machines. 

Of course, the brush conditions have something to do with 
the above limits, and many exceptions to these figures will be 
found in actual practise. Many machines are in daily service 
which are subject to more or less ring-fire, but which have never 
developed trouble of any sort, and doubtless never will. Ap- 
parently, ring-fire in itself is not harpiful, as a rule. It is only 
where it starts some other trouble tiat it may be considered as 
actually objectionable. 

The above limiting figures are interesting when compared 
with the voltages necessary to establish arcs in general. An 
alternating arc through air will not usually maintain itself at 
less than some limiting voltage such as 20 to 25 volts, corres- 
ponding to peak values of 28 to 35 volts. Moreover, an arc 
formed between the edges of two insulated bodies, such as ad- 
jacent 'commutator bars, will naturally tend to rupture itself 
due to the shape of its path. Furthermore, the resistance and 
"reactance of the short circuited path, while comparatively low 
in large machines, will tend to. limit the voltage which main- 
tains the arc. In snr>a-11 machines with relatively high internal 
drops in tlie sfoort circuited coils> the current will not reach a 


commutator vaporizing value unless the initial voltage between 
bars is comparatively high, and usually the explosive actions 
are relatively small, and, in many cases, no senous arcs will 
develop at all. Obviously, the less the local current can in- 
crease in the case of short circuits between adjacent bars, the 
higher the voltage between bars can be, without danger. In 
machines having inherent constant current characteristics, very 
high voltages between adjacent commutator bars are possible 
without serious flashing or burning. In consequence, from the 
flashing standpoint, constant current machines can be built for 
enormously high terminal voltages, compared with constant 
potential machines. This is a point which is very commonly 
overlooked in discussing high-voltage d-c. machines. 

Cbming back to the subject of arcs between commutator 
bars, these are more common than is usually supposed, for, 
in ma,ny cases, the operating conditions are such that* these 
arcs, if very small, or limited, will show no visible evidence. 
Only very minute 'particles of copper may be vaporized. How- 
ever, these minute arcs may sometimes lead directly to more- 
serious flashing. If, for instance, they occur in proximity to- 
some live part of the machine, such as an over-hanging brush 
holder which is at a considerable difference of potential from 
the arcing part of the commutator, the conducting vapor may 
bridge across and start a big arc or flash. "In one instance/ 
which the writer has in mind, a very serious case of trouble 
occurred in this way. This was a very large capacity 250- 
volt, low-speed, generator, in which the maximum volts per 
bar were not unduly high. When taking the saturation curve 
in the shop test, this machine " bucked" viciously several 
times, apparently without reason. An investigation of the 
burning indicated a possible source of trouble. The brush 
holder arms or supports to which the individual holders were 
attached, were located over the commutator about midway 
between neutral points, and, about one inch from the com- 
mutator face. This was not the normal position of the brush 
arms, as a temporary set of holders was being used for this test. 
It was noted that just before the flashovers occurred, con- 
siderable ring-fire developed. The conclusion was drawn from 
all the evidence that could be obtained, that a small arc had 
formed between bars that had reached to the brush arms, thus 
short circuiting a high enough voltage to draw a real flash. 
This happened not once but several times. The proper holders. 


were then applied, which put the brush arms in a much less 
exposed position, and not a single flashover occurred in all the 
subsequent tests and operation. In another case, a large syn- 
chronous converter carrying full load on shop test flashed over 
a number of times, apparently without sufficient cause. The 
commutation was perfect, as evidenced by the fact that there 
was no perceptible sparking. The maximum voltage between 
bars was comparatively low. At "first the flashovers were 
blamed on drops of water from the roof of the building, but 
this theory was soon disproved. An examination of the brush - 
holders showed that certain live parts, fairly close to the com- 
mutator, were at a considerable difference of potential from the 
nearest part of the commutator. There was but little ring- 
fire on the commutator, and therefore, minute arcs at first were 
not blamed for the trouble. A modified brush holder was tried 
however, with a view to decreasing the high difference of poten- 
tial between the live parts. All flashing then disappeared and 
no trouble of this sort was ever encountered in a large number 
of duplicate machines brought through afterwards. Both the 
above cases should be considered as abnormal, and they have 
been selected simply as examples of what small arcs between 
bars may do. These two cases do not in themselves constitute 
a proof of this action, but they serve to verify other evidences 
which have been obtained. * 

In view of the fact that small arcs of a non-explosive sort 
may form at voltages considerably lower than the limits given 
in the preceding part of this paper,- it should be considered 
whether such small arcs can cause any trouble if no other live 
parts of the machine are in close proximity. One case should^ 
be considered, namely, thai of other commutator bars adjacent 
to the arc. When conducting vapor is formed by the first 
minute arc, this vapor in spreading out may bridge across a 
number of commutator bars having a much liigher total differ- 
ence of potential across them than that which caused the initial 
arc. Assume, for instance, a very crowded design of high* 
voltage corrfmutator. In some cases, in order to use high rota- 
tive speeds, without unduly high commutator peripheral speed* 
the commutator bars are sometimes made very thin and the 
volts per bar very high, possibly up almost to the limit. As- 
suming a thickness of bar and mica of 0.2 inch (or 5 bars per 
inch) and a maximum volts per bar of 25, then there is an e.m f . 
of 125 volts per inch circumference of the commutator- In such 


case, a small arc between two bars may result in bridging across 
a comparatively high voltage through the resulting copper 
vapor Therefore, when considering the possible harmful effects 
of minute arcs, the volts per inch circumference of the commuta- 
tor should be taken -into consideration. The writer observed 
one high-voltage commutator which flashed viciously at times, 
apparently without " provocation " The only explanation he 
could find was that the vapor from little arcs resulting from 
ringfire was sufficient to spread all over the commutator, the 
bars being very thin and the voltage per bar very high. How-, 
ever, difficullies from this cause have not yet become serious, 
probably because no one has yet carried such constructions to 
the extreme, in practical work 

High voltage between commutator bars may result in flash- 
ing due to other than normal operating conditions. Excessive 
overloads may give such high voltages per armature coil or per 
commutator bar, immediately under the brush, that the terrific 
current rush will develop conducting vapors under the brush, 
which appear immediately in front of the brushes, as such vapors 
naturally are carried forward by rotation of the commutator. 
This short circuit condition under the brush has already been 
referred to when treating of commutation limits It was shown 
then that an inherent short circuit voltage of 4 to 4| volts' is 
permissible in good practise Immediately under the com- 
mutating pole this voltage is practically neutralized by the 
commutating pole field, but immediately ahead or behind the 
pole it is not neutralized usually, except to the extent of the 
commutating pole flux fringe. Thus, the resultant voltage 
between two bars a little distance ahead of the brush, is liable 
to be considerably higher than under the brush Assuming, 
for instance, 3 volts per bar, due to cutting the resultant field 
just ahead of the 1 brush, then with 10 times full load current, 
for example, there would be 35 volts between bars, and this is 
liable to be accompanied by highly conducting vapor formed 
"by the excessive current at the brush contact, this vapor being; 
carried forward by rotation of the commutator. Here are the 
conditions for a flash, which may or may not bridge across to 
.some other live part If the current rush is not too great, this 
flash will usually appear only as a momentary blaze just in 
front of the brush. In many cases, if this blaze or heavy arc 
were not allowed to come in contact with, or bridge between, 
any parts having high difference of potential, it would not be 


particularly harmful. In case of " dead short circuiting" of 
large moderately high-voltage machines where the current can 
rise to 25 or 30 times normal, it is astonishing how large such 
arcs or flashes may become, and to what distances they will 
reach. The arc will sometimes go in unanticipated directions. 
The conducting vapor may be deflected by magnetic action 
and by air drafts Shields or partitions will sometimes pro- 
duce unexpected results, not necessarily beneficial. Unless 
such shields actually touch the commutator f ac e so that con- 
ducting vapor cannot pass underneath them, the vapor that 
does pass underneath may produce just as harmful results as 
if the shields were not used. Trying to suppress such arcs by 
covers or shields is very much the case of damming a river at 
the wrong end in order to prevent high water 

From the preceding considerations it would appear that a 
compensated direct-current machine should have some ad- 
vantages over the straight commutating-pole type in case of a 
severe short circuit. With the lesser saturation in the com- 
mutating pole circuit due to the lower leakage, the apparent 
armature short circuit e.m.f. will usually be better neutralized 
under extreme load conditions, and thus there will be lower 
local currents in the brush contacts In addition, the armature 
flux will be practically as well neutralized behind and ahead 
of the brush, as it is under the brush, so that, with ten times 
current as in the former example, there may be only a low 
e m.f. per bar ahead of the brush, instead of the 35 volts for 
the former case Obviously, the initial flashing cause, and the 
tendency to continue it ahead of the brush, will be materially 
reduced. The compensating winding is -therefore particularly 
advantageous in very high voltage generators, in which the 
bars are usually very thin and the maximum volts per bar are 

There is a prevailing opinion that when a circuit breaker 
opens on a very heavy overload or a short circuit, flashing is- 
liable to follow from such interruption of the current In some 
cases, this may be true However, when a breaker opens on. 
a short circuit, it is difficult for the observer to say whether both 
the opening of the breaker and the flash are due to the excessive 
momentary current, or one is consequent to the other. The 
short circuit, if severe, will most certainly cause more or less 
of a flash at the brush contacts by the time the breaker is opened, 
and if this flash is carried around the commutator, or bridges 


across two points of widely different potentials, then it is liable 
to continue after the breaker opens, and thus ^ives the im- 
pression that the flashing followed the interruption of the cir- 
cuit. In railway and in mine work in particular, a great many 
flashes which are credited to overloads are primarily caused by 
partial short circuits on the system, or " arcing shorts," which 
are extinguished as soon as the main breakers are opened, so 
that but little or no evidence of any short circuit remains. 
Such a partial short circuit however, may be sufficient to open 
the generator circuit and to cause a flash at the same time. 
Not infrequently, such flashes are simply credited to opening 
of the breakers 

There are other conditions, however, where a flash is liable 
to result directly from opening the breaker on heavy overload. 
If as referred to before, the apparent short circuit e m.f. per 
brush on heavy overload is from 25 to 35 volts, then if the 
armature magnetomotive force could be interrupted suddenly, 
with a correspondingly rapid reduction in the armature flux, 
while the commutating field flux does not die down at an equally 
rapid rate, then momentarily, there will be an actual short 
circuit voltage of a considerable amount under the brushes 
which may be sufficient to circulate large enough local currents 
to start flashing.. With commutating pole machines, this con- 
dition may result from the use of solid poles and solid field 
yokes Laminated commutating poles are sometimes very much 
of ah improvement. However, the yokes of practically all 
direct current machines are of solid material, and thus tend to 
give sluggishness in flux changes. The above explains why non- 
inductive shunts, or any closed circuits whatever, are usually 
objectionable on commutating poles or their windings. 

Iri non-commutating pole machines, where the brushes are 
liable to be shifted under the main field magnetic fringe in 
order to commutate heavy loads, flashing sometimes results, 
when such heavy overload is interrupted. 

Also, if the rupture of the current is very sudden, there \ull 
be an inductive " kick " from the collapse of the armature 
magnetic field. This rise in voltage sometimes is sufficient to 
start a flash, especially in those cases where flashing limits are 
already almost reached. 

In synchronous converters, the conditions are materially 
different from d-c. generators as regards flashing when the 
load is suddenly broken. In such machines, the flash is liable 


to follow the opening of the breaker, if simply a heavy over- 
load is interrupted. This is possibly more pronounced in the 
commutating pole machine than in the non-commutating pole 
type In a commutating pole converter, the commutating 
pole magnetomotive force is considerably larger than the re- 
sultant armature magnetomotive force, under normal opera- 
ting conditions, but is much smaller than the armature magneto- 
motive force considered as a straight d-c. or a-c. machine. 
Normally the commutating pole establishes a commutating 
field or flux in the proper direction in the armature. However, 
if, for any reason, the converter becomes a motor or a generator, 
even momentarily, the increased magnetomotive force of the 
armature may greatly exceed that of the commutating pole, 
so that the commutating pole flux will be greatly increased, or 
it may be greatly reduced, or even reversed, depending upon 
which armature magnetomotive force predominates. 

The above is what happens when a synchronous converter 
hunts, and under the accompanying condition of variable 
armature magnetomotive force, the commutating pole con- 
verter, with iron directly over the commutating zone, is liable 
to show greater variations in the flux in the commutating zone 
than is the case in the non-commutating pole converter. Ex- 
perience has shown that uhen a synchronous converter carry- 
ing a heavy overload has its direct-current circuit suddenly 
interrupted, it is liable to hunt considerably for a very short 
. period, depending upon the hunting constants of the individual 
machine and circuit. Apparently, all converters hunt to some 
extent \\ith such change in load. This hunting means wide 
variations in the commutating pole flux with corresponding 
sparking tendencies. For a " swing " or two, this sparking 
may be so bad as to develop into a flash. Thus the flash follows 
the interruption of the circuit. 

Curiously, the most effective remedy for this condition is 
one ^hich has proved most objectionable in d-c. machines, 
namely, a low-resistance closed electric circuit surrounding the 
commutating pole. The primary object of this remedy is" not 
to form a closed circuit around the commutating field, but to 
obtain a more effective damper in order to minimize hunting. 
In a paper presented before the Institute several years ago,* 
the writer, showed that the ideal type of cage winding for damp* 

*Commtitating Poles in Synchronous Converters. PSage 171. 


ing synchronous converters, namely, that in which all circuits- 
are tied together by common end rings, was not suitable for 
commutating pole converters due to the fact that the various 
sections of this cage winding form low-resistance closed circuits 
around the commutating poles. This was in accord \\ith all 
evidence available to that time, and no one took exception to 
it However, later experience has shown that this was incor- 
rect, for, in later practise, it was found that the use of a complete 
cage damper of low resistance which decreases the hunting 
tendency, also greatly decreases the flashing tendency, so that 
today most converters of the commutating pole type are being 
made with complete cage dampers. Apparently, the flashing 
tendencies in converters due to hunting are much worse than 
those due to flux sluggishness. Therefore, a sacrifice can be 
made in one* for the benefit of the other. 

In the case of a dead short circuit on the d-c side of a syn- 
chronous converter, there is liable to be flashing, just as in the 
d-c machine, and the flash and the breaker opening are liable 
to occur so closely together that an observer cannot say which 
is first. 

In d-c. railway motors, flashing at the commutator is not 
an uncommon occurrence One rather common cause of flash- 
ing, especially at high speed, is due to jolting the brushes away 
from the commutator, due to rough track, etc This is espe- 
cially the case with light spring tension on the brushes. The 
carbon breaks contact with the copper, forming an arc which 
is carried around. Another prolific source of flashing is due 
to opening and closing the motor circuit in passing over a gap 
or dead section in a trolley circuit Here the motor current 
is entirely interrupted, and, after a short interval, it comes oa 
again, without any resistance in circuit except that of the motor 
itself. If the current rush at the first moment of closing is 
not too large, and if the armature and field magnetic fluxes 
build up at the same rate, then there is usually but small danger 
of a flash, except under very abnormal conditions. The rapidly 
changing field flux however generates heavy currents under 
the brushes, thus tending toward flashing. The reactance of 
the motor, especially of the field windings, limits the first cur- 
rent rush to a great extent. According to this, closed second- 
ary circuits of low resistance around either the main poles or 
the commutating poles, should be objectionable, and experience 
bears this out 


In railway armatures, as a rule, fewer commutator bars per 
pole are used on the average than in stationary machines of 
corresponding capacity, except possibly, in large capacity 
motors. This is due largely to certain design limitations in 
-such apparatus, but this has doubtless been responsible for a 
certain amount of flashing in such apparatus 

Average em.f. and " Field Form" A rather common prac- 
tise has -been to specify the average volts per bar in a given 
machine. This, in itself, does not mean anything, except in a very 
general way; for the is really fixed by the maximum volts 
per bar, as already shown, and there is no fixed relation between 
the average and the maximum volts per bar. The ratio be- 
tween these two voltages is dependent upon the field flux dis- 
tribution, that is, the "field form." In practise, this ratio 
varies over a ^ide range, depending upon the preferences of 
the designer, upon limitations of pole space available, etc. 
Also, TTvith load, it depends upon the amount of flux distortion 
of the field, which, in turn varies greatly in practise. In well 
proportioned modern machines, where space and other limita- 
tions permit, the average e.m.f. per bar is about 70 per cent 
of the maximum at no load, and about 55 per cent to 60 per 
cent with heavy load. This means that about 15 volts per 
bar, average, is the maximum permissible, in large machines 
with considerable field distortion, if a maximum of 28 volts 
per bar is not to be exceeded. On this basis, a 600-volt machine 
should therefore have not less than 40 commutator bars per 
pole. However, this is with considerable field distortion. If 
this distortion is reduced or eliminated, the average volts can 
be considerably higher, as in machines with high saturation in 
the pole faces, pole horns and armature teeth, or with com- 
pensated fields. Synchronous converters are practically self- 
compensated and can therefore have higher limits than the 
above, if the normal rated e.m.f. is never to be exceeded. How- 
ever, in 600-volt converter work, in particular, wide variations 
sometimes momentarily occur, up to 700 to 750 volts, and such 
machines should have some margin for such voltage swings. 
The ordinary 600-volt d-c. generator also attains materially 
Hgher voltages at times, which would be taken into account 
in the limiting voltage per commutator bar and the total number 
of commutator bars per pole. 

Obviously, the " fatter * the field form, the nearer the aver- 
age voltage caa apf>nwb the maximum. With an 80 per cent 


field form, instead of 70 per cent, for instance, the number of 
bars per pole can be reduced directly as the polar percentage 
is increased; and 35 bars per pole with 80 per cent would be as 
good as 40 bars with 70 per cent assuming the same percentage 
of field distortion in both cases. An increase in the polar arc 
will tend toward increased distortion, but the reduced number 
of turns per pole should practically balance this, so that, other 
things being unchanged, the flux distortion should have prac- 
tically the same percentage as before. 

In large machines of very high speeds, large polar percentages, 
that is, large " field form constants," are very advantageous, 
but are not always obtainable, due to the space required for the 
commutating pole winding. In compensated field machines, 
with their smaller commutating pole windings, the conditions 
are probably best for high field form constants, and high aver- 
age volts per bar; and thus this type often lends itself very- 
well to those classes of ma- 
chines where the minimum 
possible number of commu- 
tator bars is necessary. This 
is the case with 'very high 
speeds, and also for very high 
voltage machines. 

Usually it is considered that 
the commutating conditions JG> 

of a machine are practically the same with the same current, 
whether it be operated as a generator or motor. However, 
when it comes to flashing conditions, there is one very consider- 
able difference between the two operations. In the d-c. gen- 
erator, the field flux distortion by the armature is such as to 
crowd the highest field density, and thus the highest volts 
per bar, away from the forward edge of the brushes. In the 
motor, the opposite is the case, and therefore there is a steeply 
rising field, and a corresponding e.m.f. distribution in front of 
the brushes. As the flash is carried in the direction of rotation 
it may be seen that, in this particular, the generator and motor 
are different. 



In the preceding, certain limitations of commutation and 
flashing have been treated. There are, in addition, a number 


of other conditions which are related closely to commutation, 
and which have already been touched upon to a limited extent. 
One of these is the permissible current density in the brushes 
or brush contacts. 

As brought out before, there are two currents to be con- 
sidered, namely, the work current which flows to or from the 
outside circuit, and the local or short circuit current which is 
purely local to the short circuited coils and the brush. The 
true current density is that due to the actual resultant current 
in the brush tip or face, which is very seldom uniform over the 
whole brush tip. The " apparent " current density is that due 
to the work current alone assumed to be uniform over the 
brush tip. The current density very commonly has been as- 
sumed as the total work current, in and out, divided by the 
total brush section, and, moreover, this has been considered 
as the true current density, the local or short circuit currents 
being neglected altogether. This method of considering the 
matter has been very misleading, resulting in many cases, in 
a wrong or unsuitable size of brush being used just to meet 
some specified current density. In many of the old, non-com- 
mutating pole machines, the local currents were predominant 
under certain conditions of load, for the brushes, as a rule, had 
to be set at the best average position, so that at some average 
load, the commutating conditions would be best. At higher 
and lower loads, the short circuit currents were usually com- 
paratively large. The wider the brush contact circumferen- 
tially, the greater would be the short circuit currents and the 
higher the actual current density at one edge of the brush, 
while the apparent density would be reduced. Thus, in at- 
tempting to meet a low specified current density, the true den- 
sity would be gregtly increased. The fallacy of this procedure 
was shown in many cases in which the brush contact was very 
greatly reduced by grinding off one edge of the brush. Very 
often, a reduction in circumferential width of contact to one- 
half resulted in less burning of the brush face. The apparent^ 
density was doubled but the actual maximum density was . 
actually reduced. Many of these instances showed very 
conclusively that much higher true current densities were prac- 
ticable, provided the true and apparent densities could be 
brought more nearly together. This is what has been accom- 
plished to a considerable extent in the modern well designed 
commutating pole machine* In such machines, the current dis- 


tribution at the brush face is nearly uniform under all condi- 
tions of load It is not really uniform, even in the best machines ; 
but the variations from uniformity, \\hile possibly as much as. 
50 per cent in good machines, is yet very small compared \vith. 
the variation in some of the old non-commutating pole machines. 
In consequence, it has been possible to increase the apparent 
current densities in the brushes in modern commutating pole 
machines very considerably above former practise, while still 
retaining comparatively wide brush faces. In fact, the width 
of the brush contact circumferentially is not particularly limited 
if the commutating field flux can be suitably proportioned; 
that is, where a suitable width and shape of commutating field 
can be obtained. In many of the old time machines, an ap- 
parent density of 40 amperes per square inch under normal 
loads was considered as amply high, while at the present time T 
with well proportioned commutating poles, 50 per cent higher 
apparent densities are not uncommon However, experience 
shows that the same brushes, with perfectly uniform distribu- 
tion of current at the brush face, can carry still higher currents. 
Therefore, in modern commutating pole machines, the actual 
upper limit of brush capacity is not yet attained. But there 
are reasons \\hy this upper lirr it is not practicable. One reason 
is that already given, that uniform current distribution over 
the brush face is seldom found. This rreans that a certain 
margin must be allowed for variations. A second reason lies- 
in the unequal division of current between the various brushes- 
and brush arms. This may be due initially to a number of 
different causes. However, when a difference in current once 
occurs, it tends to accentuate itself, due to the negative co- 
efficient of resistance of the carbon brushes and brush contacts. 
If one of the brushes, for instance, takes more than its share of 
current for a period long enough to heat the brush more than, 
the others, then "its resistance is lowered and it tends to take 
still more current. If there were no other resistance in the 
current path, it is presumable that the parallel operation off 
xarbon brushes would be more or less unsatisfactory. In the 
practical case, however, instead of the operation being im- 
practicable^ it is merely somewhat unstable. Unequal division 
of current between the brushes on the same brush arms, is to 
some extent, dependent upon the total current per arm. Where 
there are many brushes in parallel and the total current to 
be carried is very large, it is obvious that one brush may take 


an excessively large current without materially decreasing the 
current carried by the other brushes. As a rule, the larger 
the current per arm, the more difficult is the problem of prop- 
erly balancing or distributing the current among all the brushes. 
Schemes have been proposed, and patented, for forcing equal 
division, but, as a rule, they have not proved very practicable, 
although some comparatively simple expedients have been 
tried out ^ith a certain degree of success. 

In the same way, the division of current among brush arms 
of the same polarity is not always satisfactory. 50 per cent 
variation of current between different arms is not very unusual, 
and the writer has seen a number of instances where the varia- 
tion has been 100 per cent, or even much more. Obviously, 
with such variation, it is not practicable to work the brushes up 
to the maximum density possible, for some margin must be 
allowed for such unbalancing. 

Experience has shown that when current passes through 
a moving contact, as from a brush to the commutator copper, 
or vice versa, a certain action take place which resembles elec- 
trolytic action to some extent, although it is not really electro- 
lytic. It might also, be said to resemble some of the actions 
which takes place in an arc. Minute particles appear to be 
aten or burned away from one contact surface, and these are 
sometimes deposited mechanically upon the opposing surface. 
The particles appear to be carried in the direction of current 
flow, so that if the current is from the carbon brush to the 
copper, the commutator face will tend to darken somewhat, 
evidently from depositation of carbon. If the current is from 
the copper to the carbon, the brush face will sometimes tend 
to take a coating of copper, while the commutator face will 
take a clean, and sometimes raw, copper appearance. As the 
current is in both directions on the ordinary commutator, this 
action is more or less averaged, and therefore is not usually 
noticed. With one polarity or direction of current, the com- 
mutator face eats away, while with the other direction, the 
brush face is eaten away and may lose its gloss. 

The above action of the current gives rise to a number of 
limiting conditions in direct-current practise. Experience shows 
that this " eating away " action occurs with all kinds of brushes, 
and with various materials in the commutator. It appears to 
be dependent , to a considerable extent, upon the losses at the 
contact surface. In other words, it is dependent upon both the 


current and the contact drop. With reduction in contact drop, 
this burning action apparently is decreased, but in commutating 
machinery, this reduction cannot be carried very far, in most 
cases, on account of increase in short circuit current, which 
nullifies the gain in contact drop. In fact, in each individual 
machine, there is some critical resistance which gives least loss 
and least burning at the contact surfaces. 

Practise has shown that this burning action is very slow at 
moderate current densities in carbon and graphite brushes 
so slow as usually' to Be' unnoted. However, if the actual 
current density in the brush face is carried too high, the burn- 
ing of the brushes may become very pronounced. With the 
actual work current per brush usual in present practise, the 
burning of the brush face may usually be credited to local cur- 
rents in the brushes This is one pretty good indication of 
the presence of excessive local currents. It also indicates the 
location and direction of such currents, but is not a very exact 
quantitative measure of them. It is not unconr.iron, in exam- 
ining the brushes of a generator or motor, to find a dull black 
area under one edge of the brush, which obviously has been 
burned, while the remainder of the brush face is brightly polished. 
In severe cases, practically as good results t\ill be obtained 
if the burned area is entirely cut away by beveling the edge 
of the brush. 

This eating away of either the brush face or the commutator, 
and the deposit upon the opposing face, leads to certain very 
harmful conditions in direct-current machinery. As stated 
before, if the true current density is kept sufficiently low in 
the contact face, the burning is negligibly small m most cases. 
However, where the current passes from the commutator to 
the brush, it is the commutator copper which eats away, while 
the mica between commutator bars doe? not eat away, but must 
be worn away at the same rate that the copper is burnt, if good 
contact is to be maintained. Let the burning of the copper 
gain ever so little on the wear of the mica, then trouble begins. 
The brush begins to " ride " on the mica edges and docs not 
make true contact with the copper. This increases the burn- 
ing action very rapidly, so that eventually the mica stands 
well above the copper face. This is the trouble usually known 
as " high mica/' It is frequently credited to unequal rates 
of wear of copper and mica. This idea of unequal wear has 
been partly fostered by the fact that with relatively thick 


mica, the action is greatly increased, or, with very thin com- 
mutator bars, with the usual thickness of the mica, the high mica 
trouble becomes more serious. In both these latter cases, it 
is the higher percentage of mica, that is, the relatively poorer 
wearing characteristics of the mica itself, which is at fault. 
But the commutator copper does not wear away. In fact, it 
is not physically possible for it to wear below the mica. It is 
" eaten away " or burned, as described above. In some special 
cases, where this burning is unusually severe, the mica apparently 
wears down about as fast as the copper, so that the commutator 
remains fairly clean and has no particularly burnt appearance, 
but grooves or ridges, showing undue wear. But this rapid 
apparent wear is a pretty good indication that excessive burn- 
ing action is present at times, usually due to excessive local 
currents. In some cases, this burning action may be present 
only during heavy or peak loads which may be so interspersed 
with periods of light running that the true wear of the mica 
catches up with the burning of the copper. In such cases, the 
commutator may have a beautiful glossy appearance normally, 
but may wear in grooves and ridges On account of this burn- 
ing action, practise has changed somewhat in regard to stagger- 
ing of brushes on commutators to prevent ridging between the 
brushes. Formerly, it was common practise to displace all 
the positive brushes one direction axially, and the negatives 
in the other direction, in order to have the brushes overlap. 
This, however, did not entirely prevent ridging, for the burning 
of the copper occurred only under one polarity. It is now con- 
sidered better practise to stagger the arms in pairs. 

With commutating pole machines, the true current densities 
in the brushes are carried up to. as high a point as the non- 
burning requirements wall permit. Reduction in local currents 
has been accompanied by increase in the work current density. 
Therefore, conditions for burning and high mica are still exist- 
ent, as in non-commutating pole machines. In recent years, 
a new practise, or rather an extension of an old practise, has 
been very generally adopted, namely, undercutting the mica 
between bars. In early times, such undercutting was practised 
to a certain extent, usually however, to overcome mica troubles 
principally. In the newer practise, such undercutting is pri- 
marily for other reasons, although the mica problem is partly 
concerned in it. During the last few years, extended experi 
ence ha$ shown that graphite brushes, or carbon brushes with 


considerable graphite m them, are extremely good for collect- 
ing current, but on the other hand, are very poor when it comes 
to wearing down the mica due to their softness or lack of ab- 
rasive qualities Due to the graphite constituent, such brushes 
are largely self -lubricating, and therefore, "ride" more smoothly 
on the commutator than the ordinary carbon brush. They are 
therefore much quieter, and this is a very important point \vith 
the present high speeds which are becoming very much the 
practise. However, by undercutting the mica, all difficulty 
from lack of abrasive qualities in the brush is overcome, and 
thus the good qualities of such brushes could be utilized. The 
advantage of self-lubricating brushes should be apparent to 
anyone who has had difficulties from chattering and vibration 
of brushes, due to lack of lubrication. Such chattering may 
put a commutator " to the bad " in a short time, and the con- 
ditions become cumulatively worse* Chattering means bad 
contact between the brush and commutator, which in turn, 
means sparking and burning, which means increased chatter- 
ing or vibration. 

The above refers to burning of the commutator face. But 
such burning also may have a bad effect on the brushes. When 
the commutator copper burns away to any extent, it may de- 
posit on the brush face following the direction of the current. 
This coating on the brush face sometimes leads to serious 
trouble, by lowenng the resistance of the contact surf ace. This 
not only allows larger short circuit current and greater heating 
of the brush, but it makes the resistance of that particular 
path lower than that of other parallel brush paths. In con- 
sequence, the coated brush takes an undue share of the total 
current, as well as an unduly large local current. The result- 
ant heating may be such that the brush actually becomes red 
hot or glows. This heating further reduces the resistance, 
and tends to maintain the high temperatures. This glowing 
or overheating very frequently causes disintegration of the bind- 
ing or other material in the brush, so that it gradually honey- 
combs at or near its tip. This action may keep up until the 
brush makes bad contact. It may be that a similar action may 
occur coincidently on other brushes, but, there is no uniformity 
about it. This action of transferring copper to the brush is 
sometimes known as " picking up copper/' It is not limited 
to brushes of one polarity, except where the metallic coating 
is caused primarily by the work current. Where it results from 


high local currents, it may be on the brushes of either polarity, 
for the local currents go in and out at each brush. However, 
according to the \\nter T s experience, this coating is more com- 
mon on the one polarity 

Glowing and honeycombing of brushes is not necessarily 
dependent upon the metallic coating on the brushes, although 
this latter increases the action Anything that will unduly in- 
crease the amount of current m any brush contact for a period 
long enough to result in heating and lower contact resistance, 
with brushes in parallel, may start this gloxving and honey- 
combing. It is not as common an action in modern machines 
as in old time ones. 

As an evidence that poor contact or high contact drop tends 
to produce burning, may be cited the fact that, in many cases 
of apparent rapid wear of the commutators, such wear has 
been practically overcome by simply undercutting the mica 
and thus allowing more intimate contact between brush and 
copper. In some instances, this also lessened or eliminated 
the tendency to pick up copper. Thus undercutting has been 
very beneficial in quite a number of ways. 

There are certain limitations in direct-current machines, de- 
pending upon the minimum number of slots per pole which can 
be used. Provided satisfactory commutating conditions can 
be obtained, it is in the direction of economy of design to use 
a relatively low number of slots per pole, with a correspond- 
ingly large number of coils per slot. This is effective in several 
ways. In the first place, insulating space is saved, thus allow- 
ing an increase in copper or iron sections, either of which al- 
lows greater output. In the second place, \vider slots arc favor- 
able to commutation. Thus the natural tendency of d-c. de- 
sign is toward a minimum number of slots per pole. But if 
this is carried too far, certain objections or disadvantages arise 
or become more prominent, so that at some point they over- 
balance the advantageous features. As the slots are widened 
and the number of teeth diminished, variations in the reluct- 
ance of the air gap under the main poles, with corresponding 
pulsations in the main field flux become more and more pro- 
nounced. These may effect commutation, as the short cir- 
cuited armature coils form secondary circuits in the path of 
these pulsations But before this condition becomes objec- 


tionable, other troubles are liable to become prominent, such 
as " magnetic noises, 1 ' etc. If the machine is of the commutat- 
ing pole type, there are liable to be variations in the commutat- 
ing pole air gap reluctance, so that it may be difficult to obtain 
proper conditions for commutation. A relatively wide corn- 
mutating zone is required if there are many coils per slot; also, 
all the conductors per slot usually will not commutate under 
equal conditions, which may result in blackening or spotting 
| of individual commutator bars symmetrically spaced around 
the commutator, corresponding to the number of slots. Innon- 
commutating pole machines, it may be difficult to find a suit- 
able field or magnetic fringe in which to commutate, and thus 
the first and last coil in each slot will have quite different fluxes 
in which to commutate. 

Depending upon the relative weight of the various advant- 
ages and disadvantages of a small number of slots per pole, 
practise varies greatly in different apparatus. In small and 
medium capacity railway motors, where maximum output in' 
minimum space is of first importance, and where noise, vibra- 
tions, etc. are not very objectionable, the number of slots per 
pole used is probably lower than in any other line of d-c. ma- 
chines, six to eight per pole being rather common. In the 
smaller and medium size stationary motors, where noise must 
be avoided, a somewhat larger number of slots is used in gen- 
eral, depending somewhat upon the size of the machine. On 
still larger apparatus, excepting possibly, small low-speed en- 
gine type generators, 10 slots or more per pole are used in 
most cases, and, in general, more than 12 are preferred. In 
the large 600-volt machines, the number is fixed partly by the 
minimum number of commutator bars per pole, and the num- 
ber of coils per slot. Assuming three coils per slot, then with 
a minimum number of commutator bars of about 40 per pole, 
the minimum number of slots per pole will be 14, and with 
two bars per slot, will be correspondingly larger. This there- 
fore represents one of the limits in present practise. 

Noise, Vibration, etc. Mention has been made of limita- 
tions of noise and vibration being reached, in considering the 
minimum number of slots. This is a very positive limitation 
in design, especially so in recent years, when everything is being 
carried as close as possible to all limits in economies in materials 
and constructions. All the various conditions which cause 
unkltte noises in electrical apparatus are not yet well known, 


and the application of remedies is more or less a question of 
" cut-and-try." 

A fundamental cause of noise in direct-current machines lies 
in very rapid pulsations or fluctuations in magnetic conditions. 
This has been well known for years, and many solutions of the 
problem of preventing such variations in magnetic conditions 
from setting up vibrations and consequent noise, have been 
proposed, but many of them appear to hold only for the particu- 
lar machine, or line of machines, for which they were devised. 
A perfectly good remedy in one machine not infrequently 
proves an utter failure on the next one. There are certain 
remedies for noise in direct-current machines which apply pretty 
generally to all machines, but, as a rule, such remedies mean more 
expensive constructions In general, large air gaps and gradual 
tapering of the flux at the pole edges tend toward quiet opera- 
tion. A large number of slots per pole tends toward quietness. 
However, the trend of design has been toward very small air 
gaps, especially in recent designs of small and moderate size 
d-c motors, also, the aim has been to use as few armature slots 
as possible Moreover, newer designs with steel or wrought 
iron frames, as a rule, have the magnetic material in the frames 
reduced to the lowest limit that magnetic conditions will per- 
mit. Also, with the general use of commutating poles, the 
tendency has been toward " strong " armatures and corres- 
pondingly weak fields, so that the total field fluxes and field 
frames are relatively small compared with the practise of a 
few years ago. With these small frames, resonant conditions 
not infrequently are encountered, especially in those machines 
which are designed to operate over a very wide range in speed. 
There is liable to be some point in the speed range where the 
poles or frame, or some other part, is properly tuned to some 
pulsating torque or " magnetic pull " in the machine. In such 
case, a very slight disturbance of a periodic nature may act 
cumulatively to give a very considerable vibration and conse- 
quent noise. 

The pulsations in magnetic conditions which produce vibra- 
tion may be due to various causes, but, as a rule, the slotted 
armature construction is at the bottom of all of them. Open 
type armature slots usually arc much worse than partially 
closed slots. Such open slots produce ** tufting " or " bunch- 
ing " of the magnetic flux l>etween the field and armature, and 
it is this bunching of flttx which usually, in one form or another. 


produces a -magnetic pulsation or pull which sets up vibration. 
This bunching of lines may be such as to set up pulsating mag- 
netic pulls at no-load as well as full load In other cases, the 
ampere turns in the armature slots tend to exaggerate or accen- 
tuate the bunching so that the vibration varies with the load. 
This bunching of the flux may act in various ways. The total 
air gap reluctance betvi een the armature and tnain poles may 
vary or pulsate, so that the radial magnetic pull between any 
main pole and the armature will pulsate in value. If the re- 
luctances under all the poles are varying alike, then these 
pulsating radial pulls will tend to balance each other at all 
instants. However, if the reluctances under the different poles 
do not vary simultaneously, then there are liable to be un- 
balanced * radial magnetic pulls of high frequency, depending 
upon the number of armature teeth, speed of rotation, etc. 
If this frequency is so nearly in tune with the natural period 
of vibration of some part of the machine, such as the yoke, 
poles or pole horns, armature core and shaft, that a resonant 
condition is approximated, then vibration and noise are almost 
sure to occur. 

Radial unbalanced pulls, as described, are liable to occur when 
the number of armature teeth is other than a multiple of the num- 
ber of poles; and the smaller the number of teeth per pole, the 
larger will be the unbalancing in general As a remedy, it 
might be suggested that the number of armature slots always 
be made a multiple of the number of poles However, there are 
several objections to this One serious objection is that, on 
small and moderate size d-c machines, the two-circuit type of 
armature winding is very generally used, and, with this type of 
v( hiding, the number of armature coils and commutator bars must 
always be one more or less in number than some multiple of the 
number of pairs of poles. Mathematically therefore, with a two- 
circuit winding, the number of slots can never be a multiple of the 
number of poles unless an unsymmetrioal winding is used, 
that is, one vuth a " dummy " coil A second objection to using 
a number of slots which is a multiple of the number of poles, 
is that there are pulsating magnetic pulls "which may be exag- 
gerated by this very construction There are two kinds of mag- 
netic pulls, a radial, which has already been considered, and a 
circumferential, due to the tendency of the armature core to 
set itself where it will enclose the maximum amount of field 
flux. Obviously, if the arrangement of slots is such that when 


one pole has a maximum flux into the teeth, another pole has 
a minimum, then the circumferential puslations in torque 
will be less than if all poles enclosed the maximum or the mini- 
mum flux simultaneously. This latter condition will be produced 
when the number of armature slots is a multiple of the number 
of poles. Therefore, in dodging unbalanced radial magnetic 
pulls by using a number of armature slots which is a multiple 
of the number of poles, the designer is liable to exaggerate the 
circumferential variations in torque or pull, so that he is no better 
off than before. This circumferential pulsating magnetic pull 
may act in various ways to set up vibration, and if there is any 
resonant condition in the machine, vibration and noise will 

Several years ago, the writer made some very interesting tests 
on a number of d-c. machines to discover the nature of the vi- 
brations which were producing noise. 
These machines had very light frames 
and were noisy, although not exces- 
sively so. The following results were 
noted : In certain four-pole machines, 
it was noted that the frames vibrated 
in a radial direction, as could be 
easily determined by feeling. How- 
ever, upon tracing around the frame 
circumferentially, nodal points were 
p IG 4 noted. In some cases, there were 

points of practically no vibration 

midway between the poles, as at A in Fig. 4. In other cases 
the point of least vibration was at B t directly over the main 
poles. Apparently, minimum vibration at A and maximum at 
B occurred when the pulsating magnetic pulls were in a radial 
direction, while, with circumferential pulls, the maximum vibra- 
tion was at A . It was also noted in some instances that a varia- 
tion in the width of the contact face of the pole against the yoke 
produced vibrations and noise, and nodal points in the yoke, 
the vibrations being a iBaximum at A . 

In still other cases in commutating pole machines, vibrations 
and noise were apparently set up by either radial or circumferen- 
tial magnetic pulsations under the cominutating poles themselves, 
as indicated by 'tfoe-f^cfc tbat mnoval of the comirrutating poles, 
or a cdnsideraMe mcr-ease m their air gaps, tended to overcome 
the noise. la s&cfo ases, tlie noise usually increased with the 
load, m ooiastant speed imchines. 


Skewing of the armature slots, or of the pole faces, has proven 
quite effective in some cases of vibration and noise. Tapered 
air gaps at the pole edges have also proven effective in many 
individual cases However, the causes of the trouble and the 
remedies to be applied in specific cases are so numerous and so 
varied that at present it is useless to attempt to give any limita- 
tions in design as fixed by noise and vibration due to magnetic 

From time to time, cases have come up where noticeable 
11 winking " of incandescent lights occur, this being either of a 
periodic or non-periodic character, the two actions being due to 
quite different causes In either case, the primary cause of the 
difficulty may be in the generator itself, or it may be 'in the 
prime mover. The characteristics of the incandescent lamp 
itself tends, in some cases, to exaggerate this winking To be 
observable when periodic, the period must be rather long, cor- 
responding to a very low frequency Periodic flickering of 
voltage may be considered as equivalent to a constant d-c 
voltage with a low-frequency small-amplitude alternating e m f . 
superimposed upon it In view of the fact that incandescent 
lamps of practically all kinds give satisfactory service without 
flicker at 40 cycles uith the impressed e m f varying from zero 
to 40 per cent above the effective value, one would think that a 
relatively small variation of voltage, of 3 per cent or 4 per cent 
for instance, would not be noticeable at frequencies of 5 to 
10 cycles per second. However, careful tests have shown 
that commercial incandescent lamps do show pronounced 
flicker at much lower percentage variations in voltage, de- 
pending upon the thermal capacity of the lamp filament. Based 
on such thermal capacity, low candle power 110-volt lamps, for 
example, should show more flicker than high candle power lamps. 
Also, tungsten lamps for same candle power should be more sen- 
sitive than carbon lamps, due to their less massive films. In 
fact, trouble from winking of lights has become much more pro- 
nounced since the general introduction of the lower-candle power, 
higher-efficiency incandescent lamps. 

In view of the fact that winking has been encountered with 
machines in which no pronounced pulsations in voltage appear 
to be possible, a series of tests was made some years ago to 
determine what periodic variation was noticeable on ordinary 


low-candle power Tungsten and carbon lamps A lamp circuit 
\vas connected across a source of constant direct e m f , and in 
series with this circuit was placed a small resistance which could 
be varied at different rates and over varying range The 
results were rather surprising in the very low pulsations in volt- 
age which showed a flickering of the light when reflected from a 
white surface With the ordinary frequencies corresponding to 
small engine type generators that is, from 5 to 10 cycles peri- 
odic variations in voltage of \ per cent above or below the mean 
value were sufficient to produce a visible wink, with 16-candle 
power carbon lamps; while 1 per cent variation above and below 
was quite pronounced With corresponding tungsten lamps, 
only about half this variation is sufficient to produce a similar 
wink. These tests were continued sufficiently to show that such 
periodic fluctuations in voltage must be limited to extremely 
small and unsuspected limits This condition therefore imposes 
upon the designer of such apparatus a degree of refinement in 
his designs which is almost a limitation in some cases. 

It is probable that non-periodic fluctuations in voltage do 
not have as pronounced an effect in regard to winking of lights 
as is the case with periodic fluctuations, if they do not follow 
each other at too frequent intervals, unless each individual 
pulsation is of greater amplitude, or is of longer duration 
Possibly a momentary variation in voltage of several per cent 
will not be noted, except by the trained observer, unless such 
variation .has an appreciable duration 

A brief discussion of the two classes of voltage variations 
may be of interest, and is given below. 

Periodic Fluctuations. As stated before, these may be due 
to conditions inside the machine itself, or may be caused by 
speed conditions in the prime mover. Not infrequently, the 
two act together. Variations in prime mover speed can act in 
two ways; first, by varying the voltage directly in proportion 
to the speed, and secpnd, by varying the voltage indirectly 
through the excitation, the action being more or less cumulative 
in some cases. Such sp^ed variations usually set up pulsations 
corresponding directly to t the revolutions per jraiimte and in- 
dependent of tbe number of pok #a th& ? machine$, 

In the machine itself, p^o^y^^i^ pf frequency lower 
than normal Creque^v^fSte !&*%* Hw*fj may be caused 
by magnetic <fes5p^^fe^|*^^^ s^j, r> by unsymmetrical 
windings, Ustt^$ ? 3fl|^ fluctua^ 


lions at a frequency corresponding to the normal frequency of 
the machine, and therefore will have no visible effect unless 
such normal frequency is comparatively low, which is usually 
the case in engine type d-c. generators. In other cases, these 
dissymmetries may give pulsations corresponding to the rev- 
olutions, and not the poles For instance, if the armature 
periphery and the field bore are both eccentric to the shaft, 
then magnetic conditions are presented which vary directly 
with the revolutions. 

However, there have been cases where no dissymmetry could 
be found, and yet which produced enough variations to wink 
the lights Usually in such cases, the number of armature 
slots per pole was comparatively small, and the trouble was 
overcome by materially increasing the number of slots per pole. 
A second source of winking has been encountered in some three- 
wire machines in which the neutral tap is not a true central 
point In such case, the neutral travels in a circle around the 
central point and impresses upon the d-c voltage a pulsation 
corresponding to the diameter of the circle Its frequency how- 
ever, is that of the machine itself and is therefore more notice- 
able on low frequency machines, such as engine type generators. 

Non-Periodic Pulsations or Voltage " Dips." In all d-c. 
generators, there is a momentary drop or " dip" in voltage with 
sudden applications of load, the degree of drop depending upon 
the character and amount of load, etc The effects of this 
have been noted most frequently in connection with electric 
elevator operation, in which the action is liable to be repeated 
with sufficient frequency to cause complaint. Various claims 
have been made that certain types of machines did not have 
such voltage dips, and that others were subject to it In con- 
sequence, the writer and his associates made various tests in 
order to verify an analysis of this action which is given below, 

The explanation of this dip in voltage is as follows. Assume, 
ior instance, a 100-volt generator supplying a load of 100 am- 
peres that is, with one ohm resistance in circuit. The drop 
across the resistance is, of course, 100 volts. Now, assume 
that a resistance of one ohm is thrown in parallel across the 
circuit. The resultant resistance in circuit is then one-half 
ohm However, at the first instant of closing the circuit through 
the second resistance, the total current in the circuit is only 
100 amperes, and therefore the line voltage at the first instant 
momentarily must drop to 50 volts. However, the em.f. 


generated in the machine is 100 volts, and the discrepancy of 
50 volts between the generated and the line volts results in a 
very rapid rise in the generator current to 200 amperes. If 
the current rise could be instantaneous, the voltage dip would 
be represented diagrammatically by a line only, that is, no 
time element would be involved. However, the current can- 
not rise instantaneously in any machine, due to its self-induction, 
and therefore, the voltage dip is not of zero duration, but has 
a more or less time interval. The current rises according to an 
exponential law, which could be calculated for any given ma- 
chine if all the necessary constants were known. However, 
such a great number of conditions enter into this that is it usually 
impracticable to predetermine the rate of current rise in de- 
signing a machine, and it would not change the fundamental 
conditions if the rate could be predetermined, as will be shown 

A rough check on the above theory could be obtained in the 
following manner, by means of oscillograph tests. For example, 
it was assumed in the above illustration that with one ohm 
resistance in circuit, an equal resistance was thrown in parallel, 
which dropped the voltage to one-half. In practise, the actual 
drop which can be measured might not be as low as one-half 
voltage, as the first increase in current might be so rapid as to 
prevent the full theoretical dip from being obtained. However, 
an oscillograph would show a certain amount of voltage drop. 
If now, after the current has risen to 200 amperes and the con- 
ditions become stable, the second resistance of one ohm is 
thrown in parallel with the other two resistances of one ohm 
each, then in this latter case, the resultant resistance is re- 
duced to two-thirds the preceding value, instead of one-lialf, 
as was the case in the former instance. Therefore, the dip 
, would be less than in the former case. Again, if "one ohm re- 
sistance is thrown in parallel ^ith three resistances of one ohm 
each, the restdtant resistance becomes three-fourths of the 
preceding value, that is, the voltage dip is still less. There- 
fore, according to the above analysis, if a given load is thrown 
on a machine, the dips will be relatively less the higher the load 
the machine is carrying. Also, if the same percentage of load 
is thrown on each time, then the dips should be practically the 
same, regardless of the load the xnadiine is already carrying. 
For example, if the machine is carrying 100 amperes, and 100 
amperes additional is tha^cmii o% ike dip shotild be the same as 



if the machine were carrying 300 amperes and 300 amperes 
additional were thrown on. 

Also, according to the above theory, a fully compensated 
field machine, (that is, one with a distributed winding in the 
pole faces proportioned to correctly neutralize the armature 
magnetomotive force) should also show voltage dips with load 
thrown on. To determine if this is so, several series of tests were 
made on a carefully proportioned compensated field machine. Two 
series of tests were made primarily. In the first, equal in- 
crements of currents were thrown on, (1) at half load, (2) at full 
load, and (3) at 1J load on the armature. In the second series 
of tests, a constant percentage of load was thrown on; that is~ 7 
at half load the same current was thrown on as in the first test, 
while at full load, twice this current, and at If load, three 
times this current was thrown on. 

According to the above theory, all these should show voltage 
dips, although the machine was very completely compensated. 
Also, in the first series of tests, the dips should be smaller with 
the heavier loads on the machine, while in the second series 
they should be the same in all tests. This is what the tests 
indicated. In the first series, the dips in voltage varied, while 
in the second series, they were practically constant. The re- 
sults of these tests are shown in the following table. (The 
oscillograph prints were so faint that it was not considered 
practicable to produce them in this paper.) 


Load on generator 

Increase in load. 

Dip in voltage 





417 Amps 

700 Volts 








200 * 








300 * 





Tests, B C and D in the table show the dips for the first 
series of tests, while B, E and F show results for second series. 
The time for recovery to practically normal voltage was very 
short in all cases, varying from 0.002 to 0.004 seconds accord- 
ing to the oscillograph curves, but even with this extremely 


short time, there was very noticeable winking of tungsten 
lamps, in practically all tests The oscillograph curves showed 
practically no change in field current, except in test A. 

The machine used in these tests was a special one in some 
ways. It was a 500-kw., 1200- volt, railway generator with 
compensating windings and commutating poles, In order to 
keep the peripheral speed of the commutator within approved 
practise, it was necessary in the design to reduce the number 
of commutator bars per pole, and consequently the number 
of armature ampere turns, to the lowest practical limit. This 
resulted in an armature of very low self induction, which was 
very quick in building up the armature current with increase 
in load, This machine therefore did not show quite as severe 
variations as would be expected from a normal low-voltage 
machine of this same construction. However, these two series 
of tests did show pronounced voltage dips which were sufficient 
to produce noticeable winking of incandescent lamps. Presum- 
ably, therefore, all normal types of generators will wink the 
lights under similar conditions. 

Data obtained on non-compensated machines of 125 and 250 
volts indicate the same character of voltage dips as were found 
in the above tests. This should be the case, for, by the fore- 
going explanation, the compensating winding has no direct re- 
lation to the cause of the dip. 

It will be noted in these curves that the voltage recovers to 
normal value very quickly. However, incandescent lamps 
will wink, even with this quick recovery, if the dip is great 
enough. There is some critical condition of voltage dip in 
each machine which would produce visible winking of lights. 
Any increments of load up to this critical point will apparently 
allow satisfactory operation. If larger loads are to be thrown 
on, then these should be made up of smaller increments, each 
below the critical value, which may follow each other in fairly 
rapid succession. In other words, the rate of application of 
the load is of great importance, if winking of lights is to be 
avoided Therefore, the type of control for motor loads, for 
instance, should be given careful consideration in those cases 
where steadiness of the light is of first importance, and where 
motors and lights are on the same circuit. 

An extended series of tests has shown that, in most cases, 
10 per cent to 15 per cent of the rated capacity of the generator 
can be thrown on in a single step without materially affecting 


the lighting on the same circuit, and provided the prime mover 
holds sufficiently constant speed. However, judging from the 
quickness of the voltage recovery, the prime mover, if equipped 
with any reasonable flywheel capacity, cannot drop off materially 
during the period of the voltage dip as shown in the curves. 
The dip in voltage due to the flywheel is thus apparently some- 
thing distinct frdm the voltage dip due to the load. However, 
if the load is thrown on in successive increments at a very 
rapid rate, the result will be a dip in voltage due to the prime 
mover regulation, although the voltage dips due to the load 
itself may not be noticeable. 

Thejabove gives a rough outline of this interesting but little 
understood subject of voltage variations. Going a step farther, 
a similar, explanation could be given for voltage rises when the 
load is suddenly interrupted, in whole or in part. This is 
usually known as the inductive kick of the armature when the 
circuit is opened This may give rise to momentarily increased 
voltages which tend to produce flashing, as has already been 
referred to under the subject of flashing when the circuit breaker 
is opened. 


This presents two separate limitations in d-c. design, one 
being largely mechanical and the other being related to voltage 
conditions. As regards operation, the higher the commutator 
speed, as a rule, the more difficult it is to maintain good contact 
between brushes and commutator face. This is not merely a 
function of speed, but rather of commutator diameter and speed 
together. Apparently it is easier to maintain good brush con- 
tact at 5000 ft per minute with a commutator 50 in. in di- 
ameter than with one of 10 in. in diameter. Very slight un- 
evenness of the commutator surface will make the brushes 
" jump " at high peripheral speeds, and the larger the dia- 
meter of the commutator with a given peripheral speed, the 
less this is. 

The peripheral speed of the commutators is also limited by 
constructive conditions. With the usual V-supported com- 
mutators, the longer the commutator, the more difficult it is 
to keep true, especially at very high speeds and the higher 
temperatures which are liable to accompany such speeds. 
Therefore, the allowable peripheral speeds are, to some extent, 
dependent upon the current capacity per brush arm, for the 
ength of the commutator is dependent upon this. The per- 


missible speed limits, as fixed by mechanical constructions- 
have been rising gradually as such constructions are improved, 
At the present time, peripheral speeds of about 4500 ft. per 
minute are not uncommon with commutators carrying 800 to 
1000 amperes per brush arm. In the case of 60-cycle, 600- 
volt synchronous converters, 5200- to 5500-ft. speeds are usual 
with currents sometimes as high as 500 to 600 amperes per arm, 
In the case of certain special 750-volt, 60-cycle converters, oper- 
ated two in series, commutator speeds of about 6400 ft. have 
proved satisfactory. These latter, however, had comparatively 
short commutators. 

For the small diameter commutators used in d-c. turbo- 
generator work, peripheral speeds of 5500 to 6000 ft. have been 
common. However, such machines usually have very long com- 
mutators and of the so-called " shrink-ring " construction. The 
brushes may not maintain good contact with the commutator 
at all times, and in a number of machines in actual service, the 
writer, in looking at the brush operation, could distinctly see 
objects beyond the brush contacts; that is, one could see 
" through " the contact, and curiously, in some of these cases, 
the machines seemed to have operated fairly well. One ex- 
planation of this is that the gaps between brushes and com- 
mutator were intermittent,, and, with one or more brush arms in 
parallel, one arm would be making good contact, while another 
showed a gap between brushes and commutator. Appar- 
ently, the commutators were not rough or irregular, but 
were simply eccentric when running at full speed and the 
brushes could not rise and fall rapidly enough to follow 
the commutator face all the time. Incidentally, it may be men- 
tioned at this point, that with the higher commutator speeds 
now in use, there has come the practise of " truing " commutators 
at full speed. This is one of the improvements which has al- 
lowed higher commutator speeds. 

The other limitation fixed by peripheral speed, namely, that 
of the voltage, is a more or less indirect one. It is dependent 
upon the number of commutator bars that are practicable be- 
tween two adjacent neutral points; or, in other words, it is 
dependent upon the distance between neutral paints. The 
product of the distance between adjacent neutral points and the 
frequency, in Alternations, gives the peripheral speed of the 
commutator, (distance between neiitcal points in feet times 
per &&&^]&$^ peripheral speed in feet per 


minute). With a given number of poles and revolutions per 
minute, the alternations are fixed Then, with an assumed 
limiting speed of commutator, the distance between neutral 
points is thus fixed. This then limits the maximum number of 
commutator bars, and therefore the maximum voltage which 
is possible, assuming a safe limiting voltage per bar From 
this it may be seen that the higher the penpheral speed, the 
higher the permissible voltage with a given frequency In the 
same way, if the frequency can be lowered (either the speed or 
the number of poles be reduced) the permissible voltage can 
be increased with a given peripheral speed. Where the speed 
and the number of poles are definitely fixed and the diameter 
of commutator is limited by peripheral speed and other con- 
ditions, the maximum practicable d-c voltage is thus very defi- 
nitely fixed. This ^s a point which apparently has been mis- 
understood frequently It explains why, in railway motors, for 
high voltages, it is usual practise to connect two armatures per- 
manently m series, also, why two 60-cycle synchronous conver- 
ters are connected in series for 1200- or 1500- volt service. In 
synchronous converter work, the frequency being fixed once for 
all, the maximum d-c. voltage is directly dependent upon the 
peripheral speed of the commutator 


The principal intent, in this paper has been to show that cer- 
tain limitations encountered in d-c. practise are just what should 
be expected from the known properties of materials and electric 
circuits. The writer has endeavored to explain, in a simple, 
non-mathematical manner, how some of the apparently com- 
plicated actions which take place in commutating machinery 
are really very similar to better understood actions found in 
various other apparatus An endeavor has also been made to 
show that a number of the present limitations in direct current 
design and operation are not based merely upon lack of ex- 
perience, but are really dependent upon pretty definite condi- 
tions, such as the characteristics of carbon brushes and brush 
contacts, etc Possibly a better understanding of the character- 
istics and functions of carbon brushes will result from this paper. 

The writer makes no claims to priority for many of the ideas 
and suggestions brought out in this paper However, much of 
the material is a direct result of his own investigations and those 
of his associates during many years of experience with direct- 
currcnt apparatus 



The following method of determining the maximum capacity 
which can be obtained with given dimensions and for assumed 
limitations as fixed by commutation, flashing and other con- 
ditions, is based upon certain formulas which the writer de- 
veloped several years ago, and which appeared in a paper before 
the Institute.* 

On page 2389 of the 1911 TRANSACTIONS of the Institute, 
the following general equation is given : 

I c W t R s TC-K r _ _ 2 Dp _ 

10* 1 C] (L L}) (025^+05) (D+P P ) 

9 + - 035 N ) + c *- (JL33 ds + - 52 + 2 - 16 s 


Where I c = Current per armature conductor. 

W t = Total number of armature conductors 
T c = Turns per armature coil or commutator bar. 
L & Li \= Width of armature core and commutating pole 

faces respectively. 
D = Diameter of armature. 
p = No. of poles. 
N = No. of slots per pole. 
d 9 = Depth of armature slot. 
5 = Width of armature slot. 
n = Ratio of width of armature tooth to slot 'at 

surface of core. 

Ci, Cz, 3, 4 are design constants. 

In order to simplify the above equation, the following as- 
sumptions are made; 

(a) No bands are used on armature core, thus eliminating 
the last term in the above equation. 

(b) Li = L, thus eliminating the first expression inside tha 
bracket in the above equation, 

B&th the above assumptions are in the direction of increased 
capacity with a given short circuit voltage, E c . 
Equation (1) then becomes, 

*Theory of Commutatang and Its Application to Commutating-Pole Machines, Page 201. 


[Sept. 16 

+ c 3 (1 33rf s + 0.52 + 2 16 s Vn) 1 (2) 

The various terms in equation (2) should be put in such form 
that limiting values can be assigned to them as far as possible. 
In order to do this the equation can be condensed and simplified 
as follows, for large machines: 

(a) Assuming parallel type windings, 

7" 2 to F 

\y t " r ^ where F& = Average volts per commutator 
l/ & 

bar or coil. 

where It = Total current. 




IE = Kilowatts X 10* = Kw 10 3 

Also, R a p = 2/, where/ = Frequency in cycles per second. 

_, , I c W t R s T, TT Kw p X 4/ r c V u . ,....- 
Therefore, - j^ - = - V & X 10 s - ' p g 

watts per pole. 

(6) Let P< = Armature tooth pitch, 

Then D 

and c 2 - (09 + 0.035 N) = c 2 -- (0.9 + .035 .Vj P t 

p 7T 

In case of a chorded winding, the term 0.035 N should be 
0.035 Ni t where NI represents the number of teeth or tooth pitches 
spanned by the coil. 

(c) In the second tenn inside the bracket in equation (2), 

the ratio =- can be transformed into an expression containing 

P if as follows: 

E =B t S t C P R,W a 


B t = Flux density in armature teeth. 

St Section of iron in armature teeth. 

C p = Field form constant (percentage polar area). 

R s = Revolutions per second. 

W s = Wires in series. 

S t = N T p L c^ where T = Width of tooth, and c* = the 
ratio of actual iron to the core width L. 

P 2 
As an approximation, T X9 = - (This is a fairly close ap- 

proximation within practical limits in the usual armature con- 
structions) . 

Then, S,= 


r -S = 1 

This can further be condensed as follows. 

W.= Tc-22-,*rLAR t p-2f 

t L 1QS v *> 

Therefore, = = ^ 

(d) The expression (1.33 d, + 0.52 + 2.16 $ Vn) can be 
modified as follows, 

Vw" = ^1 T = ^(^ = ^ on the basis that T xs = ^~ 


Then, 2.16 sVn 1.08 P t approx. 
and, (1.33 d, + 0.52 + 2.16 sVnj = (1 33 d s + 0.52 + 1.08 P t ) 

Substituting all the above transformations in equation (2) 
we get, 

4 ( - 9 + 35 Nl} NPt 


E c V b 10 5 

Kw p = ^ ^., 

P (4) 

4c 2 (0.9 +0 035tfi)jyyP< a + ~^^r (1 33d. + 0.52+ 1 08 P t \ 

ntLpL\ C 5 1 c [ 

Maximum Kilowatts per Pole. Differentiating (4) to obtain 
P t for maximum Kw pt 

c Nf (0.9 + 0.035 tf * ' 6 (1 33 <f s + 52) 

* c #* Cp J\ C$ 
TT Ci V b 10 s 

If Pi in equation (5) could be derived and then substituted 
in equation (4), then for any assumed value of E c and with the 
other terms given limiting values, an expression for the maximum 
kilowatts per pole could be obtained. The writer has not been 
able to solve this directly in any sufficiently simple manner, 
although a complicated approximate expression can be obtained. 
However, for practical purposes, the solution for an3 r given con- 
ditions can be obtained by trial methods and the results plotted 
in curves. 

For instance, in equations (4) and (5), the following terms 
may be given limiting values for a given class of machines and 
for a specified voltage 

T r - Turns per coil 

c<i = End flux constant. 

Af = Number of slots per pole A r i = No. of teeth spanned 

by coil. 

GZ = Brush short circuit constant. 
Vi = Average volts per bar. 
C p = Field form constant With max. volts per bar fixed, 

then V max. X C p = V b . 
Bt = Flux density in teeth. 
& = Ratio of actual iron width to core width L. 

Also, type of armature winding can be fixed and departure 
from full pitch winding, or amount of chording can be given. 


There will then remain for any assumed value of E c , the terms, 

Kw p = Kilowatts per pole. 

P t = Tooth pitch. 

/ = Cycles per second. 

d s = Depth of armature slot. 

All four of these latter terms are in equation (4), and the last 
three in equation (5). Therefore, assuming the depth of slot, 
equation (5), the values of P t for different frequencies may be 
determined by trial methods. The corresponding values of 
P t , f and d 3 can then be substituted in equation (4), and the 
kilowatts per pole thus determined. Tables or curves can then 
be prepared giving the kilowatts per pole for different frequencies 
and for different assumed slot depths. 

A series of such tables have been worked out for a specified 
set of Conditions as given below. The assumed limiting con- 
ditions were as follows: 

E c = 4.5, that is, one turn per coil parallel type winding 

is assumed, 
e m f =600 volts. 
C p = 0.68 
Vb = 14.3. No of commutator bars per pole = 42. No 

compensating winding is used. Therefore, Vb. 


" 42 

14.3 = 

and max. volts per bar at no load = 
21. Allowing 25 per cent increase for 

flux distortion, and increased voltages at times, 
gives 26 3 at full load. 

r.> = 1,25 for average constructions. 

<-, = Vanes with the number of coils per slot and the aver- 
age number of bars covered by the brush, but as- 
suming 2 bars covered, then Cz = 0.4 approx, 
with 1 slot chording, and with either 2 or 3 coils 
per slot. 

B t - 150,000 lines per sq. in. on the basis of actual iron 
and all flux confined to the iron. 

r & = 0.75. This allows for 90 per cent solid iron and j 
of the total width taken up by air ducts (aboul 
|" duct for each 2" of laminations). 




14 f 

(Two cases have been assumed, one with 3 coils 

per slot and 14 slots per pole, and the other with 
2 coils per slot and 21 slots per pole. 

14 Slots per Pole. Substituting the above values in equations 
(4) and (5), then for 14 slots per pole equation (4) becomes, 

Kw, = 3767 E c [ 

and equation (5) becomes 

jP? = 18 36 (2.5 d, + 1) + 19 P t 



Incidentally, equation (6) can be simplified to a certain extent 
by partially combining with equation (7), giving the following 
equation : 


Kiu f = 99 EC 

3.725 (2.5 d. + 1) - 

Equation (8), of course, can only be used with the values of 
P t determined from equation (7). 

Three values for d s were chosen, 1 in., 1.5 in., and 2 in., which 
cover the practical range of design for large d.-c. generators. 
Frequencies from 5 to 60 cycles were also chosen. The corres- 
ponding values for Pt and Kw p are tabulated below, 




J, =1* 

d = 15" 

d -o* 

per sec. 

/>! K~> 

Pt Kp 

Pi *, 


2 85 in. 670 

3 08 m 647 

3 255 in 620 


2 20 453 

2 362 428 

2 504 407 


1 685 2Q9 

1 828 282 

1 945 266 


1 455 235 

1,575 219 

1 680 208 


1 302 197 

1 417 183 5 

1 515 173 


1.20 173 5 

1 305 160 

1 398 151 5 


1 I2o 153 

1 226 143 5 

1 310 135 

21 Slots Per Pole. Substituting the proper values in equations 
(4) and (6) for 21 slots per pole, and, one slot chording, and then 
solving for Pt and Kw p for the same slot depths and frequencies, 
the following table is obtained: 





d s = 1 in. 

<f <r =* 1 5 in. 

d s = 2 in 

per sec. 

P t Kw p 

Pt Kw p 

Pt Kw p 


1 985 in. 576 

2 14 in 542 

2 27 in 515 


1 53 380 

1 56 333 

1 77 338 


1 185 249 

1 29 J,52 

1 36 214 


1 022 195 

1 12 181 

1 192 368 


922 16,i 

1 003 130 

1 077 141 


830 142 

932 131 

997 123 


796 126 

874 117 

936 110 


Two cases only need be considered, namely 25 and 60 cycles. 
For these two cases, more definite limits can be given than for 
the above rather general solution for d-c. machines. 

25 Cycles. Let N ~ 21, and NI = 20; also, assume two 
coils per slot for 600 volt machines. 

ri = 1.0. 

c z = 0.37 

ft = 165,000 

C p = 0.7 

Then for assumed values for depth of slot of 1 in., 1.5 in , 
and 2 in., and for E c = 4.5, the following values of P t and 
Kw p are obtained: 


Depth of 






per pole. 

1 in 



1 5 

1 19 



1 275 


60 Cycles. Let N = 15, and 
per slot for 600 volts. 

c 3 = 0.4 
B t = 150.000 
C p = 0.66 

14, Also, assume 3 coils 



Then assuming s bt depth of 1 in , 1.25 in., and 1 5 in., and 
EC = 4.0, the following values of P t and Kw p result: 


Depth of 





per pole. 

1 in 

1 14 


1 25 

1 195 

137 5 

1 5 

1 24 


The above tabulated results agree pretty well with practical 
results obtained in large generators and converters. There are 
so many possible variations in the limits assumed that only 
general results can be shown For instance, in Table I, a 
constant limiting induction in the armature teeth of 150,000 
lines per sq. in. is assumed. With low frequencies this can be 
increased, while with frequencies of 50 to 60 cycles, somewhat 
lower inductions will be used. Also, the commutation con- 
stant C\ which is dependent upon the number of bars covered 
by the brush is naturally subject to considerable variation. 

The results obtained are predicated upon parallel types of 
windings and a minimum of one turn per armature coil. If 
types of windings having the equivalent of a fractional number 
of turns per coil less than one, prove to be thoroughly satis- 
factory for large capacity machines, then the above maximum 
capacities can be materially increased. However, accepting 
the results as they stand, the limits of capacity as fixed by 
commutation are in general about as high as other limitations 
will allow 




FOREWORD About ten years ago, the author found that there was 
considerable misunderstanding regarding the regulating charac- 
teristics of commutating-pole machines, and the conditions 
which were to be met in parallel operation. In consequence, he 
prepared this brief article for the use of the engineers of the 
Westinghouse Electric & Manufacturing Company. It has 
proved so satisfactory that it has been kept in publication ever 

This paper was written before the term " comrnutating 
pole" was adopted to replace the term "interpole" which is 
found throughout the article. (Eo.) 

THE inherent regulation characteristics of the armature of a 
direct-current machine has much to do with its parallel 
operation with other machines. When two direct-current ar- 
matures are coupled in parallel and delivering load to the same 
external circuit, it is necessary, in order to obtain stable conditions, 
for each armature to tend to "shirk" its load; that is, it must 
naturally tend to transfer load to the other machine. This 
tendency to shirk may be either in bad speed regulation due to the 
prime mover which drives the armature, or in the drooping voltage 
characteristics of the armature itself. A drooping speed character- 
istic indirectly produces a drooping voltage characteristic in the 
armature and therefore both causes lead to the one characteristic, 
namely, drooping voltage, as the condition for stable parallel 
operation. This drooping voltage characteristic must be the in- 
herent condition. In practice, the voltage at the armature 
terminals frequently rises with increase in load, but its rise is due 
to some external condition, such as increased field strength. 

Direct-current machines, as hitherto ordinarily constructed, 
naturally give drooping voltage characteristics in the armature 
windings. If two such armatures are paralleled they tend to 
divide the load in a fairly satisfactory manner provided then- 
prime movers regulate similarly in speed. If means are applied 
for giving a rising voltage characteristic to the machines, such as 
series coils in the field, then the armature terminals must be 
paralleled directly in order to maintain stability. If, for instance, 
the armatures are not paralleled directly, but the paralleling is 



done outside the series coils, then the operation will be unstable 
unless the machines still have drooping voltage characteristics 
If they have rising characteristics, then parallel operation is im- 
practicable. If either machine should take an excess of load, its 
voltage would rise, while that of the other machine would fall due 
to decreased load. This condition would naturally force the first 
machine to take still more load and the second one to take still less 
This condition would continue until the first machine actually fed 
current back through the other machine and it would be necessary 
to cut them apart to avoid injury. However, by paralleling the 
two armatures inside the series coils, that is, between the series 
coils and the armature terminals, this unstable condition is avoided 
due to two reasons, first, the inherent drooping voltage character- 
istics of the armatures, and, second, the fact that the series coils 
are paralleled at both terminals, thus forcing them to take pro- 
portional currents at all times and thus compounding both ma- 
chines equally. 

If direct current machines are so designed or operated as to 
give rising instead of drooping armature characteristics, then 
parallel operation is liable to be unstable. This condition could 
be obtained in ordinary machines by prime movers which tend to 
speed up with increasing load, thus producing rising voltage on the 
armature. Ordinarily, such speeding up of the prime mover would 
have to be rather large, as the normal drooping characteristics of 
the ordinary armature is fairly large. However, prime movers of 
this character are comparatively rare. 

A second condition which can give a rising voltage is found 
not infrequently in the interpole type of direct-current machine. The 
interpole generator is similar to the ordinary type of generator, 
except that midway between the main poles small poles are 
placed which carry windings or coils which are connected directly 
in series with the armature. The winding on the interpoles is 
connected directly in opposition to the winding in the armature. 
The maxirnurn magnetizing effect of the armature winding is found 
at the points on the armature corresponding to the coils which are 
being commutated. The interpole is intended to be placed 
directly over these points and the interpole winding normally has 
such a value that it not only neutralizes the magnetizing effect to 
the armature winding at these points, but it also sets up a small 
magnetic field in the opposite direction whicl^ assists in the com- 
mutation of the armature coil. Therefore the interpole winding 


must have a number of ampere turns equal to the maximum ampere 
turns in the armature winding, plus the excess ampere turns 
necessary to produce the required commutating field strength. 

When this interpole winding is placed directly over the com- 
mutating position of the armature winding it should have prac- 
tically no effect on the armature characteristics. If, however, the 
interpole winding is not placed over these positions it will have an 
effect on the voltage characteristics of the machine, tending to 
either raise or lower the voltage, depending upon the position of 
the interpole with respect to the commutating position. The 
commutating points on the armature depend directly upon the 
brush position If the brushes are rocked backward or forward 
from the point corresponding to the mid position between the poles 
then the position of the commutated armature coils moves back- 
ward or forward with the brushes. As the commutating pole is 
fixed in position it is evident that the relation of the commutating 
pole to the coils undergoing commutation can be changed by the 
different brush settings. Herein lies a possible trouble in parallel 
running, for the commutating points can be so shifted, with 
respect to the commutating pole, that the armature winding 
voltage characteristics can be made to rise instead of droop. As 
explained before, this is an unstable condition for parallel operation. 

This condition can be illustrated in the following manner: 


Let Pig. 1 represent two main poles, and interpoles, with the 
brushes set in a position corresponding to the middle point of the 
interpole. The polarity of the interpoles and main poles is in- 
dicated in this figure. The polarity of any interpole, when the 
machine is running as a generator, is always the same as the 
polarity of the main pole immediately in front of the interpole. 
When the brush is placed in a position corresponding to an exact 
intermediate point in the interpole it is evident that the aramture 


coils lying between two commutating points, that is, the winding 
between a and b in Fig. 1, is acted upon by induction from the 
main pole and by half the induction from the interpoles adjacent to 
the main pole. However, as these two interpoles are of opposite 
polarity, and the induction is the same from each, it is evident that 
they have equal and opposite effects on the armature winding 
between a and b, and therefore do not affect its voltage 

In Fig. 2 the brushes are given a slight back lead so that the 
commutation is under the trailing magnetic flux from the inter- 
pole. It is now evident that between a and b the induction is 
from the main pole and from one interpole principally. With the 
back lead at the brushes, this interpole is the one immediately 
behind the main pole and therefore of the same polarity. This 
interpole therefore becomes a magnetizing pole and adds to the 
e. m. f . generated between a and b As the strength of this inter- 
pole is zero at no load and rises with load, it is evident that it tends 
to give an increased voltage between a and b as the load increases 
and thus tends to produce a rising voltage characteristic instead 
of a drooping one. The ampere turns in the interpole, as stated 
before is considerably greater than in the armature, but ordinarily 
the effect of these ampere turns is almost neutralized by the 
opposing effect of the armature winding However, with the 





H \- 

FIG. 2 

back lead, as indicated in Fig. 2, the opposing effect of the armature 
winding is shifted to one side of the interpole and thus the inter- 
pole ampere turns become more effective in actually magnetizing 
the armature, but become less effective in creating a commutating 
field for the coils which are now being reversed by the brushes. On 
account of this less effective field it may be necessary in practice 
to still further increase the ampere turns on the commutating poles 
in order to bring the trailing magnetic fringe up to a suitable 
value for producing proper commutation. It is evident, that this 


increased ampere turns on the commutating pole increases the 
induction under other parts of the commutating pole as well as 
under the trailing tip, and this increase under the other parts of 
the pole still further increases the voltage between a and b. 

With a back lead therefore the interpole may have the same 
effect as the series winding on the main field; that is, it may 
compound the machine so that the voltage at the terminals is rising 
instead of falling, even without any true series winding on the 
main poles. The machine therefore becomes an equivalent of a 
compound wound machine and if there is no equalizer between the 
interpole winding and the armature terminal, the generator may 
be unstable when paralleled with other machines. 

Take the case, next, where the brushes are given a forward 
lead, as shown in Fig. 3. Comparing this with Fig. 2, by the same 
reasoning it is evident that the interpole is now opposing the 
effect of the main pole, in the winding between a and b. The 
interpole therefore tends to produce a drooping voltage charac- 
teristic and has just the opposite effect of the series winding. In 
this position of the brushes the interpole winding tends to give 
good characteristics for parallel operation, but as the effect of the 
interpole is in opposition to the main pole it is evident that more 
series winding is required on the main field in order to over- 
compound the machine as a whole. Also, with the brushes in this 
position the interpole is not as effective in producing gpod com- 
mutation and therefore more ampere turns are required on their 
interpole winding. Therefore, both the interpole winding and the 
main series winding must be increased when the brushes are given 
this forward postion. However, parallel operation should be 

It is evident, therefore, from the above considerations, that 
for best results the brushes should be so set that the true point of 
commutation comes midway under the interpole. If this position 

'FIG. 3 


is found exactly, then the interpole should have practically no 
effect on the voltage characteristics of the armature, and parallel 
operation with other generators should be practicable A very 
slight forward lead is favorable to paralleling, but lessens the 

As a back lead at the brushes, when the machine is acting as a 
generator, tends to improve the compounding and lessens the 
series winding required on the main field, it might be suggested 
that this gives a cheaper and more efficient machine and that 
therefore this arrangement should be used, with some means added 
for overcoming the unstable conditions of paralleling. One 
means proposed for this is an additional equalizer connected 
between the interpoles and the armature terminals. This has 
been used in one or two instances, but in principle the arrange- 
ment is inherently wrong. When the interpole windings are 
paralleled, then the currents in them must divide according to 
their resistances. This condition would not be objectionable 
provided the armature currents also always varied in the same 
proportion. With slow changes in load this condition might be 
obtained. However, there are conditions of operation where the 
armature currents will not rise and fall in proportion and there- 
fore the interpole windings, with this arrangement, would not 
always have the right value to produce the desired commutating 
fields. By rights, each armature should be connected directly 
in series with its own interpoles and the currents in the two should 
rise and fall together for best results. This condition will not be 
obtained when an equalizer is connected between the armatures 
and interpoles. This solution of the problem should therefore be 
avoided in general. 

All the above leads to the fact that very accurate brush set- 
ting is required on interpole types of machines, and furthermore, 
when such setting is once obtained it should not be capable of 
ready adjustment or change. For this reason interpole machines 
should not have any brush rocking gear. In machines where such 
gear is present it would be better, in general, if the brush rocking 
mechanism were removed after the proper setting of the brushes is 
once obtained, and means should be employed for locking the 
brushes in this correct position. 

The correct setting of the brushes is rather difficult to obtain 
in many cases. Where the armature conductors can be traced 
from the commutator bars back under the poles, it is feasible in 


general to locate the correct setting by the position of the com- 
mutated coil with respect to the interpoles. In standard practice 
the throw or span of the coil is made, as nearly as possible, equal to 
the pole pitch. In a parallel type of winding where the number of 
slots is an exact multiple of the number of poles, the space of the 
coil can be made exactly equal to the pole pitch. In this case if 
the winding can be traced through, the brushes can be so set that a 
coil or turn exactly under the middle of the commutating poles 
has its two ends connected to the two adjacent commutator bars 
which are symmetrically short-circuited by the brush; that is, the 
insulating strip between these two bars should be under the 
middle of the brush. To carry this out properly it is necessary to 
trace the conductor, with absolute exactness, through the slots 
When there are several separate turns side by side in one slot, it is 
advisable to select a middle, or approximately middle, turn for 
determining the brush setting. 

In the case of a 2-circuit or series winding, it is more difficult 
to determine the brush setting by tracing out the coils, for the 
number of slots in such windings is usually not an exact multiple 
of the number of poles and therefore the span of the coil is not 
exactly equal to the pole pitch. In this case the position of the 
coil must be averaged; that is, one edge or half of the coil may be 
slightly ahead of the middle point of its interpole, while the other 
half is slightly behind the middle of the interpole. Even if the 
position of the coil is properly fixed it is not easy to fix exactly 
the corresponding brush setting, as the two commutator bars to 
which the coil is connected do not lie adjacent to each other, as in 
a parallel type of winding, but are two neutral" points apart. Also, 
the number of commutator bars is not an exact multiple of the 
number of poles (except in some rare cases where there is an idle 
bar) and therefore they do not have a symmetrical relation to the 
brushes. The best that can be done therefore is to average the 
brush position as well as possible. 

If the winding is chorded; that is, if it has a span considerably 
shorter than the pole tip, then its position will have to be averaged 
in the manner described above. 

In some cases it is not practicable to trace out the coils in the 
above manner, as the end windings may be so covered that it is 
not possible to trace an individual coil from the commutator to the 


On later machines it is the practice to put a mark or "cross" 
on the tops of two adjacent armature teeth The top conductors 
which lie in the slot between these two teeth are connected to 
commutator bars which are also marked with a cross at their 
outer ends. In this way it is possible to trace from the commut- 
ator to the slots. 

When an interpole generator is running alone, or where it is 
operating properly with other machines, and the commutation is 
satisfactory, it is unnecessary, of course, to look into this question 
of locating the best brush setting. In those cases, however, where 
the machine does not parallel properly with others, and it is 
evident that the brush setting is wrong, then if the above procedure 
cannot be followed, a better brush setting can be found by deter- 
mining the voltage characteristics of the armature This can be 
done by operating the machine with various loads with the series 
winding cut out of circuit. If, under this condition, the voltage 
either rises with increase in load, or droops but very little, then it 
is evident that a greater forward lead would improve the opera- 
tion. The brushes could then be shifted slightly forward and the 
regulation noted. After a brush setting has been obtained which 
gives a considerably larger drop in the voltage, then parallel oper- 
ation should again be tried with this brush setting, the series coils, 
of course, being connected in circuit. After proper paralleling is 
obtained, then it may be necessary to re-adjust the strength of the 
commutating field. If the machine has had a considerable back- 
lead before and is shifted to the no-lead condition, then it may be 
necessary to weaken the interpole winding somewhat. If the new 
brush setting however, should correspond to as much forward 
lead as it had back lead before, then the interpole strength may not 
require readjustment and the commutation may be just as good 
as before. After the proper conditions have been obtained, the 
brush holder position should be marked so that it can be readily 
found again if necessary. 

There is another feature wherein an interpole machine is 
different from the non-interpole type, namely, in the amount of 
series winding. In the non-interpole type the brushes are usually 
given a very considerable forward lead. In consequence of this 
forward lead a part of the armature ampere turns are actually 
effective in demagnetizing the field, and extra series turns are 
necessary simply to overcome this demagnetizing effect, without 
accomplishing any useful result. 


On the interpole type, however, with the brushes set property 
there is no lead at the brushes and therefore none of the armature 
turns are tending to directly oppose the main field. In conse- 
quence of this the number of series turns may be reduced and the 
resistance of the series coils is correspondingly reduced. When 
operating interpole machines in parallel with other types it may 
be necessary to increase the resistance of the series circuit in order 
that the interpole machine may take its proper share of the current 
through the series coils. This result is obtained best, in general, 
by a resistance connected in series with the series coil and not by a 
shunt connected across the series coils of the other machines. A 
shunt across a series coil of one machine is, in reality, a shunt 
across all the machines which are operating in parallel, and it may 
be more effective, in one machine simply because of the resistancr 
of the leads connecting the various machines. These statements 
apply to other types of machines as well as the interpole. 


FOREWORD This paper was prepared for the American Institute 
of Electrical Engineers at the request of the Power Station 
Committee of the Institute. It was presented in January, 1913 
It contains a quite complete description of the two principal 
types of turbo alternator rotors up to that time. In the latter 
part of the paper, it takes up the problems of ventilation, tem- 
perature and insulation from the turbo-generator standpoint. 
Attention was called to the high temperature liable to be en- 
countered in very wide core machines, such as turbo-generators, 
due, to a certain extent, to mechanical limitations. The neces- 
sity for the use of mica in such windings was also brought out. 


THE real problems in the design of turbo-alternators did not 
really develop until the high-speed, large capacity units came 
into demand. In the earlier work, the difficulties in design were 
mostly those due to lack of experience and to insufficient knowl- 
edge of the possibilities of materials, etc. As more data were 
obtained, the speeds and capacities were gradually increased 
until with the present large capacities and high speeds, a number 
of conditions are encountered which may be considered as true 
physical limitations. 

The principal difficulties in the design of the earlier machines 
were found in the permissible weight on bearings, undue noise 
due to the open construction of the machines, and the troubles 
incident to the through-shaft construction of the rotor. 

The bearing problem was eliminated by securing more complete 
data, which showed that the possibilities in this feature had 
hardly been touched upon. 

The solution of the noise problem was largely one of enclosing 
the machine. The noise was practically eliminated, but the 
greater problem of ventilation then developed. 

In overcoming the difficulties of the through-shaft construc- 
tion, the first great advance was made in the direction of larger 
outputs at higher speeds. In very high-speed machines r the 
diameter of the shaft in the rotor core is necessarily small. As 
the overall diameter of the core is comparatively small, it fol- 
lows that, after allowing for the slot depth, and the metal in the 




core necessary to withstand the high rotative stresses, there is 
left but little available space for the shaft. About 600 kv-a. 
capacity at 3600 rev. per min. was the limit with this construction. 
The first great advance in this problem was made by the intro- 
duction of rotors without the through-shaft. By this means, 
the parts of the shaft adjacent to the rotor core proper, could 
be very much heavier 1 than with the through-shaft type, 
and this combined, with the solid rotor core, gave great stiffness 
or rigidity compared with the former through-shaft type. This 
allowed much larger cores, with correspondingly increased out- 
puts. The two-pole parallel slot type of rotor with bolted-on 
shaft construction, as described later, was apparently a leader 
in this respect, due t6 mechanical, rather than electrical, char- 
acteristics, WKen this type had proved to be a successful one, 
the possible capacities of two-pole .3600-revolution, 60-cycle 
machines at once jumped from 600 to 1000 kv-a., and this was 

FIG. 1 

quickly followed by 1500, 2000 and 3000 kv-a. units at 3600 
revolutions. Since then, the increase in capacity at this speed 
has been more gradual, but has been carried up to 5000 kv-a. 
at present, with possibilities of a 6250 kv-a. unit: 

The radial slot type of rotor, also described later, when con- 
structed with its core and shaft in one piece, quickly, followed the 
parallel-slot type in the above growth, and may eventually catch 
up with its only rival in the two-pole, 60-cycle field of construc- 

About the same timfe that the through-shaft type was super- 
seded in the two-pole, 60-cycle machine, a corresponding change 
was made in the two-pole, 25-cyde, and in four-pole .rotors for 
both frequencies, so that, at the present time, practically no- 
designs for the highest speed machines use the through-shaft 
type of construction. This latter, however, has been retained 
in some of the more moderate speed large capacity units. 



On account of the high, rotative and peripheral speeds, the 
general design of large capacity turbo-generators turns upon the 
type and construction of the rotor, rather than the stator. 
Various designs and types of rotors have been developed but, 
with rare exceptions, only two general types are now built in 
this country. These may be designated as the radial-slot and 
the parallel-slot types. Each has a number of advantages over 
its rival and each has given good results in practice. 


In the radial slot type, as usually constructed for high-speed 
machines, the core and shaft are forged in one piece in the Smaller 
and more moderate sizes, but may be built up of a number of 
separate plates or disks bolted rigidly together in the larger sizes. 
In this type, the core is cylindrical in all cases, and in the outside 
surfaces are radial slots, usually arranged in groups, in which the 

FIG. 2 

exciting windings are placed. While all radial slot types of 
rotors bear a general resemblance to each other, yet there are 
marked differences in the method of forming the slots and teeth 
which constitute the outer surface. In some types the solid 
rotor core has radial slots milled or slotted in the main body of 
the core. In other cases the slots are formed outside the main 
core by inserted teeth, usually with overhanging tips, between 
which the exciting coils lie. These two general constructions 
are illustrated in Fig. 3. Examples of the inserted-tooth con- 
struction are found in the large moderate speed rotors of one 
American company, and in somewhat higher speed machines of 
a German company. However, with the advent of the high- 
speed, high-capacity machines, the milled-in construction of 
the radial slots appears to be taking the lead, due to certain 
mechanical limitations in the inserted-tooth types. 

On account' of the radial slots and the usual concentric arrange- 



men! of the exciting coils, the field or exciting turns cannot be 
assembled and insulated before placing on the core, except in 
the inserted-tooth type of construction. With the milled-in-slot 
type, the field conductors, usually of flat strap, are dropped into 
the slot one at a time, with insulation between individual turns. 
For ease of winding, the ends are usually allowed to overhang 
the core, and require a very ample outside support in the very 
high speed machines This is illustrated in Fig. 4. The com- 
pleted coils are usually held in place by strong non-magnetic 
wedges in the tops of the slots. These wedges are usually carried 
by overhanging pole tips, in the inserted-tooth type, or by grooves 
in the sides of the slots in the milled-slot type The design of 
the supports for the overhanging end windings has furnished one 
of the difficult problems in this type of construction Examples 

FIG. 3 

FIG. 4 

of radial slot end windings, and of the rotor complete, are shown 
in Figs. 5 and 6. 

This general construction of the radial slot type of rotor is 
obviously applicable to machines of any number of poles. With 
a two-pole machine there will be only two groups of coil slots and 
two groups of concentric coils, while with four poles or six poles 
there will be four or six groups respectively. It is evident that, 
with this construction, a cylindrical rotor is obtained, regardless 
of the number of poles. It is also evident that the problem of 
supporting the end windings becomes an increasingly difficult 
one, as the number of poles is decreased and the span of the 
end windings is correspondingly increased. 

The support over the end ^findings usually consists of a heavy 
ring which, in very high-spefed machines, must consist of material 



FIG. 5 


having extra good physical characteristics, for this ring must 
not only be able to carry itself, but must also carry the weight 
of the underlying end windings which it supports. In the German 
inserted-tooth rotor, the end windings are supported by steel 
bands of many layers, instead of the solid steel ring. In some of 
the lower speed radial slot machines, such as one American type 
with inserted teeth, the end supports are of ring form usually 
made in a number of sections, which are bolted to an inner shelf 
by numerous bolts extending from the outer ring between the 
coils of the end windings to the inner shelf. While this construc- 
tion is satisfactory for the more moderate peripheral speeds, yet 
with the much higher speeds in some of the later practise, this 
construction has been superseded by a solid ring type of support. 


In the parallel-slot type of rotor, the slots for the exciting coils, 
for any number of field poles, lie in planes parallel to one another 
and to the rotor axis. The arrangement is illustrated by Fig 7. 
As usually constructed, the slots are cut across the ends of the 
poles, as well as in the sides, so that the exciting coils are em- 
bedded in metal throughout their length. The object of this 
general arrangement of parallel slots is to facilitate the winding 
of the exciting coils. The rotor can be placed upon a turn-table, 
or similar device, and rotated, to wind the coils in place under 
tension. Two or more coils can be wound at the same time, as 
is actually done in practice. As the coils can be wound under 
tension, and as the conductors usually consist of thin flat strap, 
which can be wound in very tightly, the resultant winding is a 
very substantial piece of work. The finished winding is sup- 
ported by metal wedges over the coils. 

It is obvious that, with this construction, no external support 
is required for the end windings, as the field core proper furnishes 
the necessary support. It is largely on account of this feature 
of well supported end windings that the parallel-slot type took 
a leading position during the growth of .the larger two-pole, 
60-cycle alternators. With the radial-slot type, the support 
of the end windings presented a more difficult problem in the 
large capacity, high-speed, two-pole machines, which, however, 
is being gradually solved. 

In the two-pole, parallel-slot construction, in order to utilize 
the available winding space to advantage, it is necessary for the 
windings to cover the central portion of the core end where the 



FIG. 7 

FIG. 8 


shaft is usually attached, as shown before m Figs 7 and 8 There- 
fore, with this construction, a separate " head " or driving flange 
must be bolted to the core- at each end, this head carrying the 
shaft, as shown in Fig 8. To avoid magnetic 'shunting of the 
field flux, this driving head must be made of non-magnetic 
material, usually of some high grade bronze, to which the shaft 
is attached in such a way as to keep the magnetic leakage as 
low as possible. This makes a good strong construction, but is 
necessarily rather expensive, due tp the bronze driving heads. 
As these cost but little more for a long rotor than for a short one, 
the construction therefore tends toward relatively long, small 
diameter cores in order to lessen the relative dimensions of the 
bronze heads. 

In two-pole, single phase machines of this construction, the 
copper cage damper for suppressing the armature pulsating re- 
action on the field, is comprised partly of these bronze heads, 
which form the " end rings " for the copper bar& embedded in 
the slots in the rotor face. 

In the four-pole, parallel-slot machine, no bolted-on driving 
leads are necessary, for the core proper and the shaft may be 
:ast, or forged, in one piece, or in two or more pieces, which are 
Doited or " linked " together to form a solid core. The principal 
lifference between the, two-pole and the four-pole parallel slot 
Constructions, is that the latter must have salient or projecting 
Doles, in order to utilize the parallel construction for the slots, 
vhile the two-pole machine is preferably made cylindrical. Fig. 
illustrates this feature. 

It is evident that there is considerable available space lost by 
fie openings between the projecting poles, while the sections of 
he poles themselves are cut down very materially by the slots 
Dr the exciting winding. The limitations therefore in such a 
otor are in the magnetic section of the field poles and in the 
vailable copper space, and in these features the four-pole parallel 
!ot rotor is inferior to the radial slot type. In the two-pole 
lachine, however, the difference between the radial slot and the 
arallel slot is not nearly so pronounced,' as is indicated in Fig. 10 
here the two arrangements are shown on one core for compari- 
)n. It may be seen from this that, in the iwo-pole form, the 
*o constructions approach each other, to a certain extent, 
>me of the slots in the parallel construction being radial, while 
,hers depart but little from the radial. One disadvantage in the 
ro-pole, parallel-slot type, however, lies in the smaller amount of 



copper space which is obtained, for the slot space must necessarily 
cover a less proportion of the total circumference than is permis- 
able with the radial slot type. This winding space is limited by 
the physical requirements as regards bending and breaking 

strains in the overhanging tip a 
in Fig. 10. In the radial slot 
type, the slot space has no such 
limitation. Also, on account of 
the grouping of the field copper 
into a narrower zone in the 
parallel-slot type, the heat con- 
duction from the copper presents 
a more difficult problem than in 
the radial type. 

At first glance, it would appear 
that the effective length of the 
field core in the parallel-slot type 
is very considerably diminished by the slots across the ends 
of the core. However, this is only an apparent effect, for the 
true length of the core should be taken as that inside of the 
winding slots, and it should be considered that the additional 

FIG. 10 

PIG. 11 

length of the core at the pole face is in the nature of a coil 
support which takes the place of the separate support in the 
racial slot type. Therefore, if over-all lengths, including 
rotor coil supports, are compared in the two types, there 
is but little difference, as indicated by Kg. 11. However, 


if the armature core is made of the same width as the pole face 
in both types of rotors, then in the parallel-slot type it will b 
materially greater than in the radial, for the over-hanging pol 
tips of the parallel-slot machine are also eifective magneticall 
in furnishing flux to the armature. Therefore, as regards th 
stator, this tends toward ?i wider core in the axial direction, an< 
a shallower depth of iron back of the armature slots, as indicatfe* 
in Fig. 12. Also, on account of the relatively larger polar sur 
face, in the parallel slot type of rotor, the magnetic flux densit; 
in the air gap is usually rela tively smaller than in the radial slo 
type, which conduces towards a larger depth of air gap. Alsc 
on account of the larger polar surface, the available space fo 
armature slots and teeth is correspondingly increased. There 
fore, this type of construction is better adapted for the straigh 
air gap method of ventilation, as will -be described later. Th 
greater section available for slots and teeth at the stator pol 
face permits a large number of ventilating ducts. The relatively 
large depth of gap allows a large amount of air to be fed througl 
the air gap to the ducts. Therefore, the " radial " type of stato 
core ventilation has been used very largely with this type o 
rotor construction. In the parallel-slot type of rotor, it is obviou 
that, due to the large polar surface compared with the minimun 
section of the field core, a limit in design is found in the magneti< 
saturation in the field core itself. 

In the four-pole parallel-slot rotor, the field section is mor< 
limited than in the two-pole machine, due to the fact that con 
siderable magnetic space is lost by the notches between the pro 
jecting poles. However, in this type of construction, the aii 
gap method of ventilation is relatively easy, due to the fact thai 
these interpolar spaces furnish easy access of the ventilating aii 
to the stator ventilating ducts. In consequence, the problerr 
of ventilation is usually not a serious one in this type of rotor 
Due to the polar projections, however, the tendency to noise is 
obviously greater than in either the radial-slot type or the two 
pole parallel type, which are always cylindrical. 

Nothing has yet been said as to the peripheral speeds obtained 
in some of the actual designs of th;* higher speed generators. 
These, in themselves, indicate some of the limitations which now 
confront the designer. 

In the 5000 kv-a., two-pole, 3600-rev. per rnin., 60-cycle 
generator already referred to, which is of the parallel-slot rotor 
construction, the rotor diameter is 26 in. (66 cm.) This gives a 



FIG. 12 

FiO. 21 


peripheral speed of 408 ft. (124.3 m ) per second, or approximately 
24,500 ft (7468m) per minute The core is designed for .a 
very considerable margin of safety, and is actually tested at 
overspecds which give practically 30,000 ft. (9144 m ) peripheral 
speed at the surface pf the core 

In certain 19,000 kv-a , 62J-cycle, four-pole, 1875-rev per min. 
machines now being built, which are of the radial-slot rotor 
construction, the rotor diameter is 49 in. (124 4 cm ) This gives 
a peripheral speed of 24,000 ft. (7315 m.) per minute. This 
compares with a speed of 21,600 ft. (6583 m.) in a 21,000 kv-a. 
two-pole, 1500-rev per min., 25-cycle, radial-slot machine also 
being built, the rotor core of which is shown in Fig. 12. Obviously 
the mechanical limitations are being more closely approached in 
the 60-cycle machines, up to the present capacities. 

If, a comparison is made between the above 5000 and 19,- 
QOO kv-a. rotors, with their parallel and radial type construc- 
tions, it is found that their limitations lie in quite different 
features. In the radial-slot type, the core stresses are much 
lower than in the rother, but the supporting end ring is an im- 
portant problem, requiring for its solution, a very high grade 
steel for the material of the ring. In the parallel-slot rotor, the 
maximum stresses are in the core itself, principally in the parts 
which overhang the slots at the sides and ends of the core. In 
the radial slot core, there are no such overhanging masses. In 
both construction's, the core material is purposely made of re- 
latively soft steel, having a high percentage elongation, the ob- 
ject being to obtain a material which can yield sufficiently to 
transfer the strains from local higher points, to adjacent lower 
parts, and thus equalize them, to a great extent. 

The smaller diameter rotor cores are made of steel forgings, 
in one piece. The larger cores are made up of thick steel plates 
assembled and bolted together to form a solid jnass comprising 
the core and shaft extensions. By this disk construction, com- 
mercial material is used which is of uniform quality clear to the 
center of the disks. The fiber of the material is in a direction 
best suited to the directions of stress. With corresponding sise 
disks made in one piece, the outside, to a certain depth, can be 
given fair physical characteristics, but the center is liable to be* 
glass-hard, as found by experience. However, this may not 
be a prohibitive condition in machines of more moderate per- 
ipheral speeds. Herein lies one great difference between American 
and European limitations. In American practice, 60-cycles, 


calling for 3600 and 1800-rev. per mm machines, is the standard 
frequency, while in Europe, 50-cycles is standard, giving 3000 
and 1500-rev. per min. machines. These lower speeds make 
an enormous difference in the possibilities of design and construc- 


On account of th very great capacities, at high speeds, now 
being obtained in turbo-generator practice, a number of problems 
are being encountered, the solutions of which are producing 
more or less radical changes, both in design alid in practise. 
Softie of the limitations now encountered are in the relatively 
high temperatures in certain parts, high losses in a relatively 
small space, the difficulty of ventilation, due to the requirement 
of endrmous volumes of cooling air through limited openings 
or- passages, the type of insulation, fire risks, regulation and 
short circuit conditions, etc. 

A number of these limiting conditions, such as the temperature, 
ventilation, losses, and insulation, are so closely related to each 
othfer, that ft is difficult to describe any one of them in detail, 
without including the others to a considerable extent. 


Ip the general problem of ventilation, four conditions must 
be considered, namely, the total loss, or heat, developed, the 
surface exposed for dissipating this heat to the &ir, the quantity 
of air required to carry away the heat, and the temperature 
of the cooling air. 

In the conduction of heat from the surface of a body into the 
air, the quantity of heat per unit $rea which can be dissipated 
depends upon the difference in temperature maintained between 
the surface of the body and the body of air to which the heat is 
conducted. The heat dissipated raises the temperature of 
the adjacent air a certain amount, and thus tends to reduce 
the temperature difference, utoless the air is renewed with suffi- 
cient tapidity. On the other hand, if the quantity of air is so 
great, in proportion to the heat dissipated, that there is but 
little rise in the air temperature, then any increased amount 
of air over the surface will represent practically no gain in 
ventilation. In other words, when the amount of air passed 
oyar a surface is sufficient to take up the heat dissipated from 
the surface without an undue rise, then a. further quantity of 
air is wasteful, and it may even be considered as indirectly 


harmful, in those cases where the total quantity of air is limited 
This has a direct bearing on the size of ventilating ducts or 
passages in a machine. If the air path through a duct is relatively 
long, then a considerable width of duct may be required in order 
to get the necessary quantity of air through it. On the other 
hand, if the air path is very short, then a very narrow duct 
may be most effective, for a wider duct may allow more air to 
pass through than can be utilized in taking up the heat. 

No matter how thoroughly the ventilating air is distributed 
through the heat-generating body, or however effective the 
heat-dissipating surfaces may be, the total air supplied must be 
ample in quantity, or its temperature will be raised an undue 
amount. As the surfaces to be cooled must always have a 
higher temperature than the cooling air, any considerable rise 
in the latter will have a direct influence on the ultimate tempera- 
ture which may" be attained by the body to be cooled. Con- 
versely, if an ample quantity of cooling air is supplied, but the 
heat-dissipating surfaces are insufficient, the ultimate tempera- 
ture of the body will also be affected. 

In large capacity, high-speed turbo-generators, the problem 
of ventilation is one of the most difficult ones encountered, The 
trouble lies principally in the large total loss expended in a very 
limited space. The difficulties of the problem may be illustrated 
by the following example : 

Assume, in a 1500-rcv. per min , 25-cycle, 15,000-kv-a 
machine, a total efficiency of 96 5 per cent, including air friction 
loss inside of the machine. This means a total loss in the machine 
of 565 kw., which is not excessive for this capacity, but is very 
large for the limited space in which it is developed A very 
large volume of cooling air is required for carrying away the 
heat due to this loss. A simple approximate rule for determining 
the quantity of air required is that an expenditure of one kw, 
in one minute will raise the temperature of 100 cu. ft (2.8 cu. 
m.) of air 18 deg. cent. Therefore, 565 kw. loss would require a 
supply of ventilating air of approximately 50,000 cu.'ft. (1416 
cu. m.) per minute for a rise of the out-going air of 20 deg. above 
that of the incoming air. Assuming a velocity of 3000 ft (914 m,) 
per minute, this would mean, with a cylindrical ventilating 
channel, a diameter of 56 in, (142.2 cm ), which is greater than 
the rotor diameter itself. However, as the cooling air ordinarily 
\vould be supplied to both sides of the machine, the ventilating 
passage need only be half the above section for each side. 


Obviously, such passages arc prohibitively large, and much 
greater air velocities through the machine proper are necessary 
Velocities as high as 5000 to 6000 ft. (1524 to 1828 m ) per 
minute are common, while, in some cases, more than 10,000 
ft. (3048 m.) per minute has been required m certain constricted 
sections of the air path inside the machines Therefore, no 
matter how the problem is considered, it may be seen that 
the above condition of the enormous volume of air required, 
makes the problem, of ventilation a difficult one 

There are several methods of ventilating large turbo generators, 
depending upon the system of applying the air. There is, first, 
the radial system, in which practically all the cooling air passes 
out radially through ventilating ducts in the stator core. This 
radial system of ventilating can be subdivided into two alterna- 
tive methods, depending upon whether the air is partly or 
wholly supplied through passages in the rotor, or through the 
air gap alone. These two methods are illustrated m Fig 13 
The straight air gap arrangement may require a relatively large 
air gap, combined with very high velocity of the air along the 
gap, while the other method permits a considerably shorter gap. 
The straight air gap method of ventilation is used, to a 
considerable extent, in all 60-oycle machines of two-pole con- 
struction, while it is practically the only one that has been used 
with the parallel-slot type of machine with either two or four 
poles In this parallel-slot type of rotor, however, the air gap can 
be relatively larger than the radial-slot type of rotor, as explained 
before, which compensates, to some extent, for the necessity of 
depending upon this method entirely. In the four-pole parallel- 
slot rotors, the interpolar spaces are also effective. Moreover, 
with parallel-slot rotors in general, the openings from the air 
gap into the stator ventilating ducts can usually be somewhat 
larger in total section than with the radial type of rotor, as also 
described before. However, the relatively greater axial length 
of the core of the parallel slot type of rotor increases the length 
of the constricted air passages along the air gap in the two-pole 
machines, which is a material disadvantage 

The straight air gap type of ventilation has proven astonish- 
ingly effective in cooling the rotor in both the radial and parallel- 
slot types of rotors, and with either type there is usually no 
great difficulty in forcing through enough air to cool the rotor 
core in a fairly effective manner. It must be considered, however,, 
that the total rotor loss in large turbo-generators is possibly 



only 10 per cent of the total loss which must be taken care of, and 
a relatively small proportion of the total ventilating air may 
suffice to cool it. According to actual measurements, corrobor- 
ated by general experience, the cylindrical surface of the rotor 
core can give off four or five watts per square inch (6 45 sq. 
cm.) to the cooling air, with a temperature rise of the rotor 
surface of about 35 to 40 deg. cent, above the cooling air. To 
those who have had experience with dissipating heat from electric 
apparatus, this result will appear to be extremely good. 

The real difficulty with the air gap method of ventilation, 
is not so much in getting enough air through for cooling the 

FIG. 13 

FIG. 14 

rotor itself, but it is in the much larger quantity required for 
the stator. For instance, a one-inch (2.54 cm.) depth of gap 
from iron to iron) with a 50-in. (127-cm.) diameter of rotor, 
means a total section of air path into the gap (counting both 
ends of rotor) of 314 sq. in. or 2.18 sq. ft, (0.19 sq. m.)- At a 
velocity of 10 000 ft. (3048 m ) per minute, this allows a flow 
of only 21 800 cu. ft. (617 cu. m) per minute, which will not 
take care of a large machine, from the present standpoint of 
possible capacities with the above diameter of rotor. By 
additional openings in the rotor core, this might be increased to 
30000 cu. ft. (849 cu. m.) per minute, but even this is still 


much less than a machine, with a 50-in. (127-cm.) diameter of 
rotor, would require if built for capacities otherwise possible. 
Therefore, on account of this limitation in the amount of cooling 
air, other means of ventilation have received much considera- 
tion. Two other general systems of ventilation, in addition to 
the gap method, have been used, namely, the circumferential 
method, and the axial The former has been developed and 
applied more extensively in the past, but the latter contains 
possibilities which are bringing it rapidly to the front. 

In the circumferential method of ventilation, air is supplied 
to-one or more points on the outside circumference of the stator, 
and is forced circumferentially around through the air ducts to 
suitable outlets, also on the outside surface. Air gap ventilation 
is usually combined with this circumferential method, partly to 
cool the rotor. The general arrangement is indicated diagram- 
matically in Fig. 14, in its simplest form, namely, with one inlet 
and one outlet diametrically opposite. A serious objection to 
this method of ventilation is found in the limited section of the 
ventilating path. Assuming, for example, a depth of stator 
core of 20 in. (50.8 cm.) outside the armature slots and a total 
of 40 f-in. (9.5 mm.) ventilating ducts, or a total effective duct 
space of 15 in. (37.1 cm.) width then this gives a total section 
of ventilating path of 20 X 15 X 2 = 600 sq. in., or 4. 16 sq. ft. 
(0.386 sq. in.). On account of the relatively great length of the 
ventilating path, air velocities of more than 6000 to 7000 ft. 
(1828 to 2133 m.) are not desirable or economical, but even 
with 10,000 ft. (3048 m.) velocity, the total quantity of air 
would be only 41,600 cu. ft, (1166 cu. m.) per minute. Further- 
more, this method is handicapped in machines with very high- 
speed rotors, by interference between the radial and the cir- 
cumferential systems of ventilation, so that the full benefit of 
either is not obtained. Below a certain rotor velocity, apparently 
the circumferential action can predominate, and the method is 
fairly effective up to the permissible air capacity of the stator 
ducts ; but at very high speeds the radial ventilation may very 
seriously interfere with the other, so much so, that the radial 
ventilation alone, even with its very restricted gap section, may 
give as good results as the two methods acting together. 

To avoid this interference, various methods have been devised, 
such as closing part, or all, of the radial ventilating ducts at the 
air gap to keep the radial effect from interfering with the other. 
One arrangement which has been used in Europe to a considerable 


extent is indicated in Fig 15. In this, the alternate radial air 

ducts are closed at the outside surface, while all are closed at 

the air gap- The air enters by the ducts open at the back of 

the machine, flows both circumferentially and toward the' gap, 

and crosses over to the immediate ducts by axial opeiiings back 

of the armature teeth, and then along these ducts to the outlet. 

This scheme is effective in principle, but is uneconomical in the 

sense that less than the total section of stator duqts is useful, 

as regards the quantity of air which can be carried. There is 

usually one large central duct to allow an outlet for the rotor 

ventilating air. This particular arrangement of the stator also 

uses axial ventilation in crossing over from oiieset of ducts to 

the other, which is an effective arrangement. 

A modification of the simple circumferential method of 




FIG. J5 

ventilation is to admit air to the back of the stator at two oppo- 
site sides of the machine, and deliver it at two outlets at inter- 
mediate points on thej surface, as shown diagrammatically in 
Fig. 16. By this means, the cross section of the ventilating 
path is doubled and the length is .halved. Also, the interference 
of the radial ventilation with the circumferential will be less 
harmful. A serious disadvantage in the Circumferential venti- 
lation in general is that the ventilating path is relatively long, 
especially where there is but one inlet and outlet, and therefore 
the cooling air at the outlet of the channel may be considerably 
hotter than at the inlet, with consequent less effective cooling 
action. This means points of local higher temperature in the 
core, due to the method ,of ventilation. In the radial type of 
ventilation, the coolest air is applied near the seat of the highest 
losses, namely, at the armature teeth, and immediately back of 



them, and the air, as it becomes heated, passes over the outer 
part of the iron which has a diminished loss, and therefore 
normally less heat to dissipate. Therefore, the effect of the in- 
creased temperature of the cooling medium is offset by the lower 
Loss and .consequent less necessity for ventilation, in the part 
where the air is hottest. The radial system of cooling is therefore 
theoretically the most effective, but practically, the difficulty 
s in applying it, due to the limited air passages available. 

BoUk the circumferential and the radial methods of cooling 
arre subject to one serious defect, namely, most of the generated 
heat in the stator iron must be conducted across the lamina- 
tions to the air ducts. The rate of conduction across the" lamina- 

FIG. 16 

tions is only from 1 per cent to 10 per cent as 'great as along the 
laminations themselves, according to various authorities. 
Therefore, if the heat could all be conducted along the lamina- 
tions to the ventilating surfaces, apparently much more effective 
heat dissipation could be obtained, provided sufficient surface 
is exposed to the air, and an ample quantity of air supplied. 
This has led to the development of the axial system of ventila- 
tion, as distinguished from the radial and circumferential. 
In this method, a large number of axial holes are provided in 
the stator core which may extend uninterruptedly from one 
side of the core to the other, or they may extend from each side 
to one or more large central radial channels which form the outlet. 
The usual numerous radial ducts are omitted, or may be con- 


sidered as combined in one central channel. This general 
arrangement is illustrated in Fig. 17. The rotor cooling is 
accomplished by air along the air gap, and through the rotor 
core to the large central duct. In this method of ventilation 
therefore, there is a combination of two types, namely, the axial 
and the air gap y but there is not the interference between the 
two, that is sometimes found where the circumferential method is 

From the preceding, it may be seen that the problem of 
putting a sufficient quantity of air through the machine is an 
extremely difficult one. In addition, in very large machines, 
the problem -of supplying the required quantity of air from a 
suitable blower forms another serious problem. In smaller 
capacities, and in slower speed machines, it has been the usual 
practise to attach blowing fans to the rotor shaft or core, as 


FIG. 17 

part of the outfit. There is no particular difficulty in this 
arrangement, except, possibly, in the high-speed construction 
of the fans required for 60-cycle, two-pole machines. Such 
fans can supply an amount of air whieh is limited by the diameter 
and other dimensions of the fan itself. 

Assume, for example, that by lengthening the rotor core, or 
by other modifications in the construction, the capacity of the 
machine can be doubled, and therefore double the quantity of 
air is required for cooling. If the limit of the fan design or 
operation was reached before, then obviously some radical 
change is required with the new capacity of the machine. This 
condition apparently has been reached in some of the later 
practise in large, high-speed turbo-alternators. One ot>vious 
solution of this difficulty lies in the use of separate slower speed, 
large diameter, fans or blowers. This may appear to be a step 
backward, but when the above conditions and limitations are 


taken into account, it is not so. The " tail " must not be 
allowed to '"wag the dog;" the blower, which is an adjunct, 
must not be>allowed to dominate the construction of the machine 
itself. Moreover, there are a number of meritorious features in 
the use of a separate blower. In the first place, it can be made 
somewhat more efficient than the high-speed, rotor-driven fans. 
Again, with a suitable means to drive, variable speeds, and 
therefore different air pressures, can be obtained. This feature 
may prove to be very desirable or advantageous under peak, 
or overload, or emergency conditions. 

One further condition keeps cropping out in the general 
problem of ventilation, -namely, that of filtering or washing, or 
otherwise cleaning the ventilating air With 50,000 to 75,000 
cu. ft. (1415 to 2122 cu. m.) of air per minute passing through 
a large machine, obviously in a year's time, an enormous quan- 
tity of foreign matter is carried through the machine with the 
ventilating air. A deposit of a very small per cent of this in the 
machine will probably be disastrous. In fact, however, the 
high velocity of the air through the machine serves to keep the 
air passages clear if no oil or moisture is allowed to enter. That 
a large amount of foreign matter does go through the machine 
is very soon shown in case a little oil is allowed to get into the 
ventilating passages. This oil catches the dirt and in a short 
time the ventilating passages may be very materially obstructed. 

On account of the deposit of dust, etc. in the ventilating 
passages, it is necessary to clean certain types of machines at 
more or less frequent intervals, and it is advisable to clean all 
types occasionally. With some systems of ventilation, where 
such cleaning is difficult, or almost impossible, such as that 
shown in Fig. '16, provision must be made for cleaning the air 
before it enters the machine. With the particular construction 
shown in Fig. 16, air filters are almost always supplied. In the 
American types of construction, however, such filters have not 
yet been used, except in a< more or less experimental manner, 
due probably to the greater accessibility of these machines as 
regards cleaning. But such filtering processes possess consider- 
able merit in general. One modification which is being agitated 
at present i? that of washing, instead of filtering, the air. This 
serves the double purpose of cleaning and cooling the air, and 
in very hot weather, when the available capacity of the machine 
is at its minimum, this cooling effect may mean a reduction of 
6 to 10 deg. in the tetnperatuxe 6f the machine. 



In the general problem of temperatures in electrical apparatus, 
it is not the rises, but rather the ultimate or limiting temperatures 
which are of first importance. Furthermore, the real limitation 
in ultimate temperature does riot lie in the copper and iron, 
but in insulating materials used; and only insofar as the tem- 
peratures of the former affect the latter do they concern the 
general problem. However, as insulating materials in themselves 
are not usually sources of heat, but as they receive most of their 
heat from adjacent media, such as iron or copper which may be 
generating loss, the real temperature problem, as regards insula- 
tion, resolves itself into the consideration of that of the adjacent 
materials. Therefore, it is one which, for its full analysis, 
requires a knowledge of the sources and amounts of heat gener- 
ated, and its conduction and distribution to other parts 

Broadly speaking, there is always a flow of heat from points 
of higher to those of lower heat potential and the amount of 
flow is also a function of the quantity of heat generated, the 
section and length of the paths through which it can flow, and 
the specific heat resistance of the various materials which con- 
duct the heat. In an electric generator, for example, heat is 
generated in large quantities in the armature teeth and in the 
armature core It is also generated in the armature coils when 
the machine is carrying load Part of the armature copper is 
buried in the armature slots where it is almost surrounded by 
iron, which, in itself, develops a loss, while another part, such 
as the end windings, may be surrounded by, and thoroughly 
exposed to, the ventilating or cooling air In such end portions, 
the flow of heat will usually be from the inside copper, directly 
through the insulation to the cooling air The amount of heat 
which will flow from the copper through the insulation, depends 
upon the temperature differences between the copper and the 
outside surface of the insulation, upon the cross section of the 
path of flow, upon the thickness and 4 ' make-up " of the material, 
and upon the heat-conducting properties of the insulation itself 
There is also a considerable temperature gradient from the 
outside surface to the air. If the surrounding air is not renewed 
with sufficient rapidity, the flow of heat from the insulation to 
the air may raise the temperature of the adjacent air, so that 
the total temperature drop is decreased, and the amount of heat 
dissipated is correspondingly reduced. 

In the armature core, the problem is much more complex 


In the copper buried in the armature slots, there are usually 
three paths along which the heat can flow. First, it may flow 
from the copper directly through the insulation to the fron, 
-provided the adjacent iron temperature is lower than that of 
the copper. Second, it may flow lengthwise of the copper to 
the end windings to be dissipated directly into the air from that 
portion ot the winding, as described above Third, in the case 
of open-slot machines, one edge of the coil may be exposed to 
the air in the air gap, and there may thus be a direct conduction 
of the heat through the insulation to the air in the air gap. This 
latter case, however, only holds for the upper coil, or that next 
to the gap, in the case of two coils per slot, which is the most 
common construction In the bottom coil, the only means of 
conduction in the buried portion of the coil, are to the adjacent 
iron or lengthwise to the end windings, or to the adjacent upper 
coil, which, however, would normally have at least as high 
temperature as the lower coil Therefore, the two effective 
paths should be considered as through to the iron and thence 
to the air, and lengthwise of the copper to the end windings and 
to the air. It is the relation of the various factors of these two 
paths that control the actual temperatures. 

It has usually been considered that, in the buried copper, the 
greater portion of the heat is conducted directly into the sur- 
rounding iron. This, however, is only partially true, depending 
upon many features in the construction and type of apparatus, 
The heat conductivity of copper is, roughly, about six times that 
of laminated iron lengthwise of the sheet, which is possibly ten 
to twenty times as great across the laminations. In an armature 
which is comparatively narrow and which has very open, well 
ventilated end windings, a relatively small difference in tempera- 
ture between the copper at the center of the core and that in the 
end windings, may cause a relatively large flow of heat from the 
buried copper to the end copper. Therefore, in certain design?, 
a great part of the armature copper heat may be dissipated 
through the end windings, and not through the armature core, 
especially in those cases where the armature core in itself has a 
considerable temperature rise. There even might be no con- 
duction of heat from the copper to the iron, or there may be 
conduction from the iron to the copper; for it the copper is at 
the same temperature as the iron at the center of the core, for 
instance, then at each side of the center, or as the edges of the 
core are approached, the copper temperatures will be relatively 



lower than at the center, and therefore lower than the adjacent 
iron, on the assumption that the iron temperatures would be 
practically constant over the full width of the core The con- 
ditions would therefore be as represented in Fig. 18. The solid 
line a in this figure represents the iron temperature at uniformly 
40 deg. cent, rise, and the dotted line b represents the copper 
temperatures ir6m the center of the core to the edges. Th* tem- 
peratures at tne center being assumed the same for copper and 
iron, obviously there will be a flow for heat from the iron to the 
copper near the edges or the core. The effect of this additional 
heat carried out by the copper would be such as to tend to increase 
the temperature of the copper at the center of the core by " bank- 
ing up f ' the copper heat. 

Again, if the temperature of the copper at the center is materi- 


FIG. 18 

FIG 19 

ally higher than that of the surrounding core, the conditions 
may be as represented in Fig. 19 In this case, assuming the 
core at constant temperature, there will be heat flow from the 
copper to the iron at the center of the core, and from the iron to 
the copper at the edges 

This study of the problem leads to certain very curious con- 
ditions which are sometimes found in large machines. At no- 
load, for instance, with practically no copper loss present, and 
with high iron loss, there may be a very considerable flow of 
heat from the armature teeth through the insulation into the 
copper, and thence to the end windings and to the air. In this 
way the temperature -of the armature teeth at no-load, and with 
normal voltage generated, may be considerably reduced by con- 
duction of the iron heat into the copper, while the copper itself 


may show a very considerable temperature rise. When load is 
placed upon such a machine, sufficient to raise the temperature 
of the copper up to that of the iron in the armature teeth, the 
latter is actually increased in temperature, due to the prevention 
of the heat conduction into the copper. In this way, therefore, 
the copper may apparently heat the iron, although there is no 
direct flow of heat from the copper to the iron, but the reverse 
flow is prevented. 

In high-voltage windings requiring thick insulation, the temp- 
erature drop from the copper to the outside may be relatively 
large; that is, with a given difference of temperature between 
the copper and the surrounding air," a relatively small amount 
of heat may be conducted through the insulation. Experience 
shows that the amount which can be conducted is a function of 
the quality of the material, the way it is built up, its thickness, 
and also the pressure upon it. It is almost impossible, in a 
machine in service, to calculate exactly the flow of heat, even if 
all the temperature conditions are known, for the insulating 
material itself is one of the variables in the problem The ability 
of the insulation to conduct heat will change with operating 
conditions, to some extent, as, for instance, it may tend to ex- 
pand somewhat under heat, and thus change its heat conducting 

In the armature iron, the problem of heat conduction is just 
as complicated as in the armature conductor. The principal 
sources of heat lie in the armature teeth and in the armature 
core back of the teeth. As a rule, the loss in the portion of the 
core immediately back of the teeth is relatively greater than at 
a greater depth, for the magnetic fluxes which cause the tempera- 
ture rise, generally crowd close to the teeth, so that the density 
is higher at such parts. 

The heat from the armature teeth can be dissipated along 
several paths. It can flow lengthwise of the laminations to the 
end of the tooth and into the air gap, where the ventilation is 
usually fairly good, but the tooth surface exposed is relatively 
small. In the second place, it can flow back along the lamina- 
tions to the armature core where it can spread out through a path 
of much greater cross section and be conducted partly to the back 
part of the laminations, and partly transversely to the ventilating 
ducts. A third path from the armature teeth is across the 
laminations of the teeth, to the neighboring ventilating ducts. 
This latter path, however, must necessarily be relatively poor 


in conductivity per unit section of path, compared with the 
others, but offsetting this, it is frequently of much greater cross 
section and of relatively small length. In passing from plate 
to plate, the heat must pass through the insulating varnish, 
or other material used, which is of relatively high heat resistance 
compared with the iron itself. Nevertheless, in machines with 
radial ventilation, a very considerable portion of the heat due 
to the tooth loss is carried transversely thrpugh the plates to 
the air in the ventilating ducts, simply because that is the path 
of lowest total heat resistance, everything considered. In mariy 
cases, the temperature in the core back of the teeth may be as 
high as that of the teeth, themselves, so that the only flow 
possible is across the laminations to the air ducts, or lengthwise 
to the tip of the teeth in the air gap. Therefore, the question 
whether the armature teeth may be hotter than the armature 
core, or whether the flow of heat is from the teeth to the core, 
or from the core to the teeth, is a very involved one; and yet 
upon this question depends, to a great extent, the temperature 
rise in the buried armature copper. If the armature core is 
normally hotter, than the teeth, and a considerable amount of 
heat in the teeth is carried away by the buried copper at no 
load, then it may happen that when carrying heavy load, the 
heat in the teeth will rise very considerably above the no-load 
condition, and it may actually so " bank-up " that there is still 
more or less flow from the iron to the copper, eVen with load. 
With such a condition, therefore, the outside of the insulation 
may reach a higher temperature than the inside, while in those 
cases where the temperature of the copper rises above that $f 
the iron of the armature teeth, the inside of the insulation 
will be hotter. Therefore, the temperature to which the insula- 
tion is liable to be subjected appears to be largely a problem 
for the designer to determine from his- calculations, based upon 
accumulated data and experience. This is especially the case 
with very wide armature cores and large, heavily insulated 
armature coils, such as found in large capacity, high speed 
turbo generators. In such machines, experience has shown 
that various temperature conditions may be found, depending 
upon the location and relative values of the losses in the different 
parts and the means for conducting away the heat. Tests 
have shown that, rq. some cases, the armature iron at the center 
of the core is considerably warmer than the armature copper, 
while in other cases the opposite is found to be true. 


In such apparatus, the temperatures actually obtained are 
liable to be materially higher than the usual methods of measure- 
ment will indicate. These temperatures are inherent to the 
conditions of design and cannot be avoided economically, in 
certain types of apparatus, such as turbo-generators. In such 
machines, the limitations in speed, strength of material, etc. 
force the designer to certain proportions which preclude larger 
dimensions, or lower inductions in the iron, or lower densities 
in the copper, or increased ventilation. In such apparatus 
therefore, the development apparently lies in the direction of 
insulations which will stand the higher temperatures which may 
be' obtained. 

These conditions of higher temperatures in some parts of 
the machine, than indicated by the usual tests, have been 
recognized for years by designers and manufacturers of large 
electric machinery. A rough indication of these temperatures 
can be obtained by exploring coils or thermo-couples suitably 
located. However, it is evident that such coils, if located next 
to the copper, will not give the correct temperature measurement 
if the flow of heat is from the iron to the copper, while a coil 
next to the iron will not give the correct result .with the flow 
from the copper to the iron. Experience has shown that the 
temperatures, in corresponding positions around the core, may 
not be uniform, due to local conditions. In consequence, it is 
not practicable to actually determine the true temperatures of 
all parts of the insulation on commercial machines, except by 
measurements of a laboratory nature, which wtould involve such 
a number of separate readings as to be commercially prohibitive.' 

On account of the higher temperatures which may be found 
in such apparatus, and the difficulty of making exact measure- 
jmcnts, except by laboratory methods, manufacturers very 
generally have adopted the use of mica as an insulating material 
on the buried part of the coils. Experience has shown that such 
material, when properly applied, can safely stand temperatures 
of at least 125 deg. cent. How much more has riot yet been 

Of such machines it may be said that the manufacturer, with 
his 'guarantee of 40 deg. cent, by thermometer, actually builds 
for possible temperatures of 70 to 90 deg, cent, in some parts of 
the machine, for he expects to find fairly high temperatures in 
some cases with exploring devices. The usual guarantee trf 40 
deg. cent, therefore should be considered as only a relative indi- 
cation of a safe temperature in such apparatus. 


If, for instance, the exploring coils should show 70 deg. cent, 
maximum rise under running conditions, and the permissible 
ultimate temperature of fibrous or tape insulation is assumed 
as 90 deg. cent, for continuous operation, then obviously, with 
air at 40 deg. cent, the insulation would be considered as insuffi- 
cient from point of durability, except for intermittent service, 
such as overloads, and such limited conditions. Plainly, the 
insulation, for such temperatures, should be of mica, or equiva- 
lent material, for which 125 deg. cent, has been found to be safe. 

Furthermore, it may be stated that with such mica insulation, 
a turbo generator which shows 75 deg. cent, rise by exploring 
coils, or thermo-couples, has, in fact, more margin of safety 
than the ordinary varnished-tape-insulated low-voltage machines 
of any type, which show 40 deg. cent, rise by thermometer or 
SO deg. cent, rise by resistance. 

The foregoing aims to bring out clearly that the temperature 
problem is a most complex one, in all electrical apparatus, and 
especially so in turbo-generators. It indicates that no simple 
temperature test can show all the facts, and that all commercial 
methods must be considered as approximations. It also shows 
the absurdity of classifying a piece of apparatus as good or bad, 
respectively, according to whether it tests possibly one or two- 
degrees below or above a specified thermometer guarantee. 
Also, following out the above principles on heat flow, various 
fallacies in temperature measurements might be noted. For 
example, it is usually assumed that, after shutdown, if a grad- 
ually rising temperature is shown, this is a more accurate indica- 
tion of the true temperature. But this may be entirely wrong as 
regards windings. If, for instance, the core back of the armature 
slots is materially hotter than the armature teeth while carrying 
load, then, upon shut-down, with the air circulation stopped r 
the teeth will rise to approximately the same temperature as the 
core back of the teeth, and there may be a flow of heat into the 
coils, which condition may not have existed while carrying load. 
A thermo couple on the coil or in the teeth would thus indicate 
a false temperature rise after shut-down. This is cited simply 
as one of many instances, to show the possibilities of entirely 
wrong conclusions which may be reached in the problem of 


The one fundamental condition which must be considered in 
the insulation problem, is the durability of the material itself, and 


this must be viewed from two standpoints, the mechanical, and 
the electrical. From the mechanical standpoint, the material 
may have its insulating qualities impaired by the action of 
mechanical forces which tend to crack, or crush, or disrupt the 
material itself, or it may be affected by being permeated by 
foreign materials or substances, or it may be injured by such 
overheating as will partially or wholly carbonize it, or render it 
brittle or otherwise unsuitable for the desired purpose. 

From the electrical standpoint, it may be weakened by 'deter- 
ioration of the quality of the insulating material itself or some 
of its component parts, which may be due to heating; or oxida- 
tion, or many other causes. 

The effect of mechanical injury, such as cracking crushing 
or overheating, on the insulating qualities, will depend upon 
many conditions. In some cases, with relatively low voltage, 
any effective mechanical separation of the parts is sufficient for 
electrical purposes. For higher voltages, continuity of the separ- 
ating insulating medium is necessary. 

Experience has shown that, for moderate voltages, tempera- 
tures which may injure, or even ruin, the insulating material, 
from a mechanical standpoint, may not seriously affect its 
insulating qualities. Many insuhating materials of a cellulose 
nature will still retain good insulating qualities if maintained 
at temperatures as high as 150 deg. cent, for such long periods 
that the material itself semi-carbonizes. Under such high 
temperature conditions, however, it becomes structurally bad, 
that is, it may become so brittle that it tends to crumble, or 
powder, or flake off, and thus its value as an insulation is im- 
paired by displacement of the material itself. In low voltages, 
therefore, it is not a deterioration in the insulating qualities, 
but rather a mechanical breakdown of the material itself, which 
is liable to cause trouble. With high voltages, however, the 
conditions may be quite different. With some insulating mater- 
ials, the dielectric strength may be so affected by long continued 
high temperatures that the insulating quality becomes insufficient. 
This has a direct bearing on large capacity, high-voltage turbo- 

In the problem of insulation, certain difficulties have been 
encountered in large turbo-generators, which, while they would 
have developed eventually in other large machines, yet became 
apparent more quickly and preeminently in the turbo type, due 
to the abnormal conditions in its design. The two most promi- 


nent difficulties were, first, that of relatively high temperature 
in the buried copper, already described, and second, the destruc- 
tion of the insulation by reason of static discharges between the 
coils and the armature iron 

Due to the fact that the ultimate temperature reached in such 
machines not infrequently exceeds the safe limits for insulation 
of the fibrous or cellulose type, such insulations will show 
deterioration eventually in their insulating qualities and their 
durability. In consequence, with the advent of the larger 
machines, it became necessary to return to the use of mica for 
insulating purposes on the buned part of the coil This type 
of insulation in the form of mica wrappers, had been used 
extensively on some of the earlier large capacity, slow-speed 
generators, but it had not been adopted on large turbo-generators, 
due principally to the difficulty in applying the very long 
wrappers for the straight part of the coil. However, when the 
gradual deterioration of the fibrous type of insulation was noted 
in large turbo-generators, the mica wrapper type of insulation 
was again taken up and, after considerable experiment, was 
applied successfully for the outside insulation on the straight 
parts of the coils This use of mica overcame the deterioration 
in the insulating qualities of the outside insulation, but for 
a while it was considered that a fibrous type of insulation was 
still effective between turns in those coils where there were 
two or more turns in series per coil. As stated before, the 
insulating qualities of many fibrous materials will stand up 
astonishingly well under low voltages, when the material is 
apparently so greatly heated that it is practically carbonized, 
Therefore, temperatures which did not carbonize, but simply 
browned, or darkened, the material, had not been considered 
dangerous, and undoubtedly many thousands of electrical 
machines of all kinds are today in operation, in which the 
insulation is in this condition, and in which no trouble need be 
expected. For this reason, little or no trouble was expected 
between turns on the turbo-generators. However, a new con- 
dition was encountered in large capacity machines, namely, the 
insulation between turns, when it became dry and brittle at the 
higher temperatures, was liable to be injured by the terrific 
shocks to which the coils were subjected in such machines, in 
case of a short circuit across the terminals. The insulation would 
be cracked, or so distributed that short circuits would occur later, 
without apparent cause. These short circuits on large machines, 


most often appeared as breakdowns to ground, even with the 
mica wrapper insulation on the outside of the coil Incidentally, 
several cases were discovered where arcs had occurred inside 
the coils between adjacent turns, and where they had not yet 
broken through the outer insulation to ground. For many 
months the writer, with his associates, followed up this matter, 
examining all available coils and windings. Eventually the 
conclusion was reached that many of the breakdowns to ground 
had actually started between turns on the inside of the coil. 
Moreover, as a corroboration, it was noted that in machines 
with one conductor per coil, the breakdowns were practically 
negligible. This investigation led to the practise of insulating 
the individual turns, in each coil, from end to end, with mica 
tape. After the adoption of this practice, it is noteworthy that 
the breakdowns to ground practically disappeared, although 
the outside insulation to ground had not been changed in type 
or thickness. 

Many improvements have been made in recent times in the 
application of this mica insulation. One of these is the Haefely 
process, developed in Europe, but now being used extensively 
in this country. By this process, the mica wrappers are s'o 
tightly rolled on the cofl that practically a solid mass of insula- 
tion, of minimum thickness and greatest heat conductivity is 

By means of the mica insulation, the temperature difficulties 
in general have been entirely overcome, and a durable and non- 
deteriorating construction, from an insulation standpoint, has 
been obtained with the temperatures which appear to be more 
or less inherent in the large, high-speed turbo-generators. 

The second trouble, namely, that due to static discharges 
between the annature copper and the iron, was also encountered 
to a certain extent, on some of the earlier machines. It was found 
that these discharges were apparently " eating " holes, or even 
grooves, through the outside insulation of the armature coils. 
This effect was most pronounced at the edges of the air ducts 
and at the ends of the armature core, where edges were presented 
by the iron. After a long period, the holes or grooves would 
become so deep that the insulation was weakened or ruined. 

This was a very disturbing condition, when it was once fully 
recognized and appreciated. Again, a comprehensive investi- 
gation was made to discover a cure for this difficulty. Various 
types of machines and windings were examined. It was noted 


"that the action was a function of the voltage of the machine, 
"but was noticeable, in some cases, at relatively low voltages. 
During the course of the investigations, it was noted that where 
mica, wrappers were used with an outside layer of tape, the " eat- 
ing away " extended only through the outside wrapping in as 
far &s the mica, and that no 'apparent effect at the tnica was 
visible. Even when examined , with a very powerful microscope, 
no evidence of any puncture of the mica was found, in any case 
These investigations naturally led to th,e- conclusion that the 
#iost suitable remedy for the trouble was the use of mica insula- 
fton, which was also a remecly for the temperature conditions 
This is one of the rare cases in large turbogenerators where two 
desirable conditions do not conflict With feach other. The rnica 
insulations on the buried part of the coil has now been very 
generally adopted in this country on high-voltage ipachines, 
whether of- the turbine-driven, or aiiy other type 

This, static trouble was considered so serious at one time that 
low voltage practice with step-up transformers was adopted 
by some manufacturers as the safest course, until something 
positive in the way of a femedy was proved out. This trouble 
promised to be one of the most serious encountered in high- 
voltage generator work, and even threatened to revolutionize 
practise in winding generators for the higher voltages. However, 
as consistently advocated by the writer, the us of mica, suit- 
ably applied, appears to have entirely overcome this trouble, as 
evidenced by several year's experience, and 'all indications now 
are that there need be no fear from static discharges on windings 
of 11,000 and 13,000 volts, Even in the 11, 000- volt New Hatfejn 
generators with one terminal grounded, which gives the equiva- 
lent of a I9 t OOO-volt T three-phase winding with the jieutral 
grouiided, the mica insulation appears to-be successful and dur- 


In most of the early turbo-generators, the rotor winding 
in the slots was insulated with fibrous material!'" fish paper " 
and " horn M fiber having the ^reference. 'One of the difficulties 
in the rotor is, that the insulation between the winding and the 
slot is liable to be crushed or cracked by the higlji centrifugal 
forces. In the earlier insulatibns, before fish paper was used, it 
was found that even at very moderate temperatures, the insula- 
tion got dry and brittle, and cracked readily. Fish paper, or 
horn fiber, was then adopted pretty generally. Such material 


apparently stood much higher temperatures than the ordinary 
fibrous insulations. However, experience also showed that 
eventually this also became brittle, arid was liable to be cracked, 
and then displaced, due to the centrifugal forces. There is 
always the possibility of a sinaH amount of movement in the 
field coils when a machine is being brought up to speed, and this 
movement, in itself, may eventually damage the insulation if 
it is at all brittle. 

As the capacities and speeds of turbo-generators were increased 
and the space limitations for the rotor windings became more 
pronounced, the resulting higher normal temperatures led to 
the adoption of mica for the insulating material in the slots with 
either mica or asbestos for the insulation between turns. As 
the voltage between adjacent turns is always extremely low, 
what is needed -is really a durable separating medium, rather 
tharf an insulation, this medium being one which will not become 
crisp or brittle at fairly high temperatures. Asbestos has 
served for this purpose very effectively, and even has some 
advantages over mica, as the latter must be applied in relatively 
small pieces in the form of strap or tape, and the individual 
pieces are more readily displaced or shifted than is the case 
with asbestos. Some very severe tests have been made in 
order to determine the possibilities of such rotor insulation 
In one case, a turbo rotor thus insulated was run at full speed for 
over 40 hours, with such a current that the rise by resistance 
in the rotor copper was about 250 deg. cent It was the in- 
tention to continue this test very much longer, but the conduc- 
tion of heat from the winding to the core, and thence through 
the shaft to the bearings, was such that finally the bearings 
became overheated and gave out. After this test, the winding was 
carefully dismantled, and no evidence of any injury^ to the 
insulation could be discovered. Of course, such temperatures 
a*e not recommended in turbo rotor practice, but this was 
simply an attempt to find' a temperature limitation. If a 
designer wants to find the facts in any apparatus, he will obtain 
the most valuable information if he operates the apparatus 
up to the point of destruction. He thus fixes a limit which 
he must keep below. 

The use of thica, or mica and asbestos, on turbo rotors has 
been very .generally adopted in this country at the present time, 
and it may fee said that, within the writers experience, no case 
of .destruction of these windings through heating, has 


come to his notice, although a great number of them have 
been in service for a relatively long time. In many of the older 
machines with fish paper insulation in the rotors, the conditions 
of ventilation and the normal ratings of the machines were 
such that the maximum temperatures in the rotor windings 
were relatively much less than in present practise It may there- 
fore be said that the use of mica in the rotor has been largely 
due to the introduction of the larger capacities and higher speeds. 


The total iron and copper losses in a large, high-speed turbo- 
alternator are in general no higher than in a corresponding 
capacity low-speed machine 

As far as the iron losses are concerned, no further comment 
need be made than that the magnetic flux densities in general 
are somewhat lower than in lower speed machines of same 
frequency, and therefore the losses per unit volume of material 
are no larger. 

The total armature copper losses in turbo-alternators, as a 
rule, are considerably smaller than in corresponding capacity 
machines of the moderate or low-speed types. This is due partly 
to the use of a smaller total number of conductors, and partly 
to a lower current density in the armature conductors. As 
brought out before, in a narrow core machine, a considerable 
portion of the buried copper heat may be conducted lengthwise 
of the conductor into the end winding, and there dissipated 
into the air In the turbo-generator, with its much wider 
core and greater distance from the buried copper to the end 
windings, a smaller percentage of the buried copper heat will 
be conducted into the end windings. To partly compensate 
for this, it is usual to work the armature copper in the turbo-gen- 
erators at a lower current density, and therefore at a relatively 
lower total copper loss. This is somewhat of a handicap in the 
economical design of the generator, as extra space is thus required 
for the armature winding. In some of the earlier machines, the 
armature conductors were made of solid copper bars of relatively 
large section, partly for stiffening or bracing the end windings, 
as will be referred to later With these solid conductors there 
was a very considerable loss in the buried copper due to eddy 
currents. To compensate for this, the armature conductors 
were made very large in section, so that the current density, 
due to the work current alone, was very low compared with 


practise in other types of machines. On account of the com- 
paratively large section of armature conductors, the conduction 
of heat from the buried copper to the end windings was relatively 
large. In some of these earlier, large capacity machines, 
the nominal current density in the armature conductors was 
so low, and the section of conductors so great, that the total 
buried copper loss, due to the work current, could be carried 
from the buried paft of the coils into the end windings with a 
comparatively small (drop -in temperature, so that, t if there had 
been no eddy currents present, the buried copper would have 
shown less rise .than the iron. Any considerable rise which 
occurred was thtis chargeable to eddy currents in the buried 
conductors, rather than to the work current. While such con- 
struction was fairly effective for the purpose, yet it was decid- 
edly uneconomical in design, as indicated before. In fact, 
with later proportions and methods of design, the safe oiitputs 
of some of the earlier machines could easily be 50 to 75 per cent 
greater, largely on account of elimination of eddy currents and 
improvement in methods of dissipating heat from the end wind- 
ings. In many of the older machine's, the ventilation of the 
end windings was not nearly as effective as in modeni types, 
due principally to the form and arrangement of the end con- 
nectors, tlsually $ir spaces were allowed between adjacent 
coils although, in some instances, these were so small as to 
give but little benefit. Moreover, in many cases, the type of 
fend winding employed rendered these air spaces between coils 
rather ineffective, unless special means were taken td deflect 
the air between the coils. With later constructions, the end 
windings lie more or less agross the path of the ventilating air, 
and there are ample openings between the coils, so that a very 
considerate part of the ventilating air will actually pass between 
the cO^ls of thfe end windings in such a way as to give the niaxi- 
nuih possible ventilation. When it is considered that the total 
armature copper loss may be only 20 per cent of the total stator 
loss, it will be seen that an excessive amount of air is Hot re- 
quired when the end windings are properly arranged for most 
effective ventilation. 

Much effort has been expended in eliminating or reducing 
th eddy current, losses in the buried copper of large turbo- 
generators, as well as in othef types of large capacity alternators. 
These eddy currents aare due to two sources, namely, the alter- 
nating magnetic flux across the slots djie to ttte armature ampere 


turns per slot, and secondly, the magnetic fringing from the 
rotor pole face into the open armature slots. In some instances, 
tests have indicated that the local e.m.fs. set up in the armature 
conductors by the flux through the slot opening is very consider 
ably greater than those due to the flux across the slot. Obviously 
with partially closed slots, this fringing into the top of the slot 
should be practically absent. 

The simplest remedy for the eddy currents set up by these 
local e.m.fs. is to sub-divide the conductors into a number of 
wires or conductors in parallel, so arranged or connected that 
the local e.m.fs. oppose and, to a great extent, balance each 
other. This opposition may be obtained by special arrangement 
of the conductors in each Individual slot, 6r parallel conductors* 
in the two halves of a complete coil may be connected in oppo- 
sition to each other. Some of these arrangements do not com- 
pletely balance the opposing e.m.fs., but they include the resist- 
ance of the complete coil in the eddy current circuit, so that the 
eddy losses are not only very materially reduced, but they are 
distributed over the entire coil, including the end windings, which 
condition, in itself, represents a very material improvement. 


An important problem connected with the insulation of 
large turbo-generators, is found in the fire risk, or danger of 
destruction of th3 end windings due to starting an arc at some 
point. On account of the tremendous ventilation in such 
machines, a fire, if once started, may quickly ruin the entire end 
winding. An extended investigation was made, with a view 
to providing an insulation which would not burn rapidly. 
Among other tests, the end windings were finished on the out- 
side with an asbestos covering or tape However, such tape 
requires some sort of sealing varnish, or material to fill its pores, 
to keep it from absorbing moisture or oil. The tests showed that 
if a fire was once started, combustion would be maintained by 
the gases liberated by the u gasification " of the varnishes and 
other material in the end windings, whether the coil was covered 
with asbestos or not. No covering which was tested appeared 
to be very effective. Although some outside covering might be 
found which would be slightly effective in preventing fire from 
starting so readily* yet, if once started, it appears that a fire 
can very easily maintain itself in such machines. Eventually, 
the conclusion was reached that the safest course would be to 


provide suitable closing doors or valves in the air inlets to com- 
pletely shut off the incoming air to the machine. In addition, 
suitable doors on the air outlets, where they can be applied, 
should also be helpful, by retaining the smoke and burnt gases 
inside the machine, which thus assist in smothering the flames. 
The uSe of fire extinguishers of the gaseous type will usually 
be rather ineffective, unless the incoming air and ventilation is 
practically cut off. For instance, with 60,000 cu. ft, (1698 cu. 
m.) of air per minute passing through a large machine, the 
addition of a little gas for extinguishing the fire would hardly 
make any impression. In one instance, in attempting to extin- 
guish a fire, an effort was made to feed the gas in against the 
ventilating pressure of the fans. Obviously, this would not 
work, and then a hose was used in order to get enough pressure 
to counteract the fan action. Although the fire was extinguished, 
the resultant effect of fife and the high pressure water was that 
new insulation was required. 

It has been known for many years to designers, that alterna- 
ting current generators can give, at the instant of short circuit, 
a much greater current than that which they will give on con- 
tinued short circuit. The first emphatic evidence of this, in the 
writer's experience, was in connection with the first Niagara 
generators in 1894. Upon short circuiting one of these machines 
at 'full speed and normal voltage, the results indicated a current 
tush so great that it was apparent that it was limited only by 
the anhature self-induction, and not by the so-called synchronous 
reactance. Later, after being put into actual commercial 
service, it was found necessary to brace the end windings on these 
machines. However, at that time, no suitable instrument, 
such as the oscillograph, was available for determining the 
conditions on short circuit, and the phenomena did not permit 
of fjuich 6*p'erimental investigation. 

Similar evidence was found from time to time, as in the first 
Manhattan Elevated engine type generators, which bent their 
$nd windings out of shape on a dead short circuit. But the real" 
possibilities for trouble in this matter did -not develop until 
the large capacity turbo-generators came into use. In these 
^machines, the armature ampere turiis per pole are so high, 
Compared with moderate speed alternators, that the stresses 
due to the stray magnetic fields oo^xort circuit are much greater 



than the natural rigidity of the end windings will withstand. 
The manufacturer of such apparatus, without data of any 
quantitative value at hand, did not fully recognize the Veal 
weakness in the end windings until disaster overtook them. Even 
then it was a long and difficult undertaking to overcome the 
trouble. All kinds of designs of etid supports, and various ar- 
rangements of end windings were tried, with more or less success. 
But each new -step in the increase in capacity opened up the 
problem a&ain. It was soon noted that those armature windings 
which were made up of cable or small wires, suffered most on 
'short circuit, and for awhile therq was a tendency on the part 
of some manufacturers to .use heavy, solid conductors to give 
rigidity in the end windings. This was effective within certain 
limits, but was very expensive from the design standpoint, as 
on account of eddy currents in the buried copper It was neces- 
sary to work at a very low current density, which was not 
economical in winding space. 

FIG. 20 

In this country, the types of armature windings finally 
narrowed down to the open-slot construction, usually with an 
upper and lower coil per slot, with the end winding arranged 
in two layers, similar to d-c. armature windings, or the common 
induction motor primary windings. This turbo end winding 
was extended at various angles to the axis of the machine froih 
almost parallel up to 90 deg., as shown in Fig. 20. The principal 
survivor of these types, is one which extends at some angle 
between 30 and 60 deg. to the axis. There are several reasons for 
this, first, it allows a very 'substantial bracing to be applied to 
the end windings. Second, the stray fields around the end 
windings do not, to any extent, cut the adjacent solid parts, 
such as the end housings, stator and end-plates, etc. An angular 
position of approximately 45 deg. seems to be a good compromise 
on these points. Ample supports, as shown in Pig. 21 can be 
applied for bracing the windings against movement in any 


direction. Such end windings are usually braced against metal 
supports attached to the stator end-plates. The coils are so 
clamped to the racks, and are so braced against each other that 
the windings will sustain a dead short circuit across the terminals, 
even in the largest capacity machines, without injury. 

On some recent large turbo-generators the end windings have 
been further strengthened by double metal racks between the 
two layers of windings, so arranged as to securely key these two 
layers to one another at certain points. Moulded mica troughs 
are placed around the coils as an extra insulation from the metal 
racks. By this keying of the two layers to one another, the 
winding as a whole is stiffened, quite irrespective of any other 
clamping arrangement. In fact, this is practically equivalent to 
putting the end windings in rigidly held slots, thus approaching 
the conditions which obtain in the buried part of the coil. 

In order to limit the momentary short circuit current, the 
armature reactance is now usually made as large as the condition 
of the design will permit. This naturally means high ampere 
turns per pole, which in turn means high synchronous reactance, 
and consequently poor inherent regulation of the machine, 
especially on inductive loads. This can be illustrated by the 
following example: Assume a 5000-kw. unit of an earlier design, 
which can give 25 times full load current on momentary short 
circuit. By certain improvements in the design of the armature 
coils, such as the use of deeper slots, better subdivision of the 
copper to eliminate eddy currents, improved ventilation and 
'conduction of heat, etc., the capacity of the machine is assumed 
to be increased to 10,000 kv-a., the number of armature turns 
remaining the same as before. It is evident that when short 
circuited, the revised machine will give the same total current 
as on the former rating, which, however, is only 12 times 
the rated current on the new capacity basis. Obviously, the 
end winding stresses are no greater than before, although the 
nominal capacity has been doubled, and if it -were possible to 
satisfactorily brace the end windings with the former rating, 
the same bracing should be effective on the new rating. This 
illustrates, roughly, what is taking place in later designs, although 
the steps in the change may not be just those mentioned. Again, 
in the above example, it is obvious that, with the now rating, 
the inherent regulation at full load is the same as at 100 per 
cent overload on the old rating, which means that it is relatively 
poor. Another way to express this is, that the old fating might 


give 2| times full load current on steady short circuit, while the 
new rating gives 1 \ times. 

This condition of poorer regulation is inherent in the newer 
practise, bttt is apparently acceptable to the users of such 
apparatus, for a variety >of reasons wliich do not come within 
the province of this paper. 


The foregoing covers, in a general way, fnany of the problems 
encountered in large turbo-generators, and defines the situation 
as it stands at present. 

It anay be suggested, in connection with the temperature 
problem, that the high temperatures obtained are due to forcing 
the construction too far; but, in answer, it may be stated that 
it is forced no further in this feature than in many others. The 
whole design has been carried far beyond the most economical 
construction, from the generator standpoint alone. In fact, the 
whole machine is more or less a compromise between desirable 
conditions as a generator, and most economical conditions as 
part of a combined turbine and generator unit. It may e 
added that the ultimate limits in construction and capacity will 
be obtained only when the steam turbine conditions are satis- 
fied, and there are indications that possibly this result is being 
approached now with the present high speeds. 

There is one small consolation in all the confusion of develop- 
ment which has attended > the turbo-generator work, in the few 
years it has been with us, namely, the question of choice of 
speed has been practically eliminated. For 25 cycles, there 
remains only one speed, namely 1500 revolutions, with two 
poles, from the smallest unit up to 251000 kv-a. as a possible 
upper limit. For 60 cycles, up to 5000 kv-a , two-pole machines 
at 3600 revolutions are being furnished, while from this 
capacity up to 20,000 kv-a. four poles may be used. 

It will be evident to any reader of this paper, that the designers 
of large turbo-alternators 'have had a strenuous time during the 
past few years very much more so than is indicated herein, for 
their successes, rather than their failures have been discussed. 
In fact, much of the time they have been working ahead of their 
data and experience. In presenting this situation from the 
design point of view, it is hoped that a better and clearer under- 
standing of the turbo-generator ^problem will be obtained by all 
who are interested in such apparatus. 


FOREWORD In 1911 and 1912, a revision of the standardization 
rules of the American Institute of Electrical Engineers was being 
made. The problem of temperature guarantees was referred to 
a sub-committee, consisting of Dr. Steinmetz and Mr. Lamme. 
It was decided by the Standards Committee to hold a mid- 
winter convention of the Institute in February, 1913. In order 
to furnish a basis for discussion of the temperature problem at 
this convention, the sub-committee on temperature collaborated 
in the preparation of this paper. 

It may be noted that later information has modified some 
of the figures for temperature limits. (ED.) 

THE problem of permissible temperature limits in electric 
apparatus is largely that of the durability of the insulation 
used. As this may consist of materials of widely varying heat- 
resisting qualities, the probem resolves itself into one of con- 
sideration of the properties of the materials themselves. 

The durability of insulation may be considered from two stand- 
points, the mechanical and the electrical. Temperatures which may 
ruin the insulation, from a mechanical standpoint, may not radi- 
cally effect its dielectric strength. This is particularly true with 
moderate voltages where the insulation serves largely as a separat- 
ing medium. The purpose of the insulation usually is two-fold: 
First, it must serve to separate, mechanically, the electric conduc- 
tors from each other, and from other conducting structures, and 
second, it must withstand the voltage between the electric con- 
ductors and between the electric circuits, and other con- 
ducting parts. In lower voltage apparatus, usually only the 
former function applies, as the mechanical separation is more 
than sufficient to withstand the voltage used. The dielectric 
strength of the material is, however, of first importance in high 
voltage apparatus. 

A great majority of the electrical "breakdowns" on low 
voltage apparatus is due to mechanical weaknesses, as far as the 
temperature problem is concerned; that is, high temperatures 
may make the insulation brittle, or crisp, so that it may flake off, 
or powder, or crack, or be crushed by mechanical action, thus 
allowing the conductors to make contact with each other or with 
adjacent conducting material. 


The " life of insulation " is an indefinite term and must bo de- 
fined in time, mechanical strength, absence of foreign materials 
of a conducting nature, etc. Almost all insulating materials 
will be somewhat affected in time, and many of them tend to be- 
come dry and brittle. The rate at which deterioration occurs 
\\ith any given material, is some complex function of the tem- 
perature and of other conditions. 


Insulations may be classified under three headings, depend- 
ing upon their heat-resisting properties. However, all such 
classifications must be relative, for no absolute limit can be fixed, 
as there is no definite point at which injury or destruction can be 
said to take place. 

The usual insulating materials can be considered as included 
m three general classes: 

Class A. This includes most of the fibrous materials, as 
paper, cotton, etc., most of the natural oil resins and gums, etc. 
As a rule, such materials become dry and brittle, or lose their 
fibrous strength under long continued moderately high tempera- 
ture, or under very high temperature for a short time. 

Class B, This includes what may be designated as heat-re- 
sisting materials, which consist of mica, asbestos, or equivalent 
refractory materials, frequently used in combination with other 
supporting or binding materials, the deterioration of which, by 
heat, will not interfere with the insulating properties of the final 
product. However, where such supporting or binding materials 
arc in such quantity, or of such nature, that their deterioration 
by heat will greatly impair the final product, the material should 
be considered as belonging to class A 

Class C. This is represented by fireproof, or heat-proof 
materials, such as mica, so assembled that very high tempera- 
tures do not produce rapid deterioration. Such materials are 
used m rheostats and in the heating elements of heating 
appliances, etc 

All the above are relative terms. The first class, for instance, 
represents materials which are really more or less heat-resist- 
ing, but which deteriorate at lower temperatures than those in 
the second class, which are defined as heat-resisting. Also, the 
fireproof materials of the third class are not strictly heat-proof 
or fireproof, but will simply withstand very high temperatures 
for rclativelv long pcnods without undue deterioration 



In class A, the materials appear to have a very long life (or an 
almost indefinitely long life, aside from mechanical conditions) 
if subjected to ultimate temperatures which never exceed 90 
deg, cent. Also, they appear to have a comparatively long life, 
even at ultimate temperatures as high as 100 deg. cent. At 
materially higher temperatures than 100-deg. cent., the life is 
very greatly shortened, and temperatures of 125 deg. cent, will 
apparently ruin the insulation, from a mechanical standpoint, in 
possibly a few weeks, if such temperature is maintained steadily 
However, for low voltages, the insulating qualities may still be 
very satisfactory, even at this temperature, and therefore the de- 
struction of the insulation is purely one of injury or breakdown 
from the mechanical standpoint, as stated before. Tempera- 


W 6 









ft 75 100 125 130 


FIG. 1 

tures as high as 160 deg. cent, on such insulations for a con- 
siderable period may not entirely destroy their insulating qual- 
ities, althoiigh, mechanically, such temperature? appear to be 
impracticable, except for very short periods. 

In order to illustrate the relation between the possible life 
and temperature of class A insulation, Fig. 1 is shown. This 
must not be taken as representing actual results, but is simply in- 
tended to illustrate, in a very approximate manner, the very great 
shortening of the Kfe of insulation by increase in temperature. 

It may be assumed that at very high temperatures, the insu- 
lation will have practically the same life, in actual hours of high 
temperature operation, whether the temperature is applied con- 
tinuously or intermittently. For example, if an insulation has 
10,000 hours Kfe with a certain high temperature continuously 



applied, it is assumed that it will also stand the same tempera- 
ture for 10,000 hours in short periods, provided the intermediate 
temperatures are low enough to represent an indefinitely long 
life. It is probable that under the intermittent condition, the 
life will really be slightly greater, due to the fact that depre- 
elation will be largely mechanical, and the insulation may " re- 
cover/* in some of its mechanical characteristics after each period 
of high heating. 

If 3 therefore, higji temperatures 'are reached intermittently, 
with intermediate periods of lower value but still high enough 
to shorten the life of the insulation, it may be assumed that the 
total life o the insulation is the resultant of the life under the 
two temperature conditions. 













50 100 150 200 250 

FIG. 2 

In heat-resisting materials, such as those of class B tempera- 
tures of 125 deg cent are comparable with 85 deg. cent or 90 
deg cent in class A, and 150 deg cent in the tormer is comparable 
with 100 deg. cent in the Utter Pig 2 illustrates very approxi- 
mately the life-temperature curve of such insulations As in Fig 
1 , this should not be taken as an exact representation of the actual 
life Due to the greater heat-resisting qualities of such materials, 
it appears that relatively higher temperatures'are not as quickly 
harmful as in the first class 

In class C materials, it is difficult to give any reasonable indi- 
cation as to the limits of temperature, except that very 'high 
temperatures, (practically up to the point of incandescence) are 
found in some heating appliances. 




As the insulation, in itself, is not usually the seat' of generation 
of loss or heat, it is the temperature of adjacent materials which 
must be considered in defining the conditions in the insulation. 
The temperatures of the adjacent materials should therefore be 
considered only in so far as they affect the insulation itself, and 
where such temperatures do not affect the insulation, or the life 
of the apparatus, or its normal perfomance, they are immaterial 
Considering the influence of the temperatures of the adjacent 
media, the direction and amount of heat flow must be taken into 
account, as the maximum temperature in the insulatiori is de- 
pendent upon these. In the case of armature windings, for 
instance, the heat flow may be from the buried portion of. the 

coils toward the end windings. It also may be from the buried 
copper through the insulation to the armature teeth, or there may 
be a reverse heat flow from the iron to the copper, depending 
upon the various factors of construction, heat conductivity of 
the materials, amount of heat generated in the various parts, 
ventilation, heat dissipation etc. 

Depending upon conditions of heat flow and distribution, 
various methods of temperature determination may be used. 
No method is accurate, unless all the conditions of heat flow are 
accurately known, which is never the case in commercial ma- 

The difficulties in the problem of commercial temperature 
determination are illustrated by Fig. 3. 


In the figure, a represents the temperature inside an armature 
coil, b the temperature between the insulation and the iron of an 
armature tooth, c that in the body of the tooth, and d that in the 
body of the core at some point back of the coils and teeth. "Lei 
the temperatures at no load be represented on the ordinate A 
Then, at some load, represented by ordinate B, the relations 
of the various temperatures have changed. At C, D and E, 
there are still greater changes, depending upon the heat genera- 
tion and distribution. If the rated capacity of the machine is 
at E, for instance, then the armature copper is hotter than the 
iron, while if rated at B, the reverse would be true. Obviously, 
no rule can be formulated to cover these various conditions in 
different machines, nor even in a given machine, unless all the 
heat generation, distribution, and dissipation characteristics are 
known. Obviously, as far as the insulation is concerned, the 
temperatures of a and b are the only ones which need be consid- 

All temperature determinations of a commercial nature, are 
necessarily approximations, or relative indications, upon which 
proper margins must be allowed for the ultimate temperature 
possibly attained. Therefore, in apparatus where there are 
liable to be discrepancies of 10 dcg. between the measurable and 
the actual idtimate temperatures, a limit of 90 deg cent, should 
be allowed by conventional temperature measurement on insu 
lations in which 100 deg. is set as the maximum temperature with 
a reasonable length of life. 

The conventional methods of temperature measurement, as 
by resistance, and by thermometer, do not usually give the maxi- 
mum temperature, but give either the average, or the outside sur- 
face, values, and, when measuring the temperature by these 
methods, which are the only ones generally applicable, an allou - 
ance must be made m windings for possible local higher 
temperatures. These methods apply especially to those ma- 
chines of moderate or low voltages in which the insulation is 
relatively thin, so that the heat gradient from the inside copper 
to the outside surface is small. Also, they apply particularly to 
those machines m which the conditions of ventilation are not nor- 
mally difficult, and in which a fairly thorough distribution and 
dissipation of heat occurs among the various parts, such as in 
ordinary direct-current armatures, induction motors primaries, 
stators and rotors of moderate, speed alternators m which the 
width is relativelv small compared with the diameter, etc 


As the ultimate temperatures obtained by the apparatus de- 
pend upon its rise above the room temperature, or that of the 
cooling medium, and as such temperatures may vary over a wide 
range, it is not practicable to specify or guarantee ultimate tem- 
perature of apparatus without also specifying the. elements upon 
which it depends This, therefore, results in specifying the 
temperature rise in relation to that of the cooling medium. 

While most apparatus operates at materially lower cooling 
temperature than 35 deg. cent, to 40 deg cent,, yet such tem- 
peratures are sometimes reached for considerable periods of time 
in steam stations, and it appears therefore as justifiable to choose 
the permissible temperature rise, such that, at room temperature 
of 35 deg. cent, to 40 deg. cent , an ultimate temperature of 85 
deg. cent to 90 deg. cent, by conventional methods of measure- 
ment, is not exceeded This means, therefore, a temperature 
rise of 50 deg. cent, with conventional methods of testing, such 
as by increase of resistance, or by thermometer, in those insula- 
tions which can stand a continuous ultimate temperature of 
100 deg cent with a comparatively long life. This allows an 
excess of 10 deg. cent, to 15 deg. cent, for local spots, or for the 
temperature gradient through the insulation. A less allowance 
should be made for this difference when methods of temperature 
measurement other than the conventional are used, and which 
approach more closely to the highest temperature actuallv at- 

When the above temperatures are liable to be materially 
exceeded for long periods, heat-resisting insulation of class B is 
recommended With such materials, a temperature of 125 deg. 
cent is comparable with 85 deg cent to 90 deg. cent, in the 
materials of class A Therefore, on this basis of a room tem- 
perature at 40 deg cent or 45 deg cent., rises of 85 deg. cent or SO 
deg cent should not be considered harmful However, in 
those special cases where the conventional methods may not 
sufficiently approximate local high temperatures, as may be 
the case in large turbo-generators, or in wide core alterna- 
tors of large capacity, the rises. of 80 deg cent, or 85 dc 
cent should not be specified by resistance or thermometer, 
but preferably some lower temperature such as 50 deg cent 
thus allowing a very considerable margin for local higher tem- 
peratures In such apparatus with the higher temperature*, 
which require class B insulation, there fe liable to be less uniform- 
itv of heat distribution 


If special methods of temperature measurement, such as ex- 
ploring coils or thermo-couples are used in such apparatus, the 
temperature limit of 125 deg cent, should be considered, and not 
the conventional 50 deg cent rise. In those machines of this 
class which have relatively thick insulation, and consequently 
may have a high heat gradient between the copper and the iron,, 
(depending upon how much heat is flowing from the copper to- 
the iron) an ultimate temperature of the inside insulation of 
150 deg cent is considered as the limit, this being comparable 
with 100 deg. cent with insulations of class A. 

In certain classes of apparatus which are artificially cooled by 
air from outside the room, the cooling is accomplished partly by 
dissipating heat to the artificial air supply, and partly by dissi- 
pation into the surrounding room. If the temperatures of the 
cooling air and of the room are widely different, the resultant of 
the two temperatures should really be taken as that of the cool- 
ing medium. 

The variation of the temperature rise has heretofore been 
considered as having a definite relation to the temperature of the 
cooling medium. However, it appears that it does not follow 
any definite simple law, but it is sometimes positive and some- 
times negative, so that no satisfactory correction for room tem- 
perature is possible at present. It is therefore desirable to make 
the temperature tests at a room, temperature as nearly as pos- 
sibte to some specified reference temperature, so as to make any 
temperature correction negligible The reference temperature 
in the guarantees should therefore be such as Can easily be secured ; 
that is, it should be the average temperature of the places at 
which the apparatus may be operated This is from 20 deg. 
:eiit to 25 deg. cent , and as it is easier to raise than to lower the 
room temperature, the upper figure is advisable as a reference 
i*alue. This 'reference temperature therefore should be chosen 
is 25 deg cent., which is in accordance with the previous A I E E 


In the conventional methods of temperature measurement, 
>y thermometer, and by resistance, many conditions should be 
aken into account, and good judgment is required, in all cases, 
>r fallacious conclusions may be obtained 

There are many conditions which affect both the accuracy of 
he resistance and the thermometer methods of measuring tern* 
>eraturc The resistance method measures only the average 


temperature rise, and not that of local hot spots. However, it 
measures the internal temperature of windings, and therefore no 
correction is required for the temperature gradient through the 
outside insulation The proposed margin between the result 
by the conventional method, and the actual temperature can 
therefore be allowed, in the resistance measurement, as the dif- 
ference between the warmer and the average temperatures m 
the windings. In the resistance metho<J of measurements, the 
rate of transfer of heat from one part of the winding to another 
will not greatly affect the result, as the measurement indicates 
an average temperature, which is that obtained if the heat were 
equalized throughout the winding. However, the rate of flow 
of heat from the windings through the outer insulation to other 
parts, will affect the temperature measurement by resistance, and 
preferably the measurement by this method should be taken 
during operation in those parts where this is practicable, as in 
field coils, and some other instances. In those parts where the 
resistance cannot be measured during operation, this should be 
done as quickly as possible after shut-down, and the time taken 
to shut down the apparatus should not be unduly long. Prefer- 
ably, during shut-down of rotating apparatus the normal current 
should be maintained on the apparatus until at least a relatively 
low speed is obtained. This would represent only an average 
condition, as the ventilation at lower speed is very greatly de- 
creased, while the losses in the windings will remain normal, 
thus tending to give an increased temperature in the windings. 
It would be difficult to fix any definite rule which would give the 
exact temperature conditions during shut-down. 

In the measurement of temperatui*e by thermometer, con- 
siderable judgment is required Wherever possible, the tern- " 
perature should be taken during operation, but the thermometer 
with its pad should be so placed that it does not interfere with 
'the normal air circulation. In thermometer readings, as usually 
obtained on windings, the heat gradient thrdugh the insulation 
must usually be allowed for, this being 10 deg. to 15 deg as 
previously defined However, depending upon the method of 
taking the temperatures, this allowance should vary over a con- 
siderable range, depending upon whether or not the method of 
measurement approximates the actual internal temperature 
For instance, the total heat gradient from the inside copper to 
the outside air will be that through the coil insulation, plus the 
thick covering pad over the temperature bulb If the gradient 


through the covering pad is very large compared with that 
through the insulation, the thermometer may indicate almost 
exactly the internal temperature of the copper; that is, the heat 
gradient through the insulation to the thermometer, may be rela- 
tively small compared with the total gradient to the air. This 
is particularly true where the thermometer rests on a metallic 
seat which covers a considerable portion of the coil surface. In 
this case, the heat which affects the thermometer bulb will pass 
through a relatively large section of surface, with a correspond- 
ingly small drop in temperature, so that the bulb more closely 
approximates the temperature of the inside copper. 

Where there is local heating in the windings, and a consequent 
liability of rapid transference of heat to other parts, the results 
obtained by the thermometer method will vary to some extent 
with the rapidity with which the actual measurement is made; 
that is, the more quickly the thermometer can be brought up to 
the full temperature, the more accurately the temperature of 
the hottest part is determined. With a very rapid method of 
measurement, it may be possible to measure practically the in- 
ternal temperature of the copper of the winding before any great 
heat transference or dissipation has occurred. In such cases, 
obviously, the full allowance for the usual temperature margin 
should not hold. It should be fully understood that it is the 
ultimate temperature, and not the temperature rise, which 
should be considered as the limiting condition, and that the 
measured rise, plus the allowances for temperature gradient, 
plus the measured room temperature, is simply an indication of 
the possible ultimate temperature. By whatever method the 
temperature measurement is made, in all cases the results may 
TDC considered as more or less approximate, and in the end, it is 
the manufacturer who must supply the necessary margin over 
the approximate measurement, in order to make the machine 

A "blind adherence to some particular rule or method of taking 
temperatures, may lead to fallacious results in some instances. 
In armature windings, in particular, incorrect readings may be 
obtained after shut-down. For example, if the armature iron 
back of the armature teeth were hotter than the armature teeth 
and coils during operation, then the temperature to which the 
insulation is subject during operation may be considerably lower 
than that in the hottest part of the machine, due to the ventila- 
tion conditions when running. However, upon shut-down, the 


temperature at the insulation may rise to that of the hottest 
part of the machine, and therefore a false temperature, by any 
method of measurement, might be indicated. 


That with class A insulation, 90 deg. cent, be taken as the 
ultimate temperature limit, as indicated by conventional methods 
of measurement, or those which give similar results, and that 
100 deg. cent, be considered as the maximum ultimate tempera- 
ture permissible in the insulation, where a comparatively long 
life is a requirement. 

That 40 deg. cent, be taken as the limiting temperature of the 
cooling medium, or room, and that, therefore, 50 deg. cent, be 
the permissible rise by conventional methods of measurement, 
with class A insulation. 

That 25 deg. cent, be taken as tine reference air temperature. 
With the permissible 50 deg. cent nse, this gives 75 deg. cent, 
as the average operating condition, by conventional methods of 
measurement, or 85 deg cent, actual temperature, when the 
usual margin represented by the temperature gradient is added. 

An exception to the rise of 50 deg. cent, can be made in those 
cases where space or weight limitations are such that higher 
temperatures, with consequent reduced life, are commercially 
economical, such as in railway motors. In such cases, with class 
A insulation , a nse of 65 deg. cent, with reference air at 25 deg 
cent, is at present accepted as good practice. 

With class B insulations, 125 deg cent be taken as the ultimat 
temperature limit, as indicated by conventional methods of 
measurement, or by equivalent methods, and 150 deg. cent, be 
considered as the maximum ultimate temperature permissible 
in the insulation It follows therefore that 80 deg cent, to 85 
deg. cent, nse is allowable, with such insulations, by the usual 
methods of measurement 

No temperature correction should be made for variation of 
the cooling temperatures from the reference temperature of 25 
deg. cent 

When the method of temperature measurement shows the 
highest temperature actually obtained in the insulation, the maxi- 
mum temperatures specified for the given type of insulation 
should hold. 

In the final "decision on questions of temperature rise, the ulti- 
mate temperature should be the basis, rather than the rise. 


FOREWORD This paper was presented at the Chicago Section meet- 
ing of the American Institute of Electrical Engineers, November 
27, 1916. A number of papers by the author dealing with the 
temperature problem had appeared t bef ore, but the purpose of 
this paper was to put the subject in more definite shape and 
bring it more nearly up to date. During the discussion of the 
paper, considerable new data was presented by the author, and 
it has, therefore, been included in this reprint. 

This paper was listed for a second presentation before a 
regular meeting of the Institute at Schenectady in April, 1917, 
with a view to obtaining a further discussion, particularly by 
engineers on design work. This meeting was cancelled due to 
the declaration of war. (ED.) 

THE laws governing heat flow and temperature distribution 
are so similar, in many respects, to those governing electric 
current flow and electric potentials, that it is rather surprising 
that the former have received so little attention in comparison 
with the latter. Some of the laws of heat flow are so well recog- 
nized that their application to the problem of temperature dis- 
tribution in electric apparatus should have been a leading feature 
in the early developments in such apparatus; whereas, on the 
contrary, it is only recently that very careful study has been, 
made of such application. 

One object of this paper is to indicate, in a comparatively 
simple manner, some of the conditions which fix the tempera- 
tures in different parts of electric apparatus. The explanations 
given cannot be considered as new or novel in substance, but are 
merely the application of fairly well known principles of temperature 
and heat flow to electrical machinery. Before going into the general 
problem, certain simple conditions may be stated, such as: 

1. The heat flow between two points is proportional to their 
temperature difference and to the heat resistance of the path or 
paths between .them. Note the resemblance to Ohm's law. 

As a corollary to the above, it should be evident that between 



two points at the same temperature, there should be no flow of 

2. The total temperature drop between any two points or 
media of different temperatures will be the same through all 
paths of heat flow. 

3. There are no true non-conductors of heat, and, conversely, 
no perfect conductors 

4 Heat conduction and electric conduction bear some quan- 
titative relation to each other, in the broad sense that all electric 
insulators are relatively poor heat conductors, while good electric 
conductors are correspondingly good heat conductors. There 
is apparently no rigid relation between the heat resistance and 
electric resistance of the various materials used in electric ma- 
chinery, but the general relation holds and there are apparently 
no radical exceptions 

5 The rise in temperature at any point, due to generation 
of heat, is dependent (a) upon the total heat generated, and (b) 
upon the amount of heat which can be earned away along all 
available paths per decree of temperature difference. The tem- 
perature will rise until the heat dissipation equals the heat 

6 There are two ways to lessen the heat flow along any path, 
(a) By interposing higher heat resisting materials, (b) By 
lessening the temperature difference, as by raising the tempera- 
ture of the part through which the heat is to be conducted. 
Conversely, the heat flow can be increased along any path by 
the use of better heat conducting materials, or by paths of lower 
heat resistance, and by lessening the temperature of any part 
to which the heat is to flow. 

What makes the problem unduly complicated, in electrical 
machinery, is the fact that there are several different sources 
of heat generation, which may be, and often are, all active at 
the same time. Moreover, the heat losses may be distributed 
through the various heat conducting paths in such a way as to 
render any calculation very difficult and more or, less inexact, 
except in a general way. For example, there is heat generated 
by losses in the copper conductors, obeying one law; while there 
is heat generated in the iron parts under a quite different law, 
and there may be heat generated by windage and friction, 
according to a third law. As these different losses may act in 
different parts of the heat conducting circuit, it should be evident 
that the problem of determining the exact heat distributions, 



and the temperature, is a very complex one. Such a determina- 
tion is in the province of the expert analytical designer of such 
apparatus, but certain general conditions are of interest to all 
users of electric apparatus. 

Consider first the general conditions of heat dissipation from 
an armature coil In Fig. 1 is represented an armature slot with 
the surrounding iron, and with two separate "coils" per slot, as. 
is now the most common practise. Let it be assumed that the 
point a represents the "hot spot", or part at highest temperature- 
in the apparatus. The heat from this part can flow along two 
general paths, namely, longitudinally through the copper" con- 
ductor itself to the end windings, and thence to the air, and. 

laterally through the insulation to the surrounding iron, or to 
the ventilating ducts. From the iron the heat flow is then 
through various paths to the external cooling air. 


Considering first the longitudinal conduction of heat in the 
coil, then starting at the point a, the first unit of length con- 
ductor will have a certain loss. If the heat generated by this 
first unit loss were all that need be considered, then the drop in 
temperature, from the point a to the end windings, would be 
simply a function of the heat-conducting properties of the con- 
ductor itself. But the next unit length is also generating its 


own unit loss, so that the heat flow from the second to the third 
unit length is due to two units loss; in the same way, the flow- 
to the fourth unit length will be due to three units loss, etc 
Therefore, the temperature drop, or temperature difference per 
unit length of conductor, increases more rapidly as the point a 
is departed from, and if it is at a considerable distance from 
the end winding, and the losses per unit length are compara- 
tively high, a very high temperature may be required at a to 
conduct all the heat longitudinally to the end windings, In 
very wide core machines the longitudinal drop may be so great 
that the temperature at a in practise will be so far above that 
of the surrounding iron, that a very large percentage of the 
actual heat is conducted laterally through the insulation to the 
iron, even if the iron is at a comparatively high temperature, 
However, in narrow cores, the drop to the end windings may" 
be, in some cases, so very low, possibly 5 to 10 degrees, that 
with good heat dissipation from the end windings themselves, 
the point a may have, for instance, an actual temperature of 
40 deg. cent. If the iron next to a a'so has a temperature of 40 deg. 
cent, then there would be no flow of heat from a to the iron. Fxtr- 
thermore, in such a case, as the iron temperature over the whole 
width of the core may be lairly uniform, and as the copper 
temperature decreases from a to the end windings, obviously 
as we 'depart from the point a, there would be heat flow from 
the iron to the copper, and thus the windings would tend to 
cool the core. This is frequently., the case with light loads on 
a machine, for in sucji conditions the coil loss is low, while the 
iron loss remains fairly constant for all loads In such case 
there may be heat flow from the iron to the copper along the 
whole length of the buried portion of the coil At some higher 
load, the copper loss varying as the square of the load, the in- 
creased longitudinal drop will bring the copper temperature 
above that of the iron so that the heat flow is from copper to- 
iron, This condition is illustrated by Fig 2 

It must be recognized that the lateral flow of heat, from the 
coil to the iron, reduces the longitudinal drop, such reduction 
depending upon the relative percentages of Jieat flow along the 
t\\o paths ^ It must also be borne in mind that in order to have 
such longitudinal heat flow, the end windings must be able to 
dissipate their own heat at lower temperature than would b^ 
attained at a, or in the core If the end windings have little or 
no ventilation, or heat dissipating capacity, then their own 



generated heat may bring their temperatures higher than those 
of the armature iron so that the heat flow actually may be from 
the end windings toward a, and then laterally through the in- 
sulation to the core In such case, the hottest spot will be in 
the end winding rather than in the buried part of the coil Obvi- 
ously when such condition occurs there is no possibility of either 
the end windings or the buried part of the coil being cooler than 
the iron, for the heat flow throughout is toward the iron 


Considering next the lateral flow of heat through the insulation 
to the iron, the amount of heat conducted is a function of the 

temperature difference and the 
resistance of the conducting 
path. Or, in other words, if a 
given amount of heat is to be 
conducted through a path of 
given resistance, the tempera- 
ture in the heat generating part 
iron Temp's will rise until the required heat 
copper Temp's. i s conducted away 

Light Load 

Iron Temp's 

Copper Temp s. 

Medium Load 

Heavy Load 

Width olCore* 

Copper Temp s 
Iron Temp's. 

^30C Drop 

" 70C -4CTC Rise . 
in Iron with Air at 30*0" 

FIG. 2 

FIG. 3 

To illustrate this problem more concretely, let Fig. 3 represent 
the temperature conditions in a section of an armature Assum- 
ing, for example, the temperature of the copper inside the coil 
insulation as 100 deg cent., the iron temperature as 70 deg. 
cent. , and the air temperature as 30 deg cenfy , then the following 
-conclusions may be drawn. 

(a) From the outer coil (the one next to the air gap) through 
the wedge to the air gap, the temperature drop will be 100 
30 = 70 deg. cent. Obviously, any temperature measurement 
made outside the wedge, next to the air, will approximate the 



temperature of the air and not of the copper Any temperature 
measurement made beneath the supporting wedge will measure 
some intermediate temperature between the copper and the air 
If the temperature drop through the wedge should be equal to 
that through the insulation, then a measurement underneath 
the wedge should show half the temperature drop through in- 
sulation and wedge, and obviously, the measured temperature 
would be far below that of the copper. 

(b) If the temperature is measured at the outside of the coil, 
between the iron and the insulation, it would approximate the 
average of the temperatures of the iron and of the outside 
.insulation, or practically the temperature of the iron If the 
iron should be at different temperatures at the sides of the slot 
and at the bottom, then obviously different readings would be 
obtained, depending upon the location of the measuring device. 
It is evident that such temperature measurements give no in- 
dication whatever as to the true internal temperatures of the 
coil, for the heat flow and the resistance of the insulation are 
nowise involved in the measurement. 

(c) At a point a, between the two coils, there should be but 
little heat flow through the insulation, unless the copper is 
comparatively narrow. If there is but little heat flow through 
the insulation at this point, then eventually the temperature at 
the point a must rise to approximately that of the copper in the 
two coils. Therefore, a measuring device located at a will 
approximate the temperature of the copper itself, and is, in 
general, a good indication of the h&t spot at that part of the 
winding Therefore, as a practical method of temperature 
determination, a thermo-couple located at a is about the most 
satisfactory device that we have. However, the location of the 
point a along the slot is also of importance on account of the 
longitudinal flow of heat in the conductor and the consequent 
temperature drop In other words, the direction of heat flow 
in the coil itself, musl be taken into account Therefore, 
thermo-couple located as above, is only satisfactory when the 
general location of the hot spot is known beforehand This is 
usually determined, in a general way, for a given type or line 
of machines, by locating several thermo-couples along the slots 

With narrow slots and comparatively thin conductors, and 
especially with very heavy insulation, there is some flow of heat 
through the insulation which lies between the two coils, this 
heat passing out sidewise to the iron In such case, the point a 


may be of somewhat lower temperature than the copper. It 
may happen also, in some cases, that, due to unequal losses and 
heating of the two coils in the same slot, one is at a higher tem- 
perature than the other In such case, due to the heat flow 
between the coils, the temperature indication at a will not show 
better than an average of the two temperatures Furthermore, 
if the temperature at c , in a coil subdivided into many insulated 
conductors, is materially higher than at b y then the temperature 
indication at a may not be a close approximation to the maximum 


In the ordinary armature, after the heat passes from the 
copper to the iron, there is still quite a problem involved m the 
dissipation to the surrounding medium, which is usually the air 
The direction of the heat flow to the iron will depend, to a con- 
siderable extent, upon the arrangement and location of the heat 
dissipating surfaces There are two general paths of heat con- 
duction in all armature cores; namely, a flow along the lamina- 
tions to where their edges come in contact with the air or with 
other material, and a flow across the laminations toward heat 
dissipating surfaces The flow along the laminations may be 
calculated with fair accuracy. Across them it is difficult to 
determine such flow, largely because the laminations are in- 
sulated from each other by materials which are poor conductors 
of heat Also such flow is affected not only by the insulation 
between laminations, but by the perfection of contact In other 
words, the heat flow may be affected by pressure. According 
to the various figures available, the heat flow per unit volume 
of material along the laminations is from ten to one hundred 
times as great, for a given temperature difference, as across 
them Obviously, therefore, heat dissipation from the iron by 
flow across the laminations should be considered relatively in- 
efficient, yet in the vast majority of rotating machines the heat 
dissipation is largely across the laminations. The reason for 
this is that by placing ventilating passages or ducts, parallel with 
the laminations, at frequent intervals in the core, the cross 
section of the heat path in the intervening iron sections, may 
be made very large compared with the heat to be dissipated, 
so that the density of flow is very low By the same procedure 
the length of the heat path is made quite short Thus in practice, 
the temperature drop through the laminations themselves may 
be made relatively small compared with other drops However, 


not all the heat in the iron passes across the laminations to the 
ventilating ducts, for where the length of the path, along the 
laminations to any heat dissipating surface, is not large, a very 
considerable amount of the heat may be dissipated from the 
edges of the laminations themselves. In fact, in certain types 
of machines with very shallow iron cores, experience has shown 
that the ventilating ducts, parallel with the laminations, may be 
omitted, provided good ventilation is obtained over the edges 
of the laminations. It is evident, therefore, that the flow of heat 
and distribution of temperature are dependent upon the arrange- 
ment of the iron, dimensions and location of the ventilating 
surfaces etc. 


After the heat has passed from the copper to the iron, the 
resultant of the copper and iron heats must be conducted to the 
cooling medium, which is usually the surrounding air In the 
case of air, there is usually a considerable drop in temperature 
from the solid surface to the cooling air itself, the amount of 
such drop depending upon the ventilating conditions. In prac- 
tise, there appears to be a film or layer of air which adheres very 
closely to the solid surfaces. This forms a sort of heat insulating 
film, retarding the flow of heat to the cooling air In air ven- 
tilation, the effect of any considerable air movement over the 
surface appears to be that of scouring this hot film away from 
the surface and replacing it with a film of cooler air Merely 
scouring or rubbing the hot film away from the surface is not 
particularly advantageous unless some means is furnished at 
the same time for supplying an ample quantity of cooler air to 
take the place of the removed hot film. Rapid air circulation, 
by means of a supply of air from the outside, appears to accom- 
plish both results in one operation. Thus, one of the principal 
actions of air ventilation appears to be that of scouring away the 
hot contact film, while a second action is to carry the hot air 
away without mixing it with the incoming cooler air Whatever 
portion of the dissipated heat is absorbed by the incoming cool- 
ing air adds that much to the temperature of the air itself and 
eventually to that of the apparatus to be cooled. Thus mixing 
the outgoing with the incoming air makes a sort of Siemens* 
regenerative furnace and the machine bocomes cumulatively 
hotter t and hotter until the dissipation through other paths be- 
comes equal to the heat generated. In such cases the ventilation 


of the machine may only be useful in equalizing or redistributing 
the temperatures in the various parts. 

From the preceding analysis, it would appear that the tempera- 
ture at the hottest part of the coil is fixed principally by the heat 
flow through the copper, and its surrounding insulation, directly 
to the air, and by the flow from the copper to the iron, and from 
the iron to any exposed air surfaces, and then to the air. Along 
the first path, there are three principal temperature drops, 
namely, in the copper itself, then through the insulation, and 
then from the outside surface of the insulation to the air. Along 
the second path, there are also three temperature drops; namely, 
from the copper through the insulation to the iron, then from 
the iron to the exposed air surfaces, and then from the surfaces 
to the air. Along the first path each part of the copper path is 
generating its own heat, to be conducted away, in addition to 
that which is to be conducted from other parts of the path. In 
the second path, each part of the iron path may be generating 
its own heat, which adds to that coming from other parts. 
The relative amount of heat conducted along each path is de- 
pendent upon so many conditions, which vary with the load, 
that no one but an analytical designer backed by experience 
could even approximate the values by calculation. However, 
it should be obvious that any measuring device applied to the 
outside or cooling surface does not, and cannot, directly approxi- 
mate the temperature of the hottest part, except in those rare 
cases where the hottest part is dissipating heat directly to the 
air. This is true only in very special cases such as series coils 
of bare strap, etc. In any coil or part of the apparatus which is 
heavily insulated, that is, which is covered by poor heat conduct- 
ing materials, an external temperature measurement is an ex- 
tremely poor indication of the true internal temperature, unless 
many other conditions are known which may give an indication 
of the internal temperature drops- In different types and con- 
structions of rotating apparatus, hot spots may hold quite 
different relative positions with respect to the cores and wind- 
ings, so that no reasonable rule can be made to cover all cases. 
Moreover, in some classes of apparatus, it is not practicable to 
make any temperature roeasurements until after the apparatus 
is shut down, and this introduces otter very important errors 
which should be considered, such as cooling effects as a whole, 
during the period of shut-down, equalization of temperature 
due to internal condttction, etc. 



When there are hot spots, or zones, or areas, of different tem- 
peratures, in an armature winding, for instance, such difference 
in temperature is maintained by the continual generation of 
heat in the various parts But the moment that such generation 
of heat is stopped there is immediately a tendency for equaliza- 
tion of temperatures by flow of the stored heat from the hotter 
parts to the cooler. In good heat conducting materials, as copper, 
such equalization may be very rapid, so that a temperature 
indicating instrument of a sluggish type may not indicate any- 
thing like the true maximum temperature of the spot where it 
is placed, if applied after the load is removed, especially if the 
rate of heating of the thermometer bulb is much less than the 
rate of heat transfer from one part of the winding to another 
If located on a hot spot, the reading may nse to some interme- 
diate value and then drop off as the hot spot cools by heat con- 
duction to other parts. If located upon a cool spot, it may rise 
slowly for a considerable period, due partly to sluggishness of 
the thermometer and partly to the cool spot rising in temperature 
by conduction of heat from some other part. The conditions 
are so varied that no reliable conclusions can be drawn, from the 
action of the "thermometer alone, in regard to the coolest or 
hottest spot, 

A second condition which tends to make such temperature 
measurements fallacious, lies in the cooling action in the interval 
between load removal and shut-down to take temperature 
measurements In apparatus which depends upon a high degree 
of artificial cooling, such cooling effect may be very considerable. 
This is particularly true of high speed machines which require 
considerable time to come to a standstill. It is, therefore, de- 
sirable m such machines to obtain all possible temperature read- 
ings at normal speed and with load In rotating field machines, 
this is, to a certain extent, practicable, but in most rotating 
armature machines, the armature temperatures usually are not 
attainable until the machine is brought to a standstill, and even 
then some error may result from sluggishness or delay in taking 
the readings. One method which has been proposed at times, 
for lessening the sluggishness, is to heat the thermometers up to 
practically the normal operating temperature of the part to be 
measured, while the machine is still carrying load. At the moment 
of shut-down the heated thermometer is applied. This, to a 
certain extent, removes the factor of sluggishness in the ther- 


mometer itself, but is only a partial compensation It must be 
considered that the outside of the insulation is at lower tempera- 
ture than the inside, and that, therefore, the body of the insula- 
tion itself must ha^e its temperature increased by flow of heat 
from other parts. 

In the older methods of determining temperatures, it was 
assumed that the thermometer readings, obtained on a winding, 
for instance, was a true indication of the temperature of the 
winding as a whole The manufacturers of electrical apparatus 
long ago recognized the fallacy of this method, as they had found 
from bitter experience that there were liable to be hotter parts 
in the machine than any thermometer readings would indicate. 
They, therefore, designed machines with regard to the possible 
hot spot temperatures as encountered in service, rather than 
any temperature which the exposed parts of the machine would 
show Thus in designing a certain machine for safety at the 
hottest part, not infrequently the exposed parts- of the winding 
would show, by thermometer, comparatively low temperatures, 
such as 25 cleg, to 33 deg cent, rise. Therefore, as the observable 
temperature readings came so low it became the fashion to call 
for 35 deg. cent guarantees and, in many cases, the operating 
public lost sight of, or perhaps never knew, the real meaning of 
such low temperatures. Among the designers of electrical 
machinery, it was recognized that a temperature rise of 35 deg. 
cent in itself was absurdly low, but that the object in operating 
at such low temperature on a part which could be measured was 
simply to protect the machine in some inaccessible hotter part, 
where the temperature could not be measured. From the present 
viewpoint, it is astonishing what reliance has been placed upon 
temperature readings in the past. For example, if a 40 deg. 
cent machine showed 41 5 deg cent rise on test, it was unsafe, 
while if "it showed 38.5 deg. cent, rise, it was good. We now 
recognize that neither of these temperatures have any controlling 
value, unless many other conditions are known To the ex- 
perienced man they simply mean that compared with the other 
machines of similar constructions and characteristics, which have 
proved satisfactory in service, they are reasonably safe. To the 
designer they mean that when proper corrections have been 
made for the various internal temperature drops, the highest 
temperature attained, at any point; Vill be within the limits of 


durability of the insulating material used. The whole problem 
is a good deal like that of a determination of the voltage generated 
in a given power-house, by measuring the voltage at the end of 
a transmission-line. If we know all the constants of the line, 
and know the current flowing, etc., we can figure back to the 
generated voltage. Otherwise the voltage at the end of the line 
means but little. However, we know that if the system is de- 
signed with reasonable regard to economy in general, there may 
be from ten to twenty per cent, voltage drop from power-house 
to the end of the line. Therefore, by adding an approximate 
correcting factor to this voltage, we can make a reasonable 
estimate of the generated voltage. In the same way in electrical 
apparatus of certain types, a reasonable internal temperature 
drop may be approximated, which added to the observable tem- 
perature, gives a fair approximation to the hottest part, but 
the result is an approximation and must be recognized as suck. 
Primarily, the manufacturer must make a safe machine for a 
specified service regardless of the temperature guarantees, and 
the temperature measurements made on most classes of apparatus 
should be considered simply as rough approximations to indicate 
that the manufacturer has made a reasonable attempt at a safe 
machine. This may seem a rather bald statement, but never- 
theless it is a fair statement of the case. 



It has been shown in the preceding that the usual observable 
temperatures are in most cases only crude approximations to 
the real temperature conditions. It may now be shown that 
even the observable temperatures, obtained by the usual means, 
are in themselves only crude approximations in many cases. 
Take, for instance, the determination of temperature by in- 
crease in resistance; when the coil is heated its temperature may 
not be, and very frequently is not nrnforrn throughout the coil. 
As an extreme example, if one-fifth of the coil length has a tern 
perature of 80 deg. cent., while four-fifths of it has a rise, of 
30 deg. cent-, then the increase in resistance of the coil as a whole 
will correspond to a rise of 40 deg. cent. Thus, by increase of 
resistance, the temperature may be more than safe, while 
actually one-fifth of the coil is far above the safe temperature 
for ordinary fibrous insulations. In other words, the resistance 
method gives only average results and may be very misleading. 
However, in those cases where it is known, by past experience 


and otherwise, that there is very little liability of hot-spots, the 
resistance method of determining temperature is often quite 
satisfactory However, the method is limited to comparatively 
few types of windings. 

Considering next the thermometer method of measurement, 
the theory of this is quite simple, but apparently it has been very 
much misunderstood. In windings, except in rare cases, the 
thermometer is not applied directly to the heat generating 
material itself, but is applied outside of an insulating covering 
Usually the temperature drop through this insulating covering 
does not receive any consideration, and yet everything depends 
upon this. Assume, for example, an insulated coil, thermometer 
and covering pad, as shpwn in Fig. 4, Assuming the copper 
inside the coil as being of uniform temperature, and the cooling 
air at a ajid b as also at a uniform, but much lower, temperature 
than, inside the coil ; then the temperature drop from the copper 
to & will be the same as through the insulation, thermometer 

FIG. 4 

bulb and covering pad to the air at a. Obviously if the tempera- 
ture drops through the insulation and through the pad are equal,, 
then the thermometer bulb will show a midway temperature. 
This is, of course, assuming that the surface drop to the air, 
previously referred to, is very small, or that it is included as part 
of the drop through the pad- Obviously, if the drop through the 
covering pad is made very much higher than that through the 
insulation proper, then the thermometer bulb more closely 
approaches the copper temperature. Thus it is seen that all 
kinds of results may be obtained, depending upon the relative 
drops through the pad and through the insulation. In a low 
voltage machine, with relatively thm insulation, the p,ad may 
take most of the drop. With very heavy insulation, the pad may 
take proportionately less and the thermometer reading departs 
accordingly from the copper temperature. It might be sug- 
gested that a big thick pad of very poor heat conducting material 
might be used. This apparently would tend toward more ac- 


curate temperature readings, but, on the other hand, harmful 
effects may be introduced by the use of a large pad. The resis- 
tance to heat dissipation being increased in the area covered by 
the pad, obviously less heat will be carried away at this point 
and, therefore, the heat generated under the pad must be con- 
ducted to adjacent parts of the coil. This means an increased 
temperature at this point, due to the use of the pad. Again, the 
use of the pad, in some cases, may affect the normal ventilation 
of certain parts of the coil not directly covered by the pad. For 
instance, if there is a ventilating space between two adjacent 
armature coils, through which air is normally driven, a pad which 
covers this space even partially may create more or less of an air 
pocket, and thus materially affect the heat dissipation, and the 
temperature directly under the pad. Experience has shown that 
both of the above conditions are obtained when good judgment 
is not used in the application of the covering pad. This, of course, 
applies particularly to those cases where temperature readings 
are obtained while the machine is in operation. Of course, after 
shut-down, most questions of ventilation and of generation of 
higher temperature under the pad need not be taken into account. 
There are so many conditions entering into the interpretation 
of the thermometer and resistance methods of determining 
temperature, that in certain classes of apparatus it has been 
very desirable to find more accurate methods. One of these 
is in the use of so called resistance coils. In this method a coil 
of fine wire of a known temperature co-efficient, and of known 
resistance at a given temperature, is placed at the place where 
the temperature is to be measured, and the temperature rise is 
determined from the increased resistance of the coil One serious 
objection to this arrangement, is that the resistance coil must 
have considerable length and breadth so that it really indicates 
the average temperature of a considerable area instead of a point. 
When placed between two coils, as indicated in Fig. 5, it usually 
occupies so great a proportion of the slot that it indicates an 
average temperature considerably lower than at a Furthermore, 
on account of the length of such coils, there may be a consider- 
able difference between the temperatures at the two ends Thus 
the resistance coil, like the resistance measurement of the wind- 
ings themselves, gives an average result, but this average ma\ 
be limited to a comparatively small area, whereas, in the resist- 
ance method in general the indicated rise is an average of the 
whole winding. However, in 'the resistance method, the tern- 



perature of the conductors themselves is measured, whereas, 
with the resistance coil the temperature measurement is outside 
the insulation The resistance coil method is, therefore, a rel- 
ativ^ly crude approximation, although when brought out it was 
really an important step in advance In, its early application, 
many misleading results were obtained, due largely to lack of 
understanding of the principles governing temperature distribu- 
tion and temperature drop, In some cases, the resistance coil 
was placed under the wedge as at b in Fig 5 In other cases, 
the coil was placed at the side of the slot next to the iron, or at 
the bottom* Very rarely was it placed midway between the two 
coils, probably because this was a more difficult application and 



Resistance Coil 

r"ft"^: Resistance 


FIG 5 

also because the greater accuracy of such location was not rec- 
ognized From the use of resistance coils many good engineers 
drew the conclusions that the upper limit of permissible tempera- 
ture for fibrous insulations was only 80 deg to 90 deg cent , 
because with the coils located in certain ways and places, de- 
terioration of insulation at some other point was liable to begin, 
if the above temperatures were exceeded The error was in not 
recognizing the temperature drop between some hotter spot and 
the average location of the resistance coil When this condition 
was recognized the results obtained by resistance coils became 
more consistent with the facts 

A later development than the resistance coil is the thermo- 
couple as a practical device for measuring temperature One 


great advantage of the thermo-couple is its very small size, so 
that it can indicate the temperature at practically a point instead 
of a very considerable area Moreover, as it is a zero current 
method of measurement, when used with a potentiometer no 
question of size or length of the connecting leads need corne up 
The thermo-couple is so small and has so little mass, that it can 
follow very quickly any temperature changes where it is located 
If properly placed it furnishes the most accurate temperature 
indicator which we now have, as it can be located in all sorts of 
normally inaccessible places However, its use is practically 
limited to stationary apparatus, In rotating apparatus, or 
rotating parts it can be used only after shut-down, which intro- 
duces errors, as already shown. 


In all the preceding considerations it has been assumed that 
the copper inside the coils itself is at a uniform temperature, m 
any given unit of length. This is practically true, provided the 
coil is made up of a single conductor, or of a relatively few con- 
ductors with only a moderate amount of insulation between them 
When several coils or conductors are placed side by side, as in 
Pig. 6, it would appear at first glance that the middle coils should 
heat much more than the outer ones But, in reality, unless 
there are many layers of coils, the temperatures of the different 
coils will not vary greatly from each other For instance, in 
Pig. 6, the 'heat generated in the middle conductor is only 'one- 
third that of the total generated in the coil, and yet the two 
side surfaces through which this heat passes to the adjacent coils 
aggregate almost as much as the total outside dissipating surface 
of the whole coil, through which. all the lateral heat flow is dis- 
sipated. Considering further that the insulation between the 
middle coil and its neighbors is relatively thin compared with 
the outside covering, it is obvious that the temperature drop 
from this coil to the adjacent ones will be comparatively small,- 
possibly not over ten per cent of the drop through the outside 

However, with a large number of coils side by side, the condi- 
tions become cumulatively worse. Here, the drop from the 
center conductor to the next one, may be small But the 'drop 
from the second conductor to the third is considerably greater 
due to the heat of two conductors being transmitted. Prom the 
third to the fourth thefe is a drop corresponding to the losses 



of three conductors, etc Thus, there is a gradually increasing 
temperature drop from the center of the coil toward the outside 
surface, and if the coil be very deep, that is, if it consists of many 
insulated layers, the sum total of the drops may be quite large 
Or, putting it in another way, with a comparatively deep coil, 
the temperature rise from the outside surface of the coil itself 
toward the center will be very rapid at first, and gradually taper 
off, as indicated in Pig. 7. This is indicated very clearly m the 
case of an over-heated field of coil of fine wire. Here the first 
outside layers will usually be found in a fairly good condition, 
but at a comparatively little distance inside the coil there may 
be severe roasting or evidence of overheating, which may be 

FIG. 6 

Temp at Center 
N Temp at Edge 

FIG. 8 


FIG. 7 

almost as bad as at the center. (See Pig. 8.) In such case, the 
temperature measurement on the outside of the coil is no satis- 
factory indication of the hot-spot temperature A temperature 
measurement by resistance, while a closer indication than that 
by thermometer, also may be very misleading It may be stated 
that modern design tendencies are toward comparatively shallow 
field coils, largely on account of this condition 


The whole object of this paper is to show the problem of tem- 
perature distribution and temperature measurement, as it 
actually is. It is the writer's desire to show that no hard and fast 
rules can be made for determining the facts in the case, and that 


the best rules and methods now practicable are only approximate. 
The present limitations set for insulating materials are much 
higher than were considered practicable only a few years ago. 
This is not because the limits have been raised, but because, 
through a better understanding of the facts, the real upper 
limits of temperature as fixed by durability of insulation, are 
now known to be considerably higher than was believed to be 
the case only a short time ago If the real limits were in accord- 
ance with former beliefs, then all the evidence of the more accu- 
rate modern tests and data would indicate that the vast majority 
of the existing electrical machines should have "roasted out" 
comparatively early in their operation The higher temperature 
limits were there, but were not recognized Now we recognize 
them and attempt to make reasonable allowances for differences 
between the measurable temperatures *and the actual hottest 
parts. The present method may be crude, but we are not going 
at it blindly, as was formerly the case. Formerly the manu- 
facturer took the real responsibility for making a machine that 
was safe for the service, whatever the guarantees called for. 
Today the responsibility is still his, but he is attempting to 
educate the public to a knowledge of his real problems, and to a 
recognition that temperature determination is far from being 
an exact art. There is no sharply defined line between good, and 
bad in the insulating materials as affected by temperature, con- 
sequently there is no sharp line between safe and unsafe 


I have made tip a sketch which brings out much better than 
any description, some of the fundamental differences between 
Class "A" and Class "B " insulations. These might be called the " 
time-temperature curves for these insulations. These must be 
considered as approximations only, as, from the very nature of the 
materials themselves, no exact curves are possible. The important 
feature to be considered in the curves, is the general shape rather 
than any absolute values. 

We have made a great many temperature tests of insulations 
to determine their durability; also we have made examinations of 
a very large number of windings which have been in service for 
many years, but for which we had only approximate data as to 
temperatures. Obviously it is impracticable to carry on an ac- 
curate life test covering a long period of years, so what we did in 


most of our tests, was to carry the temperatures up to such 
points that destruction was either reached or indicated in a 
comparatively limited period of time. 

Curve A indicates approximately the durability of class A 
insulations for various temperatures This should be recognized 
as being approximate, but it is optimistic rather than pessimistic. 

Curve B applies to well built class B insulations, as now fur- 
nished by some of the electrical manufacturing companies. 
Such insulations contain a large percentage of heat resisting 
materials with a comparatively small percent of binding material 
and the insulation is applied so tightly that deterioration or de- 
struction of the binder does not appreciably loosen up the true 
insulating material 

Considering curve A, taking 105 deg. cent., as the ultimate 
temperature limit for long life without undue deterioration, then 
with a very slight increase in temperature, say to 115 deg. cent., 
the life is shortened very much, and at 125 deg. cent, such in- 
sulation is good for only a very few months at the most. At 
150 deg. cent, it has an exceedingly short life. 

Next considering curve B, our available data indicate that for 
over twelve months operation at 200 deg. cent., the insulation is 
in first class shape; in fact, much better than class A insulation 
at 110 deg. cent., for the same length of time. At 300 deg. 
cent, for six months, the insulation really shows better than class 
A insulation at 115 deg. cent, for the same length of time, and, 
at 400 deg. cent., the class B insulation for three months is better 
than class A insulation at 125 deg. cent, for the same length 
of time. If we now assume the continuous life for the class B 
insulation as 150 deg. cent., then it is seen that a 33 percent 
increase in temperature for one year is no more harmful than a 
5 percent increase in temperature over the 105 deg. cent, for 
class A insulations for one year. Also a 100 percent increase in 
temperature above its continuoiis limit for six months is com- 


parable with a 10 per cent increase in temperature for class A 
insulation for the same period. For still higher temperatures 
the percentage is far more in favor of class B. 

What I want to bring out in particular by means of this dia- 
gram, is that the factor of safety for overloads is vastly greater 
for class B than for class A insulations, on the basis of continu- 
ous life being taken as ISO deg. cent, and 105 deg. cent., respec- 
tively. Part of this difference is inherently in the characteris- 
tics of the materials themselves, but no doubt part of it is due 
to the fact that the arbitrary ISO deg. cent, limit set for properly 
built class B materials is considerably too low in comparison 
with 105 deg. cent, for class A. But, whatever the explanation, 
the difference is there. 

In regard to the very high temperatures for class B insulations, 
such as 300 deg. and 400 deg. cent, shown in curve B, attention 
should be called to the fact that unless there is an exceedingly 
high temperature drop through the insulation itself, any outside 
supporting layer or wrapper of fibrous materials is liable to be- 
come unduly heated and may disintegrate. Therefore, while the 
insulation proper might stand 400 deg. cent., for instance, yet if 
this was continued for any considerable length of time, so that 
the outside supporting material became excessively heated, such 
material would have to be of something else than the usual treated 
tape or fibrous wrappers. However, it so happens that very 
high temperatures are rarely attained in practice, except in the 
case of armature conductors buried in slots. In such case the 
surrounding iron assists very materially in cooling the finishing 
wrapper on the coils, unless the high temperature is maintained 
for a very considerable period. 

Some are inclined to look askance at mica at 150 deg. to 200 
deg. cent., but it must be remembered that in certain heating 
apparatus mica is used up to 500 deg. cent, and, in some cases, 
even up to 750 deg. cent. Practically all micas will stand up 
to about 600 deg. cent , without undue deterioration, and some 
grades will stand up to 1000 deg cent. From this viewpoint, 
the temperature of 150 deg. to 200 deg. cent, in armature coils 
appears to be very low and the whole matter turns upon the way 
such mica is used. If the percentage of mica in the insulation is 
relatively high and the mica is put on so tightly that the binding 
material can disintegrate and loosen up and yet the natural elas- 
ticity or springiness of the mica can hold the insulation tightly in 


place, then such insulation can stand very high temperature with-* 
out injury. But, if the mica is wound or placed so loosely that this 
disintegration of the binding or supporting material allows the mica, 
part to loosen up materially, then the insulation qualities may still 
be very good from the dielectric standpoint, but may be in such 
poor shape mechanically that vibration or shocks may shift it 
or displace it sufficiently to injure it as an insulator. This de- 
fect is a mechanical one and not in the quality of the material itself. 
Mr. Junkersfeld has spoken of some of his early experiences 
with high temperature, and he mentioned that the data which 
he and his associates obtained have had a marked nfluence in 
leading the manufacturers toward better grades of insulation. 
This is no doubt correct, but I wish to call attention to the fact 
that the manufacturers were also following this matter inde- 
pendently of the operating companies, with the same end in 
view. For instance, the company with which I am associated, 
insulated the 1894 Niagara generators with mica. We did not 
know whether such insulation was required, but we thought t 
was good material and so put it on. Later tests showed that 
this was a very fortunate decision, and now 7 after twenty years 
of operation, this insulation is still in very good shape, although 
subjected to very much higher temperatures than originally con- 
templated, 150 deg. to 200 deg. cent, being not uncommon ac- 
cording to later tests. Also in 1898 and 1899 the large engine- 
type Manhattan Railway generators had mica insulation, in the 
form of wrappers, on the armature coils. Following this, mica 
insulation was used for quite a number of years, mostly on large 
high- voltage alternators. About 1904 we built some large capac- 
ity 60-cycle turbo-generators on which we used mica wrappers 
on the armature coils. In service, one of these machines was 
injured from some mechanical cause and we had to rewind it, 
One of the fads about this time was special oiled-linen tape 
insulation, and quite a pressure was brought to bear upon us to 
rewind this machine with such oiled tape. With this insulation 
the armature broke down in a comparatively short time (within 
a few months, if I remember rightly). When the coils were 
removed, the outside layer of insulation next to the iron was 
found to be apparently in fair shape, but next to the copper the 
insulation showed indications of being excessively heated; in 
fact, it was badly carbonized in some places. We then reinsti- 
lated with tnica aad the machine was operated for many years 
without trouble. Here was a direct comparison between class A 


and class B insulations. I do not know how hot those coils ran, 
but, judging from the appearance of the oiled-tape insulation, 
it must have been materially above 125 deg. cent. Here was a 
fortunate instance where the machine was first insulated with 
mica tape and then afterwards insulated with fibrous materials, 
so that actual comparison was obtained with the two materials 
This was ten to twelve years ago, so that it cannot be said that 
experience showing the relative merits of these two types of insu- 
lation is only of recent date. 

In the same way similar experience was obtained with field 
insulation. Practically all our early turbo-generator fields were 
insulated with fibrous sheet materials. Numerous instances oc- 
curred where such insulations deteriorated so much that re- 
winding was required. This led to numerous tests for tempera- 
ture. In some of these earlier machines there was evidence of 
practically uniform overheating throughout the whole winding, 
thus indicating practically uniform temperature. In such cases 
it was comparatively easy to approximate the ultimate temper- 
ature from readings of the field currents and the field volts, thus 
obtaining the increase in resistance and from this the temperature 
rise. Such tests soon developed the fact that temperatures of 
110 deg. to 125 deg. cent were not uncommon on the earlier 
turbo-fields, while with the increased capacities and higher 
speeds, toward which we were continually tending, the indica- 
tions were that still higher temperatures would be attained. 
This led to the development of mica insulation for the field 
windings of turbo-generators In 1906 and 1907 a number of the 
earlier hot fields were rewound with mica and such fields have 
been operating up to the present time, or until discarded in 
favor of larger units. The record with these mica insulated 
fields has been extremely good. In some of the tests which we 
made on these earlier machines to determine the suitability of 
mica for field insulation, we carried one field up to 250 deg. cent, 
for forty-eight hours, and would have continued the test very 
much longer, but the conduction of heat from the core through 
the shaft to the bearings was sufficient to overheat them How- 
ever, at the end of this test the insulation was found to be in 
absolutely good condition. This was a very mild test, in view 
of our later investigations on mica, but at that time it was con- 
sidered wonderful* I am simply bringing up such points to 
indicate that mica has been used quite extensively on turbo- 
generators for many years. 


FOREWORD The material given was first presented in a discussion 
at one of the meetings of the Association of the Edison Illumin- 
ating Companies, September, 1915. It was afterwards revised 
for publication in the Electric Journal. (ED.) 

IN the artificial cooling of power-house and sub-station apparatus 
especially that of large capacity, a number of conditions have 
developed from time to time which have given trouble or which 
have provoked more or less discussion. A number of these points 
are here presented briefly. 


There is a very definite physical relation between the heat 
which must be dissipated from a machine, the resultant temper- 
ature rise, and the quantity of air which is passed through the 
machine to carry away the heat. The law is that one kilowatt of 
loss dissipated into the air will raise 100 cu. ft. of air 18 degrees C. 
in one minute. Therefore, if there is a definite loss to be 
dissipated by the ventilating air and a desired limit to the permis- 
sible rise in the temperature of the air leaving the machine, then 
there must be a definite volume of air per minute through the 
machine. The problem then resolves itself into getting this air 
through the machine or apparatus. An allied problem lies in the 
means for getting the heat from the copper or iron to the air. 


The pressure required for a given quantity of air is dependent 
upon the size of air passages or apertures, upon the shapes of the 
ducts, i. e , number of bends, abruptness of bends, length of ducts, 
etc., and upon the velocity of the air. Too sma.11 passages means 
high velocity of the air, with consequent high pressure required. 
However, with many classes of artificially-cooled apparatus, the 
space available for air passages is comparatively small at some 
places, so that the air velocities are very high. This means 
ventilation losses, but in many cases these are unavoidable with- 
out radical changes in design whfth, in themselves, would mean 
increased losses of other sorts equal to, or greater than, the possible 
reduction in ventilation loss. 



Usually, the greater part of the pressure developed by the 
ventilating fans is used up in the ducts or passages through the 
machine itself. However, not infrequently, part of the pressure 
is taken up by restrictions of some sort in the inlet or outlet con- 
duits If the machine is self-cooling in the sense that the rotor 
carries its own fan, then such restrictions in the conduits may 
very seriously affect the quanity of air which passes through the 
machine. It is, of course, possible to make the ventilating fans 
of greater capacity, just to take care of such contingencies, but this 
would be penalizing good engineering to take care of bad, for the 
losses due windage would then be unduly large where proper con- 
duits are furnished. Cases have been noted where as much as 30 
to 40 percent of the available pressure has been taken up by 
improperly designed inlet pipes. 

Where the ventilating fans are driven independently of the 
main rotor, that is, by motors, it is practicable to vary the air 
pressure to suit the requirements, provided the driving motor can 
have its speed adjusted over a suitable range. This makes the 
ventilation more or less independent of restrictions. It has the 
further advantage of allowing the quantity of air to be varied to 
suit the load conditions. By this means an increased quantity 
of air, but with correspondingly increased windage loss, is avail- 
able with heavy loads, while less air with lower losses may be used 
at lighter load This arrangement is somewhat more efficient 
than the self-cooled arrangement with the fans on the main motor 
shaft, principally because of the excessively high fan speeds com- 
mon in the latter case, especially on turbo-generators in which fan 
speeds far above efficient operation are used. In fact, in high- 
speed turbo-generators fan efficiencies of 20 to 30 percent are not 
unusual. Much better efficiencies can be obtained from separate 
slower speed fans (50 to 60 percent). However, the reduction in 
windage losses is not proportional to the increased efficiency of 
the fans, for part of the windage loss is due to ''churning" of the 
air passing over the rotor, and this will be present regardless of the 
method of supplying air to the machine and is, to a certain extent, 
a function of the quantity of air which passes through the air-gap. 
This part of the windage loss is, therefore, greater in those machines 
where all the cooling air passes through the air-gap than is the case 
with those types where a considerable part of the air passes di- 
rectly through the armature core, as in axially ventilated stators. 
In either type of ventilation, variation in the quantity of air with 
load is advantageous. 



This was a matter of little importance in the days of small 
capacity per unit-area of generating room. However, in these 
latter days, where capacities of from three to ten times those of 
former days are developed in the same space, the question of 
disposal of hot air from the machines is becoming important. Large 
turbo-generators may require from 50,000 to 100,000 cu. ft. of air 
per minute. Five or six such machines in one generator room, 
operating at full load, means 250,000 to 500,000 cu. ft. of air per 
minute pouring into the room, this air being from 20 to 25 degrees 
C. above the normal air temperature. Obviously some provision 
should be made for getting this hot air out of the room. In the 
average generator room, the total cubical capacity of the room will 
be only five to ten times the total volume of air passing through 
the machines per minute. This gives a good quantitative idea of 
the extent of ventilation required in order to prevent undue tem- 
perature rises of the room air. In some of the more modern 
stations provision has been made for exhausting the hot air from 
the machines into the boiler room. 


The enormous quantity of air required for ventilating large 
turbo-generators brings up the question of amount of dirt carried 
into the machines by such air. Assume, for instance, 60,000 cu. 
ft. of air per minute through a given machine. This weighs ap- 
proximately 4800 pounds. The usual turbo-generator will, there- 
fore, pass through itself, each thirty to forty minutes, a weight of 
air equal to its own total weight. Or, presenting the matter in 
another way, 45,000,000 cu. ft. of air passes through in twelve 
hours. Assuming as a rough approximation, that only one 
hundredth-millionth of the volume of air consists of dust or foreign 
particles, then the above means that 0.45 cu. ft. of dust passes 
through the machine in twelve hours, or 45 cu. ft. in 100 days of 
twelve hours each. If the air inlet is in a dusty place, the above is 
not at all an impossibility. Of course, a considerable part of this 
dust will go directly through the machine, but in the air swirls 
and eddies inside the machine some of it win be deposited, and 
eventually this becomes a considerable handicap to the ventilation. 
This dust acts harmfully in two ways. In the first place, it may 
partially close the ventilating passages and thus decrease^ the 
quantity of veatilatiQ^ air> la the second place, it may form a 


coating upon the heat-radiating surfaces so that the cooling air 
cannot come directly in contact with such surfaces. Ordinarily, 
in dissipating heat from a surface to the air, a thin film of hot air 
adheres to the surface, and the heat is conveyed from the surface 
through this film to the moving air. With high-velocity air 
striking the surface, this film of hot air is scoured away from the 
surface, so that new air continually comes in contact with the 
surface. If however, a coating of dust, or of other heat-insulating 
particles, gathers on the surface, then the ventilating air cannot 
come in direct contact with the surface and the heat-dissipation is 
at a lower rate. Dirt is particularly liable to adhere to the sur- 
faces in case minute particles of oil are carried into the machine. 


Air washers are now being installed very generally in large 
generating plants to clean the air which passes through the ma- 
chines. These washers have beneficial results in two ways. In 
the first place, they clean the air, thus preventing, to a great 
extent, the deposit of dirt in the machine. In the second place, 
they cool the air in hot weather, thus directly improving the 
capacity of the machine by allowing a greater temperature in- 
crease without exceeding a specified limit of temperature. A 
number of attempts have been made to cool turbo-generators by 
means of water in suspension in the ingoing air, that is, by "fog." 
Such methods as yet have shown no particular promise. 

Instances have occurred where fine, dry snow has been drawn 
into artificially-cooled apparatus and there melted, with the 
formation of water on the windings. This can only happen in 
cold weather and could be avoided in several ways. An opening 
from the inside of the building could be provided in the air intake 
so that the ventilating air is not taken from the outside. Another 
method would be to use the air washer at such times, by which 
means the snow would be abstracted. It is usually not con- 
sidered advantageous to operate the washers during extremely 
cold weather, but in case of incoming snow there may be con- 
siderable advantage in operating the washers. 


A number of instances have been noted where, with the in- 
coming air at an extremely low temperature (near zero F.), the 
end windings of turbo-generators were found covered with a film 
of water. In one case this proved disastrous. Apparently this 


was not condensation of water from the air, in the ordinary sense 
for the armature windings were much hotter than the incoming air. 
One explanation is that this is due to "frozen fog," or ice particles 
in suspension, which melt when they come in contact with the 
heated parts of the machine. One remedy for this trouble is in 
the use of doors from the interior of the building to the inlet pipe, 
which can be opened in extremely cold weather to admit warmer 


When a fire is started in such apparatus the artificial ventila- 
tion tends to spread the fire very quickly, especially in turbo- 
generators. Various remedies have been proposed, such as 
firedoors or dampers in the inlet conduits, the use of "fireproof 1 
end windings, chemical extinguishers, etc. Firedoors have 
proven only partially effective, possibly due to the fact that it is 
difficult to shut off the air completely. For instance, in a ma- 
chine taking 50,000 cu. ft. of air per minute, if 99 percent of the 
air is shut off, the remaining 500 cu. ft which can pass through the 
machine may be sufficient to maintain quite a destructive blaze. 
Furthermore, there are liable to be small leaks around the end 
housings of the machine which will admit a little air. 

It has also been suggested that equally good results would be 
obtained by enclosing the outlets from the machine in case of fire, 
thus retaining the products of combustion inside the machine, 
these forming a fairly good fire extinguisher in themselves. 

As regards chemical extinguishers, these are practically useless 
unless the incoming air can be almost completely shut off. With 
50,000 cu. ft, of air, for instance, passing through a machine, the 
small amount of gas which the extinguisher could furnish would be 
so diluted as to be worthless. One point to keep in mind in apply- 
ing extinguishing gases is that they must be applied at the incoming 
side of the fan. 

In some cases of fire the operators have turned on water from 
high-pressure mains, on the theory that, while water may ruin the 
insulation, yet fire may result in still greater damage. 

Various attempts have been made to produce * 'fireproof" 
insulations for the end windings of turbo-generators. The diffi- 
culty lies in the fact that available fireproof materials, such as mica 
and asbestos, cannot be used alone. Mica requires some support- 
ing or binding material, while asbestos requires some filling 


varnish in order to obtain suitable insulating quality. It is these 
binding or filling materials that are the real source of trouble, for 
these give off gases if the temperature is sufficiently high, and these 
tend to maintain or increase the blaze. Thus the outlook is not 
very promising. 

When the initial blaze is produced by a short-circuit or arc 
inside the machine, a sudden interruption of the excitation, by kill- 
ing the voltage, may extinguish the arc before a general conflagra- 
tion is established. If the excitation is from a motor generator 
across the terminals of the machine, then a short-circuit in the 
machine may automatically shut down the exciter set. In the same 
way, if the ventilating fan is driven by a motor across the terminals 
of the machine, the ventilation may decrease automatically. 

The above covers various suggested methods for preventing 
damage by fire inside such apparatus, All of them are admittedly 
defective, but each of them possesses some merit. A simple satis- 
factory method of fire protection for such apparatus is much to be 


Recent high-speed turbo-generator rotors are all of the cylin- 
drical type, with relatively smooth exterior surfaces. Neverthe- 
less, due to their enormously high peripheral speeds and the great 
quantity of air through the air gaps, there is always very consider- 
able noise developed inside the machines themselves. As such 
machines are always very completely enclosed, except through 
their outlet and inlet pipes or openings, these latter are usually 
responsible for any complaints regarding noise. Several cases 
have developed where the inlet conduits, opening directly to the 
outside of the btiilding, have permitted undue noise. In other 
cases, sheet metal conduits have acted as sounding tubes and ap- 
parently have exaggerated the noise. Changing to plaster-filled 
expanded metal conduits has helped in some cases. In other cases, 
carrying the conduits up to the roof of the building has proved 
effective, A secondary result of this arrangement is that deaner 
air is obtained, unless the inlet is exposed to an undue amount of 
dirt from the chimneys. 


FOREWORD This paper was prepared for the eighth annual conven- 
tion of the Association of Iron & Steel Electrical Engineers held at 
Cleveland, September, 1914. The object of the paper was to 
present, in as simple form as possible, certain problems, such as 
insulation, commutation, speed control of induction motors, etc., 
which particularly concerned iron and steel electrical engineers. 


IN steel mill electrical work there are a number of subjects which 
are of very particular interest at present. In both alternating 
and direct-current apparatus, there is the general subject of insul- 
ation troubles, which is always open to discussion. In direct- 
current work, commutation and commutator troubles are subjects 
which are always with us. Also, as the induction motor is prob- 
ably used more than any other in mill work, the problem of obtain- 
ing variations and adjustments in speed with this type of motor 
has become a very important one. In alternating-current work, 
there is the question of the most suitable frequency, which has 
come up prominently in the past two or three years. While these 
various subjects may appear to be more or less disconnected, yet, 
in fact, they are already allied in mill work, and all steel mill 
electrical engineers are liable to be called upon to deal with them. 
In the presentation of these subjects, a semi-technical method 
is followed, and all mathematics, except where masked under some 
other form, are omitted. The various subjects are treated in the 
order of their convenience, and without regard to their relative 


Practically all electrical apparatus uses insulation in one form 
or another. Such insulation in general constitutes the weakest 
part of the machine, both mechanically and electrically. Insofar 
as the generation or utilization of energy is concerned, its functions 
are passive, it serving merely as a protection. But in another way, 
its functions unf ortunately are not passive, namely, in its effect on 
heat flow and dissipation. In most cases, the parts which have to 
be insulated are heat-generating. This is especially true of the 



windings of electrical apparatus. Experience shows that aH 
electric insulators are heat insulators to a great extent, and ex- 
tremely good heat insulators in the case of the most practicable 
materials. It is well known that the best way to apply heat in- 
sulations is in the form of superimposed layers, and this happens 
to be the most practicable way of applying most electrical insula- 
tions. It is also weU known that air pockets in heat insulations 
improve their heat-insulating qualities. It is partly on this 
account that, in the application of electric insulations, air pockets 
are avoided as much as possible, and endeavor is made to fill such 
pockets with varnishes or impregnating gums which act as better 
heat conductors than air or gases. In general, it may be said that 
heat is transmitted more effectively by conduction through solid 
bodies, or between solid bodies in contact, than by convection 
through gaseous bodies. Therefore, the more solid, or the better 
filled is the insulation, the better it will conduct heat as a rule, and, 
in fact, there is not such a great difference between the heat-con- 
ducting qualities of the various commercial insulations, on the 
basis of equal solidity. The principal differences are found in the 
ways the materials are applied. While some materials may con- 
duct heat two or three times as well as others, yet this difference is 
very small compared with the difference in heat-conducting 
qualities between ordinary insulations and any of the so-called 
electrical conductors, such as metals. For instance, a difference 
of temperature of 1C. between the opposite sides of an inch cube 
of copper will allow a heat flow 2500 times as great as with a cor- 
responding cube built up of oil tape. And an inch cube of wrought 
iron, which is considered a poor electrical conductor, will conduct 
about 400 times as much heat as the block of insulation. There- 
fore, when compared with electrical conductors, we may say that 
the heat-conducting qualities of the usual built-up insulations are 
fairly uniform. 

The heat-conducting ability of insulation is a function of the 
thickness or distance the heat has to traverse, just as in all other 
bodies. Therefore, when a heat-generating body is covered with 
insulation, it is desirable to make such insulation as thin and 
compact as possible, where it is desirable to keep the temperature 
as low as possible. This is an elementary fact which has been very 
much neglected and overlooked in the past. 

In electrical apparatus, it may be said that it is not the tem- 
perature in the heat-generating body itself which is harmful, but 


it is the effects of such temperature upon the enclosing or con- 
tiguous insulation which must be taken into account. Most of the 
flexible insulations in every-day use do not have high heat-resisting 
characteristics. The effect of the heat usually is more harmful 
to the mechanical characteristics of the material than to the elec- 
trical characteristics. Most fibrous insulations, when exposed to 
fairly high temperatures for long periods, or exceedingly high 
temperatures for much shorter periods, show a tendency to become 
very brittle, and, in time, they may even carbonize to a greater or 
less extent. However, for moderate voltage stresses, even this 
very dry or semi-carbonized condition of the insulation does not 
appear to seriously affect its insulating qualities. The real harm 
lies in deterioration or possible injury of its mechanical properties 
that is, it may become so brittle that it will not stand mechan- 
ical shocks or vibrations, and may crack or scale off so that its 
insulating qualities are impaired simply through mechanical 
defects. Here is where certain filling or impregnating varnishes 
or gums are particularly useful. As the fibrous insulation tends 
to become brittle at high temperatures, the varnish or gum may 
tend to soften at the same temperature, and thus conteract, to a 
certain extent, the brittleness of the fibrous material itself. A 
second function of such gums or varnishes is to act as fillers for all 
spaces and interstices, and thus to assist in conduction of heat, 
but, still more, to act as a cushioning material to keep the con- 
ductors from vibrating under shocks, etc. Of course, the impreg- 
nating gums or varnishes have a certain value as insulating 
material, but probably the above functions are of far greater 
value. For instance, the ordinary cotton covering on a wire will 
stand far more abuse when treated with some kind of gum or 
varnish than when used in the dry condition, for, in the former 
case, the individual fibers of the covering are actually pasted in 
place, and are therefore much less liable to be separated and thus 
allow metal parts to come in contact. Usually what is required 
between adjacent conductors in a coil is a positive mechanical 
separation of a very limited amount. In many cases, if the bare 
conductors could be maintained at a distance apart corresponding 
to the thickness of the usual cotton covering, this would be suf- 
ficient for protection against the voltages between the wires, 
TJie layer of insulation on -the wires themselves furnishes the sim- 
plest and easiest method of obtaining this mechanical separation, 
and the varnish or gum treatment makes this separating medium 
of more mechanical and durable construction, and, at the same 
time, improves the heat-conducting qualities. 


There are limits to the heat-resisting qualities of all practic- 
able insulations. Ordinary fibrous materials of a cellulose nature 
or base, will stand about 95 C, to 100C. without becoming too 
brittle to be durable. However, the same materials, when treated 
with suitable varnishes or gums, apparently stand temperatures of 
about 105 C. without undue deterioration mechanically. At 
this temperature the material does not appear to carbonize, and 
the varnish or gum assists in maintaining mechanical continuity 
of the material. At materially higher temperatures, deterioration 
gradually takes place at a rate depending upon the actual tem- 
perature attained. Even at 150C., treated fibrous materials may 
have a total life of several months before the material becomes 
unsuited for its purpose. If such high temperature exists only 
for short periods, and during the remaining time the insulation is 
subjected to relatively low temperatures, then the life of the 
apparatus measured in years, may be fairly great. In other 
words, high peak temperatures may not be very harmful, pro- 
vided the sum total of such peak periods does not add up to as 
long a period as required to injure the materials if maintained at 
the same peak temperature steadily. However, the life of 
insulation does not decrease in direct proportion to the 
increase in temperature, but at a much faster rate. 

Other insulations in common use are mica, asbestos and 
certain varnishes and gums. Pure mica will stand enormously 
high temperatures, such as 700C. or even higher. Good grades 
of asbestos stand at least 400 C. as shown by actual test, and 
possibly very much higher. However, neither mica nor asbestos, 
in itself, is a good material for application to windings, due to 
mechanical conditions. In order to obtain flexibility, mica must 
be built in thin sheets and then assembled in the form of a paper 
or tape. This requires some continuous supporting base, usually 
a thin tough paper, to which the mica is attached by some form of 
binding gum. The result therefore consists of both high and low 
heat-resisting materials. If the continuity and durability of the 
resultant mica insulation, after application to a coil, is dependent 
upon the durability of the binding and supporting material, then 
such insulation is limited to temperatures corresponding to fibrous 
materials. If, however, the binding and supporting material can 
deteriorate without materially injuring the insulation as a whole, 
thai such composite insulation can stand comparatively high 
temperatures. In present practice, such temperatures are limited 


to approximately 1SOC. for steady operation, not because this is 
an actual limit, but largely because of lack of extended experience 
at materially higher temperatures. Apparently, such materials, 
when properly applied, will stand 300C. on peak service about 
as well as the treated fibrous materials will stand 150C, 

Asbestos, as an insulation, is pretty poor material, but as a 
mechanical separator, where high temperatures are obtained, it 
may be very effective. Due to its open fibrous character, there 
is no true over-lapping of insulating surfaces, and, to make as- 
bestos effective as an insulator, it must be filled in with some 
insulating filler or gum, in which case, the gum is the real insul- 
ator. However, asbestos may answer for a very good supporting 
material for other insulations, such as mica, when subjected to 
very high temperatures. Also, asbestos may be a suitable in- 
sulation on conductors with very low potential between them, as 
in field windings and in armature windings with low internal 
potentials. It should be considered as essentially a separating 
and supporting material, rather than as an insulation. 

In recent years, a number of synthetic resins, such as Bakalite, 
have been developed, which have a field in insulation work. 
Such materials usually have high heat-resisting qualities, but, in 
their final condition, are liable to be hard and brittle. Some of 
them are used extensively as impregnating or filling varnishes, 
and when so applied, are in fluid form, and require further baking 
to change them to the final form. Some very extravagant 
claims have been made for them by those who were not sufficiently 
acquainted with the materials and their properties. They are 
very valuable in many ways, but, like all other materials, they 
have their limitations. In their application to armature windings, 
it is advisable, in many cases, to apply the coils before being given 
the final baking on account of the greater flexibility of the unbaked 
coils. But the final treatment usually leaves the armature 
winding in such rigid condition that, in case of repairs, it may be 
necessary to completely destroy the whole winding. This looks 
like a bad feature, but to counter-balance it, it may be said that, 
for certain kinds of work, such prepared windings are less liable 
to damage, and therefore the necessity for repairs is much reduced* 
As impregnating compounds, where stiffness or rigidity is ad- 
vantageous, such materials have proved very satisfactory, but, 
where considerable flexibility is desirable, compounds of this 
nature may not pix>ve so desirable. If the impregnating material 


is brittle and is liable to cracl:, tinder stresses due to change in 
temperature, or movement, or shock, then it loses a certain part 
of its value. Where flexibility is important, gums or varnishes 
which soften with heat are desirable. 

In armature or field windings, it is very unusual to find con- 
stant temperatures throughout the whole winding, due to the 
different heat-conducting and heat-dissipating conditions in 
different parts. Therefore, the higher temperature points or ' ' hot 
spots" must be considered in fixing the insulation temperature 
limits. It is the highest temperature to which the insulation is 
subjected that must be considered in fixing the limits, and only in 
rare cases do the ordinary methods of temperature measurements 
indicate the highest temperatures actually attained. Ordinary 
thermometer measurements approximate the temperature at 
some accessible point, but this may not be, or, likely, is not the 
hottest part. A determination of the temperature by increase in 
resistance gives only an average value. Therefore, by actual 
measurement by the usual methods, the above mentioned tem- 
perature limit of 105C. for treated fibrous materials is not allow- 
able. For instance, the usual full load guarantee of 40C. rise 
with a cooling air temperature of 40C. will give 80C. as the 
temperature measured, thus allowing a margin of 25 C. for some 
higher internal temperature that is, for the hot spot. The usual 
overload guarantee of 55 C. by thermometer, with air at 40 C., 
will give 95 C. measured, or a margin of 10C for the hot spot, 
which apparently is right on the ragged edge. But then, this 55 C. 
guarantee is usually given only for overloads or intermittent 
service, and it is this condition which allows the proper margin. 
If, however, an accurate means should become practicable for 
determining the actual hot spot temperatures, then it would be 
practicable to rate machines at the 105C. measured temperature. 
As this cannot be done at present, we must fall back on a lower 
measurable temperature, and allow a suitable margin. 

In certain classes of apparatus where the higher temperature 
regions are pretty definitely known, it is possible and practicable 
in many cases, to insulate, in the hotter regions, with materials 
which have higher heat-resisting characteristics, as already de- 
scribed. This is the case in many high voltage machines, and in 
machines with very wide armature cores, such as some turbo- 
generators, high speed large capacity alternators, etc. In such 
machines, the middle part of the armature core is liable to be 


much hotter than any other part. Therefore, it is rather common 
practice in such machines to insulate the buried part of the arma- 
ture coils with composite mica insulation, which can be easily 
applied on the straight portions of the coil. On the curved end 
parts of the coils, where taped insulations are required on account 
of the curvature, much lower temperatures are usually obtained, 
and thus fibrous tape insulations are amply safe for this part. 

In apparatus subject to very heavy overloads for relatively 
short periods, excessively high temperatures may be attained by 
the copper inside the insulation, but if followed by much lighter 
load, the high temperature may drop so rapidly that no apparent 
damage occurs. Experience has shown that, not infrequently, 
local temperatures of 200C. to 300C. are attained for a very 
short time. When such temperatures occur close to any soldered 
connections, there is danger of damage due to unsoldering, for or- 
dinary commercial solders will soften at about 170C. to 180C., 
while pure tin solders will soften at about 220C. to 230C. 
Therefore, temperatures which, due to their short duration, ap- 
parently do not harm the insulation, may actually unsolder con- 

The above covers briefly the temperature part of the insula- 
tion problem. But insulating materials also serve another pur- 
pose, namely, to shield the conducting or live parts of the machine 
from other foreign conducting materials, such as dirt, grease, oil, 
water, etc. Oils and greases are usually considered as non-con- 
ducting, but when they are liable to carry with them conducting 
materials, such as copper and carbon particles, they become con- 
ductors in effect. Also, ordinary dust, or dirt, or deposit from the 
air, is a relatively poor conductor* but conducts far better than 
the usual insulations, and is therefore, to a certain extent, danger- 
ous. As a conductor, water is considered as comparatively poor, 
and yet no one would class it as an insulator. Both water and oil 
may be directly harmful and may be indirectly injurious by their 
actions upon the insulating materials themselves. In the case of 
cloth tape insulators, the cloth may be considered as simply form- 
ing a K base or reinforcing structure for the insulation proper, which 
is usually some varnish or oil compound. The insulation value 
of the material depends principally upon the continuity of the 
layers of varnish or oil. The cloth structure itself has no true 
continuity. In applying such tapes or insulations, the layers over- 
lap each other in such a way as to give the best sealed circuit. 


During the process of taping, the surface may be varnished re- 
peatedly to further seal the overlapping joints, and to obtain 
greater continuity of the insulating film. Also, in the composite 
mica insulations, the mica laminas are very thin and arranged in 
a number of layers in such a way as to overlap as completely as 
possible to form insulating films. The binding material between 
layers or films is largely for the purpose of sticking or binding 
the mica laminae to each other. Therefore, with either type of 
insulation, continuity of the insulating film, is the first requi- 
site, and any action or treatment which tends to break the 
films will naturally tend to weaken the insulation. In high 
voltage armature coils, in particular, it is of utmost importance 
that the completed coils should not be sprung or bent to such 
an extent that the insulation films are liable to be cracked or 
"buckled" at any point, for this immediately produces a local 
weakness. In all cases, extreme care should be taken in handling 
such coils, especially in placing them on the cores. Moreover, in 
machines which are liable to carry excessive currents, even momen- 
tarily, and which are thus liable to distorting magnetic stresses, 
the windings must be so braced that movements sufficient to 
crack or buckle the insulation are not permitted. There is but 
little real flexibility in such insulations when built of any con- 
siderable thickness. Insulation on cables might be cited as an 
exception, but here the insulating varnishes are soft and possibly 
semi-viscous, so that a certain amount of bending does not break 
the insulating films. To maintain this condition of soft flexible 
insulation, cables are guaranteed usually only for very low maxi- 
mum temperatures, compared with the temperatures usually 
found in electrical apparatus. 

The continuity of the insulating films may be injured in other 
ways than by bending. If, for instance, a newly insulated coil 
which has been insufficiently baked or " seasoned," is subjected 
to a comparatively high temperature for even a short time, 
certain volatile matter in the varnishes may be given off in the form 
vapor, and these vapors may force or break their way through the 
insulating films. The writer has in mind one case where a taped 
insulation was used on a rewinding, with the shortest possible 
time for the baking before applying the coils to the machine. The 
insulation tests were high, but a heavy load was thrown on the 
machine at once and carried for several hours. At the end of this 
rim, the insulation test showed that the insulating material had 


deteriorated very greatly, so much so that the machine was 
considered to be in a dangerous condition. Upon removing some 
of the coils, an examination showed what looked like little vol- 
canoes all over the surface of the insulation. Further investiga- 
tion showed that these were real volcanoes, for the high inter- 
nal temperature had vaporized some of the original solvents which 
had not been entirely removed from the varnish, and such vapors 
had actually erupted through the superimposed strata represented 
by the insulating films or varnishes. Therefore, at each one of 
these points of eruption, there was a breakdown of the insulating 
strength of greater or less depth, depending upon where the vapor 
was formed. This is cited simply as a very good illustration of 
what can happen in "green" insulation. 

Another source of difficulty which is not unusual, is that due 
to water, or oil, or other foreign materials getting into the insula- 
tion. Submersion of electrical apparatus, due to floods, is not 
uncommon in industrial plants, due to their proximity to rivers, 
in many cases. In some cases, experience has shown that a 
flooded machine can be dried out with apparently no harmful 
after effects, while in other cases, it has been found almost hope- 
less to save the apparatus. This depends to some extent upon the 
kind and character of the insulation and the means for getting rid 
of the water without injuring the insulation itself. If water has 
percolated into the coil and becomes sealed or trapped inside, then 
high internal temperatures obtained by any means may simply 
vaporize the water without getting rid of it. If the insulation is 
porous, the water may be driven off readily. If the drying heat is 
applied from the outside, then, before the center is heated suffic- 
iently to vaporize the water, the outside insulating films may seal 
together under the higher outside temperature, so that the internal 
vapors cannot escape except by disrupting the film If, on the 
other hand, heat is applied from the inside, by means of current 
for instance, and the heating is too rapid, vapor may be formed 
more rapidly than it can percolate through the insulation, and it 
may injure the insulation in escaping. fc Also, in the case of elec- 
trical heating, non-uniformity of temperature must be taken into 
account. For instance, the armature winding of a high voltage 
alternator might be operated on a short circuit for the purpose of 
drying out. The drying out current may be so high that the 
center of the armature core is considerably above 100C. or the 
boiling point of water, while the end windings may be 30 percent 


or 40 percent cooler. In such case, the water in the hot part of the 
coils is simply vaporized and driven to the end windings and there 
condensed. This is not an unusual condition in drying out high 
voltage windings which contain moisture. One instance may be 
cited, where, several years ago the power house of the Westing- 
house Electric & Manufacturing Co. was flooded for several days, 
and several large 2200 -volt turbo-generators were partly sub- 
merged. One of these machines was dried out on short circuit for 
about a week at a temperature of possibly 120C. inside the coil. 
At the end of this time, no leak to ground showed and the machine 
was put in service. A few weeks afterwards, a short circuit 
occured inside one of the coils, in the end winding. When dis- 
mantled, this coil was found to be sopping wet in the end portion, 
although the buried part of the coil was fairly dry. The baking 
process had simply distilled the water from the center to the end 
parts. An ftTa.Tnina.tinn of others of the submerged coils showed 
the same condition. It is possible that untaping of the end wind- 
ing sufficiently to have allowed the escape of vapor would have 
allowed this machine to dry out properly, but apparently this 
would not be the case unless the end windings in themselves could 
have been brought up to a temperature considerably above 100C. 
and this might have meant 1SOC. in the buried portion, which 
would probably have been injurious, except to mica insulations, 
which did not happen to be on these machines. Furthermore, it 
is not always easy to get rid of moisture, even at 100C. with 
fibrous insulations. One very effective manner of doing so is by 
means of a vacuum. Experience has shown that if apparatus to 
be dried out is heated to tne boiling point, in a vacuum, the moist- 
ure usually is removed very completely. For most effective re- 
sults the water should be vaporized, for, under some conditions, 
and with some materials, the force of capillarity may approximate 
IS Ibs. so that a good vacuum alone may not be able to overcome 
the capillary action. From the scientific standpoint, the use of 
vacuums in drying goes much further than the above. For ex- 
ample, the boiling point of water is very much reduced in a 
vacuum, so that materially lower temperatures may be used for re- 
moving water than would otherwise be the case. For rapid drying 
under ordinary air pressures, considerably over 100C. is needed, 
while in a fairly good vacuum, 100C. or less, may allow a very 
rapid evaporation of moisture and a correspondingly rapid and 
thorough drying. 


In the same flood which submerged' the generator above re- 
ferred to, vast quantities of other apparatus of various types and 
designs were also flooded, and, in drying out this apparatus, a 
great deal of valuable experience and data were obtained. A 
summation of this and other experience may be of value and 
interest, and is therefore given below. 

Low voltage alternating-current windings, such as induction 
motors and alternating-current generators for 600 volts and less, 
dried out very readily by the application of current to the windings. 

In general, low voltage, direct-current armature windings were 
dried out by the application of current or by baking in ovens. How- 
ever, there was great difficulty in drying out commutators, and 
eventually the only real satisfactory way proved to be by heating 
them in a vacuum. Therefore, the finaldrying out of direct-current 
armatures was principally by vacuum. 

The complete drying out of field coils was very difficult, either 
by current heating or by ovens. However, the outside of the coils 
could, in many cases, be dried sufficiently to show practically no 
ground, while the inside of the coil was still wet. In most cases, 
field coils could be operated in this condition and could eventually 
dry themselves out. This would probably be satisfactory for 
drying out individual machines, but was not considered satis- 
factory for stock apparatus. Vaccum drying under high temper- 
ature proved most satisfactory, and this was adopted. 

High voltage windings for generators and transformers were 
dried out in vacuum, no other methods proving entirely satisfac- 
tory, except in individual instances. 

It may be borne in mind that his was a situation where super- 
ficial correction was not permissible. During the various tests 
and methods which were carried out, searching investigations of 
the results were made in order to determine the sufficiency of the 
method used. Field coils and armature coils were opened up at 
various stages of the process for examination. For instance, one 
of a lot of street railway armatures which were dried in an oven 
until apparently all right, was dismantled for exa.mina.tion. The 
windings appeared to be fairly well dried out, but upon opening 
the commutator V-ring, very considerable- moisture was found 
tinder the commutator bars and IB the mica bushing. Apparently, 
oven baiing would not remove this satisfactorily. The commut- 
ator was then sealed tightly and the armature was then put in a 
vacuum oven and dried for a few hoars. After this all water had 


disappeared from the commutator. Another commutator was then 
opened and purposely filled -with water and then closed and sealed 
as tightly as possible before placing in the vacuum oven. After 
an over-night's treatment, the inside of the commutator was found 
to be entirely free from moisture. This test illustrated the ability 
of the vacuum oven to remove water. It was then adopted very 
generally for drying out such apparatus as was liable to have 
water sealed or trapped inside the insulation. It must be under- 
stood, however, that certain kinds of apparatus were dried out 
just about as well using temperature alone. In these cases, how- 
ever, as intimated before, the vaporized water could readily 
escape to the air. 

There is one condition, however, where even vacuum oven 
drying may not produce the desired result, for the operation of 
drawing off the water may injure the insulating varnish films. To 
illustrate, some years ago one of the large power plants at Niagara 
Falls was flooded to a considerable depth by an ice jam, which 
backed the water up into the power house. The machines were 
flooded to a depth of twenty or thirty feet for a period of several 
days, and the windings were pretty thoroughly impregnated 
throughout with water. Strenuous attempts were made to dry out 
these windings by heating to temperatures of 12SC. or higher. 
The end windings were untaped at points to allow the moisture to 
escape. Also, attempts were made to create a vacuum around 
the machines by means of air-tight covers or casings and vacuum 
pumps, but this latter was not very satisfactory. After a few weeks, 
apparently but little progress had been made. A chemist then 
advanced the suggestion that, IE linseed oil varnishes had been 
used in the insulation, then, under the conditions of flooding 
which had occurred in this plant, the varnish itself would have 
absorbed water, and he was of the opinion that heating alone, 
unless carried up to the destructive point, would not drive off 
this water. Investigations were then made along this line, and 
it was actually found that the varnished films were thoroughly 
filled with water, and moreover, this water could not be removed 
without more or less injury to the film itself. For moderate or low 
voltage machines, apparently, the removal of the water would not 
injure the insulation sufficiently to prevent operation, but in high 
voltages, such as 6600 volts or higher, the insulation would be left 
in a relatively weakened and unsafe condition. In the machines 


in question, it was found" advisable to remove the insulation en- 
tirely and replace with new. 

As a rule, field coils can be dried out in a fairly satisfactory 
manner by heating with current for a sufficiently long period. 
When a field coil is thoroughly wet inside, its resistance may fall 
considerably, due to low resistance between turns and layers, but 
when current is applied, there is but little danger of burnouts, as 
the leakage of current through the insulation is distributed over 
such large surfaces that there is no danger of burning at any one 
point, unless there is some defectively insulated point in the coil. 
Therefore, after the coil is sufficiently dried so that its leakage to 
ground, or any metal supports, is sufficiently low to be safe, then 
usually the coil can be put in operation and allowed to dry out in 
regular service. If, however, the field coil rotates and is subject 
to centrifugal or other forces, the wet condition of the internal in- 
sulation may allow internal distortions or movements which might 
cause partial short circuits. 


In the practical design and operation of electrical apparatus, 
there is no problem which is apparently more enshrouded in mys- 
tery than that of commutation. Theoretically this problem has 
been treated in various ways and analyzed to various degrees, but 
the practical results not infrequently disagree with the theoretical, 
principally because the latter are predicated upon conditions which 
are not, or can not, be obtained practically. Moreover, even when 
the problem can be correctly analyzed, and a proper solution 
indicated, it may not be practicable or feasible to apply such 
solution. In other words, the theory may show just where a 
trouble lies, but the application of a suitable remedy is another 

The difficulty is that the theories of commutation are built 
upon many conditions which are inter-dependent. But many of 
these conditions differ, in different types, designs or sizes of 
machines, and, even in a given machine, a change in one condition 
may greatly modify other conditions. For instance, the local or 
short-circuit currents which are present in the coils short-circuited 
by the brushes during the operation of commutation, have a pre- 
ponderating influence on the commutation, and yet, these local 
currents are greatly influenced by many conditions, such as the 
dimensions and grade of brush, condition of contact surface, 


rigidity and type of brush holder, etc. Obviously, with such 
variable elements entering into the problem, any rigid analysis is 
most difficult. In such cases, the theory is valuable in locating 
any probable causes of difficulty. 

A great variety of conditions or phenomena are encountered 
in commutating machinery, which require more or less knowledge 
of the theory of commutation in order to understand them. For 
instance, the causes for sparking, flashing, burning of brushes, 
undue wear of commutator copper, etc , are all directly related to the 
commutation problem. Even questions of composition, or grade 
of brushes, type of brush holder, etc., are related problems. 

As it is the writer's purpose to treat the above points from 
the practical side, he considers it advisable first to give a brief 
explanation of the commutation problem from the standpoint 
which he has found to be simplest and most illuminating. 


A direct-current armature, when carrying current, becomes 
a true electro-magnet, with the poles located on the armature at 
positions corresponding to those coils which are in contact with 

PIG. 1. 

the brushes. If the armature were surrounded by a smooth ring 
of iron, (Fig. 1) then a magnetic field would be set up between the 
armature and the surrounding ring, this field having maximum 
values at points corresponding to the brush contacts and zero 
values midway between. The magnetic field would rise, from 
the zero points, uniformly to the highest values, because the arma- 
ture winding, which is a magnetizing winding, is uniformly dis- 
tributed over the armature core. If, now, deep notches were cut 
in the surrounding ring at points corresponding to the highest 
field, (Fig. 2) there would be comparatively weak field at these 
points, due to the high reluctance of the gap or notch. The ex- 


ternal ring in this case represents the field structure of an ordinary 
D. C. machine, and, in such machines, the armature when carrying 
current actually tends to set up magnetic fields in this manner, and 
the coils in contact with the brushes are practically always located 
in the space between poles, that is, in the notches in the sur- 
rounding ring in the above illustration. Also, the coils in contact 
with the brushes are those in which the current is reversed in 
direction when passing under the brushes, or, in other words, are the 
commutated coils. Therefore, it may be seen that the commutated 
coils always lie in what would be the strongest magnetic field set 
up by the armature winding, if there were no polar notches. But, 
even with such notches or interpolar spaces, the armature winding, 
when carrying load, tends to set up some magnetic field, part of 
which is in the space over the armature, and part of it across the 
armature slots. This latter flux from slot to slot, is not influenced 
by the notches in the surrounding iron, that is, by the interpolar 
spaces. In addition, the armature winding sets up a magnetic 
field around the end windings. 

PIG. 2 

The annature winding, when carrying current, therefore 
always tends to set up a magnetic field, through which the arma- 
nire conductors rotate and generate e. m. f.'s, just as when they 
cut the mai'r> field set up by the field windings. These coils short- 
circuited by the brushes also generate e. m. f.'s and as their ter- 
minals, which are conomutator bars, are connected together or 
short-circuited by the brushes on the commutator, the e. m. f.'s 
generated by cutting the armature flux tend to cause local or short- 
circuit currents to flow in such coils. Such currents will be known 
hereafter as the local or short-circuit currents, to distinguish them 
from the useful or work currents of the annature. 

As an armature coil carrying current in a given direction, 
approaches and passes under the brush, the current should die 


down to zero value at the midpoint of the brush and should rise 
to full value in the opposite direction by the time the coil leaves 
the brush. This would be a theoretically perfect condition, but 
is very difficult or almost impossible to attain in practice. The 
coil, while under the brush, has an e. m. f . generated in it due to 
cutting the armature field, as already described, and a local 
current circulates. This short-circuit current normally adds to 
that of the work current before reversal, and thus tends to main- 
tain it right up to the moment that the coil passes out from under 
the brush. The reversal of the current in the short-circuited coil 
is thus accomplished almost instantaneously instead of gradually. 
If, however, this local current could be generated in ike opposite 
direction, then it would tend to oppose the work current as the coil 
came under the brush, so that the resultant current would first die 
to zero and then rise in the opposite direction, if the short-circuit 
current were just large enough; so that the work current would 
simply replace the short-circuit current in direction and value as the 
coil passes out from under the brush. Therefore, if the short- 
circuit current were of exactly the right value, the resultant 
effect would be the same as if the work current alone were present 
and this current died down to zero value as the coil passed the 
middle of the brush, and rose to full value in the opposite direction 
by the time the coil left the brush. In other words, a local current 
of the proper direction and value in the commutated coils will give 
theoretically ideal commutation. Such local current therefore, 
might be designated as the commutating current. In practice 
it is found however, that a current somewhat higher than the 
ideal value gives the best general results as will be explained 
later. In any case, however, this local current, to be effective, 
must be in the reverse direction to that which would normally 
be set up by the armature coil cutting the magnetic field due to 
the armature winding itself. This means therefore, that where 
commutation is accomplished by means of short circuit or local 
current, an external field in the opposite direction to the armature 
field is necessary for setting up such local currents. This result 
may be accomplished in several ways. The brush may be rocked 
forward or backward until the commutated coil comes under the 
magnetic field, or fringe of the field, set up by the main field 
winding. If rocked in one direction (forward in the generator, 
backward in the motor) the direction of the main field will be in 
opposition to the armature field. Obviously, if shifted into a 


strong enough external field, the armature field may be com- 
pletely neutralized at some given point, such as that of the short- 
circuited coil, or the resultant field might be even strong enough 
in the opposite direction to set up the desired local or short-circuit 
current in the commutated coil. Under this condition, ideal corn- 
mutating conditions should be obtained. However, as the 
armature magnetic field at this point tends to vary with the 
armature current, while the external field tends to remain constant, 
it is evident that the ideal resultant field will only obtain at one 
particular load. Therefore, only an average condition can be 
obtained in this way. However, by shifting the brushes backward 
or forward under the external field, the proper local or commutating 
currents should be obtained for any given load; but brush shifting 
is not usually considered a very practical method of operation, 
although required by many non-commutating pole machines in 
service. What is needed is an external field directly over the 
commutated coils which is always of the proper strength to set up 
commutating currents of the right direction and of the proper 
value with respect to the work currents, so that the resultant in 
the short-circuited coil will give the effect of the work current 
reversing at the middle of the brush at all loads. To accomplish 
this, an external field to produce this local current should always 
be in opposition, to the armature magnetic field, should be some- 
what greater in value, and should vary in proportion to the 
armature field, that is, to the armature current. This result is 
accomplished by the now well-known commutating pole, which is 
simply a small pole placed over the commutated coil, and usually 
excited by a winding directly in series with the armature, but 
having a somewhat greater magnetizing force than the armature 
winding. The function of this pole is solely to set up in the 
armature coil a local or commutating current of the proper di- 
rection and value. 

A condition which makes the problem of commutation very 
much easier to solve is the use of a relatively high resistance in the 
short-circuited path of the coil which is being commutated. Due 
to the extremely low resistance of the ordinary armature coil, a 
relatively low magnetic field set up by the armature would generate 
enormous local currents in the short-circuited coil if the resistance 
of the coil itself were the only limit. These .currents might be ten to 
twenty times as great as the normal work, current, and would add 
seriously to tfce difficulties of commutation. Even if a, commut- 


ating field were present, this would have to be proportioned and 
adjusted so accurately, to get the right value of the commutating 
current, that the construction would be almost impracticable. 
But if considerable resistance, compared with the coil itself, 
could be interposed in the short circuited path, this obviously 
would so greatly assist in fixing the value of the short-circuit 
current that undue refinement and adjustment would not be 
necessitated. Let us suppose, for instance, that the short circuit 
e. m. f , in the commutated coil is two volts, and a copper brush of 
practically zero resistance at the contact is used, then the resist- 
ance of the short-circuited coil itself limits the current which 
flows. Let us assume that this gives a local current of ten times 
the value of the work current. Now, if, instead of the zero resist- 
ance brush contact, one of about ten times the resistance of the 
coil is used, then the total resistance in circuit becomes over ten 
times as great, and the short-circuit current is cut down to a 
value comparable with that of the work current. Obviously this 
in itself would represent an easier condition of reversal without 
any external reversing field, and, with such a field, extreme ac- 
curacy in proportions and adjustment are not necessary to give ap- 
proximately the right value of the local current which assists in 
commutation. Therefore, a relatively high resistance brush, 
that is, one with contact resistance high compared with the re- 
sistance of the coil is of very material aid in commutation. This 
is wherein the carbon brush is such a successful, or even necessary 
adjunct of the commutating machine. It serves such a vital 
purpose that it may be said that, without the carbon brush or 
its equivalent, the electrical industry would never have made 
anything like the progress it has made. 

The principal function of the carbon brush being that of 
limiting the local current, it might be assumed that the advan- 
tages might be increased indefinitely by further increasing the 
resistance, but there are usually limits to all good things. The 
carbon brush increases the resistance in the path of the local 
currents, but it also increases it in the path of the work current. 
If the resistance is carried too high, the losses due to the work 
current may constitute a more serious objection than the local 
currents. Consequently, practice is one continual compromise on 
this point. In those cases where the short-circuit current is 
normally relatively small due to low value of the armature mag- 
netic field, it is obvious that a lower resistance in the short-circuit 


path can be used, or, in other words, a low resistance carbon 
brush is practicable, with consequent low loss due to work cur- 
rent. In other cases with higher inherent local current, higher 
resistance carbon will give better average results. It is thus 
obvious that one grade of carbon brush is not the best one for 
different machines unless they all have similar inherent commut- 
ating conditions. It is exceedingly difficult to give equal com- 
mutating characteristics to machines of different sizes and types, 
and, in most cases, even of the same type or line. In non-com- 
mutating pole machines, the grade of the brush is of more im- 
portance than in the commutating pole type, for in the latter, we 
have a means, in the commutating pole strength itself, of modi- 
fying or controlling the value of the local current. But the best 
commutating pole gives only average correction, that is, average 

Fig. 3. 

value of the desired local current, and the resistance of the brush 
must be depended upon to take care of pulsations or irregularities 
in the local current, acting to smooth them out to a greater or less 
extent. Thus a fairly high resistance brush is required on the 
commutating pole machine, but its resistance tisually can be 
lower than that required in the non-commutating pole type. 

The above gives a crude idea of the phenomena of commuta- 
tion. However, there are a number of very closely related condi- 
tions, such as burning and blackening of the commutator, causes 
and effects of high mica, effect of under-cutting of mica, rapid 
and undue wear of commutator copper and brushes, etc., which 
can be explained more or less directly by the above theory. 

Blackening of the commutator, high mica, and rapid wear of 
the commutator copper and brushes may aH be credited to actual 
burning, or something similar to electrolytic action, occurring under* 


the brushes. It is not usually the bright sparks at the brush tips 
which cause trouble, but it is frequently on unnoted sparking under 
the brush face. These sparks may be very minute, so much so that 
they would naturally be assumed to be harmless. The apparent 
electrolytic action under the brushes may be really a similar 
sparking action which cannot be observed. Experience has 
shown that usually there is a tendency for minute particles of the 
conducting material to be burned or carried away from the con- 
tact surfaces, (Fig. 3) depending on the direction of the current. 
These particles appear to travel in the direction the current i& flow- 
ing, but they do not always deposit on the opposite contact sur- 
face. If the current is from the brush to the commutator copper, 
the brush surface tends to be eaten away, while with the current 
from the copper to the brush, the copper eats or burns away. 

Fig. 4. 

With ordinary current densities and very low loss in the brush 
contact, this action seems to be very minute, but it appears to in 
crease rapidly with increased loss at the brush contact. There- 
fore, high current density in the brush contact may produce 
this action. This does not mean high density due to the work 
current alone, but means the high actual density, due to both 
work and local currents. In non-commutating pole machines 
and in commutating pole machines with bad adjustment, the 
local current under the brush may exceed in value the work 
current. As this adds to the work current at one edge of the 
brush and subtracts at the other edge, the result will be greatly 
increased density and high watts tinder one part of the brush. 
This may result in burning away one edge of the brush surface, 
and is frequently observedin examination of brushes. This usually 
is most noticeable where the current passes from the brush to the 
commutator, but at the holders of the other polarity, a similar 
action is tending to burn away the commutator copper. How- 
ever, the commutator mica does not tend to burn away, and there- 



fore, if the mica does not wear down mechanically at the same rate 
that the copper burns away, eventually the mica stands an in- 
finitestimal amount above the copper (Fig. 4) and the brushes will 
make a decreasingly good contact with the copper itself. This 
increases the loss at the brush contact and increases the burning 
action which results in still poorer contact, so that the results 
become cumulatively worse. If, however, these periods of burn- 
ing are intermittent, due to variable load conditions, and there is 
considerable operation at lighter loads or non-burning conditions, 
the mica may wear down somewhat and the commutator and 
brushes may polish sufficiently during these periods to mask the 
direct effects of the burning. But the results may show in 
grooving and apparent rapid wear of the commutator face. There 
may thus be burning without blackening, or without direct evi- 


Fig. 5. 

dence of high mica. If, however, the burning peiods exceed the 
polishing, then visible evidence of burning and high mica may be 
found. The brushes may also show this burning and, in some 
cases, may honey-comb badly at one edge, or even over the whole 
surface. Where one edge burns over a very distinct area, (Fig. 5) 
it usually may be assumed that the burnt area could just as well 
be cut away, with but little harm to the operation, and possibly 
some good, for the fewer the commutator bars that the brush spans 
the lower will be the total short-circuit current, as a rule. And, 
moreover, by doing away with the localized burning tinder the 
brush, it may be assumed that the commutator burns less also. 
However, cutting away part of the brush face will crowd the work 
current into the remaining portion, so that burning in this portion 


may be increased. Therefore, narrowing the brush face is not 
a general remedy for the trouble, but is successful in some cases. 

Burning of the commutator may also be coincident with a 
deposit of copper on the brush face. This is usually known as 
"picking up copper. " Apparently, in some cases, where the 
copper is burnt away from the commutator face, it actually 
collects or deposits on the brush face, forming low resistance spots 
or surfaces. This gives the equivalent of low resistance brushes, 
with consequent increase in local current and still greater burning 
action. Moreover, with several brushes in parallel, any one 
brush with copper on its face, may tend to take more than its 
share of the total work current, which tends to cause further heat- 
ing and burning. Increased heating in itself will cause a greater 
tendency of the brush to take an undue proportion of the current, 
for carbon brushes, unfortunately, have a negative coefficient of 
resistance. This means that if any brush carries more than its 
share of current and becomes heated thereby, its resistance is 
reduced and it tends to take still greater current. This is particu- 
larly the case with a large current per brush arm, with a large 
number of brushes in parallel on each arm. If one brush, for any 
reason, such as picking up copper, takes an undue share of the 
current, it may take an increasing share until the contact surface, 
or the whole commutator end of the brush, may become red hot 
and slowly disintegrate. Such action, if continued, will event- 
ually so destroy or injure the brush contact that it carries a de- 
creased current, the resistance increases, and eventually the current 
falls, not only to normal, but probably far below normal value, 
due to bad contact, and the other brushes must then assume an 
excess. If other brushes repeat this action, then the condi- 
tions become increasingly bad for the remaining brushes, and 
they may repeat the same action. In time, all the brushes may 
thus become badly burned or honey-combed. 

Such conditions are sometimes very difficult to correct. 
In some instances, higher resistance brushes bring improve- 
ment, while in other cases, lower resistance brushes are better. 
If, for instance, the local currents are relatively small, and the 
burning or picking up of the copper is due principally to the work 
current, then a lower resistance brush may actually reduce the 
watts at the contact, even though the local current may be in 
creased. If, on the other hand, the local currents are high and 
are principally responsible for the burning action, then a still 


higher resistance may actually reduce the loss. Thus it may 
be seen that, in many of these commutator and brush prob- 
lems, it is impossible to make a definite statement regarding 
the effect of any given make of brush unless one knows what 
is actually taking place in the machine. And every time the 
brush is changed, the conditions change, for they are more or 
less inter-dependent. 

Another remedy for some of the above troubles is under- 
cutting the mica between commutator bars. This does not re- 
move the primary cause of the trouble, namely, the large local 
currents or high current density on the brush contact, but it 
lessens the harmful effect of these by allowing the brush to main- 
tain better contact with the commutator copper, thus reduc- 
ing the contact losses. In this way the burning action can be 
diminished, in most cases, to such an extent that the commut- 
ator face will polish, and this in turn will enable the brush to 
make better contact with the copper. Undercutting in general 
is advantageous where the commutator mica takes up a large 
percent of the total surface, such as 20% or more. The larger 
the percentage of mica, the less liable it is to wear away as rapidly 
as the copper, and the greater the liability of the brushes being 
lifted above the copper, with consequent burning and blacken- 
ing. Not only must the percentage of mica be relatively small, 
but the actual thickness between two adjacent bars must also be 
limited, unless the mica is undercut. The general practice at 
present with non-undercut commutators is about 1-32 in. thick- 
ness between bars; and even considerably less than this, as^low 
as .018" to .020" is not unusual in small machines which are 
not undercut. Where commutators axe undercut, there is a pos- 
sibility, or liability, of carbon dust collecting in the slots and short- 
circuiting between bars, unless the peripheral speed of the com- 
mutator is sufficient to keep the slots dear. Therefore, in slow 
speed commutators which are undercut, it may be necessary at 
times to brush out or dean the slots. In high speed commutators, 
or in variable speed machines which intermittently reach high 
speeds, there is usually but little difficulty from carbon collecting 
in the slots. Obviously, where a commutator is to be slotted, 
there is no necessity for maintaining a minimum thickness of 
mica, as the limitation in such cases is in the width of the slot, 
which may be as much as 1-16 in. in some cases. Wide slotting, 
however, is liable to produce brush chattering in some cases. 


Slotted commutators, while advantageous in some ways, 
present certain operating objections in others. Except where 
the brushes cover several bars, the slotted commutators are liable 
to produce more or less chattering of the brushes, unless fre- 
quently lubricated. Therefore, with such commutators, self- 
lubricating brushes are recommended. Such brushes usually 
contain, or consist of graphite, and, in consequence, generally 
they are of lower resistance than ordinary carbon brushes, and 
therefore are not as useful in assisting commutation. In com- 
mutation pole machines where the resistance of the brushes is of 
relatively less importance, graphite brushes are often very satis- 
factory. As such brushes are usually soft in texture, they are 
not well adapted for wearing away the mica in commutators 
which are not undercut. 

In the application of the commutating pole to direct-cur- 
rent machines, certain conditions have arisen which are pecul- 
iar to this construction. For instance, according to the theory 
already given above, the flux of the commutating pole should 
arise and fall directly in proportion to the armature current. 
This means that there must be practically no saturation in the 
commutating pole circuit. Probably many of the early diffi- 
culties with commutating pole machines were due to a lack of 
appreciation of this point. Also, in machines with rapidly 
changing current, the commutating pole flux should change 
at the same rate as the armature current or the proper local cur- 
rent conditions for commutation are not obtained. This means 
that the commutating field winding should not be adjusted or 
varied in strength by means of a non-inductive shunt, as is com- 
mon practice in adjusting series field winding. As the com- 
mutating pole winding is normally inductive, then in the case 
of sudden change in current, an improper proportion of the 
current will flow through a non-inductive shunt at the time 
that the armature current is changing. Either no shunt should 
be used, or, if one is necessary, it should have the same induct- 
ance as the commutating field circuit. The former is prefer- 
able but requires most accurate designing. 

Another requirement in commutating pole machines, is 
accurate setting of the brushes. As a certain definite value of 
the local or commutating current is desired in the short-circuited 
coil, it is obvious that the coil must be short-circuited by the 
brushes at some very definite position with respect to the com- 



mutating pole. This is especially true in reversing machines. 
Otherwise, the commutation in the two directions would not be 
alike. Furthermore, an incorrect setting of the brushes, with 
respect to the commutating pole, will have a slight effect on the 
inherent regulation of the machine. In a direct-current generator, 
for instance, with the brushes set so that commutation is exactly 
central to the commutation pole, the magnetic flux of these poles 
has no resultant effect on the generated e. m. f. of the armature 
winding as a whole, and therefore has no effect on the regulation. 
But if the brushes are shifted ahead of this correct point in a gener- 

Fig. 6. 

ator, (Fig. 6) part of the commutating pole flux becomes effective 
in generating e. m. f., and in opposition to th.Q armature e. m. . 
A back lead in the same way would tend to increase the armature 
e. m. f . Thus the inherent regulation of the generator is affected by 
the brush setting. In a motor with commutating poles, a for- 
ward lead of the brushes tends to increase the counter e. m. f . 
generated and thus tends to lower the speed with increase in 
load, while a back lead tends to increase the speed. With per- 
fect setting of the brushes with respect to the commutating 
poles, and an adjustment of the commutating pole strength 
just sufficient to give the theoretically ideal short circuit or 
commutating current, the commutating 'poles should have prac- 
tically no effect o*r the speed/ But in actual practice, in direct- 
current motors, it is found better to actually over-compensate 


slightly that is, to make the commutating pole slightly stronger 
than the ideal value. This increases the local or commutating 
current above the ideal value so that commutation is well com- 
pleted before the coil leaves the brush. This gives less spark- 
ing, or "blacker*' commutation, at the brush edge, and ap- 
parently is more satisfactory from the operator's standpoint. 
But this over-compensation has an effect on the speed character- 
istics of a motor. It means that the zero point of the current 
is shifted backwards, and the resultant effect is similar to shift- 
ing the brushes backwards, and it therefore tends to speed up 
the machine, as described before. But this speeding-up ten- 
dency will vary with the load, as the commutating pole strength 
increases with the load. If the motor normally has a "flat" 
speed curve, this increase may be sufficient to bring the speed 
with load above the no-load speed, and this is an unstable condi- 
tion in the operation of constant speed motors. For instance, if 
a motor with rising speed characteristics has a high inertia load 
suddenly thrown on it, the heavy current required will tend to 
speed up the machine, and thus take a still heavier current. 
But a drooping speed curve has the opposite effect. Therefore, 
in motors built for general market conditions, where the load 
conditions may be of any nature, it is desirable that so-called 
constant speed motors should always have slightly drooping 
speed characteristics at least. But commutating pole motors, if 
designed along ideal lines, that is with high armature magneto- 
motive forces and with comparatively flat inherent speed 
characteristics, are liable to be affected, in speed, to a certain 
extent, by overcompensation of the commutating pole. In some 
cases, this effect may be so small that it does not over-balance 
the inherent droop in the speed curve. In other cases, it may 
more than over-balance, so that the actual speed curve rises with 
load. This is particularly noticeable in adjustable speed ma- 
chines for a wide range in speed. Such machines have full field 
strength at lowest speed, and here the effect of the local or com- 
mutating current on the speed may be very small. At three or four 
times speed, however, the main field is very weak, while the com- 
mutating current is practically the same as at low speed, and 
therefore has three or four times the effect in increasing the speed. 

Therefore, such motors are liable to have rising speed curves 
at Ijigher speeds, although they may be slightly drooping at 
the lowest speeds. 


Obviously, as this effect is a function of the armature cur- 
rent, it should be corrected by means of the armature current. 
This is readily accomplished by adding to the main field wind- 
ing a very small winding in series with the armature, and con- 
nected to magnetize in the same direction as the shunt winding 
on the field poles. While this might be looked upon as a series 
winding, yet its function is that of compensation for commu- 
tating pole action. The ampere turns in this compensating 
winding are normally very small, being just sufficient to bal- 
ance the effect of the excess short circuit current in the corn- 
mutated armature coils. In adjustable speed machines, at low 
speed and full field strength, this small compensating winding 
has but very little effect, as it is so small in proportion to the 
shunt ampere turns. At the highest speeds, however, where the 
shunt ampere turns are very low and the rise in the speed curve 
is liable to be greatest, this compensating winding has the most 
effect. It therefore tends to produce proper compensation at 
all speeds and loads. 

Another phenomenon which has appeared in direct cur- 
rent machines, and which, at times, has been falsely credited to 
commutating conditions, is that of "pitting," or "eating away" 
of the mica between commutator bars. Nearly all manufac- 
turers and operators have encountered this difficulty at some 
time or other. This has also been credited to high voltage be- 
tween bars, too thin mica, quality of carbon brush, use of lubri- 
.cants, etc. Some years ago, the writer and his associates made 
an extended study of this matter, based upon a very large number 
of cases of actual trouble. The results which were derived from 
the general practical data from machines in actual service were 
so conflicting that no positive conclusions could be drawn di- 
rectly from such data. However, eventually the evidence pointed 
to oil as apparently one of the fundamental conditions in this 
trouble. Extended tests were then made to determine the 
effects of oil on the mica, and the results indicated very clearly 
that those insulations which absorb oil were liable to pit or eat 
away in time. Apparently where the oil could dissolve out the 
binding material in the tnica^ ininute particles of carbon or copper, 
wotild be disseminated through the mica, thus decreasing its re- 
sistance locally. Combustion of these particles would usually be- 
noticed as "ringfire" around the onimmtator. Ring-fire is almost 
always due toincaadesc^it carbon particles scrapedoff the brushes, 


but is not usually harmful if the mica adjacent to the spark does 
not burn or deteriorate. Experience shows that the ordinary good 
grades of mica plate are not affected by such ring-fire, and it is 
only when foreign conducting particles penetrate into the plate 
that this burning may gradually eat away the mica. It was found 
that some binding materials used in building mica-plate were 
much more soluble in oil than others, and it was noted that in 
those plates with soluble binders, the pitting was most pro- 
nounced. In fact, in those grades of mica plate, where the 
binder was practically insoluble in oils, no pitting was found, 
even under very extreme conditions of test This led then to 
one solution of the pitting trouble, namely, the use of what 
might be designated as insoluble binders in the mica plate, with 
very tight construction of commutator, so that oil could not 
penetrate along the sides of the plate, and with care in prevent- 
ing oil from getting on the commutator. With the first two con- 
ditions, the latter should not be so important, yet one never 
knows whether the first two conditions are perfect, especially 
after a machine has been in operation for a long period and has 
been subjected to severe changes in temperature at the com- 

After getting at the probability of oil as a pause of pitting, 
many careful examinations were made of pitted mica, and in 
general, there was evidence that the mica binder had been at- 
tacked by oil. In some instances where the operators were ab- 
solutely sure that there had never been any oil on the commut- 
ator, careful chemical and microscopical analysis showed the oil- 
In some cases the mica was actually spongy with oil, and yet 
it was claimed that no evidence of oil had ever been noticed on 
the commutator. 

Much* depends upon the grade of mica used in the plate 
for building up in commutators. Certain micas seem to wear 
much faster than others, and yet be just as good insulators. The 
well known amber micas seem to be by far the most successful 
for this purpose. Many attempts have been made to use cheaper 
grades of white mica, and, in some cases, with good success, but 
the difficulty is that it is not uniformly successful, and trouble 
from high mica may develop only after a large number of ma- 
etjines have been put on the market. Many costly experiences 
of this sort have made the manufacturers very conservative in 
this matter* It takes so long to find whether a new mica is good 


or not, that it Is questionable iii most cases whether it should be 
tried out at all. 

In the early days, the commutator mica was made com- 
paratively thick, and was punched out of solid material. When 
the slotted types of direct current armatures came into use, 
with their greater sparking tendencies, the old solid thick mica, 
used with surface-wound armatures, immediately showed trouble 
due to high mica. This soon led to thinner mica, which helped 
the trouble somewhat. Then somebody discovered that, by 
splitting the mica into very thin plate and reassembling, without 
binder, the results were still better, as this split-up mica seemed 
to wear or flake off at the edges much better than solid mica. 
Then someone discovered that stall better results were obtained 
by splitting or flaking the mica and building up into plates, with 
a suitable binder Since that time, practically no radical im- 
provements have been made in commutator mica, except in the 
binding material possibly, and in the better choice of the grades 
of mica used, but the mica-plate of today in general is very similar 
to the mica-plate of IS or 20 years ago. Of course, refinements 
in manufacture have occurred, which, in most cases, however, 
have had but little effect on the quality of the product. At 
present, it does not look as if a more suitable material can be 
found for this purpose Many attempts have been made to 
substitute other materials, but these have only proved successful 
in certain applications. The built-up mica possesses certain 
physical qualifications which have not been obtained with any other 
material. For instance, under heating and cooling, the commut- 
ator changes in dimensions circumferentially, as well as axially, 
and under this action the mica undergoes much more compression 
at times than at others. Therefore, a material of a more or less 
elastic nature is required between bars, in order to avoid per- 
manent compression, with resulting eventual looseness. Struc- 
turally the mica-plate meets this condition very well. In the 
second place, the material between bars should be one which 
wears down properly and yet does not have any cutting or grind- 
ing action on the commutator and brushes as it wears off. Mica 
apparently meets this condition, while many other mineral com- 
pounds, such ate asbestos, appear to have a grinding action. In 
the third place, the material should be more or less heat and 
spark-resisting. Again, it should be a non-absorbent of oil. 
tnica-plate seems to be the oaity material so far which 


meets all the requirements for general purposes. Hard, inelastic 
materials of various sorts have been tried and have not proved 
successful. Asbestos in sheets and plates and in conjunction 
with other materials, has not proved very satisfactory. Fibrous 
and cellulose materials have given good results in some cases, 
but are not sufficiently heat and oil-proof for general purposes. 
The only material departure has been made in micas used with 
undercut commutators. In such cases, white micas and others 
which have all the good characteristics of the amber micas, except 
their wearing characteristics, can be used, for the undercutting 
eliminates the necessity for good wearing qualities. At the same 
time, undercutting, as explained before, is advantageous in other 

There are a number of mechanical conditions entering into 
the practical side of the problem of commutation. As shown 
before, it is very important to maintain good contact between 
the brush face and the commutator in order to keep down losses 
and burning action. Moreover, good contact in general should 
mean good contact over the whole brush face in order to keep down 
the current density. Therefore, if the brushes chatter or vibrate 
in their holders, or have a rocking action tending to give alternate 
""heel-and-toe" contact, obviously the operation is liable to be 
affected thereby. Vibrating brush holders, vibrating brushes 
and chattering or movements of any kind with respect to the 
commutator face, are objectionable and, not infrequently, very 

Vibrating brush holders may be due to various causes, which 
do not show up on the manufacturer's test. The machine may 
be so located that its environments are to blame for vibration. 
Bad gearing may cause chattering. The foundations or sup- 
ports may not be as substantial as on the shop test, so that some 
small disturbance may be exaggerated and produce vibrations in 
parts or in the whole machine. Sometimes the brushes may 
chatter due to lack of lubricant, and this may set the whole brush 
holder structure into vibration. Whatever the cause, it is always 
best to stop such vibrations as far as possible, especially in ma- 
chines handling large currents. 

Vibration of brushes in their boxes may be due to badly 
fitted brushes, (Pig. 7) or, on machines which have long been 
in operation with heavy currents per brush, the brush boxes 
may be eaten away inside so that they are not of uniform di- 


mensions. Low resistance shunts on carbon brushes are for 
the purpose of carrying away the current from the brush by 
some other path than through the sides of the brush box. How- 
ever, due to the raking or dragging action of the commutator 
on the brushes, they are liable to bear rather heavily against 
one side of the box, especially at the lower edge next to the com- 
mutator. Some current will naturally pass to the box, and this in 
time will tend to burn away the boxes and the carbons. How- 
ever, as the carbons burn away, they are eventually replaced 
by new ones; but the boxes are seldom replaced, and in time 
they may burn away so that they are larger next to the commut- 
ator. The brush then fits tightly only at the top and is free to 
move or vibrate or chatter at the commutator end, which is just 
the place where such movement should be avoided. Attention 
is called to this action in particular on account of the carelessness 
often exhibited in regard to the shunts on the brushes. 

Fig. 7. 

Chattering is not infrequently due to lack of lubrication 
on the commutator or in the brushes. When the commutator 
gets too dry and has a high polish, a radial, or almost radial 
brush may vibrate or chatter just as a pencil does when moved 
along a pane of glass. If the commutator is lubricated with 
oil to overcome this, care should be taken not to use an excess 
of oil or the mica may absorb it. Frequently immediately after 
oiling, a commutator shows ring-fire, which is due to combustion 
of minute particles of carbon and oil over the top of the mica. 
Sometimes chattering is best overcome by the use of self -lubri- 
cating brushes. In undercut commutators, the slots are liable 
to cause chattering with non-lubricated brushes, giving out a 
noise of a pitch comparable with the product of the revolutions 
by the number of commutator bars. As oil-lubrication should 
be used with caution on undercut machines, practice now usually 
calls for some form of self-lubricating brush, partly or wholly 
graphite, as has been referred to under "under-cutting. 1 * 

In communicating pole machines it is especially important 
that the brushes should not have any heel-and-toe movement, 


for when the brush makes contact at one edge or the other, the 
result is equivalent to rocking the brushes backward or forward, 
which, as explained before, is particularly objectionable in such 

In direct-current machines, burnt or black spots will some- 
times develop on the commutator at points removed from each 
other a distance corresponding to that between holders of the 
same polarity. This is sometimes very bothersome, and the 
cause of the difficulty is not always easy to find. Any condi- 
tion which produces one bad spot may tend to produce similar 
spots symmetrically displaced around the commutator. When 
one spot is formed, and this spot passes under one brush arm, 
the brush contacts at this arm are naturally poorer and the other 
brush arms of the same polarity tend to take the load, and the 
current density in their brushes is correspondingly increased 
during this short period. If there is any tendency toward high 
mica, for instance, then the increased current at these points will 
produce increased burning away of the copper and burnt spots 
may develop. If once developed they may gradually travel 
around the commutator until the whole commutator is black. 
A local or high mica strip may be the initial cause of the trouble, 
or a rough spot on the commutator may give the same result. 
But very often, resultant high mica, following the initial cause, 
tends to spread the trouble. As soon as such black spots are 
noted, further trouble frequently can be headed off by scraping 
or cutting the mica below the copper surface in the burnt regions. 

One of the most severe conditions that any direct-current 
generator can encounter is a dead short-circuit across its ter- 
minals, or in the immediate neighborhood of the machine. Very 
few machines outside of those of comparatively small capacity 
and of low voltage, can stand such short-circuit without severe 
flashing. Tests have shown that moderately large direct-cur- 
rent generators will give, at the moment of short-circuit, from 
20 to 30 times full load current. No ordinary commutating 
machine can be built to take care of such a current rush, and 
vicious arcing and flashing generally results. This is an inherent 
condition in the design. No responsible manufacturer who 
knows his business will guarantee to overcome this character- 
istic. However, fortunately, the great majority of short-circuits 
on direct-current power systems occur at some distance from the 
generator, and moreover, in many cases, such short-circuits are tnade 


through arcs rather than by dead contact, so that the generators 
do not get the maximum possible current rush. 

It might be suggested that quick-acting circuit breakers 
would take care of such extreme conditions by opening at the 
loads for which they are set. But this setting is that at which 
the tripping mechanism works, and if the current rises rapidly 
enough, it may be far in excess of the tripping value by the time 
that the breaker actually opens or ruptures the circuit. In fact, 
this is just what happens in the case of a severe short-circuit. 
Oscillograph tests have shown that the current "rush" on short- 
circuit may reach its maximum value in one-fiftieth of a second, 
or even less, while the ordinary commercial circuit breakers 
seldom operate in less than one-tenth of a second, which, in 
reality, is pretty rapid action for a mechanical device There- 
fore, it will have to be an extremely rapid-acting breaker which 
can get the circuit open before the short-circuit current has risen 
to several times the full load value. 

One other subject might be considered under commutation, 
namely, the influence of the commutating characteristics upon 
the permissible range of speed variation and speed adjustment in 
direct-current motors. There are two general methods for ob- 
taining speed variation in such apparatus, namely, by variation 
in the e. m. f . supplied to the armature terminals, and by varia- 
tion in the field strength or flux. 

In the early development of adjustable speed motors, the 
first of the above methods was used almost exclusively, largely 
on account of the fact that the commutation problem was more 
easily handled with this method. As the motor could be given 
full field strength much of the time, and as the reduction in field 
strength was not great under any conditions, fairly good com- 
mutation was obtainable in general. Where constant torque was 
required, this method was fairly satisfactory and economical 
However, where constant horse-power was required, obviously, with 
this method, the armature current had to be increased directly as 
the armature voltage and speed were reduced. Thus the armature 
had to be designed for a voltage capacity corresponding to the 
highest voltage, and a current capacity corresponding to the 
lowest voltage. Thus it became quite large for a given horse- 
power rating, and was therefore very tineconornical in material. 
However, the larger currents at lower vblta^s did not tepresent 
such a hardship m confutation, for %M& increases in current 


were accompanied by corresponding reductions in speed, which, 
made commutating conditions proportionately easier, Thus 
the commutating problem was not so serious with this method 
of speed control. However, for constant horse-power service, 
it was obvious, early in the development, tliat the most econ- 
omical arrangement as regards material, would be the use of a 
constant voltage across the armature, thus requiring constant 
armature current, speed control being obtained by variation in 
the field strength. But this meant that the field had its full 
strength at the lowest speed, and the field flux would be decreased 
directly as the speed was increased. This was the ideal arrange* 
ment, but, unfortunately, commutating conditions were very 
difficult at the weaker fields, that is, at the higher speeds. In 
consequence, a number of more or less freakish designs were put out, 
with the idea of overcoming the commutation troubles, by using 
variable field speed control. Some of these designs were satis- 
factory from the operating standpoint, but this method of speed 
control did not reach its full development, except in the com- 
mutating pole type of machine, thus showing that the commut- 
ation problem was a serious one in this method of operation. 
With properly proportioned commutating poles, the commutating: 
conditions are practically independent of the speed, so that, 
with their use, the real limitations in speed range are found in 
other conditions, such as the instability of very weak fields, etc. 

It is evident from the above that, as regards speed regu- 
lation in direct-current work, a constant field motor is at a serious 
disadvantage compared with one in which the field strength can 
be varied. In fact, this holds true for alternating-current as 
well as for direct-current motors, as will be shown later. The 
alternating-current induction motor as essentially a constant field 
machine, and normally operates at constant speed, on a given fre- 
quency. In the constant field direct-current motor, it was shown 
that speed variation is accomplished by variation in the voltage 
applied to the armature. The analogous condition in the induction 
motor would be in variation in the frequency applied, and not 
in the voltage. This, then leads up to the next subject, namely* 

Repeating preceding statements, the induction motor is 
primarily a constant speed machine when supplied with constant 
frequency and e, m. f ., which is the standard condition in all 


alternating power service. When running at full speed, the 
secondary frequency and e. m. f . are both very small, the frequency 
being only such as will generate enough e. m. f. to send 
the required secondary currents through their own windings 
when closed upon themselves. If the secondary resistance 
is increased, with a given current flowing, the secondary 
e. m. f. must be proportionately increased, and the secondary 
frequency, to generate such e. m. f., must also be correspondingly 
increased. This secondary frequency, which represents the 
departure from synchronous speed, or the "slip," is therefore 
always proportional to the secondary e. m. f . Therefore, speed 
regulation of the motor, by means of the secondary circuit, means 
corresponding, or proportional variation in the secondary frequency 
and e. m. f ., and all methods of speed regulation or adjustment of 
induction motors through secondary control are based upon fre- 
quency and voltage variation in the secondary circuits. 

All methods of speed regulation of induction motors may be 
classified under two general heads; (1) primary circuit control, 
and (2) secondary circuit control. 


Two general methods are practicable, namely, variation in 
the number of primary poles, and variation in the frequency 
supplied to the primary. The former method is very limited in 
the range of control which is practicable. Usually, two operat- 
ing speeds can readily be obtained, while three or four lead to much 
added complication, and more than four speeds does not appear to 
be commercially practicable except in very special cases. For 
fine graduations in speed, pole changing is apparently out of the 
question. c 

By suitable change in the primary frequency supplied to the 
motor, any desired speed, or speed range, is obtainable. But in 
general, the problem of furnishing this variable frequency is just 
as serious as that of speed adjustment of the motor itself on a 
fixed frequency. In other words, it takes the difficulty away 
from the motor and transfers it elsewhere, but does not eliminate 

There are various ways of generating variable frequency. 
For instance, an alternator may be driven by an adjustable speed 
motor. This should be af direct-current motor, for, if an alter- 
nating motor is used, the problem of varying its speed is just 
the same as that of the induction motor which is to be regulated. 


Another way is to connect the alternating-current generator to 
an adjustable speed prime mover, such as an engine or water- 
wheel. Such methods of regulation require one generating 
outfit for each motor to be regulated, except where two or more 
motors are to be regulated over the same range at the same time 
The method in general is very seldom used. 

Other possible methods of regulating the primary frequency 
lie in frequency changers of certain types by which a given fre- 
quency can be converted to any other frequency by commutation 
of alternating current. Various types of such machines are 
possible but they possess certain very objectionable limitations, 
in that they must commutate currents of frequencies approxim- 
ating those of the primary supply system At 25 cycles, this may 
be practicable, in some cases, but on 60 cycle supply circuits it is 
out of the question. One other serious objection to regulating 
the primary frequency is that the frequency controlling device 
must have a capacity equal to that of the motor to be regulated, 
that is, the entire input of the induction motor must be handled 
by the frequency regulator. 

In general, therefore, regulation of induction motor speed by 
change in primary frequency is not advisable, and appears to be 
practicable only in certain very special applications. This then 
brings us to the alternative of secondary circuit control. 


As brought out before, any speed variation of an induction 
motor with unchanged primary frequency means accompanying 
change in the secondary frequency and voltage. At standstill, 
the secondary voltage is a maximum, and the secondary frequency 
is 100 percent of that of the primary, that is, it is the same as 
the primary. At true synchronous speed, the secondary voltage 
and frequency are zero. At any intermediate speed, the secondary 
voltage generated and the secondary frequency are respectively 
equal to the standstill voltage, and the standstill or primary 
frequency, multiplied by the slip, in percent, the slip being the 
drop from synchronous speed. It should be noted that the second- 
ary generated voltage is mentioned, for this is not the same as the 
secondary terminal voltage, due to a certain internal drop in the 
windings when current is flowing. This internal drop is usually 
small compared with the secondary standstill voltage, usually 
being from 2 percent to 3 percent except in small motors, and 


therefore may be neglected in any general discussion not involv- 
ing exact calculations. 

All methods of secondary circuit control in induction motors 
include some methods of regulating or controlling the secondary 
voltage and frequency. The simplest practical device is the use 
of resistance inserted in the secondary circuit. In order to get 
the required current, for a given torque, through such resistance, 
the voltage must be increased and this requires increase in the 
secondary frequency, that is, drop in speed. But with a given 
resistance, if the load or torque is varied, the secondary current 
must vary, which means corresponding variation in voltage and 
secondary frequency, that is, in speed. Therefore, speed 
regulations by secondary resistance means variable speed with 
variations in torque; and constant speed with variation in torque 
is only obtainable by varying the resistance inversely with the 
current, in order to obtain a constant voltage drop. Such method 
of speed regulation is therefore satisfactory only to a limited 
extent. Moreover, this method of speed regulation is unecon- 
omical, in that there is a rheostatic loss practically proportional 
to the drop in speed below synchronism. At half speed, for 
instance, half the output of the motor is wasted in resistance. 

Obviously what is needed is some arrangement which will 
absorb the required secondary voltage in other than resistance f 
and which will automatically hold such voltage constant, with 
varying current, in those cases where constant speed character- 
istics are required for each speed setting The difficulty in ob- 
taining such a device is not simply in the voltage range required, 
but is largely on account of the range in frequency necessary. 
Therefore, all such devices must be of adjustable frequency, and 
therein lies the true difficulty, just as in the case of frequency 
changers in the primary circuit, as already referred to. The 
difficulty, however, is not nearly as serious in the case of regula- 
tion of the secondary circuit, for the variable frequency device 
needs to be of a total capacity corresponding to the slip, in 
percent. Furthermore, where the departure from synchronism 
is not large, the actual frequency in the frequency controlling 
machine is so low that commutator type alternating-current machines 
are permissible up to relatively high capacity. The problem there- 
fore resolves itself into one of variable frequency, just as in the 
case of primary circuit j^gulatibn, as already referred to, except 
that the frequencies and capacities dealt \ffith usually are very 


much lower in the case of secondary circuit control The problem 
is simply easier, but not of a different nature. 

In all methods of rating by secondary control, the regulating 
device must absorb power corresponding in percent practically 
to the secondary terminal voltage, or the secondary frequency, 
or slip. This power must be utilized if economical operation is 
required. There are three general methods by which it can be 
utilized, namely, it may be transformed to mechanical power 
and assist in driving the motor shaft, or it may be transformed 
to the primary or line frequency and fed back into the line, or it 
may be transformed to direct current for use in some other part 
of the system. Combinations of these three methods may be 
used. For instance, this secondary power may be transformed 
to direct current and then be transformed to mechanical power 
by means of a direct-current motor connected to the induction 
motor load. Or, it may be transformed to direct current and 
then re-transformed to the primary frequency and fed back into 
the line. 

Three types of variable frequency devices have been pro- 
posed for absorbing the secondary terminal voltage, namely, 
A. C. commutator motors, rotary converters, and commutator 
type frequency changers. In the first named, the A. C. commut- 
ator motor either delivers its power directly to the shaft of the 
induction motor, or to an A. C. generator which returns it to the 
line, or to a D C. generator which delivers its current to some 
D. C. system or load where it can be utilized. In the second type 
mentioned, a rotary converter absorbs at its collector rings the 
secondary terminal voltage and transforms it to a proportional 
direct-current voltage. The direct-current power is then fed into 
a direct-current motor connected with the induction motor load, 
or is transformed to the primary frequency by a suitable motor- 
generator set. In the third type, a commutator type frequency 
changer transforms the secondary terminal voltage and frequency 
to a proportionate voltage at the primary frequency, and, by 
means of suitable transformers, the secondary power is then re- 
turned to the primary supply circuit. 

Each of these arrangements possesses some advantages over 
the others, and also some disadvantages. The A. C. commutator 
motor is a relatively expensive type of machine, especially for 
very low speeds. Therefore, when the induction motor to be 
regulated is of comparatively low speed, placing the commutator 


motor on the induction motor shaft means a relatively expensive 
commutator machine. In such cases, it may be advisable to 
either gear it to the load or connect it to a generator which returns 
power to the line or delivers it to another system. By such 
means, a smaller and higher speed commutator type A. C. motor 
may be used, but at a certain expense in auxiliary apparatus. For 
a frequency of 25 cycles, the A. C. commutator motor does not 
present any undue inherent difficulties if the speed range of the 
secondary control is not too large. With 50 percent drop in 
speed, for instance, the frequency handled by the A. C. commut- 
ator motor is only 12J^ cycles. But with a 60 cycle supply 
system, a speed range of 50 percent means that the A. C. 
commutator motor must handle 30 cycles, which is a much more 
difficult and expensive proposition. 

With the rotary converter speed regulation, no new or diffi- 
cult problems are involved, either in the transformation or utiliza- 
tion of the secondary power. Where the induction motor speed 
is not too low, a direct-current motor connected to the shaft may 
utilize the direct-current power from the rotary converter. How- 
ever, unlike the A. C. commutator scheme above described, the 
rotary converter arrangement makes its best showing in connec- 
tion with 60 cycle supply systems,- for, with the higher frequency, 
the secondary frequency of the main motor is correspondingly 
higher for the same speed range, which allows the use of a relatively 
smaller rotary for the same percentage of power transformed. 
To illustrate On a 25 cyde supply system, with 30 percent speed 
range, the maximum secondary frequency is 7 J^ cycles. A 4-pole 
rotary operating at this frequency will run at 225 r.p.m. ; that 
is, at this speed, it transforms or utilizes 30 percent of the power 
of the induction motor. Considering now, 60 cycles, with the 
same speed range, the secondary frequency becomes 18 cydes, 
and a 4-pole rotary of 30 percent of the motor capacity will 
operate at 540 r.p.m. on this frequency. Obviously, a rotary 
converter of much smaller dimensions can be used, than in the 
former case. The auxiliary means for absorbing the direct cur- 
rent power from the rotary converter can be practically the same 
for either frequency. Therefore, with this method, 60 cydes 
makes the better showing. 

In the third scheme, (Fig. 8) the secondary frequency of the 
induction motor is transformed directly to the primary frequency 
in a single machine. The auxiliary means for utilizing the trans* 



formed power consists of suitable stationary transformers with 
tap for varying the voltage. As the frequency changer is a rather 
unusual device, a brief description of its principle may not be 
out of place at this point. As usually built, it consists of an 
armature like that of a rotary converter, equipped with both 
commutator and collector rings. Unlike the rotary, the field 
may consist of a simple " keeper" or ring, (Fig. 9) without wind- 
ings, which encircles the armature. Also, unlike the rotary con- 

Fig. 8. 

Fig. 9. 

verter, the commutator is equipped with a double or triple set 
of brush holders for handling polyphase current. The ordinary 
spacing of the D. C. holders on a direct current rotary would 
correspond to one ph&oe of the frequency changer. The arma- 
ture can be driven by any suitable small-capacity, adjustable 
speed device. Practically the only load carried by the driving 
device consists of brush friction and windage. 

If such a machine has its collector rings connected to the 
main supply system through suitable transformers, a rotating 
field will be set up in the armature core (and keeper) which travels 
around the core at a speed corresponding to the frequency divided 
by the number of poles, just as in an induction motor. If, now, 
the core is rotated mechanically in the opposite direction, at a 
speed equal' to the frequency divided by the number of poles, then 
the magnetic field set up in the core will stand still in space and 
could be replaced by an external field excited by direct current, 


just as in a rotary converter. Under this condition, the brushes 
on the commutator would tend to deliver current, that is, alter- 
nating current having zero frequency. Under this condition, 
the external stationary keeper has zero frequency in it, while the 
armature core has normal frequency. Assume now that the core 
is rotated either faster or slower than synchronous speed. The 
magnetic field set up by the armature winding will travel back- 
ward or forward in space at speed corresponding to the departure 
of the core from synchronism and the brushes on the commut- 
ator will tend to deliver alternating current at a frequency pro- 
portional to the departure of the armature core from synchronous 
speed. Thus by varying the speed of the armature core from 
synchronism, any desired frequency can be obtained at the commu- 
tator brushes But the voltage at the commutator brushes is 
practically equal to the voltage at the collector rings, regardless 
of the speed of rotation, and by varying the voltage supplied to 
the collector rings, the voltage at the commutator can be varied 
independently of the frequency, which is dependent solely upon the 
spead of the armature. Thus, independent control of the voltage 
and frequency is obtainable, which makes the device quite flexible 
in its application. But such a device has other very desirable 
characteristics. As it is primarily one form of rotary converter, we 
should naturally expect that it would show some of the small-cop- 
per-loss characteristics of the rotary converter. Analysis, however, 
shows that it goes even further than this. In a 6-phase rotary 
converter, for instance, the armature copper loss averages about 
26 percent of that of a corresponding direct-current winding, due 
to part of the alternating current being fed directly through to the 
direct-current circuit without transformation. But as there is 
transformation from one kind of current to another, the operation 
is incomplete, and there are certain transformation losses which 
are especially large at and near the so-called tap coils, which are 
connected to the collector rings. But in a 6-phase frequency 
changer of the above type, the transformation is from 6-phase 
alternating current to 6-phase current of another frequency, and a 
still larger percent of the current passes through without transform- 
ation than is the case in a 6-phase rotary converter. In consequence, 
the frequency changer has only about two-thirds as much copper 
loss as a rotary converter; that is, for 6-phase, it is less than 18 
percent of that of a corresponding direct-current machine. More- 
over, the tap-coil losses of the fotary converter are practically 


absent. It thus becomes an extremely effective transforming de- 
vice, as far as frequency is concerned. 

Such a device is also very economical as regards iron losses- 
As it generates voltages, when connected to the secondary ter- 
minals of an induction motor, which are proportional to the speed 
range, obviously, with moderate speed variations, the magnetic 
flux in the armature core will be small compared with what would 
be necessary to generate full secondary voltage. Also, even this 
reduced induction is at a comparatively low frequency in the 
surrounding ring or keeper, and is only at full frequency in the 
armature core proper. Evidently therefore, the armature core 
and armature teeth sections can be made relatively small where 
the range of speed adjustment is small, such as 25 percent to 35 
percent from synchronism. This small average core loss, together 
with the very small copper loss tends toward a very economical 
construction of machine. 

In such a frequency changer, the problems of commutation 
are very similar to those in the A. C. commutator motor, and at 
25 cycles the design becomes simpler and easier than at 60 cycles. 
The machine is independent of the speed of the induction motor 
to be controlled, which is not the case with the A. C. commutator 
motor in its simplest application, namely, direct connection to the 
main motor shaft. Such frequency changer in its simplest form 
may be arranged to be self -compensating, and the conimutating 
conditions can be brought well within those of well-proportioned 
A. C. commutator motors. 

In the application of these various speed regulating devices, 
two power conditions should be given consideration, namely, 
whether the motor outfit is to develop constant horse power at the 
shaft, or constant torque, with the developed power varying in 
proportion to the speed. In most cases, in steel mill work, con- 
stant torque is all that is necessary, while in a few special cases 
constant horse power may be desired. 

Where constant torque is preferred, the frequency converter 
should prove to be most desirable in many ways, particularly on 
account of its flexibility in application, so that a few suitable 
sizes should cover range of application. In this feature, and in a 
number of others, it has the advantage over the rotary converter 
or the alternating-current commutator motor schemes. 

Where constant horse-power is required, it is questionable 
whether any one scheme has the advantage in all cases. Where 


the induction motor speed is fairly high, and the frequency is low, 
the A. C. commutator motor directly connected to the induction 
motor shaft is a good arrangement, as only one regulating ma- 
chine is required. If, however, the speed is so low that the A, C. 
commutator motor connection to the main shaft is inadvisable, so 
that power must be returned to the supply system, then this 
arrangement will not compare favorably with the frequency 
changer scheme. The rotary converter scheme, delivering 
direct-current power to a motor on the induction motor shaft, 
also makes a good constant power outfit, but where the speed is 
too low to allow an economically proportioned direct-current 
motor, the scheme is also at a disadvantage compared with the 
frequency changer. But where the frequency changer is used 
with constant horse-power, the main induction motor must be 
large enough to deliver the rated power to the shaft at the lowest 
speed, the excess power being transferred to the line by the fre- 
quency changer. If, however, the main motor is operated above 
synchronism at its highest speed, by an amount corresponding to 
the slip below synchronism at its lowest speed, then the increase 
in capacity of the main motor and of the frequency changer, to 
give constant power at the shaft, will be only about half as much 
as if all the speed variation were below synchronism. This brings 
up a point not yet brought out, namely, that some of these ad- 
justable speed devices allow operation of the main motor above 
synchronous speed. This is particularly true in those cases where 
the speed-regulating or auxiliary apparatus can impress its own 
frequency upon the secondary of the main motor, and where such 
impressed frequency can be independently controlled. In such 
cases, by gradually varying the frequency down to zero and then 
up in the opposite direction, the main motor can have its speed 
varied through the synchronous position. 

Various other methods of speed regulation have been pro- 
posed, but most of them have not yet seen actual test. Several 
schemes have been proposed for utilizing mercury vapor rectifiers 
for transforming the secondary current of the induction motor to 
direct current. This is one case where a frequency-changing 
controlling device does not form a fundamental part of the con- 
trol. On the other hand, such method of control is as yet more 
theoretical than practical, and moreover, the mercury rectifier is 
not yet in general use for power service. At best, therefore, this 
method is one of the future. 



Correction of power factor in induction motors, by means of 
a low frequency exciter in the secondary circuit is feasible. In 
connection with the above described adjustable-speed devices, it 
may be said that almost all such devices can be designed to correct 
power factor, as well as to produce change in speed, and, in many 
cases, this involves practically no extra complication. For 
instance, in the frequency changer method, where the voltage can 
be varied independently of the frequency, an increase in the 
frequency changer voltage without change in speed would simply 
mean the transfer of wattless current from the supply system to 
the secondary circuit of the induction motor, and this replaces 
the primary wattless magnetizing current. By proper voltage 
adjustment, the primary wattless component could be reduced to 
zero, or even changed to a large leading, instead of lagging, com- 
ponent, with consequent change in power factor from lagging to 
leading of any desired value. This simply illustrates the general 
method of power factor correction by all these various devices. 

Fig. 10. 

While on the subject of power factor correction, it may be 
stated that only two general methods of power factor correction 
are practicable namely, by means of static condensers connected 
across the system, and by means of rotating condensers of some 
form. (Fig. 10). 

The static type of condenser is commercial on a small scale, 
and, possibly, may become so on a large scale in the not far dis- 
tant future. Large capacity static condensers can be built at 
present, but possibly not at a cost which will compete with the 
rotating type. 


Rotating condensers may be divided into two sub-classes, 
namely, synchronous and non-synchronous. The synchronous 
type is well known commercially. Usually it is simply a syn- 
chronous motor with over-excited field. It may or may not 
deliver power as a motor while acting as a condenser. 

The non-synchronous condenser is simply a non-synchronous 
or induction motor with its secondary excited, instead of its 
primary. When acting as a condenser, the secondary is over- 
excited. It is therefore somewhat similar to the synchronous 
coiidenser, low frequency alternating current, instead of direct 
current being used for excitation. The condenser also may act 
as a motor delivering power. This type of condenser has not 
been used to any extent in this country. 

From the foregoing discussion of speed control, it is apparent 
that the frequency of the supply system has an important bearing 
on the induction motor problem in general. There are only two 
accepted standard frequencies in general use in this country, 
namely, 60 and 25 cycles, and both are in use in central power 
stations. The tendency for mills and factories to purchase power 
from such central power stations, instead of generating it them- 
selves, appears to be increasing. This therefore, leads to another 
subject of direct interest to steel mill engineers, namely, 


This question is not limited to mill work, but has become 
a very general one in the whole electrical business. Some years 
ago a committee of steel mill engineers decided upon 25 cycles 
as a standard frequency for steel mill work. The reasons for this 
decision were amply sufficient at that time and still hold good to 
a certain extent. But, in more recent times, the general tendency 
of the large central station or power companies toward 60 cycles, 
together with the sale of such power to steel mills and other in- 
dustrial plants, has changed the situation somewhat. At the 
time that the steel mill committee recommended in favor of 25 
cycles, there was an apparent tendency of the large power com- 
panies toward this frequency. But, as intimated, that tendency 
is now reversed, partly -due to improvements in certain types of 
apparatus, such as rotary converters. It therefore may be 
pertinent to discuss this subject of frequency more fully, in view 
of its possible influence on -mill work. 


The principal loads to be handled by general power or central 
station alternating current plants are: first, lighting, including 
arc, incandescent, etc.; second, motor power service; and, third, 
direct-current service for various purposes, such as railway, etc. 

In general, there is no particular question regarding the 
better frequency for lighting service, for 60 cycles, for direct use 
in both arc and incandescent lamps undoubtedly gives better 
results than 25 cycles. 

When it comes to motors, either synchronous or induction, 
60 cycles present more advantages in general, except for very low 
speeds, and, even in this case, with synchronous machines, the 
choice is in doubt. In the case of induction motors, however, 
there are certain fields where 25 cycles will show better results. 
This is in very slow speed work, or very slow speed in proportion to 
the capacity. It is a rule in practically all types of generators 
and motors that the greater the number of poles, the greater must 
be the total magnetizing ampere turns. In windings excited 
by direct current, the number of exciting turns may be increased 
with increase in the number of poles, at a certain expense in cop- 
per, so that the actual exciting or magnetizing current may not be 
excessive, even with a very large number of poles that is, in very 
slow speed machines. But in induction motors, the same turns 
are used for magnetizing and for generating counter e. m. f . The 
latter condition usually so fixes the number of turns, in a given 
capacity and speed of machine, that the actual magnetizing 
current increases very greatly with increase in the number of 
poles, that is, with decrease in speed, so that, with a large 
number of poles, this magnetizing current becomes so large in 
comparison with the work current that the characteristics of the 
machine are very seriously affected. This increase can be limited 
to a certain extent by increasing the dimensions of the machine, 
that is, its cost. Herein is where 25 cycles may give consider- 
able advantage over 60 cycles. For instance, a 4-pole, 25 cycle 
motor will have about the same speed as a 10-pole, 60 cycle 
motor. The 4-pole motor should, and usually does have smaller 
magnetizing current than the 10-pole. However, the 4r-pole 
machine for the same speed should require more material than 
the 10-pole, on account of higher magnetic flux conditions. There 
fore, if the 10-pole machine were made of larger dimensions than 
the 4-pole, but utilizing the 4-pole magnetic material, its magnet- 
izing current might be made fairly comparable with that of the 


4-pole machine. However, with the same total useful material, 
but arranged in larger dimensions, the idle material, such as 
frame, supports, etc., will be somewhat greater in the 10-pole 
machine of the same speed, and therefore, in general, for equal 
speed and equal characteristics, the 60 cycle induction motor 
should cost more than the 25 cycle. However, for general power 
distribution with relatively small motor capcities, it is not correct 
to compare a 10-pole 60 cycle motor with a 4-pole, 25 cycle; for, 
in most cases, 60 cycle motors of higher speed can be used, such 
as eight, six and four-pole, giving respectivley 900, 1200 and 1800 
r.p.m., neglecting the small slip. These higher speed and smaller 
number of poles in general more than offset the advantages of the 
25 cycle, 4-pole, 750 r.p m. motor as regards cost and character- 
istics, and at the same time, the greater choice in speeds is very 
advantageous. In 25 cycles, the highest speed is 1500, 
with, two poles, and experience has shown that, in size and con- 
struction, a 2-pole induction motor has very little advantage over 
a 4-pole, except possibly in very small capacities. Therefore, 60 
cycles, with its 4-pole 1800 r.p m., 6-pole 1200 r.p.m., 8-pole 900 
r.p.m. motors, has a decided commercial advantage over the 25 
cycle system with its 2-pole 1500 r.p.m., and 4-pole 750 r.p.m. 

However, when we compare, for instance, a moderate capacity 
12-pole, 250 r.p.m., 25 cyde with a 30-pole, 240 r.p.m., 60 cycle 
motor we may find the advantage considerably in favor of the 25 
cycle, so much so that if all the motors to be used in a given 
plant were of this speed or lower, and there were no other offsetting 
advantages for 60 cycles, such as lighting, etc., then the proposi- 
tion ; would look like a good one for 25 cycles. However, if only a 
small percentage of the total load is represented by such low speed 
motors, then the 60 cycle supply may make otherwise a sufficiently 
good showing to warrant its use. If, however, we go to the extreme 
case of moderate, or even very large, capacity motors at 75 to 100 
r.p.m., then we run into almost prohibitive constructions with 60 
cycles, either in size or in operating characteristics. At 60 cycles, 
such motors are liable to have such low power factors that the 
actual current taken by the motors is so large compared with the 
work current that, even with poor performance, a very large 
motor is required for a given capacity. In 25 cycles however, such 
motors can make a very much better showing. Therefore, at the 
present time, 25 cycles prepresents the most suitable frequency 


for such motors. However, hope may be extended for the 60 
cycle motor. If such motors are to be operated at constant 
speed, or even tinder variable or adjustable speeds, as has been 
described under an earlier subject, it is possible and practicable 
to overcome the difficulty of the poor power factor and large 
current from the supply system by connecting a special low fre- 
quency exciter in the secondary circuit of the induction motor, 
which will supply the magnetizing current to the secondary 
instead of the primary, just as in the non-synchronous type of 
condenser already referred to. This does not eliminate the 
magnetizing current in the motor, but simply puts it in the 
secondary circuit. 

The above is a considerable digression from the central station 
problem, but it has a direct bearing on the purchase of power by 
fmills from central stations. From the above, it is obvious that 
or the general sale of motor power to varied industries, the 160 
cycle central station has a direct advantage over the 25 cyce, 
in the great majority of service. 

When it comes to the question of delivering direct current 
from an alternating-current system, the 25 cycle system, in con- 
nection with rotary converters, is generally assumed to have 
considerable advantage over the 60 cycles. However, even that 
advantage is disappearing, due to recent advances in the design 
of high speed apparatus for converting from alternating to direct 
current. Where motor generators are used, 60 cycles in general 
allow a more satisfactory choice of converting set; for, in many 
cases, for a given capacity, the 60 cycle set can be, given a some- 
what higher speed than the 25 cycle. Therefore, the advantage 
of 25 cycle, if such exists, must lie in rotary converters. But 
recent advances in 60 cycle rotary converter construction have 
made the 60 cycle rotary a strong, and pretty reliable competitor 
of the 25 cycle rotary, so much so that, at the present time, quite 
a number of electric railways have shut down their own D. C. 
generating stations, and are buying power from 60 cycle central 
stations through 60 cycle rotaries. This development has removed 
one of the most serious handicaps of the 60 cycle system, so that 
the present tendency of central station work, and even power 
transmission, is strongly toward 60 cycles. The steel mill en- 
gineers should therefore keep this tendency strongly in mind. 


FOREWORD-^-This raper was presented at the twenty-eighth annual 
convention of the Association of Edison Illuminating Companies 
at Hot Springs, Va., September, 1912. At that time, the syn- 
chronous booster type of converter was becoming well estab- 
lished and it was the author's purpose to show in this paper some 
of the conditions of commutation which were encountered in the 
synchronous type of machine. (ED.) 

EXPERIENCE shows that the rotary converter is one of the 
most satisfactory and reliable of the various types of rotating 
electrical machinery. In its efficiency of transformation, its com- 
mutation and temperature characteristics, and its operating 
characteristics in general, it makes an extremely good showing. 
Furthermore, it is a type of machine which has not changed greatly 
in the last decade. The more recent developments have been 
principally in the use of commutating poles and in a very material 
increase in the rotative speeds, these two features, however, being 
closely allied, as will be shown later. 

In considering the various characteristics of the rotary con- 
verter, such as its current and voltage capacities, e. m. f . regula- 
tion, commutation and the use of commutating poles, maximum 
speeds permissible with a given output, etc., certain fundamental 
conditions or limitations in the design, are of controlling import- 
ance. In order to obtain a fuller understanding of the possibilities 
and capabilities of such apparatus, a brief consideration of these 
fundamental conditions will be given. 


One condition of controlling importance in all commutating 

machinery is the commutation. If the machine does not com- 

mutate well, then perfections in other features are overshadowed. 

High efficiency, low temperature rise, and low first cost, do not 

outweigh bad operation at the commutator. 

In the ordinary commutating machine, the armature winding, 
when carrying current, sets up local magnetic fields, or fluxes, 
across which the armature conductors cut and thus generate 



e. m. f.'s, just as when they cut across the main field fluxes. These 
local fields, due to the armature ampere turns, usually have peak 
values at those points on the armature where one or more armature 
coils are short-circuited by the brushes on the commutator. The 
conductors or turns which are thus short-circuited, have certain 
voltages generated in them, and the brushes are short-circuiting 
across these voltages. It may thus be said that there is a certain 
short-circuit voltage per armature coil, or between adjacent com- 
mutator bars, which may be called the inherent short-circuit e, m. /. 
per bar. If the brush is wide enough to cover several bars, then it 
short-circuits the voltages of several bars. The average value of 
this may be called the inherent brush short-circuit e. m. f. The value 
of this latter is of utmost importance in commutating machinery, 
for it is upon this, and the resistance of the brush, that the amount 
of short-circuit, or "local," current depends. 

When the work current, or that which flows to the external 
circuit, passes from the commutator to the brush, it should be 
distributed evenly over the whole brush contact, providing there 
are no disturbing conditions. On the basis of uniform distribu- 
tion of current over the brush contact, the minimum current 
density at the brush contact would naturally be obtained, which 
would be an ideal condition in many ways. This ideal distri- 
bution of the work current over the brush contact area will be 
called the apparent current density in the brush, to distinguish it 
from the true current density, which is due to the resultant current 
in the brush, which is always greater than the work current. 

The principal cause of the difference between the true and the 
apparent densities in the brush lies in the local or short-circuited 
current, due to the brush short-circuit voltage just described 
This local current distributes over the brush contact according to 
the short-circuited voltages under the brush contact, and is thus 
practically zero at the middle of the brush, and maxitnum at the 
edges, flowing from the commutator to the brush at one side of 
the mid-point, and from the brush to the commutator at the other 
side. It thus adds to the work current at one side of the brush, 
and subtracts from it at the other side, and, not infrequently, the 
local current is so great, relatively, that the resultant current at 
one brush side will be several times that due to the work current, 
while at the other edge it will actually be in the opposite direction. 

This condition can be illustrated by Figs. 1, 2 and 3. 



Fig. 1 represents the conditions where only the work current 

The height ab, which is uniform, represents the value of the 
work current. 

Fig. 2 represents the conditions under the brush if only the 
local current is considered (on the assumption that the field due to 
the armature work current is present, but the work current itself 
is absent). 

ac represents the maximum current in one direction at one 
edge of the brush, while de represents an equal and opposite current 
at the other edge. 



Fig. 3 represents the conditions when both currents are 
present. At one edge of the brush the current is excessive com- 
pared to the work current, while at the other edge the current is in 
the opposite direction. Obviously, the part of the brush between 
d and f in this figure is not only useless, but is worse than useless, 
for it not only does not carry any current into the armature, but 
actually adds to the current carried by the part between a and /. 
Therefore, if the part between d and/ were actually cut away, the 
remaining part between a and / would not be worked as hard as 
before. This diagram represents a- somewhat extreme condition, 
but is not an unusual one, as experience has shown, for in a great 
many commutating machines in actual service, improved results 
have been obtained by narrowing "the brush contact a certain 


amount. Obviously, the apparent current density in the brush 
would be represented by the height ab, while the true current 
density would be represented by the maximum height ag, in Fig- 3, 
which may be several times as great as the height ab. 

It is evident from consideration of the above figures that the 
conditions would be greatly improved by any reduction in the 
value of the local or short-circuit current. Narrowing the brush, 
as mentioned above, is, to a certain extent, effective. This reduces 
the local current, but at the same time it reduces the effective path 
for the work current. Another partial remedy would be in the use 
of higher resistance at the brush contact, such as is furnished by 
certain makes of brush. This would reduce the local current 
without reducing the area of the brush contact, but at the same 
time it introduces resistance in the path of the work current, which 
is practically equivalent to reducing the area of the path. It is, 
therefore, to a certain extent, equivalent to narrowing the brush. 
A third and more satisfactory method is to reduce the inherent 
short-circuit voltage across the brush, while at the same time 
retaining the full width of the brush. This, however, is a question 
of design and the proportioning of the machine itself, and obviously 
such modification cannot readily be supplied to a machine al- 
ready constructed. This method of correcting trouble will be 
referred to again. 

The above figures illustrating the effect of the local current, 
do not make the story quite as bad as it actually is. If the brush 
contact resistance in a given brush were of constant value, irres- 
pective of the current in it, then the above illustrated conditions 
would hold. But the brush resistance is actually variable in 
effect; that is, at ordinary working current densities, the e. m. f. 
drop across the brush contact does not increase directly with the 
current, but at a much less rate. This, therefore, is equivalent 
to a decrease in the resistance of the brush contact with increase 
in current and, unfortunately, this decrease is very pronounced, 
even within the limits of permissible current densities. Therefore, 
with local current in the brush, giving high densities at the outer 
edges, the resistance of the brush may be so reduced as to give even 
worse distribution than indicated by Fig. 3. 

Cotisidering next, actual permissible brush drops, it may be 
noted that, as the local current enters at one side of the brush and 
leaves at the other side, the contact resistance in series with the 


local'cttrrent path is twice the ordinary contact resistance between 
brush and commutator. From an examination of a large amount 
of data on brush drops, it appears that, with the ordinary com- 
mercial brushes, there is about 1 to 1.25 volts drop between the 
brush and the commutator when carrying currents of 30 to SO 
amperes per square inch. With the brush contact resistance 
indicated by these drops, it is evident that with a short-circuit 
voltage of 2 to 2 ^ volts across the brush, a local current could 
flow which would have a value at the brush edges corresponding 
to a current density of 30 to 50 amperes. Assuming a short-cir- 
cuit voltage which would give a density of 50 amperes per square 
inch at the edges, then with a work current flowing which also 
gives an apparent current density of 50 amperes, the resultant 
density at one brush edge would become zero, while at the other 
edge it would become 100 amperes per square inch. With brushes 
having a low contact resistance the conditions would be worse, 
and there would be a current of negative direction at one edge of the 

In practice, an inherent brush short-circuit e. m. f . of 2 to 
volts is very seldom found, as it is too low a value for commer- 
cial designs. However, with much higher short-circuit e. m. f .'s 
the conditions would obviously be very much worse than indicated 
above, and yet in commutating machines of the non-commutating 
pole type, inherent short-circuit e. m. f.'s of 4 or 5 volts would be 
considered relatively low, and even 7 or 8 volts would not be 
considered unduly high in some cases. Evidently, with such 
e. m. f.'s actually across the brush, the local currents in the brush 
should be excessive and there should be severe sparking and 
burning at the brushes and commutator. However, this im- 
possible condition is overcome to a considerable extent by generat- 
ing an opposing voltage in the short-circuited coils. This result 
is obtained in non-commutating pole machines by shifting the 
brushes toward one of the pole corners to such an extent that the 
short-circuited coils are cutting across a small part of the main 
field flux, which thus generates a small e. m. f . in them. The shift 
of the brushes must always be in such a direction that this e. m. f ., 
due to the main field, is in opposition to the short-circuit e. m. f . 

To illustrate the above case, let it be assumed that the in- 
herent short-circuit voltage across the brush at full load, with no 
lead at the brushes, is six volts. This, if not partially neutralized, 
would generate an unduly high local current, so that the operating 


conditions would be comparatively bad. Then, assume that the 
brushes are shifted so that the short-circuited coils are cutting 
across a main field flux sufficient to give three volts. As this is in 
opposition to the normal short-circuit e. m. f ., the resultant short- 
circuit e. m. f . will be equal to 6 3 = 3 volts, which would not 
be anything like as bad as before. If, now, the load is removed 
from the machine, the brushes still retaining their lead, the three 
volts due to the main field will still be generated in the short- 
circuit armature coils, and there will be a no-load short-circuit 
e. m. f. of three volts, which would set up a local short-circuit 
current. However, as no work current is present under this con- 
dition, the short-circuit current could obviously be practically as 
great as the maximum value of the resultant current at the full 
load conditions. Therefore, if three volts short-circuit e. m. f . is 
permissible at full load, then four or five volts would be permissible 
at no load with practically the same commutating conditions as at 
full load. Therefore, the brush could be shifted forward into a 
field representing four volts, for instance, and thus at no load the 
short-circuit voltage will be four volts, while at full load it would be 
6 4 = 2 volts. Therefore, by this means an impossible com- 
mutating condition, represented by no lead at the brushes, be- 
comes a possible and practicable condition by giving a certain 
amount of lead. On non-commutating pole machines where a 
slight amount of lead is almost always required, a resultant short- 
circuit e. m. f . of three volts across the brush may be permissible, 
in some cases, at full load, but this cannot be assumed to be true in 
all cases, for there are other conditions, besides commutation, which 
are dependent upon the amount and distribution of currents in the 
brush. Of these other effects, the principal ones may be classified 
as, burning of the commutator and brush faces, high mica, and 
picking up of copper. 


An elaborate and long extended series of tests has shown that 
when a relatively large current passes from a brush to a commutator 
or collector ring, or vice-versa, there is a tendency for undue 
"wear," as it might be called, of either the commutator or brush 
face, depending upon the direction of current. If the current is 
from the commutator to the brush, then the commutator face 
"wears" or is "eaten" away, while with the current from the 
brush to the commutator, the brush shows increased wear. This 


is not a true mechanical wearing away of the commutator or 
brush, but is more like an electrolytic action, except that usually 
the particles taken from one surface do not deposit on the other. 
This rate of wear, as shown by test, is a function of the current 
density, the area of surface through which the current passes, and 
the contact drop. It is not directly proportional to the contact 
drop, or the current, but increases in a much greater proportion 
than either, or possibly even more rapidly than the product of the 
two. However, this is difficult to determine definitely, for with 
the wear once started, the trouble tends to accentuate itself. In 
other words, this wear will increase the contact drop and in turn 
the increase in contact drop will exaggerate the wear, so that the 
action is cumulative. This wearing action is apparently very slight 
in amount at true brush densities of SO to 60 amperes per square 
inch, with carbon brushes, and if the commutating characteristics 
are very good, even much greater true densities are practicable, 
possibly up to 100 amperes per square inch. If the apparent 
density could be brought up to the true density; that is, if no 
current but the work current were present, then this high current 
density in the brush might be utilized in well designed machines, 
but this implies the absence of all local currents, also, perfect 
division of the current between the various brushes and brush 
arms, as will be referred to later. These two conditions are rarely 
attained in practice, and it would probably be dangerous to 
attempt apparent densities of 100 amperes per square inch in the 
ordinary carbon brush; but with commutating pole machines, 
where an opposing e. m. f . is generated in the short-circuited 
armature coils, the condition of relatively small local currents can 
be obtained by very careful proportioning of the commutating pole 
field. This means therefore that higher current densities in the 
brushes are feasible in commutating pole machines in general than 
in the non-commutating pole type. This has a direct bearing on 
the synchronous converter problem, as will be shown later when 
considering high speeds and marimum outputs with a given 
number of poles. However, the condition of perfect division of 
current between the different brushes has not been obtained in any 
simple, practical manner, and therefore some margin in brush 
current density must be allowed, even in commutating pole ma- 



When the maximum current density in a brush contact is 
comparatively high, due to local currents or other causes, the 
commutator and brush "wear" may be relatively rapid compared 
with the mechanical wear due to friction of the brushes on the 
commutator. Under this apparent wear the commutator copper 
will be slowly eaten away by the current, but the commutator mica 
will not be materially affected. The mica must wear down by the 
mechanical friction of the brushes, If the "eating away" of the 
copper exceeds the mechanical wear of the mica, then a condition 
is reached which tends to increase the def ect. As soon as the copper 
face is burned even an infinitesimal amount below the mica, the 
brush face tends to "ride" on the mica and thus has a reduced 
contact on the copper surface, or even none at all. This condition 
increases the burning action and eventually results in the so-called 
"high mica" where there is an actual gap between the brush and 
the commutator face Such a condition, once started, does not 
tend to cure itself, except tinder certain special conditions of 
operation. This high mica is frequently charged to the use of 
"hard" mica, which tends to produce a similar condition. 

In some cases, this trouble from high mica may not be due to 
either excessive local currents or hard mica, but may be due to a 
relatively high proportion of mica to copper surface. Where 
comparatively thin commutator bars are used on a machine, the 
thickness of mica between the bars is not reduced in proportion, 
so that the percentage of mica may be relatively high. In con- 
sequence of this high percentage, the mica itself does not wear as 
rapidly as where a less total amount is used, while the copper may 
eat away at the same rate. This may therefore tend toward high 
mica, even where the local currents are relatively small. This 
condition of high percentage of mica is found particularly in high 
voltage machines where the number of bars is necessarily great 
and the thickness, of each bar correspondingly small. On the 
other hand, with low voltage machines, the percentage of mica is 
relatively less, but other conditions may enter which partly 
neutralize this advantage. With lower voltages, for a given 
capacity, the current is correspondingly greater and, with a given 
contact drop, the losses are correspondingly increased and the 
tendency to produce noise by the brushes is also greater. To 
overcome these objectionable features, a soft, low resistance brush 
is frequently used. This, however, increases the tendency for 


local currents and thus increases the copper wear, while at the same 
time a softer brush has less grinding action on the mica. There- 
fore, the low-voltage machine may also tend toward high mica. 

A common, and very effective, remedy for this tendency 
toward high mica is to "undercut" the mica so that, everywhere 
on the brush wearing surface, it lies slightly below the copper 
surface. This does not remove the initial cause of the trouble, 
namely, the tendency to eat away the copper surface. But it 
must be considered that this initial tendency is usually very slight, 
and that the major part of the wear is due to the lessening of the 
contact between the brushes and the copper, thus increasing the 
burning tendency. In consequence of undercutting the mica, the 
brush can always maintain good contact with the commutator 
face, and thus the actual burning may be so slow as to be practically 
negligible. The true gain from undercutting the mica thus lies 
in the maintenance of more intimate contact between the copper 
and the brush. 

This eating away of the commutator face may occur in service 
and yet the commutator may polish beautifully. This is found 
in some cases where the burning action is pronounced, and yet the 
conditions of operat? jn are such that the mica can be worn down 
mechanically as rapidly as the copper bums away. This is not 
infrequently the case with machines where there are heavy peak 
loads of relatively short duration, followed by very much longer 
periods of operation with but little load. Under such conditions 
the burning action during the peak loads, with a consequent 
tendency to high mica, is hidden by the grinding action of the 
brushes on the mica during the long periods of operation at light 
load, so that the mica is kept practically flush with the copper and 
the copper surface is polished. That real burning is present is 
often indicated, in such machines, by relatively rapid wear on the 
commutator in grooves when the brushes are not well staggered. 


Another condition which sometimes accompanies high current 
density in the brushes, is the so-called "picking up of copper." 
Apparently, under some conditions, particles of copper, eaten 
away from the commutator face, will collect on the brush face. 
This may result in glowing at the brush contact, eventual burning 
away or "honey-combing" of the brush surface and general 
trouble at the commtitator. This difficulty is possibly largely 
cumulative in its action. A slight qoating of copper, or copper 


"spots," may form on a brush. This gives a more intimate, or 
lower resistance, contact with the commutator face. With many 
brushes in parallel, an undue percentage of the total current may 
then pass through this one point, or brush, or low resistance con- 
tact, and the current density at this point may even become so 
great that the burning will be excessive. The resistance of the 
carbon brush, in itself, does not help this condition, for, un- 
fortunately for this case, carbon has a negative coefficient of 
resistance so that heating lowers its resistance and thus accentuates 
the unequal division of current. One remedy for this condition 
is a more uniform contact resistance between the brush and the 
commutator. Experience has shown that undercutting the mica 
will frequently overcome this difficulty of picking up copper, 
particularly so if the machine can be "nursed" until the com- 
mutator face acquires a glaze. In some cases, a different grade 
of brush will be an improvement, but it is generally difficult to 
predict the most suitable brush, unless the inherent commutating 
characteristics of the machine are well known. This picking up 
of copper appears to be, to a great extent, a function of the cur- 
rent density, and is apparently somewhat of an electrolytic action, 
the copper eating away from the commutator and depositing upon 
the brush. Whatever tends to materially reduce the tendency 
for the commutator face to eat away, also tends to reduce the 
picking-up effect. 

The foregoing features, while apparently minor in nature, are 
all of fundamental importance in commutating machinery in 
general, and particularly so in the case of commutating-pole 
rotary converters, especially in those commutating-pole rotaries 
which have what might be called self-contained or "auto" regula- 
tion of voltage, such as those with synchronous boosters, or with 
regulating poles. 


In the direct-current generator of large capacity and high 
speed the commutating pole has proved to be a real necessity. 
In such machines, due to the reduced number of poles and high 
armature ampere turns per pole, and consequent large fields or 
fluxes set up by the armature, together with the high speed, the 
inherent short-circuit voltages across the brush have reached 
excessive values, such as 12 to 14 volts at normal load. Such 
e. m. f /s, unless largely neutralized, would obviously set up exces- 


sive short-circuit currents under the brush. As a resultant short- 
circuit voltage under the brush of about 2 volts or less at full load 
is desirable, it is obvious that some such device as the commut- 
ating pole, which introduces an opposing e. m. f. in the short- 
circuited armature coils, is practically a necessity; and, further- 
more, this opposing e. m. f . must vary practically in proportion to 
the load, in order to keep within the permisssible short-circuit limits 
across the brush at all loads. Shifting the brushes forward into 
an active field to neutralize 12 volts, for instance, is obviously 
impracticable, for if a sufficient opposing e m.f., such as 10 volts, 
is thus introduced into the short-circuited coils at full load, then 
it is so large that it will give prohibitive currents at no load if the 
same brush lead is maintained. Therefore with such a machine 
of the non-commutating pole type, the brushes must be shifted 
with the load, which, in many cases, is not a practicable condi- 
tion. Consequently, the commutating pole, with its neutralizing 
e. m. f . varying in proportion to the load, is a necessary dev