!)^^)iJiasa'^i^tJia;xga!Jia!>sai®j!ji«!«<)i«<):«i
^lln
Kansas (Ettg
publtr ICtbrarg
This Volume is for
REFERENCE USE ONLY
c
1U29— 6m— H
From the collection of the
re
THE BELL SYSTEM
TECHNICAL JOURNAL
A JOURNAL DEVOTED TO THE
SCIENTIFIC AND ENGINEERING
ASPECTS OF ELECTRICAL
COMMUNICATION
EDITORIAL BOARD
Bancroft Gherardi F. B. Jewett
H. P. Charlesworth W. H. Harrison E. H. Colpitts
L. F. Morehouse H. D. Arnold O. B. Blackwell
Philander Norton, Editor J. O. Perrine, Associate Editor
INDEX
VOLUME IX
1930
AMERICAN TELEPHONE AND TELEGRAPH COMPANY
NEW YORK
« ^ J w • • •
•J
. •• •
Mr 1 6]
(iS7H77
THE BELL SYSTEM
TECHNICAL JOURNAL
VOLUME IX, 1930
TABLE OF CONTENTS
January, 1930
Telephone Communication System of the United States —
Bancroft Gherardi and F. B. Jcivett 1
Structure and Nature of Troostite — Francis F. Lucas 101
Radio Broadcasting Transmitters and Related Transmission Phe-
nomena — Edzvard L. Nelson 121
Wire Line Systems for National Broadcasting — A. B. Clark 141
Notes on the Heaviside Operational Calculus — John R. Carson ... 150
Contemporary Advances in Physics, XIX — Karl K. Darroiv 163
Wave Propagation Over Continuously Loaded Fine Wires —
M. K. Zinn 189
Theory of Vibration of the Larynx — R. L. Wegel 209
April, 1930
Developments in Communication Materials — William Fondillcr . . . 237
Transoceanic Telephone Service — Short- Wave Transmission —
Ralph Bown 258
Transoceanic Telephone Service — -Short-Wave Equipment —
A. A. Oszvald 270
The \\^ords and Sounds of Telephone Conversations — Norman R.
French, Charles W. Carter, Jr., and Walter Koenig, Jr 290
The Reciprocal Energy Theorem — John R. Carson 325
The Approximate Networks of Acoustic Filters — W. P. Mason . . 332
Contemporary Advances in Physics, XX — Karl K. Darrow 341
Motion of Telephone Wires in Wind — D. A. Quarks 356
Economic Quality Control of Manufactured Product —
W. A. Shewhart 364
Optimum Reverberation Time for Auditoriums —
Walter A. MacNair 390
3
BlilJ. SYSTEM TECHNICAL JOURNAL
July, 1930
Radio Telephone Service to Ships at Sea —
William Wilson and Lloyd Espcnschicd 407
A General Switching Plan for Telephone Toll Service —
//. S. Osborne 429
Image Transmission System for Two- Way Television —
Herbert E. Ives, Frank Gray and M. W. Baldwin 448
Synchronization System for Two-Way Television — H. M. Stollcr 470
Sound Transmission System for Two-Way Television —
D. G. Blattner and L. G. BosHvick 478
Transmitted Frequency Range for Telephone Message Circuits —
W. H. Martin 483
Some Recent Developments in Long Distance Cables in the United
States of America— .4. B. Clark 487
Phase Distortion in Telephone Apparatus — C. E. Lane 493
Measurement of Phase Distortion — H. Nyqnist and S. Brand .... 522
Effects of Phase Distortion on Telephone Quality —
John C. Steinberg 550
Long Distance Cable Circuit for Program Transmission —
A. B. Clark and C. W. Green 567
October, 1930
Chemistry in the Telephone Industry — Robert R. Ullliains 603
The Trend in the Design of Telephone Transmitters and Receiv-
ers— fi^. H. Martin and W. F. Davidson 622
Mutual Impedances of Ground-Return Circuits —
A. E. Boiven and C. L. Gilkeson 628
A Survey of Room Noise in Telephone Locations —
W. J. Williams and Ralph G. MeCnrdy 652
Contemporary Advances in Physics, XXI — Karl K. Darrozv 668
A Study of Telephone Line Insulators — L. T. IVilson 697
The Transmission Characteristics of Open-Wire Telephone Lines —
E. L Green 730
Transients in Parallel Grounded Circuits, One of Which is of In-
finite Length — Liss C. Peterson 760
Impedance Correction of Wave Filters — E. B. Payne 770
A Method of Impedance Correction — //. W. Bode 794
Index to Volume IX
Approximate Networks of Acoustic Filters, The, JV. P. Mason, page 332.
Auditoriums, Optimum Reverberation Time for, Walter A. MacNair, page 390.
B
Baldwin, M. W., Herbert E. Ives and Frank Gray, Image Transmission System for
Two-Way Television, page 448.
Blattner, D. G. and L. G. Bostivlck, Sound Transmission System for Two- Way
Television, page 478.
Bode, H. W., A Method of Impedance Correction, page 794.
Bostzmck, L. G. and D. G. Blattner, Sound Transmission System for Two-Way
Television, page 478.
Bowen, A. E. and C L. Gilkeson, Mutual Impedances of Ground-Return Circuits —
Some Experimental Studies, page 628.
Boivn, Ralph, Transoceanic Telephone Service — Short-Wave Transmission, page
258.
Brand, S. and H. Nyquist. Measurement of Phase Distortion, page 522.
Broadcasting, National, Wire Line Systems for, A. B. Clark, page 141.
Cable Circuit, Long Distance, for Program Transmission, A. B. Clark and C. IV.
Green, page 567.
Cables, Long Distance, in the United States of America, Some Recent Develop-
ments in, A. B. Clark, page 487.
Carson, John R., Notes on the Heaviside Operational Calculus, page 150.
Carson, John R., The Reciprocal Energy Theorem, page 325.
Carter, Charles W., Jr.. Norman R. French, and Walter Koenig, Jr., The Words
and Sounds of Telephone Conversations, page 290.
Chemistry in the Telephone Industry, Robert R. Williams, page 603.
Clark, A. B., Wire Line Systems for National Broadcasting, page 141.
Clark, A. B., Some Recent Developments in Long Distance Cables in the United
States of America, page 487.
Clark, A. B. and C. JV. Green, Long Distance Cable Circuit for Program Trans-
mission, page 567.
Contemporary Advances in Physics, XIX. Fusion of Wave and Corpuscle The-
ories, Karl K. Darrozv, page 163.
Contemporary Advances in Physics, XX. Ionization of Gases by Light, Karl K.
Darrozv, page 341.
Contemporary Advances in Physics, XXI. Interception and Scattering of Elec-
trons and Ions, Karl K. Darrozv, page 668.
Control, Economic Quality, of Manufactured Product, W. A. Shezvhart, page 364.
Corpuscle Theories, Fusion of Wave and, Karl K. Darrozv, page 163.
Darrozv, Karl K.. Contemporary Advances in Physics, XIX. Fusion of Wave and
Corpuscle Theories, page 163.
5
BELL SYSTEM TECHNICAL JOURNAL
Darrotv, Karl K., Contemporary Advances in Physics, XX. Ionization of Gases
by Light, page 341.
Darrozv, Karl K., Contemporary Advances in Physics, XXI. Interception and
Scattering of Electrons and Ions, page 668.
Davidson, W. F. and IV. H. Martin, The Trend in the Design of Telephone Trans-
mitters and Receivers, page 622.
Developments in Communication Materials, William Fondillcr, page 237.
E
Economic Quality Control of Manufactured Product, W. A. Shewhart, page 364.
Effects of Phase Distortion on Telephone Quality, John C. Steinberg, page 550.
Electrons and Ions, Interception and Scattering of. Contemporary Advances in
Physics, XXI, Karl K. Darrozv, page 668.
Espcnschicd, Lloyd and William Wilson, Radio Telephone Service to Ships at Sea,
page 407.
Filters, Acoustic, The Approximate Networks of, W. P. Mason, page 332.
Filters, Wave, Impedance Correction of, E. B. Payne, page 770.
Fondiller, William, Developments in Communication Materials, page 237.
French, Norman R.. Charles W. Carter, Jr., and Walter Koenig, Jr., The Words
and Sounds of Telephone Conversations, page 290.
Frequency Range for Telephone Message Circuits, Transmitted, W. H. Martin,
page 483.
G
General Switching Plan for Telephone Toll Service, A, H. S. Osborne, page 429.
Gherardi, Bancroft and F. B. Jezvett, Telephone Communication System of the
United States, page 1.
Gilkeson, C. L. and A. E. Bozven, Mutual Impedances of Ground Return Circuits —
Some Experimental Studies, page 628.
Gi-ay, Frank, Herbert E. Ives, and M. W. Baldzmn, Image Transmission System
for Two-Way Television, page 448.
Green, C. JV. and A. B. Clark, Long Distance Cable Circuit for Program Trans-
mission, page 567.
Green, E. L, The Transmission Characteristics of Open-Wire Telephone Lines,
page 730.
H
Heaviside Operational Calculus, Notes on the, John R. Carson, page 150.
Image Transmission System for Two-Way Television, Herbert E. Ives, Frank
Gray, and M. W. Baldzvin, page 448.
Impedances, Mutual, of Ground-Return Circuits — Some Experimental Studies, A.
E. Bozven and C. L. Gilkeson, page 628.
Impedance Correction of Wave Filters, E. B. Payne, page 770.
Insulators, Telephone Line, A Study of, L. T. Wilson, page 697.
Ionization of Gases by Light, Contemporary Advances in Physics, XX, Karl K.
Darrozv, page 341.
Ives, Herbert E., Frank Gray, and M. W. Baldzvin, Image Transmission System
for Two-Way Television, page 448.
BELL SYSTEM TECHNICAL JOURNAL
Jewett, F. B. and Bancroft Gherardi, Telephone Communication System of the
United States, page 1.
K
Koenig, Walter, Jr., Norman R. French, and Charles W. Carter, Jr., The Words
and Sounds of Telephone Conversations, page 290.
Lane, C. E., Phase Distortion in Telephone Apparatus, page 493.
Larynx, Theory of Vibration of, R. L. Wegel, page 209.
Loaded, Continuously, Fine Wires, Wave Propagation Over, M. K. Zinn, page 189.
Long Distance Cable Circuit for Program Transmission, A. B. Clark and C. W.
Green, page 567.
Lucas, Francis F., Structure and Nature of Troostite, page 101.
M
McCurdy, R. G. and IV. J. Williams, A Survey of Room Noise in Telephone Loca-
tions, page 652.
MacNair, Walter A., Optimum Reverberation Time for Auditoriums, page 390.
Martin, W. H., Transmitted Frequency Range for Telephone Message Circuits,
page 483.
Martin, W. H. and W. F. Davidson, The Trend in the Design of Telephone Trans-
mitters and Receivers, page 622.
Mason, W. P., The Approximate Networks of Acoustic Filters, page 332.
Materials, Communication, Developments in, William Fondiller, page 237.
Measurement of Phase Distortionj H. Nyquist and 5". Brand, page 522.
Method of Impedance Correction, A, H. W. Bode, page 794.
Motion of Telephone Wires in Wind, D. A. Quarks, page 356.
Mutual Impedances of Ground-Return Circuits — Some Experimental Studies, A. E.
Bozvcn and C. L. Gilkcson, page 628.
N
Nelson, Edivard L., Radio Broadcasting Transmitters and Related Transmission
Phenomena, page 121.
Networks of Acoustic Filters, The Approximate, JV. P. Mason, page 332.
Noise, Room, in Telephone Locations, A Survey of, W. J. Williams and R. G.
McCurdy, page 652.
Notes on the Heaviside Operational Calculus, John R. Carson, page 150.
Nyquist, H. and 6'. Brand, Measurement of Phase Distortion, page 522.
Open-Wire Telephone Lines, The Transmission Characteristics of, E. I. Green,
page 730.
Optimum Reverberation Time for Auditoriums, Walter A. MacNair, page 390.
Osborne, H. S., A General Switching Plan for Telephone Toll Service, page 429.
Osivald, A. A., Transoceanic Telephone Service — Short-Wave Equipment, page 270.
Payne, E. B., Impedance Correction of Wave Filters, page 770.
Peterson, L. C, Transients in Parallel Grounded Circuits, One of Which is of In-
finite Length, page 760.
7
BELL SYSTEM TECHNICAL JOURNAL
Phase Distortion in Telephone Apparatus, C. E. Lane, page 493.
Phase Distortion, Effects of, on Telephone Quality, John C. Slcinberg, page 550.
Phase Distortion, Measurement of, //. Nyquist and S. Brand, page 522.
Physics, XIX, Contemporary Advances in. Fusion of Wave and Corpuscle The-
ories, Karl K. Darroiv, page 163.
Physics, XX, Contemporary Advances in. Ionization of Gases by Light, Karl K.
Darroiv, page 341.
Physics, XXI, Contemporary Advances in. Interception and Scattering of Elec-
trons and Ions, Karl K. Darroiv, page 608.
Q
Quarles, D. A., Motion of Telephone Wires in Wind, page 356.
R
Radio Broadcasting Transmitters and Related Transmission Phenomena, Edward
L. Nelson, page 121.
Radio Telephone Service to Ships at Sea, William Wilson and Lloyd Espenschied,
page 407.
Receivers, The Trend in the Design of Telephone Transmitters and, W. H. Martin
and W. F. Davidson, page 622.
Reciprocal Energy Theorem, The, John R. Carson, page 325.
Reverberation Time for Auditoriums, Optimum, Walter A. MacNair, page 390.
Shewhart, W. A., Economic Quality Control of Manufactured Product, page 364.
Ships at Sea, Radio Telephone Service to, William Wilson and Lloyd Espenschied,
page 407.
Short-Wave Equipment, in Transoceanic Telephone Service, A. A. Oswald, page
270.
Short-Wave Transmission, in Transoceanic Telephone Service, Ralph Bown, page
258.
Some Recent Developments in Long Distance Cables in the United States of Amer-
ica, A. B. Clark, page 487.
Sounds, and Words, of Telephone Conversations, Norman R. French, Charles W.
Carter, Jr., and Walter Koenig, Jr., page 290.
Sound Transmission System for Two-Way Television, D. G. Blattner and L. G.
Bostwick, page 478.
Steinberg, John C, Effects of Phase Distortion on Telephone Quality, page 550.
Stollcr, H. M., Synchronization System for Two-Way Television, page 470.
Structure and Nature of Troostite, Francis F. Lucas, page 101.
Study of Telephone Line Insulators, A, L. T. Wilson, page 697.
Survey of Room Noise in Telephone Locations, A, W. J. Williams and R. G. Mc-
Curdy, page 652.
Switching Plan for Telephone Toll Service, A, H. S. Osborne, page 429.
Synchronization System for Two-Way Television, H. M. Stollcr, page 470.
Telephone Communication System of the United States, Bancroft Ghcrardi and
F. B. Jcwett, page 1.
Television, Two-Way, Image Transmission System for, Herbert E. Ives, Frank
Gray, and ^L W. Baldwin, page 448.
8
BELL SYSTEM TECHNICAL JOURNAL
Television, Two-Way, Synchronization System for, H. M. Stoller, page 470.
Television, Two-Way, Sound Transmission System for, D. G. Blattner and L. G.
Bostwick, page 478.
Theory of Vibration of the Larynx, R. L. Wcgel, page 209.
Toll Service, Telephone, A General Switching Plan for, H. S. Osborne, page 429.
Transients in Parallel Grounded Circuits, One of Which is of Infinite Length, L.
C. Peterson, page 760.
Transmission Characteristics of Open-Wire Telephone Lines, The, E. L Green,
page 730.
Transmitted Frequency Range for Telephone Message Circuits, W. H. Martin,
page 483.
Transmitters and Receivers, Telephone, The Trend in the Design of, W. H. Mar-
tin and IV. F. Davidson, page 622.
Transoceanic Telephone Service — Short-Wave Equipment, A. A. Oszvald, page 270.
Transoceanic Telephone Service — Short-Wave Transmission^ Ralph Sown, page
258.
Trend in the Design of Telephone Transmitters and Receivers, The, IV. H, Martin
and W. F. Davidson, page 622.
Troostite, Structure and Nature of, Francis F. Lucas, page lOL
W
Wave and Corpuscle Theories, Fusion of. Contemporary Advances in Physics,
Karl K. Darrow, page 163.
Wave Propagation Over Continuously Loaded Fine Wires, M. K. Zinn, page 189.
Wegel, R. L., Theory of Vibration of the Larynx, page 209.
Williams, Robert R., Chemistry in the Telephone Industry, page 603.
Williams, W. J. and R. G. McCurdy, A Survey of Room Noise in Telephone Loca-
tions, page 652.
Wilson, L. T., A Study of Telephone Line Insulators, page 697.
Wilson, Willia})! and Lloyd Espenschied, Radio Telephone Service to Ships at Sea,
page 407.
Wind, Motion of Telephone Wires in, D. A. Quarles, page 356.
Wire Line Systems for National Broadcasting, A. B. Clark, page 141.
Words and Sounds of Telephone Conversations, The, Norman R. French, Charles
IV. Carter, Jr., and Walter Koenig, Jr., page 290.
Z
Zinn, M. K., Wave Propagation Over Continuously Loaded Fine Wires, page 189.
The Bell System Technical Journal
January, 1930
Telephone Communication System of the United States ^
By BANCROFT GHERARDI and F. B. JEWETT
This paper presents the results which have been obtained up to the
present time in developing telephone communication in the United States
of America, this development having been worked out in a form to meet the
particular conditions which present themselves in that country. The paper
first deals with a brief description of the general structure and organization
of the telephone communication system giving the organization of the
Bell System which handles the greater part of the telephone service of
the country and the reasons for and advantages of this organization. _ In
this connection some figures are presented with respect to the technical
personnel who are continuously engaged in studies to develop the art and
to provide new methods and facilities for improving the service.
Local service, that is the service within the limits of a single telephone
exchange area, is next discussed. Figures are given with respect to the vol-
umes of telephone calls handled in the Bell System, the speed with which
the connections for these calls are completed and the operating force re-
quired. Reference is also made to the standards of transmission given
and the various problems encountered in meeting these standards. Figures
are given with respect to station growth, to the increased efficiency of sta-
tion apparatus and to the improvement in types of instruments. Various
types of private branch exchanges provided to meet the needs of customers
using a large amount of telephone service are discussed. The cable plant
is considered mainly from the construction standpoint and typical illus-
trations are given of some of the construction practices. The various
types of central office switcning systems in common use are described,
including magneto, common battery and dial systems, the latter including
both the step-by-step and panel systems which ate being provided in
increasing amounts in the Bell System. The subject of buildings to house
these various equipments as well as the operating forces and headquarters
staffs in many cases is briefly discussed, also standardized layouts and
floor plans. The problem of giving telephone service in the rural com-
munities, which is a very important one in the telephone development
in the United States, is also biiefly treated.
The toll service is considered, first with respect to the shorter haul toll
business and the problems involved and then with lespect to the long
distance toll service. Figures are given showing the speed of service and
the amount of traffic handled. For the short distance toll service, two
important methods of handling tne business are described, namely, manual
straightforward tandem and dial tandem.
The long distance service, which has developed most rapidly in recent
years, is described in some detail in the paper. Among the important
features of this service is noted the recently developed method of com-
pleting toll calls with sufficient speed so that on most of the calls the calling
subscriber remains at the telephone. The various types of toll circuits
are described including open wire circuits operated both at voice frequencies
and by carrier systems and long toll cable circuits. The operation of
these long circuits requires a large number of repeaters in tandem and
the design and maintenance problems which this arrangement requires
are pointed out in the paper.
1 Presented by Dr. F. B. Jewett before the World Engineering Congress, Tokio
Japan, October, 1929.
1
1
2 BELL SYSTEM TECHNICAL JOURNAL
Information is given with respect to international telephone connections
in North America, between North America and Europe and other inter-
national connections. In covering this subject some of the important
items relating to the operation of the transatlantic radio channels are
given and reference made to the projected transatlantic telephone cable.
Various forms of special services closely allied with the message tele-
phone service are described. These include telegraph service, telephone
circuits provided for private use, foreign exchange service, telephone net-
works for program transmission to radio broadcasting stations, electrical
transmission of pictures, telephony in connection with aircraft operation,
ship to shore telephony, telephony to mobile stations such as railroad
trains, telephone services of railroads and other public utilities, telephone
public address systems and television. Reference is also made to some
of the by-products of the telephone development work which include im-
provements in submarine cable telegraphy brought about by the discovery
of the alloys known as "permalloy and oerminvar," the development work
in the reproduction of sound and in tne talking motion pictures.
In concluding, the paper points out that careful studies of the future
development of the telephone industry indicate a somewhat accelerated
rate of development of the services required to meet the demands of the
customers and a continuing very rapid technical development of telephone
plant and systems to prov-ide the necessary facilities.
In treating such a large subject in a paper of this kind it has been neces-
sary to deal with technical problems in rather general terms and as an
attachment to the paper references are made to numerous articles in the
technical press for the more technical information.
General
THE purpose of this paper is to give a general description of the
telephone communication system of the United States of Amer-
ica, outlining briefly some of the more important engineering problems
involved and indicating the service results obtained. At the begin-
ning of this paper it seems important to give a brief description of
the general structure and organization of the telephone communica-
tion system.
The commercial telephone system of the United States is entirely
owned and operated by corporations, partnerships, and individuals.
A group of 24 closely associated Bell Telephone Operating Companies
owns and operates 14.8 million telephones and the telephone lines
used for toll service within their territories. In addition there are
in the country about 4.7 million telephones owned by several thousand
independent telephone companies which have operating agreements
with the Bell Companies providing for the interconnection of lines,
thus permitting the operation of 19.5 million telephones as a single
system. There are in addition about 140,000 telephones in the coun-
try not connected with the Bell System.
The 24 Bell Operating Companies cover the entire area of the
United States and are responsible for all Bell Telephone operations
within their respective areas. A number of the larger companies are
subdivided into autonomous operating units, there being at the
TELEPHONE SYSTEM OF THE UNITED STATES
present time a total of 34 such units in the country. In many cases
the area within corporate limits or within the limits of an operating
unit is identical with that of a major political subdivision of the
United States (a State) and this simplifies the application of govern-
mental regulation. A typical organization of a Bell Operating Com-
pany is indicated in Fig. 1.
'Accounting . . General Auditor
'Building and Equip-
ment Engineer
Plant Extension Engi-
neer
Financial. . . .Treasurer f Chief Engineer J Outside Plant Engineer
Operation . . .Vice President
Board 1
of ^ President^
DirectorsJ
Personnel . . .Assistant to
President
■<
General Commer-
cial Manager
General Traffic
Manager
General Plant
Manager
Transmission Engineer
Costs and Inventory
Engineer
General Supervisor of
Methods and Results
General Employment
Supervisor
ij General Supervisors —
1 1 Other Functions
Staff Engineers
Division Superintend-
ents
Legal General Counsel
.^Secretary
Fig. 1 — Organization of typical Bell Telephone Operating Company.
In order to facilitate the best possible handling of the long distance
service between points in different operating companies and to avoid
the problems which would arise from divided responsibility, the long
distance business involving territories of two or more Associated
Companies is handled throughout the country by the Long Lines
Department of the American Telephone and Telegraph Company.
These operations are, of course, in the closest cooperation with the
operations of the Associated Companies without duplication of con-
struction or of operating effort.
An important feature of the Bell Telephone System is the general
4 BELL SYSTEM TECHNICAL JOURNAL
departments maintained by the American Telephone and Telegraph
Company, including the Bell Telephone Laboratories. These depart-
ments, constituting about 7,500 engineers, scientists, business experts
and assistants, are continuously engaged in studies to develop the
art and to provide methods and facilities for improving the service.
They also provide consulting advice to the operating companies on
all phases of the telephone business and render to them a large variety
of services. One of these services is making available to all the
companies rights under all patents necessary for the fullest and most
economical development of the business. It is the intention, in
general, that work which can best be done once for all the entire
telephone system rather than individually by the several operating
companies shall be done by these general departments and that the
specific solution of the telephone problems in each area shall be the
responsibility of the operating company involved, who, however, are
free at all times to get the advice and assistance of the general staff.
In all of the work of the general staff close contact is maintained
with the various operating telephone companies of the Bell System.
The experiences of these companies are studied and analyzed to make
available for all the companies the valuable results to be derived in
this way, and the advantages to be obtained by comparing the ex-
periences of different companies under similar conditions.
The organization of the American Telephone and Telegraph Com-
pany and the Bell Telephone Laboratories is indicated in Fig. 2.
Another very important feature of the Bell System is the very
close relation between operating and manufacturing branches of the
work through the ownership by the American Telephone and Tele-
graph Company of the Western Electric Company, Inc., and arrange-
ments between that company and the operating companies for the
supply of telephone apparatus and materials. This permits the
manufacture of apparatus and the purchase of materials from outside
suppliers to be done on the basis of the large quantities required for
the entire Bell System resulting in great economies.
The organization of the Bell Telephone System is such as to result
in close cooperation between the companies dealing with different
branches of telephone work. This brings about the conditions nec-
essary for universal service, for the development of the art along
orderly and non-conflicting lines, and for the standardization of all
apparatus, communication systems and operating methods to the
extent that such standardization is helpful.
New types of telephone plant, operating methods, methods of
maintenance and business methods are standardized by the general
TELEPHONE SYSTEM OF THE UNITED STATES 5
departments of the American Telephone and Telegraph Company,
and are adopted and placed in use by all of the Associated Operating
Companies to the extent that they apply to their local conditions.
Special arrangements are, of course, made available to meet special
requirements. The specifications for all standardized apparatus and
rPersonnel Vice President
Board
of
Director?
Operation
and Engi-
neering. .
.Vice President
and Cliief
Engineer
Assistant Vice President
Benefit and Medical
Work
J Assistant Vice President
Employee Relations
Assistant Vice President
Special Employment
and Training
Commercial Engineer
.■Assistant Vice President
Plant Operation Engi-
neer
General Operating
Results
Traffic Engineer
Plant Engineer
Legal .
f Attorney
.\'ice President J Ta.x Attorney
and General 1 Patent Attorney
Counsel |^ General Solicitor
Transmission Develop-
ment Engineer
Outside Plant Develop-
ment Engineer
Research Engineer
'Vice President
President*!
President
Development
and
Research
Assistant Vice-, Equipment Development
Engineer
Electrical Interference
Engineer
Technical Representative
in Europe
Bell Telephone
Laboratories
Chairman
of Board
President
Vice Presi-
dent
Director of Research
Director of Appara-
tus Development
Director of Systems
Development
General Patent At-
torney
Director of Publica-
tion
Personnel Director
Accounts,
Finance and
Business Vice President
'Comptroller
Treasurer
.\ssistant \'ice President
General Policies and
Contracts
Assistant Vice President
General Service Bureau
Information.
Secretary
r.^dvertising Manager
. VicePresident-i .\ssistant Vice President
L Publicity
Fig. 2 — Organization of the general departments of the American Telephone and
Telegraph Company, including Bell Telephone Laboratories.
materials are prepared by the Bell Telephone Laboratories, and the
Western Electric Company is enabled to concentrate on the task of
purchasing and manufacturing standardized supplies, materials and
apparatus in accordance with these standard specifications. Stand-
ardization also has great operating advantages in minimizing stocks
6 BELL SYSTEM TECHNICAL JOURNAL
of materials and providing interchangeability both of materials and
working forces and is very important in making possible the operation
of the entire interconnected network as a single system with uniform
grades of service.
As has been stated by Mr. Gifford, President of the American
Telephone and Telegraph Company, "The ideal and aim today of
the American Telephone and Telegraph Company and its Associated
Companies is a telephone service for the nation, free, so far as humanly
possible, from imperfections, errors, or delays, and enabling at all
times any one anywhere to pick up a telephone and talk to any one
else anywhere else, clearly, quickly and at a reasonable cost." With
this aim in view, continuous effort is made further to improve and
to extend the service within the nation and also the telephonic con-
nections to other nations. It is recognized also that changes in
business and social conditions bring about repeated changes in the
services desired by the people of the nation and in the character and
appearance of facilities furnished to them. These facts, in addition
to the onward march of the application of science, form an important
basis for the continued study by the general staff of the development
of all phases of the telephone system.
A few figures relative to the size and growth of the Bell System
are helpful in an understanding of the more specific telephone prob-
lems which are discussed below. Such figures are included in the
statistical summary appended to this paper and include data regarding
telephone messages, numbers of telephones, miles of wire and amount
of telephone plant.
In accordance with the general organization of the Bell System,
the engineering problems involved in the design, construction and
maintenance of the plant of each operating telephone company are
the responsibility of the engineering department of that company.
General studies of methods of improvement of service and the devel-
opment of new apparatus and systems of communication, together
with consulting engineering advice, are provided by the general de-
partments.
For the provision of new plant to meet additional demands for
service, in the case of the more important items, often one year, and
sometimes more, is required between the completion of detailed engi-
neering plans and completion of construction. Furthermore, to ob-
tain maximum economy it is necessary that much of the new construc-
tion provide for expected increases in demands for service for a number
of years to come. This applies particularly to telephone buildings
and to runs of underground conduit and to a lesser extent to cables,
TELEPHONE SYSTEM OF THE UNITED STATES 7
pole lines and many other very important parts of the telephone
plant. The engineering of the additions to the Bell Telephone System,
now aggregating over 500 million dollars a year is, therefore, neces-
sarily based on careful forecasts of the amount and type of business
to be expected for a number of years in the future and good engineer-
ing judgment must be applied in determining the types, quantities
and design of plant. These must take into account not only the
expected amount of service required but also expected future changes
in the character and standards of service demanded and in the appa-
ratus and materials expected to become available. In view of the
capital expended in extensions and the large amount of plant already
in service, the engineering work involved is considerable. There are
now approximately 10,000 engineers engaged in the work of the Bell
System of which approximately 6,300 are in the operating companies,
2,200 in the headquarters departments and 1,500 in the Western
Electric Company. These figures apply to men doing work of engi-
neering grade, and inclusion of assistants of all kinds, stenographical,
clerical, laboratory, etc., would more than double these figures.
Local Service
General
Service within the limits of a single telephone exchange is spoken
of as local service. This generally includes service within a large met-
ropolitan area, a city with its surrounding suburbs or a town or village.
During 1928 customers of the Bell System originated approximately
24,000 million local calls of which approximately 19,000 million
originated from manual and 5,000 million from dial telephones. This
represents an average daily usage of approximately 5.5 calls per
telephone station per day.
The speed of service is illustrated by the following average figures.
In the smaller cities with manual operation where the operator who
takes the call completes it herself without trunking, the average time
from the start of the call to the answer of the called station is 19
seconds. The corresponding figure for manual calls in large cities
based on about three million observations made in the year 1928 in
38 large cities of the country is 28.8 seconds. The same observations
indicate that when fully converted to the dial system the speed of
service in the large cities will be about 22.5 seconds.
As to the accuracy of service, 98 per cent of all calls are handled
without error. The most serious errors are those resulting in wrong
numbers. The mistakes made by the subscribers and equipment
8 BELL SYSTEM TECHNICAL JOURNAL
under the dial system are about the same in number as those made
by subscribers and operators under the manual system.
Calls resulting in busy reports amount to 10 per cent. This is
something which is not directly under the control of the telephone
company since the subscriber determines the telephone facilities which
are provided. Records are kept, however, in both manual and dial
offices of the lines responsible for the greatest number of busy reports
and efforts are made to have the subscribers take additional facilities.
Standards of transmission are applied to the design of the plant
to insure that transmission will be clear between the most remote
parts of the exchange area. This depends on the design of station
equipment, wire lines and switchboard equipment, and is expressed
in terms of the combined electric and acoustic efficiencies of the cir-
cuits from the mouth of the talker to the ear of the listener. This
overall efficiency is expressed in terms of the adjustment of a standard
reference circuit. The standards in use in the United States refer
to the maximum transmission loss permitted between any two sub-
scribers and vary in magnitude between equivalents of 18 decibels
and 22 decibels, depending on the circumstances of different cases.
In order to meet these transmission standards the Bell Companies
have standard requirements regarding the efficiency of transmitters
and receivers and other station equipment, and these are made the
basis for engineering the wire plant. Transmission losses in switch-
boards are kept as low as practicable and within specified limits.
The wire plant for subscribers lines and trunks is designed to be within
the limits required for meeting the transmission standards. If under
special conditions it appears desirable to exceed these limits, this is
done only with the approval of responsible engineering authorities.
To handle calls at the local switchboards there was in the Bell
System in 1928 an average operating force of about 122,000 young
women. In addition an average force of approximately 36,000 were
employed at the toll boards of the Bell System. This made a total
operating force of 158,000. In order to make up for losses and for
growth, 75,000 women were employed, and to select this number
approximately 300,000 applicants were interviewed.
One of the important administrative problems is the scheduling
of the operating forces so that an adequate number may be available
in each central office throughout every period of the day. A method
has been worked out whereby all types of operating work are equated
to a common unit of measurements and the number of such units
that an operator should handle to give the best service most efficiently
has been determined. Frequent counts are maintained of the num-
TELEPHONE SYSTEM OF THE UNITED STATES 9
ber of calls handled throughout each hour of the day and in this way
the forces are adjusted to the work to be done.
In order that the demand for telephone service may be met promptly
as it develops and further that plant additions may be along sound
and economic lines, calls for careful planning. To this end the funda-
mental plans prepared for the different exchange areas forecast the
telephone development from 15 to 25 years in the future. Such
fundamental plans show the proposed central ofifice locations, the
boundaries of the districts to be served by each office, and the plan
of the underground conduit system. They are based on analysis of
the existing market for telephone service; the forecasted market at
a future date, considering both growth and distribution of population;
expected changes in wage levels; estimates of the amount of service
that will be sold under probable future rate conditions; and other
factors.
Station Apparatus
One of the most important parts of the telephone plant is the appa-
ratus installed on the subscribers' premises known as the station
apparatus. Of this equipment the telephone transmitter and tele-
phone receiver are fundamentally important elements and continued
research work has been carried out to improve the efficiency, clarity
of reproduction and reliability of these instruments. As a result of
improvements in transmitters, receivers and induction coils the over-
all efficiency, for example, of the station apparatus has since 1912
increased by a factor of 6.5. At the present time commercial trans-
mitters when fully energized by direct current, are capable of deliver-
ing electrical energy in the form of voice currents 200 times as great
as the acoustic energy of the voice of the speaker by which the trans-
mitter is actuated. For the most important part of the frequency
range used in speech this ratio of output power to speech power is
considerably greater. That is to say, the transmitter acts as a high
ratio amplifier.
In the Bell System the type of station equipment most generally
in use is the desk stand. As the result of extensive development work
it has been possible to produce a hand set which has transmission
characteristics equal to those of the desk stand equipped with the
best instruments heretofore in use. The hand set development in-
volved the solution of difficult problems, the principal of which were
to prevent singing or distortion of quality on account of the rigid
connection between receiver and transmitter and to make the trans-
10
BELL SYSTEM TECHNICAL JOURNAL
mitter efficient through the wide range of positions in which it is
placed by the user.
The latest form of this instrument is shown in Fig. 3. In addition
to the usual black finish, this telephone as well as the bell box and
other station apparatus have recently been made available in five
colors, statuary bronze, old brass, oxidized silver, ivory and French
gray.
Practically all the service for business purposes is provided by
individual lines or by private branch exchanges as discussed later.
CSi^
Fig. 3 — The latest form of hand set.
For residences, howeyer, there is in the United States a large develop-
ment of two-party and four-party lines. The two-party stations are
provided with selective ringing so that each station is signaled only
for its own telephone messages, and the four-party stations are pro-
vided in some places with selective ringing and in others with semi-
selective ringing.
Party lines have furnished a satisfactory means of providing service
to small users and have been an important factor in the development
of new fields of service in residences.
To care for situations where something more than a single line
TELEPHONE SYSTEM OF THE UNITED STATES
11
with one or two telephones is needed, but where an inter-communi-
cating system or private branch exchange is not justified, so-called
wiring plans are used which provide various arrangements for associ-
ating the station equipment with the telephone lines. For the most
part the customers' needs are satisfactorily met by one of the ten
standard arrangements in general use. A specific example is that of
Fig. 4 — -Telephone booth provided for public telephone stations.
two central office lines with two main and two extension stations.
Calls to or from either telephone line may be made from any one of
the four telephones. Answering at a main station provides privacy
by cutting off all the other telephones.
There are in the United States a considerable number of extension
stations. At the present time there are in the Bell System over 1.3
12 BELL SYSTEM TECHNICAL JOURNAL
million of such stations. This number is rapidly increasing particu-
larly for residence use as people appreciate further the advantages
of having telephones in a number of convenient locations. The best
residences are more and more being equipped to have telephones
available in all parts of the house.
In order to make telephone service possible for those people whose
sense of hearing is more or less deficient, special sets are installed.
By means of a vacuum tube amplifier which the user can adjust,
the receiving may be amplified so as to bring the range up to the
point giving best results, this point depending on the degree of im-
pairment of his hearing.
Public telephone stations constitute an important part of telephone
development in the United States, there being at present more than
275,000 of such stations in service. Whereas residence and business
service is largely given by contract, the customers contracting to
pay a definite amount per month or a certain amount per call, a great
many of the public pay stations are supplied with coin boxes by means
of which the money is collected at the time the call is made.
These installations are also for the most part in booths to insure
quiet and privacy. Fig. 4 shows a form of booth furnished by the
telephone companies, provided with a seat and with lighting and
ventilation as well as a convenient location for the necessary telephone
directories.
Private Branch Exchange
For customers who use a large amount of telephone service, one
of several types of private branch exchange is provided which not
only permits distribution of the incoming calls to the particular
station desired but also makes it possible for one extension to call
another without going through the central office. In the Bell System
Private Branch Exchange Stations 2,740,000
Per Cent of Total Bell Stations 19.0
Private Branch Exchange Boards
Cordless '. 53,300
Cord 60,900
Total 114,200
Private Branch Exchange Cord Positions
Manual 68,600
Dial 1,700
Total * 70,300
Private Branch Exchange Attendants
Cord Board Attendants 75,000
Cordless Board Attendants 53,000
Total 128.000
Fig. 5^Private Branch Exchange Statistics for the Bell
System as of Feb. 1, 1929.
TELEPHONE SYSTEM OF THE UNITED STATES
13
there are 36 million telephone connections handled each day by about
128,000 private branch attendants. About 17 per cent of all local
calls originate at these boards. Equipment of both the manual and
dial type is installed, the latter being particularly adapted to extension-
to-extension calling. Further data regarding private branch exchanges
are given in Fig. 5.
The smallest manual installation is the cordless board illustrated
in Fig. 6 where connections are established by means of keys, and the
capacity is limited to three trunks and seven stations. A larger type,
Fig. 6 — -Cordless private branch exchange installation for 7 stations and 3 central
office trunks.
using cords for the completion of connections, is illustrated in Fig. 7
with a capacity for fifteen trunks and 200 stations.
For the largest private branch exchanges the equipment is of much
the same type as that used at central offices. A large switchboard
for one of the public utilities having 1,600 stations is shown in Fig. 8.
Connecting this private branch exchange with the central office there
are 148 lines and in addition 151 tie trunk lines extend to other private
branch exchanges having a business association. There are a total
of 42 switchboard positions.
14
BELL SYSTEM TECHNICAL JOURNAL
•m - ■"^^Mi
%
i ■
f 1
••'ti '■-•■••':,L^ III III! II nil
k
Fig. 7 — Private branch exchange switchboard arranged for 15 trunks and 200 stations.
Fig. 8 — Private branch exchange— 1600 stations, 148 lines to Central Office, 151
tie trunks to other private branch exchanges, 42 switchboard positions, 60 operators
and 24 hour service.
TELEPHONE SYSTEM OF THE UNITED STATES 15
At the end of 1928 in the Bell System there were about 650 dial
P.B.X. installations with about 100,000 lines. The smaller sizes of
this equipment are designed to meet the needs of the larger residences
and the larger sizes are adequate for any business office and large
industrial plant. Typical equipment arrangements are shown in
Fig. 9.
In general the private branch attendants are in the employ of the
subscribers having this type of equipment. It is essential that the
attendants be recruited and trained with the same care as central
office operators, and to this end, the telephone companies maintain
employment bureaus and training courses for the benefit of private
branch exchange attendants. Subscribers are encouraged to send
their attendants to these training courses for retraining wherever this
appears advisable.
Instructors highly trained in local and toll central office operation
and in the best methods for handling private branch exchange work
constantly visit private branch exchanges in order to be of assistance
both to the subscribers and to the attendants in giving the best pos-
sible service.
Cable Plant
While open wires are occasionally used in limited quantity at out-
lying points, 96 per cent of the exchange area wire plant is in cable.
Of this 74 per cent is underground.
An outstanding development is the steady increase in the number
of pairs of conductors which it is possible to place in a single lead
sheath of 6.7 cm. outside diameter. From the early use of 30 to 60
pair 19 gauge conductors there has been a continual increase in the
number of pairs and a decrease in the size of the wires until at the
present time 1,800 pairs of 26 gauge conductors are placed under a
single sheath for use in the denser areas. The significance of this
development is indicated in Fig. 10 which shows the year in which
each important step was taken and the relative cost per pair of con-
ductors resulting from each step in the development.
In urban development main cables, called feeder cables, usually
of the maximum size, radiate from each central office through under-
ground conduit to the various parts of the area served. These feeder
cables in general are run full size for considerable distances rather
than being diminished at branch cable points, and the flexibility thus
obtained is found advantageous for conditions in this country. Each
main feeder cable has smaller branch cables bridged to it at intervals.
The main feeder cable may continue all the way through the area
16
BELL SYSTEM TECHNICAL JOURNAL
Fig. 9 — -Typical small dial private branch exchange installation.
TELEPHONE SYSTEM OF THE UNITED STATES
17
or it may divide into two or more smaller cables branching out either
underground or aerially in the area.
A type of distributing plant located along the rear properties of a
residential block is illustrated in Fig. 11. This represents the usual
O
o
I <
a. 00
< —
a: CO
-I
u <
<
0.
cc.
UJ
a.
10
I-
tn
O
u
UJ
>
16
15'
14
13
12
n
10
1 1
1.0 MM.i
50
PAIF
100
PAIR
1
132 PAIF
?
.152 PAIf
R
IE
P
AIRI
di.
Jb P
MRt
303
PAli
b.
.c
1 Mh
A.
46
5 P
\\R-
404
fir
PAI
6 Pi
MR <
6
4 j^f^
90
9 PAIR <
—
rsTTKT
.4
MM
00
CO
OD
00
o
o
o
01
o
tn
en
10
o
(VI
01
00
(«
o>
YEAR
Fig. 10.
Fig. 11 — -Aerial cable distribution along rear property line. Poles used jointly for
electric lighting power and telephone distribution.
18
BELL SYSTEM TECHNICAL JOURNAL
type of construction followed in such areas. With continuous build-
ing construction the distributing cables are often attached to the rear
walls as in Fig. 12 or extended through the basements of the buildings.
Underground cables are carried in conduit consisting of various
combinations of multiple tile duct. A typical duct run, shown in
Fig. 12 — Distribution telephone cable attached to rear wall of building showing
terminal boxes and entrance by twisted pair into cellar.
Fig. 13, illustrates the materials and methods of construction generally
employed.
At the central office the conduit system is designed to meet the
ultimate requirements of the building and terminates in a cable vault
as shown in Fig. 14. With this entrance arrangement the main
TELEPHONE SYSTEM OF THE UNITED STATES
19
cables are spliced in the cable vault to smaller units of silk and cotton
insulated conductors which extend up through the building in slots
or ducts to the main distributing frame.
Switching Systems
The outstanding development in switching systems for a telephone
communication has, of course, been the rapid trend toward an increase
Fig. 13 — -72 duct underground cable run under construction.
in the extent to which the operations are performed by automatic
machinery. The general characteristics of the different types of
switching systems in use in the United States and their extent of
use are briefly discussed below.
Magneto
Magneto switching arrangements are used in small places and
scattered rural areas. They vary in size from an arrangement to
interconnect two or three lines up to a switchboard handling 300 to
400 lines. The average size of the magneto switchboards in the Bell
20
BELL SYSTEM TECHNICAL JOURNAL
System is 170 lines. At the end of 1928 there were about 3,500 mag-
neto offices with approximately 5,500 operators' positions, serving
1.1 million telephones.
Common Battery
At the end of 1928 there were 2,036 common battery offices, the
maximum size office serving 10,500 lines and the average being 3,700
Fig. 14 — Cable vault in Central Office, St. Louis, Missouri, showing entrance of
cables through ducts and connecting to silk and cotton insulated cables extending
up through the building.
lines. There was a total of 46,000 switchboard positions where the
subscribers' lines are answered. In addition, there were 13,000 so-
called trunk positions which are required where it is necessary to
trunk calls from one office to another in areas having more than one
central office.
The trend in development in manual switchboard has been toward
performing more of the necessary switching and signaling operations
automatically by means of somewhat more complicated circuit and
equipment arrangements and less and less by the operator. These
changes have resulted in less manual operating labor and in an im-
TELEPHONE SYSTEM OF THE UNITED STATES 21
provement in the service. Some of the more important of these
changes are as follows :
1. Automatic ringing which continues automatically at regular inter-
vals until the subscriber answers.
2. Ringing tone, very much reduced in volume, to the calling sub-
scriber automatically advising him when ringing is in progress.
,1. The audible busy signal, a tone placed on the calling line when
the called line is busy.
An important recent change in manual central ofhce equipment
relates to the trunking methods employed in completing a connection
from one central office to another. In most of the larger cities the
so-called "straightforward" method is used. With this operating
plan the number that is desired in the distant office is passed by the
originating or "A" operator to the completing or "B" operator over
the trunk that is used for completing the trunk connection. This is
in contrast to the "call circuit method" where all orders between
operators are passed over a separate wire known as a call circuit.
The principal equipment changes at the "B" positions have to do
with the different circuit plans for connecting the trunk operator's
telephone set to the trunk. This is done either by means of a key,
by means of plugging the trunk into a listening jack or automatically
by means of suitable relays. At the "A" end the principal change is
the arrangement for testing whether or not an outgoing trunk is
busy. This is done either by means of a lamp indicating a free trunk
or by a lamp or tone indicating that a part of a trunk group is free.
Dial Equipment
Dial equipment of two types known as the step-by-step system
and the panel system respectively are used in about equal amounts
in the Bell System. The change from manual to dial operation pre-
sented a very large problem from an engineering, a manufacturing,
an installing and an economic standpoint. At first the dial installa-
tions were to care largely for growth but they have been followed by
installations for the replacement of existing manual equipment where,
all factors considered, this was clearly justified. In this way an or-
derly program has been developed. Figure 15 indicates the total
number of stations on a dial and on a manual basis for each year
since 1921 and the expected program up to 1933. Under the present
contemplated dial program it is estimated that the areas employing
step-by-step equipment will be on a complete dial basis by 1937 and
that all areas employing panel equipment will be substantially com-
pleted by 1942.
BELL SYSTEM TECHNICAL JOURNAL
1921-1928 ACTUAL
1929-1933 ESTIMATED
<0
z
o
H
«0
ti.
o
ft
2
O
_J
2
20
1 8
1 6
14
1 2
1
8
.^-'
.^'^
^''
-*- rs.V-
^'"'
^-^^
^
0^
f)^"^
. —
NUAL
^^^
— — .
. STATIONS
MA
p
^^'
,.-''
**
0^
'■^
^
'^-^
-*
1921 '22 '23 '24 '25 '26 '27 '28 '29 '30 '31 '32 1933
END OF YEAR
2.7 3,6 5.6 8.7 12.5 15.5 18.7 22 27 33 38 42 47
PER CENT DIAL STATIONcs OF TOTAL
Fig. 15.
Fig. 15 — Relation between manual, dial and total stations — -Bell owned stations.
Step-hy-Step Dial System
The step-by-step system is used in the Bell System in single office
areas and in the smaller and medium sized multi-office areas where
the number of central offices is limited and consequently the trunking
problems are not complicated. Step-by-step equipment is in service
in 194 offices to which are connected 1.6 million stations.
The fundamental unit of the step-by-step dial system is the selector
illustrated in Fig. 16. This switch has a capacity for 100 terminals
placed ten on a level and ten levels high, thus making possible the
selection of any one of a hundred lines. By placing several selectors
in series a network of central offices may be built up, each office
serving 10,000 telephones.
The selectors are mounted on iron frameworks and the terminals
are cabled to cross-connecting frames so that any grouping can be
made as may be demanded by the number of calls which the par-
ticular selector handles. A typical arrangement of selectors is shown
in Fig. 17.
TELEPHONE SYSTEM OF THE UNITED STATES
23
Fig. 16 — -Typical step-by-step selector showing relays which control the circuit opera-
tion mounted at the top, and the selector banks of 100 terminals at the bottom.
Fig. 17 — Installation of step-by-step dial equipment showing selectors mounted on
iron framework.
24
BELL SYSTEM TECHNICAL JOURNAL
Panel Dial System
In the largest cities and the smaller municipalities around them
making up the large metropolitan centers, a much greater degree of
complexity is encountered in switching the calls due to the large
number of offices of varying types to which calls are destined and due
Fig. 18— Typical panel selector frame. Capacity 30 selectors in front and 30
selectors in rear. Motor driving mechanism is at the bottom of frame and controllinc
apparatus at either side.
TELEPHONE SYSTEM OF THE UNITED STATES
25
Fig. 19 — Panel dial office sender frame showing the apparatus for five senders.
26
BELL SYSTEM TECHNICAL JOURNAL
to the routings involved in order to trunk economically either large
or small volumes of trafific. The panel system was developed to
meet these requirements and is now installed in a number of such
metropolitan centers, notably, New York, Chicago, Philadelphia,
Boston, Detroit, Cleveland, St. Louis, Pittsburgh, Baltimore, San
Francisco, Buffalo, Kansas City, Seattle, Providence and Omaha.
Panel equipment is in service in 128 offices to which are connected
1.6 million stations.
Fig. 20 — ■Installation of panel dial equipment. The unused floor space is provided
for future growth.
In the panel system the fundamental switching unit is a large switch
consisting of five banks of 100 terminals each. The selectors, by
which contact is made with any one of the 500 terminals, move ver-
tically on both sides of the terminal banks. A typical panel frame
having capacity for 60 selectors is illustrated in Fig. 18.
In the panel system the selectors do not follow in synchronism with
the impulses of the dial as in the step-by-step system. Rather, a
group of apparatus known as the " sender " records the impulses and
in turn directs the operation of the several selectors in the train until
the called terminal is reached. By this means the trunking arrange-
ments and the numbering scheme can be designed independently of
each other. This, combined with the large capacity of the panel
TELEPHONE SYSTEM OF THE UNITED STATES 27
selectors, makes possible economies in interofifice trunks and a reduc-
tion in the number of selectors involved in completing a connection
in a large exchange. The selection may be either to a dial office or
to a manual office reached direct or through a tandem switchboard.
Figure 19 illustrates the apparatus for five senders. A switchroom
in a typical panel office is shown in Fig. 20.
Buildings
The buildings in the Bell System at present number about 6,000,
excluding those occupied by the Western Electric Company. All of
Fig. 21 — Telephone office in the residential district of Silver Spring, Maryland. De-
signed for small manual switchboard with a present capacity of 2,200 lines.
the larger and many of the smaller are owned by the telephone com-
panies. The range in size of the buildings is illustrated by Fig. 21
showing a small building for a single manual switchboard with a present
capacity of 2,200 lines, and Fig. 22 showing a headquarters office
and equipment building in a large city. This building has 66,000
square meters of floor space, in the lower 9 floors has space for dial
equipment to serve 100,000 telephones, and in the upper floors in-
cludes offices for 5,000 people. Figures 23 and 24 further illustrate
28
BELL SYSTEM TECHNICAL JOURNAL
U
13111
1 imiiunw
lliiliilliii!
liiiiHiiit^'
Fig. 22— Combined equipment and office building, New Yori< City, containing
headquarters of the New York Telephone Company, 31 stories, 66,000 square meters
of floor space. Lower 9 floors arranged for toll tandem equipment and for dial
equipment with an ultimate capacity of 100,000 telephones. Upper 22 floors
arranged for offices with a capacity of about 5,000 people.
TELEPHONE SYSTEM OF THE UNITED STATES
29
some of the recent combined equipment and office buildings for large
cities.
The objective in connection with buildings of all types, including
equipment and office buildings, garages and warehouses, is to provide
Fig. 23 — Combined equipment and office building at Detroit, Michigan containing
headquarters of the Michigan Bell Telephone Company. 19 stories, 28,000 square
meters of floor space. 13 floors arranged for equipment including toll board and dial
equipment to serve 60,000 telephones. 6 floors are arranged for offices.
30
BELL SYSTEM TECHNICAL JOURNAL
Fig. 24 — Combined equipment and office building at Cleveland, Ohio containing
headquarters of the Ohio Bell Telephone Company. 22 stories, 25,000 square meters
of floor space. 13 floors designed for equipment including the toll board and ultimate
dial equipment for 100,000 telephones. 9 floors arranged for office space.
TELEPHONE SYSTEM OF THE UNITED STATES
31
1
n
■H
V
5S
A"
=r 4=
ft
^^
*
1 ■
r
-■
1 r
-r
1
1
■■, 1
It
DC _ .
a
-- -
o
i
+--t
-t-r
I I
l-T
L-l
h
, II
;i II
n
V rr
-I— *-
rt
itn
^ ^
^
U u
n
tin
i! II
ii i|
II II
uu
A
T +
-(- +
^ ^
|l
n
n^
n
II
'11
^1
nn n fi
' I'
1 11
ti
ri
1 1
II
rtn
T-r
WOOy 5NllVU3dO
SOUV08K9ilMS
E Z
.J
n
o
Ii
bo
(U
l->
3
I : ***
C
d
c
P
bD
C
cS
u
&
en
Ii
o
«41
^3
0)
c
*s
-a
o
\B
■o
"rt
a
(L)
■M
en
I
I
a
<u
*j
<n
d
o
-U
c
rt
O
o
qil
•O
ii
<A
T3
C
oJ
^
--A-t---
4-t— .
.0 - i8?
bij
32 BELL SYSTEM TECHNICAL JOURNAL
buildings which adequately, economically and comfortably house the
equipment and personnel — both initially and throughout the useful
life of the building — and which at the same time are outstanding and
attractive, appropriate to their surroundings and a continuing source
of satisfaction to the communities in which they are erected.
The initial size of a building is determined by the costs and rate
of increase in space requirements, with due regard to the service
reactions caused by extensions of the building which is being used for
operating purposes. As a result of these considerations central ofifice
buildings are usually designed with a capacity of about twice the
initial requirements. In many cases space provided for later exten-
sions of equipment is used temporarily for office space. Possibilities
of future extensions of the buildings are provided for either by buying
more land than is required for the initial building, thus making pos-
sible lateral extensions, or by providing strength in the steel frame-
work for future vertical extensions. The possibility and type of
future extensions must, of course, also be taken account of in the
architectural design.
Floor plans have been developed representing for typical conditions
the best arrangements of the various types of central office equipment
both manual and dial. The use of these uniform floor plans greatly
facilitates the engineering, manufacture and installation of the equip-
ment, and results in savings of both time and money. A uniform
floor plan applying to a step-by-step dial office is illustrated in Fig. 25.
The relative location of the different frames and aisle space as well
as the unit size of the frames is fixed but the number of frames is
varied to meet the requirements of individual cases.
Except for the smallest buildings, non-combustible construction is
used, a steel or reinforced concrete frame and brick or stone curtain
walls being employed. The very small buildings, except where severe
fire exposures are encountered, are of frame or brick and joist con-
struction.
Rural Service
Surrounding the larger cities and towns there is in general a sparsely
settled district developed on a multi-party basis. Service is given
on common battery lines where the distances are not too great and
either semi-selective or code ringing is used. These lines usually
serve not more than six or occasionally eight parties and the common
battery signaling requirements limit the range to about se\'en or eight
kilometers.
TELEPHONE SYSTEM OF THE UNITED STATES 33
One of the most difficult service problems of the Bell System is
that of providing service to farming districts where the distances
between successive farms as developed in the United States is often
great. At the present time service in such farming areas is usually
provided by multi-party lines with magneto signaling. These rural
lines carry from six to as many as fifteen parties and may be as much
as sixty-five kilometers in length.
In the past the demand for service from rural districts of this nature
has generally been limited almost wholly to local service between the
rural customers and shorter haul toll service, and the present develop-
ment is a response to that point of view.
The extent of development of rural service is illustrated by a census
made in the State of Iowa in 1920, showing that of 213,000 farmers
86 per cent have telephone service. This percentage is doubtless
materially higher at the present time. There were in 1928, in the
Bell System, 6,000 offices which served lines that might be classified
as rural. Of these rural lines about 12,000 were on a common battery
basis and about 43,000 on a magneto basis. These figures do not
include the rural lines which were served by connecting companies,
the addition of which would increase the above figures many times.
The development of this type of service has to a very considerable
extent been in the hands of local groups of small local companies
because of the nature of the service.
With the present rapid development in the use of a nation-wide
toll service, there are a rapidly increasing number of stations where
rural customers will accept a higher grade of service designed for
general connection to telephones throughout the Bell System. The
development of improved rural service of this type is an important
feature of the present telephone program in the United States.
Toll Service
Service between telephones which are not in the same local exchange
area is called "Toll Service." With the exception of less than one
per cent of the telephones which are not connected to the Bell S^^stem,
toll service is offered in the United States between any two telephones
in the country and to a very large extent the toll plant is adequate
to provide good service between any two of these telephones.
An outstanding feature of the last few years has been the rapid
growth of the toll service. This is indicated in the appended sta-
tistical summary which shows that during the last five years the com-
pleted toll messages have increased by 67 per cent. During this
same period the number of telephones in service have increased by
3
34
BELL SYSTEM TECHNICAL JOURNAL
about 28 per cent. These figures show that in spite of the continued
increase in the number of telephones in service, the number of toll
messages per telephone have increased by about 30 per cent for this
period.
One important cause of the rapid increase in toll usage has been
the material improvements in toll service.
Figure 26 shows the increase in the speed of toll service since 1920
expressed in terms of the average time required from the placing of
a call to the response of the called party, or until the operator gives
UJ
H
3
Z
I
O
u
UJ
0.
o
<
a.
u
>
<
Fig. 26 — -The average time required from tliL' placing of a toll call to the response of
the called party, or until a definite report is made by the operator.
a definite report regarding the call. The service is sufficiently fast
so that on 95 per cent of the calls, the subscriber stays at the tele-
phone. This makes possible still more rapid service and simplified
operation.
There have also been very great improvements during this period
in the clearness of speech transmission. The maximum permissible
transmission loss between two subscribers on a toll connection has
been materially reduced. The toll plant and subscriber plant are
now so designed that most of the messages are handled with a maxi-
mum transmission equivalent for the longest subscriber lines of 20
to 25 decibels overall referred to the standard transmission reference
system.
TELEPHONE SYSTEM OF THE UNITED STATES
35
Short Distance Toll Service
General
To a large and increasing extent the toll messages are completed
and supervised by the local exchange operators who first answer the
subscriber's call, providing in this way toll service with the same
methods which are applied to local service and with comparable
..iT L A N T 1 C
OCEAN
159 OFFICES IN SUBURBAN TOLL AREA
SCALE :
Fig. 27 — New York suburban toll area indicating the number of offices in Metro-
politan and suburban districts. Population in the area 10.5 millions. Telephones in
the area 2.5 millions.
speed. This method of operation is used for most of the toll business
up to a distance of 50 kilometers and to a considerable extent up to
100 kilometers. The use of this method includes the extensive sub-
urban areas around the large cities. Calls handled by this method
now amount to 650 million messages a year and its increasing use has
36 BELL SYSTEM TECHNICAL JOURNAL
been one of the important ways in which increased speed of service
has been brought about.
The handling of this suburban telephone trafific adds greatly to
the complexity of the transmission, trunking and operating problem
of the larger cities. This is illustrated by Fig. 27 showing the sub-
urban toll area surrounding New York City. It will be noted that
the city itself includes 168 central offices and in the suburban areas
in the metropolitan district there are in addition 159 central offices.
In many cases of the shorter haul toll service, the volume of traffic
between two offices is sufficient to warrant direct trunks and the calls
are completed over these direct trunks by the usual local traffic meth-
ods. In order to provide an efficient trunking arrangement for the
smaller volume of traffic between widely separated offices, however,
tandem trunking arrangements are provided, by which the calls are
routed through a central switching point and from that point dis-
tributed to the terminating offices. Either manual or dial central
office equipment is used as outlined below, each type having its field
of application depending upon the amount of traffic and the portion
of traffic to and from manual and dial central offices. The trunks
to the central switching point, or tandem office, are in general of
somewhat larger gauge than interoffice trunks because of the greater
distances involved and the correspondingly more severe transmission
requirements. In some of the longer trunks telephone repeaters are
used. It has not been found generally economical to use repeaters
at the tandem switching point although this is done in certain in-
stances and it is possible that in the near future the more general
use of repeaters in this way may become an economical means of
meeting the transmission requirements.
Manual Straightforward Tandem
The manual straightforward tandem is used in those medium sized
areas in which most of the suburban calls are between manual switch-
boards. The arrangement of the equipment is shown in Fig. 28.
The tandem trunks from the originating office terminate on the plugs
located at the rear of the keyboard and the tandem completing
trunks to the various terminating offices appear in jacks in the face
of the switchboard. The completing trunks are provided with lamp
signals indicating idle trunks. When a call comes in on a tandem
trunk the operator is advised by a flashing lamp signal on that trunk
and her telephone set is automatically connected to it. The work of
the tandem operator is limited to making the connection at the tandem
board and making the disconnection when advised by lamp signal
TELEPHONE SYSTEM OF THE UNITED STATES
37
that the conversation is over. Her work is thus greatly simpUfied
and the operation of the board is extremely rapid.
Fig. 28 — Manual straightforward tandem board for handling suburban toll calls.
Trunks from originating offices terminate on plugs at the rear of the keyboard.
Tandem completing trunks appear in jacks in the face of the switchboard. Incoming
calls indicated by flashing of lamp associated with trunk cord.
Dial Tandem
Dial tandem systems have been designed for handling suburban
toll trafific in areas where a large proportion of the central offices are
of the dial system and also to facilitate handling the complex sub-
urban traffic around the large metropolitan areas, such as New York,
Boston, Chicago and Philadelphia. The type of dial equipment, in
general, corresponds with the type used for the local traffic in the
same area.
For use in connection with calls from dial offices, the dial tandem
usually requires no operators at the tandem office. The originating
operator in the dial office who handles the short haul traffic controls
the selection of the trunk to the terminating office.
In some of the large metropolitan centers, dial tandem apparatus
of the panel type is employed also for handling calls from manual
38
BELL SYSTEM TECHNICAL JOURNAL
Fig. 29 — -Panel type dial tandem office in New York City.
Fig. 30 — -Panel type dial tandem office. Arrangement of keyboard at tandem
positions. Incoming trunks appear as lamps and associated keys on upper sloping
part of keyshelf. Calls are completed by setting up the called office and called
number on keys on lower part of keyshelf.
TELEPHONE SYSTEM OF THE UNITED STATES
39
offices. With this arrangement operators are, of course, required at
the tandem board. The tandem operator receives a request from
the originating operator for the office and number called and by means
of keys establishes the connection by dial switching apparatus to the
called subscriber in case he is in a dial office or transmits the required
information to the terminating office operator in the case of a manual
office. Figure 29 shows an installation of panel tandem operators'
equipment and ¥\g. 30 shows one section of this equipment in greater
detail.
Fig. 31 — ^Step-by-step dial tandem board. Calls completed to the terminating
dial subscriber station by means of the 10 button key set shown on the keyboard.
Incoming calls automatically distributed to an idle operator.
A modified form of step-by-step tandem equipment using operators
has been installed in step-by-step areas for handling calls from manual
offices to dial offices in cases in which it was not advisable to equip
all the subscriber operators' positions with dials. This equipment
includes, in addition to the selectors, a simplified type of switchboard
as shown in Fig. 31.
Long Distance Service
General
For the longer hauls the subscriber is connected to a toll board
operator who completes and supervises the toll message. This method
is called the toll board method of operation, and is used for most all
40 BELL SYSTEM TECHNICAL JOURNAL
the messages over about 100 kilometers. For the purpose of this
paper this service will be referred to as long distance service. The
messages handled by this method total about 300 million messages a
year. The amount of long distance business at New York City, for
example, requires the use of 1,275 operators' switchboard positions.
During the past three years an important change has been generally
applied in the methods of handling long distance service. Formerly
the toll operator first receiving the call recorded the necessary infor-
mation on a ticket and forwarded this ticket to another operator
provided with facilities for completing calls to the particular part of
the country involved in each case. An increase in speed has been
brought about by providing the operators with arrangements both
for recording calls and for completing calls to all points so as to avoid,
in a large proportion of the cases, the necessity for transmitting the
information to a second operator.
By means of this change in method and other improvements, the
average speed of service for all long distance messages has been de-
creased from 6.9 minutes in 1925 to 2.6 minutes in 1928. Also in
1928, 90.7 per cent of the calls made by the customers resulted in
completed messages.
In placing a long distance call, the telephone subscriber in the
United States may give simply the telephone number and city desired.
This has some advantages in speed of service. At present, 50 per
cent of the long distance messages are handled in this way and this
per cent is increasing. About 15 per cent of the messages are handled
in this same way, the called telephone number, however, being sup-
plied by the operator. In addition, the telephone system offers, for
a somewhat greater charge, what is called a " particular person " serv-
ice. This means that the subscriber may, if he wishes, ask to talk
with a specified person at a distant point, giving such information
as he can regarding how that person may be located. The telephone
operator then undertakes to complete this message by locating the
desired party, following him up to points other than that designated,
if necessary, and if the calling subscriber wishes. The percentage
of messages handled on this basis increases with the length of haul.
When a subscriber wishes he may transmit to the telephone com-
pany in advance information regarding a number of calls which he
wishes to have completed in sequence, beginning immediately or at
a specified time. These sequence calls, as they are termed, are used
particularly in connection with selling by long distance telephone.
At the present time at the New York long distance office, for example.
TELEPHONE SYSTEM OF THE UNITED STATES
41
about five per cent of the business is in the form of sequence calls,
some of the sequence lists including as many as 1,000 calls.
Except during times of emergency conditions, it is not the practice
800
600
400
200
CO
-t
lO
U3
r^
00
CVI
(\1
(\l
fVJ
OJ
CVJ
o
0>
Ok
o^
a>
a*
Fig. 32 — Toll board messages per year in thousands, New York-Boston.
in the United States to limit the length of conversations on toll con-
nections. As a result it is very general for conversations, particularly
on longer hauls, to exceed the initial three-minute period, the average
length for transcontinental conversations being, for example, six
42
BELL SYSTEM TECHNICAL JOURNAL
minutes. Conversations which run one half hour or an hour are not
unusual, and in one case a transcontinental telephone conversation
was eight hours in all.
A striking feature of the long distance service is the more rapid
3 50
300
250
fo
■<t
!P
<o
h-
00
OJ
OJ
CM
OJ
M
C\J
O)
o>
o»
a>
O)
O)
Fig. 2>i — -Toll board messages per year in thousands, New York-Chicago.
growth of very long haul business than business of moderate length.
Figures Zl, 33) and 34, for example, show respectively the growth in
messages for the last five years between New York and Boston, 370
kilometers. New York and Chicago, 1,380 kilometers, and the trans-
continental business between New York and Chicago at one end and
TELEPHONE SYSTEM OF THE UNITED STATES
43
Los Angeles and San Francisco at the other end, an average of 4,700
kilometers. It will be noted that while the toll business as a whole
has increased as noted above, 67 per cent in this period, the New
York-Boston business has increased 62 per cent, the New York-
60
50
40
30
30
10
n
Tt
ir>
U7
r^
00
(VI
rvj
f\i
(M
(Vl
C\l
o>
0^
a>
a>
<r>
a\
Fig. 34 — Toll board messages per year in thousands, New York and Chicago to
Los Angeles and San Francisco.
Chicago business 194 per cent and the transcontinental business 430
per cent.
In the attached statistical summary is given the basis used for
determining long distance toll rates, including the practices in the
44
BELL SYSTEM TECHNICAL JOURNAL
United States in the offering of reduced rates in evening and night
hours. The effect of these reduced rates is in some cases temporarily
to slow down service at the hours when the reduced rates first go into
effect, because of the large demand for long distant business at those
hours.
Jp-f-^'^f''^' o
TELEPHONE SYSTEM OF THE UNITED STATES
45
Telephone Toll Lines
To handle the toll business of the United States has required com-
pleting a network of toll telephone lines completely covering the
country. This network is shown in a general way in Fig. 35. It
tc
u
t-
u
O
-J
O
a.
o
z
o
18
16
14
IS
\0
/
>
y
^
CABLE >/^
r
/
^
y
X
^
— —
OPEN WIRE,
CARRIER
1928 1929 1930 1931 1932 1933
Fig. 36 — Estimated toll circuit kilometers in plant Bell Operating Companies.
consists at the present time of about 14 million kilometers of wire on
about 300,000 kilometers of toll route. The toll circuits are partly
open wire, supported on insulators and are partly in cable. Both the
open wire and cable circuits are, in general, phantomed, giving three
independent circuits for each two pairs of wires, and in addition on
the open wire is superposed a considerable amount of carrier current
telephone circuits. The distribution at the present time between
46
BELL SYSTEM TECHNICAL JOURNAL
cable, open wire and carrier and the expected increase of each during
the next four years is shown in Fig. 36.
Open Wire and Carrier Circuits
The standard construction for open wire telephone circuits in the
United States is indicated by the diagram of Fig. 37 and a typical
pole line built in accordance with this construction is shown in Fig. 38.
3-C 3-C BH 3-C 3-C
PH r. -.., PH
&V& &V& srioi e^e""d"va
l-C
PH
l-C
&v & & V &
T 13 T j Jp T
l-C
PH
l-C
3J
6 Cr ti" vO
l-C l-C
PH
& v""8""'""&"^v a iff^&j f\ v"g"""""ff"v fl
Fig. 37 — Pole Line Configuration, phantomed construction,
12 inch spacing between wires of non-pole pairs.
Symbol
Facility
Total Circuits
V
Voice frequency-physical
20
PH
Voice frequency-phantom
10
3-C
Carrier system furnishing 3 telephone circuits
12
l-C
Carrier system furnishing 1 telephone circuit
12
T
D-C telegraph
40
BH
Carrier telegraph (10 channel)
40
Total telephone
54
Total telegraph
80
The wires are of copper and the sizes and weights are shown in the
following table.
Diameter — mm.
2.6
3.2
4.2
Weights — kg. per km.
47
74
118
Bronze and aluminum are not, in general, used in the United States
for telephone lines, being not as economical or as generally satis-
TELEPHONE SYSTEM OF THE UNITED STATES
47
factory as copper, taking into account transmission efficiency and
construction conditions.
The wires are placed on 10 pin cross-arms and supported, in gen-
eral, by double-petticoated glass insulators. The grouping of wires
to form phantoms is indicated in Fig. 37. This arrangement has been
found desirable for conditions in the United States and transposition
systems have been designed by which are obtained satisfactory opera-
tion of the phantoms and side circuits, the mutual induction between
the various circuits being sufficiently neutralized to prevent mutual
interference.
Fig. 38 — -Open wire pole line construction. Four 10-pin cross arms.
On the longer circuits telephone repeaters are installed at an aver-
age distance of about 175 to 300 kilometers, providing in that way for
adequate transmission efficiency.
The number of circuits which it is practicable to provide by means
of open wire lines has during the past decade been very greatly in-
creased by the extensive use of carrier telephone for superposing on
the telephone circuits additional channels of communication carried
by currents above the voice range of frequencies. These systems
now form a network covering the entire country and in some areas
a large proportion of the circuit growth on open wire lines is taken
care of by carrier systems. The systems range in length from a
minimum of 75 kilometers to a maximum of 3,800 kilometers.
Two types of carrier telephone systems are standard for use in
48
BELL SYSTEM TECHNICAL JOURNAL
the United States. One of these, designed for the longer hauls, pro-
vides on one pair of wires three telephone circuits in addition to the
voice frequency circuits. These three carrier circuits are provided
by the modulation of frequencies between about 6,000 and 28,000
CHANNEL
I
NUMBER
CHANNEL
I
NUMBER
FIRST SYSTEM
2 EAST I EAST 3 EAST
SECOND SYSTEM
^
a EAST 1 EAST
3 EAST
8
la
16
ao
24
28
32
FREQUENCY- KILOCYCLES PER
SECOND
Fig. 39 — -Frequency allocation of two long haul carrier systems. The blocks
indicate the range of the transmitted side band. The arrows are located at the carrier
frequencies and indicate the direction of transmission.
UJ
>■ UJ
O -I
m
o —
o o
o^
5e^
O z
a: u
1^ _]
z$
9.3
< u
>
UJ
Q
M
12
10
2 -
-2
400 800 laOO 1600 2000
CYCLES PER SECOND
2400
2800
3200
Fig. 40 — Average overall transmission-frequency characteristic of Long Haul Carrier
telephone system.
cycles, different frequencies being used for transmission in the two
directions. The different conversations are amplified together in a
common repeater at intermediate points and at the terminals sep-
arated by electrical filters providing for each circuit a band of approxi-
mately 3,000 cycles. The frequency allocation for two varieties of
TELEPHONE SYSTEM OF THE UNITED STATES 49
the long haul system in common use are shown in Fig. 39 and typical
transmission characteristics for a carrier channel are shown in Fig. 40.
The long haul carrier systems give very satisfactory service and
form a part of some of the longest circuits in the country. For ex-
ample, the direct circuits between New York and Los Angeles, Cali-
fornia, 5,100 kilometers in length, are made up of cable circuits from
New York to Pittsburgh connected permanently to a Pittsburgh-
St. Louis carrier system, and a St. Louis-Los Angeles carrier system.
These two carrier systems connected together total 4,550 kilometers
in length with 13 intermediate repeaters. Similarly the New York-
San Francisco circuits are in cable from New York to Chicago and
there permanently connected to the Chicago-Sacramento carrier sys-
tem 3,800 kilometers long with 10 intermediate repeaters.
The short haul carrier system is similar in its general character-
istics but is simplified and provides a single carrier circuit for each
pair of wires. In the case of both systems, single side-band carrier
suppression circuits are used.
In Fig. 37 showing the standard arrangement of open wires on pole
lines are indicated the carrier telephone channels and also the carrier
telegraph channels which can be superposed on these circuits without
mutual interference, after the installation of suitable transpositions
which have been designed to neutralize the mutual induction between
the circuits. It is noted that with this arrangement it is possible to
obtain from 40 wires 54 telephone circuits. Also 80 telegraph circuits
are obtained, used for special contract service as described later in
this paper.
On a number of the open wire toll routes carrying very long circuits,
it has become important to provide arrangements for using a larger
number of long haul carrier telephone systems, thus obtaining a
larger number of circuits. Whereas a number of arrangements using
the standard spacing of wires have been tried out, it is found ex-
tremely difficult to continue the use of phantoms and to so transpose
the wires as to provide adequate freedom of interference between
the higher frequency carrier channels if these are used on all pairs.
The difficulty of doing this is evident in considering that in order to
avoid overhearing it is necessary that the power transfer between
different circuits should not exceed one part in a million even though
they are parallel to each other for long distances.
In order to make possible the maximum use of long haul carrier
systems where desired, trials have been made with the arrangement
of wires shown in Fig. 41 and these trials have shown very satisfactory
results. With this arrangement the spacing of the two wires of each
50
BELL SYSTEM TECHNICAL JOURNAL
pair except the pole pairs is reduced to 23 centimeters and the phan-
toms on these wires are abandoned. Type "C" systems can then be
used on all of the pairs with this spacing. The result as indicated on
the diagram is that a 40 wire toll line provides 70 telephone and 80
telegraph circuits.
3-C
T T
3-C
3-C
~ ~ T T<^ T 88 T y^ T
3-C
8v8
3-C
T T
T T
3-C
T T
3-C
3-C
T T
3-C
3-C
T T
T T
.fivs i om& i ava
"^ ^^, i "^^ ° ."^1 ,<^ T
3-C
ML
T T
vv^
Fig. 41 — Pole Line Configuration, non-phantomed construction,
8 inch spacing between wires of non-pole pairs.
Symbol
V
PH
3-C
T
BH
Facility
Total Circuits
Voice frequency-physical
Voice frequency-phantom
Carrier telephone
D-C telegraph
Carrier telegraph (10 channels)
Total telephone
Total telegraph
20
2
48
40
40
70
80
Toll Cables
In spite of the great extension in the use of open wire circuits
brought about through the application of carrier systems, as indicated
above, it would be extremely difficult with the present rapid growth
in toll business to provide by open wire toll lines the large numbers
of telephone toll circuits now required on many routes. It is ver}-
fortunate that the development of means for providing satisfactory
long distance circuits through telephone cables has matured in time
to enable this method of construction to be widely used to meet the
TELEPHONE SYSTEM OF THE UNITED STATES 51
present demands. Also, the toll cables provide practical immunity
from the effects of storms, including the sleet storms, which are a
hazard to open wire construction in nearly all parts of the United
States.
The first long distance toll cables in the United States were placed
in service in 1906 between New York and Philadelphia and between
Chicago and Milwaukee. These cables were both placed underground
in multiple duct and are each about 150 kilometers long. The next
step in the extension of toll cables was the completion in 1914 of an
underground toll cable route between Boston, New York, Philadelphia
and Washington, a distance of 730 kilometers. Cable running west
from New York was completed to Chicago, a distance of 1,380 kilo-
meters, in 1925 and to St. Louis, a distance of 2,150 kilometers, in
1926. This permitted placing in service circuits entirely in cable
between New York and St. Louis.
The present major toll cable routes together with the extensions
which it is expected to complete during the next five years are indi-
cated in Fig. 42. It is to be seen that in accordance with these plans
toll cable will, within five years, extend entirely across the continent
and up and down the length of both Atlantic and Pacific Coasts, will
extend north into Canada and south almost to Mexico. In the north-
eastern portion of the country where the development is the heaviest,
there is already a multiplicity of toll cable routes and on some of
these routes the rate of growth is high enough to require additional
cables at successive intervals of one or two years. The amount of
toll cable added to the network this year will be about 8,000 kilo-
meters and this amount is expected to be increased materially in the
following years.
In the early toll cables before the extensive development of tele-
phone repeaters, it was necessary, in order to provide satisfactory
transmission, to use relatively large conductors and conductors up to
a maximum size of 2.6 mm. diameter (No. 10 B and S gauge), were
provided in the Boston-Washington cable.
With the perfection of telephone repeaters for use with toll cable
circuits, the transmission limitations on the extension of toll cable
were removed and the economy of such circuits greatly increased by
making it possible to use small conductors. The longest toll cable
circuits at the present time are carried over conductors of 0.9 mm.
diameter (19 B and S gauge). For the shorter circuits each path is
used as a two-way circuit, while the longer circuits use separate paths
for transmission in opposite directions. In order to improve the trans-
mission characteristics, the circuits are provided with loading coils
52
BELL SYSTEM TECHNICAL JOURNAL
/
V
<
O
2 ;
' \
5
^ I
' »
• \
1
5
<
5^
\
* \ z
1 »c;
A
~
>-".
<
1 —
' ^» * -; *
' 1
z
'i
z
5
1
)
1 • -K,
o
5
/
<
'
O
./
CO
ct)
a>
c
<u
3
o
ni
u
Oi
(b
^
2
Q.
E
_M
— r
c
£
a>
IJ
a>
3
o
c
c
cr
c
^
c
UJ
01
3
T3
^
xt
C
K
nj
o
UJ
O
TELEPHONE SYSTEM OF THE UNITED STATES
53
at intervals of 1,830 meters and at an average interval of about 70
kilometers are provided with telephone repeaters which renew the
power of the attenuated voice currents. A single standard full size
cable 6.7 cm. in diameter when so equipped is capable of providing
between 250 and 300 long distance telephone circuits.
The toll cable system includes various types of construction. For
the routes having the most rapid growth, multiple duct subway is
used. At the present time with the development of very heavy toll
demands in many parts of the country, this type of construction is
Fig. 43 — ^Typical aerial toll cable construction showing loading point.
being extended very rapidly on a number of important routes. Mul-
tiple tile duct with small splicing manholes located at intervals of
229 meters and large manholes for loading coil pots at intervals of
1,830 meters are generally used.
For routes on which the growth is relatively light, for example,
40 or 50 circuits a year and where underground construction is de-
sirable, two other types of construction have been used to a limited
extent. In one type the cable is placed in a single duct of fibre and
in the other type of construction cable covered with a double layer of
steel tape is placed directly in the earth. With both of these types
of construction, manholes are built only at loading points.
54
BELL SYSTEM TECHNICAL JOURNAL
In many places the character of the country is such that under-
ground construction would be very expensive. In such cases, and
in other cases where it seems desirable, aerial toll cable construction
has been used extensively in the United States. With this type of
construction the cable is suspended from a steel messenger wire sup-
ported on poles. Figure 43 shows typical aerial cable construction,
including a loading point, the pots of loading coils being supported
on an angle iron pole fixture.
Long circuits in toll cables have some extremely interesting elec-
trical characteristics. Figure 44 shows the net transmission charac-
U1
_l
m
O
u
Q
— llJ
>
UJ UJ
Q -I
o
>-
o
o
o
o
1 2
10
8
6
4
2
- ■?
\
/
\
/
\
/
\
i
/
\
^
y
\
^
^
400 800 1200 1600 2000
CYCLES PER SECOND
Fig. 44.
2400
2800
3200
teristic over a range of frequencies of a New York-Chicago toll cable
circuit 1,380 kilometers in length. It will be seen that the voice
frequencies are transmitted with nearly the same net efficiency over
a sufficiently wide band to give a high grade of telephone transmission.
The net characteristic indicated, however, is obtained by almost
wholly neutralizing with telephone repeaters the very large trans-
mission loss in the circuit. The New York-Chicago circuit, for ex-
ample, would have an attenuation loss at 1,000 cycles of about 470 db,
which means that without amplification the ratio of output power at
one end to input power at the other end of the circuit would be 10"'*^
The combined gain of the 19 telephone repeaters in the circuit is
about 461 db, giving about 9 db net equivalent, lender these condi-
tions, it is evident that a careful regulation of the circuit is essential.
For example, variations in the temperature of a circuit in the course
of a day could make as much as 30 db or 1,000 fold difference in the
TELEPHONE SYSTEM OF THE UNITED STATES 55
electric power received at the end of the circuit. To prevent such vari-
tions affecting the net equivalent the long circuits are all provided
with automatic regulators which adjust the gains of the telephone
repeaters to compensate for the effect of temperature variations on
the equivalent of the circuit.
The effects of transmission delay are also very interesting and
important. Voice waves travel considerably more slowly over cable
circuits than they do over open wire circuits. For example, the
velocity is about 30,000 kilometers per second for " longdistance"
type cable circuits as compared to nearly 300,000 kilometers per
second for non-loaded open wire circuits.
One important result of delaying the speech waves is the "echo"
effect. The transmitted currents are in part reflected at the distant
terminal due to variations in the impedance of the receiving circuit.
If the reflected currents transmitted back to the other end are delayed
enough they may be heard by the talker as echoes of his voice. They
may be again reflected at the sending end of the circuit and returned
to the listener as an echo following the directly transmitted speech.
The effects of these echoes are largely eliminated by devices known
as "echo suppressors" by means of which the transmission of voice
waves in one direction over the circuit causes interruption of the
path over which the echo currents are transmitted in the opposite
direction. However, the effectiveness of echo suppressors is limited
by the necessity that they shall not be operated by noise currents of
extraneous origin as this would interrupt conversations. The echoes,
therefore, are an important factor to be taken into account in deter-
mining the type of toll cable circuit to be provided to meet the trans-
mission limitations imposed on the long distance circuits.
In cable circuits introducing considerable transmission delay, the
fact that the delay is not exactly the same for waves of different fre-
quencies is also important, tending to give rise to what have been
sometimes referred to as "transient" effects. In loaded cable circuits
the waves of higher frequency are delayed more than those of lower
frequency because of the fact that the loading is applied in lumps.
The coils and condensers in the repeaters and auxiliary apparatus
on the other hand, tend to delay the waves of lower frequency. The
result is that the waves of intermediate frequency arrive first, fol-
lowed by the waves of higher and lower frequency. Devices known
as " phase compensators " can be used to reduce the effects, particularly
those caused by the line. To improve the situation at the low end
of the frequency scale special attention has been given to the design
of the repeaters and auxiliary apparatus.
56
BELL SYSTEM TECHNICAL JOURNAL
Still another effect of the transmission delay is to somewhat slow
up and perhaps otherwise interfere with conversations due to the
delay which is added to the ordinary time elapsing between question
and answer. F"or example, if a cable circuit is 5,000 kilometers long
Fig. 45 — Thirty 4-wire repeaters and associated testing equipment. The repeaters
are arranged in groups of 3 with a minimum of cabling, each group being associated
with a phantom circuit and its 2 side circuits.
and the voice waves travel 30,000 kilometers per second, the time
required for the waves to travel from one end of the circuit to the
other is }/^ second and to make a complete round trip, 3^3 second.
This }y^ second delay is evidently added to the ordinary time which
elapses between question and answer. In the United States cable
connections somewhat longer than 5,000 kilometers will be used in
TELEPHONE SYSTEM OF THE UNITED STATES 57
the future, while for international connections, of course, very much
longer distances than this will be involved. In the United States
considerable study is, therefore, being given to the effects of trans-
mission delay and to methods of avoiding difficulties on the very
long connections including the development of cable circuits of higher
speed.
The toll cable circuits today include two principal types, one,
discussed above, for the longer distances having a transmission speed
of about 30,000 kilometers a second, and the other for the shorter
distances, transmitting a narrower band of frequencies and having
about one-half the transmission velocity. In view of the superior
transmission characteristics of the long distance type circuits it is
the present practice in the design of new toll cable circuits in the
United States to limit the use of the short distance type facilities to
circuits about 160 kilometers in length if they are to be used for
switched business, and about 280 kilometers in length if used only
for terminal business.
Toll Circuit Equipment
The apparatus required for the operation of toll circuits has been
developed in the form of panels mounted on standard bays of angle
iron, thus bringing about a great reduction in the space required
compared with earlier forms of mounting. Figure 45 shows a bank of
30 four-wire repeaters arranged in groups of three, each group being
associated with a phantom and its two side circuits. Figure 46 shows
the panels containing complete terminal equipment for two type "C"
carrier telephone systems (six circuits) with associated testing appa-
ratus.
The equipment is housed in fire-proof buildings. Figure 47 shows
a typical telephone repeater station, this one being located at Prince-
ton on the cable route between New York and Philadelphia. This
building now contains 1,100 repeaters. Some of the telephone re-
peater stations now being built are designed for ultimate capacity
with extensions of 10,000 repeaters.
An interesting feature of the long telephone circuits is the use of
1,000-cycle current for signaling rather than the lower frequencies
which have been general in the past. This higher frequency has the
advantage of being efficiently transmitted by the telephone circuit
without change in the amplifying apparatus and hence does not re-
quire intermediate ringing apparatus. At the terminals it is rectified
and caused to operate relays which actuate the desired signal.
58
BELL SYSTEM TECHNICAL JOURNAL
Fig. 46 — -Complete terminal repeater apparatus for two long haul carrier telephone
systems (6 circuits) with associated testing equipment.
TELEPHONE SYSTEM OF THE UNITED STATES
59
Fig. 47 — Telephone repeater building at Princeton, New Jersey on New York-
Philadelphia cable route. Building now houses 1 100 repeaters. Ultimate capacity
2200 repeaters.
Fig. 48 — Earth boring machine and derrick. Will bore 60 centimeter hole 2
meters deep in loam or clay soil in about one minute and in stone or frozen soil in 5
or 10 minutes. Derrick operated by power driven winch for setting poles. Truck
provided with four wheel drive.
60
BELL SYSTEM TECHNICAL JOURNAL
Fig. 49 — Trenching maciiine. Digs trench 1.7 meters deep and 55 centimeters
wide, at speeds varying between 0.2 and 1.2 meters per minute, and is carried from
job to job upon a trailer drawn by 23^ ton truck.
i-^^Stf^^.
Fig. 50 — ^Automobile truck equipped with tracks for hauHng cable on private right
of way. With tracks, speed about 16 kilometers per hour. Can carrv 4500
kilogram reel up 40% grade. Tracks can be removed using special equipment
provided for that purpose; without tracks speed 27 kilometers per hour.
TELEPHONE SYSTEM OF THE UNITED STATES
61
Toll Line Construction
The construction of toll lines under a wide variety of conditions
has required the solution of many interesting problems. The rela-
tively high cost of labor in the United States contributes to the ex-
tensive use of labor saving machinery, a large amount of which has
been developed to meet the particular conditions of telephone con-
struction. Figures 48, 49 and 50 illustrate some of the more inter-
esting types of labor saving machinery used extensively for both open
wire and toll cable construction.
Numerous special construction problems are, of course, met in
specific situations. One of the interesting river crossings is illustrated
in Fig. 51.
m^^ »sf*:*afj v si »^;wftpi^'»il^^
■^
/tl^i liW:
'Mm^i
Fig. 51 — Special aerial cable construction across a river. Cable and messenger
secured to a catenary suspension wire. 2-spans each about 180 meters long.
Switching of Toll Circuits
As far as is economically practicable the toll business is handled
by direct circuits without intermediate switching. At the present
time this includes 80 per cent of the toll messages. Of the remaining
20 per cent, 17 per cent have one intermediate switch and 3 per cent
more than one intermediate switch.
It is the purpose of the Bell Telephone System to design the toll
62
BELL SYSTEM TECHNICAL JOURNAL
'J"Wf^.
r'
in
_aj
'S
a
g
o
u
tao
G
TELEPHONE SYSTEM OF THE UNITED STATES 63
telephone system in the United States to give satisfactory service
between any two points in the country. In order to accompHsh this
it is necessary to make arrangements for a minimum number of
switches between any two points. Also the toll circuits which will
be used as parts of the built-up connections must be designed for a
very high standard of transmission so that the overall efficiency of
the built-up connection will be satisfactory.
Arrangements have recently been worked out in the United States
for meeting requirements of switched traffic more satisfactorily than
has heretofore been possible. These arrangements may be briefly
described as follows:
At different points in the country there have been selected a group
of eight very important switching points shown in Fig. 52. These
eight regional centers will all be interconnected by high grade groups
of circuits directly, that is, without intermediate switch. Through-
out the country there are selected about 147 important switching
points known as primary outlets also indicated in Fig. 52, each of
which is directly connected to at least one of the regional centers.
Each of the remaining 2,576 toll offices in the country will be con-
nected to at least one of these important switching points. Further-
more, within limited areas, such for example as a State, all important
switching points will be directly interconnected. Within such an
area, therefore, any two toll offices can be connected together with
not more than two intermediate switches. Also, every toll office can
be connected to a regional center with not more than one switch and
through that center can reach any other toll office in any part of the
country with a minimum number of switches.
To insure adequate transmission on the switched connections, each
of the important switching points will be provided with means for
automatically inserting gain in the connection when two toll circuits
are switched together so that the overall connection may be operated
at the highest possible efficiency. This will, in general, be done by
automatic adjustment of the gain of terminal repeaters permanently
installed in circuits which must, in general, because of transmission
limitations be operated at a lower efficiency when used for terminal
business than when used as parts of a built-up connection.
Maintenance of Toll Service
With the present network of long distance lines in the United States,
it is common to have 20 or more repeaters installed on each of the
longer circuits and this number will increase greatly with the further
extension of toll cable. The maintenance of service over these long
64 BELL SYSTEM TECHNICAL JOURNAL
and complicated circuits is a very considerable problem both from the
standpoint of technique and of organization. In this paper, these
problems will not generally be discussed, but certain features will
briefly be indicated.
The service maintenance of the circuits includes periodic tests of
transmission efficiency with transmission measuring sets designed for
rapid and efficient use by the plant maintenance forces. The fre-
quency of tests varies according to the requirements of each circuit
group.
To expedite the testing and adjustment of the circuits the longer
cable circuits are subdivided into circuit units, these units usually
being in cable about 160 to 240 kilometers in length, including the
conductors and equipment involved in one section arranged for the
automatic compensation of temperature variations. When trouble
occurs on a long circuit, the circuit unit in which the trouble is located
is immediately replaced and the location of trouble within the circuit
unit then can be carried out without further interruption of service.
The responsibility for establishing and maintaining each circuit group
is given to a control office which is provided with private communica-
tion channels to all parts of the circuit.
An important feature in the maintenance of long toll circuits is
the physical relations between the telephone circuits and circuits for
the transmission or distribution of electric power. The Bell Tele-
phone System and the power companies of the United States as repre-
sented by the National Electric Light Association are very actively
cooperating in a study of the best means of so coordinating the plant
of telephone and power companies as to avoid interference under the
various types of conditions important in practice. By means of this
work it has been possible to find in every case a satisfactory solution
permitting each utility to extend and increase its service along natural
lines and providing proper protection of the telephone service.
International Connections
General
The connections between the telephone systems of the United
States and the telephone systems of other countries are indicated in
Fig. 53.
The territory of the United States has direct contact with only
two other nations, Canada on the north and Mexico on the south.
The common language and the close commercial relations between
Canada and the United States have naturally resulted in a well de-
veloped arrangement of lines connecting the telephone systems of the
TELEPHONE SYSTEM OF THE UNITED STATES
65
two countries. Telephone connection between the cities of the United
States and Mexico was not made until 1927, due to the unsettled
political conditions which obtained for some years in Mexico.
The many close commercial, political and social relations between
the peoples of Europe and America have naturally drawn the attention
of telephone men for many years to the possibility of establishing
Fig. 53.
telephone communication between these two continents. It was a
great satisfaction, therefore, to be able to inaugurate such a service
in 1927. The transatlantic telephone circuits already connect over
20,000,000 telephone stations in North America to over 7,000,000
telephone stations in Europe, thus joining together over 85 per cent
of the total telephone stations of the world.
In somewhat more detail, the present status of the connections
of the United States telephone system to the telephone systems of
other countries is as follows:
5
66 BELL SYSTEM TECHNICAL JOURNAL
Connections in North A merica
Practically all the telephone stations in Canada have communica-
tion to the telephone stations in the United States. There are ap-
proximately 100 long distance circuits extending from cities in the
United States to important Canadian centers, including Halifax,
St. Johns, Montreal, Toronto, Hamilton, Winnipeg, Regina, Calgary
and Vancouver. The remaining cities are reached either directly
or by switching through the important centers. In addition to long
distance circuits there are, of course, many short distance circuits
connecting points on opposite sides of the boundary which have local
relations with each other. The various companies and provinces in
Canada cooperate very closely with the Bell Companies in the United
States in the maintenance of international service and, in general,
telephone practices are very similar or identical in the two countries.
Telephone communication is extended from the United States to
Mexico by means of a telephone line crossing the border near Laredo,
Texas. Direct long distance circuits extend from points in the United
States to Mexico City, Tampico and Monterey and through these
centers to about one-half the telephone stations in Mexico. Local
toll circuits cross the border at a number of places.
Telephone communication was established between the United
States and Cuba in 1921 by the placing of three telephone cables
between Key West and Havana. Each of these cables furnishes one
telephone circuit and a maximum of four telegraph circuits. The
requirements for the cables were exacting since a length of about
190 kilometers is combined with a depth of water having a maximum
of 1,860 meters. Each cable consists of a central conductor mag-
netically loaded with a wrapping of fine iron wire and insulated with
gutta percha compound. A metallic return path for the telephone
currents is furnished by heavy copper tape wrapped outside of the
insulation and, therefore, in contact with the surrounding water.
Three of the telegraph circuits in each cable are obtained by using
" carrier currents " at frequencies slightly above the voice range. The
fourth is obtained by using frequencies below the voice range.
Connections to Europe
In 1915 the Bell System experiments on radio reached the point
where telephone messages were transmitted by radio from the United
States and were successfully received by engineers sent for the purpose
to Paris and to the Hawaiian Islands. While the Great War delayed
technical and commercial development, in 1923 the Bell Companies
were able to carry out a successful demonstration of radiotelephone
TELEPHONE SYSTEM OF THE UNITED STATES 67
transmission from a group of telephone officials in New York to a
group of people interested in communication assembled for the pur-
pose in London. The success of these experiments led to cooperation
with the British Post Office and the establishment in 1927 of telephone
service between New York and London. This service has now been
extended to include the greater part of the telephones of North America
and Europe.
As indicated in Fig. 53 there now exist one long-wave and one short-
wave telephone circuit between the two continents. A second short-
wave circuit will be placed in service about June 1 of this year and
a third in December. By the end of 1933 it is expected that there
will be in service between New York and London a group of six cir-
cuits consisting of three short-wave radio circuits, two long-wave
radio circuits, and one cable circuit. Our best information indicates
that the short-wave circuits will be suitable for service at least 60
per cent of the time, the long-wave, 90 per cent, and the cable, 100
per cent.
Since the beginning of 1929, the average number of messages handled
per week has been 275. For this period the average number of mes-
sages per day, omitting Saturday and Sunday, has been 44. Eighty-
nine messages were handled on Christmas Day, 1928.
Certain technical features of these circuits are particularly inter-
esting. The long-wave circuit operating at a frequency of approxi-
mately 60,000 cycles employs the "single side-band carrier suppres-
sion " method. This appears to be the only use of this method in
radio, although it is widely used in " carrier " circuits over telephone
wires. The energy saved by the suppression of the carrier and the
increased selectivity permitted by the narrow band of frequencies
which is transmitted gives this system a transmission efifectiveness
as great as a system of three or more times as much power using the
ordinary transmission method. At both ends the receiving stations
are situated as far north as can conveniently be reached and use is
made of highly directive receiving. It is estimated that at the United
States end these two factors represent an improvement equivalent to
an increase in power of five thousand times as compared to a non-
directive receiving station located at the same latitude as the trans-
mitting station.
The short-wave transmitting and receiving stations located not far
distant from New York and London employ highly directive antenna
systems. The design of such antennas must take into account eco-
nomic factors and possible reactions on receiving effects other than
power efficiency such as fading. The improvements effected by such
68 BELL SYSTEM TECHNICAL JOURNAL
systems depend on wave-length and transmission conditions. Under
favorable conditions the improvement effected at each end is approxi-
mately equivalent to a transmitted power increase of 100 times. The
most useful wave-lengths for this service have proved to be in the
vicinity of 16 meters, although wave-lengths of about 22 and H meters
are also provided to increase the amount of time these circuits are
satisfactory for service because at certain seasons and times of day
they are more effective than the 16 meters wave-length.
Service over the transatlantic facilities is carried on from 6.30 in
the morning to 10.00 at night in New York, corresponding to 11.30
in the morning and 3.00 A. M. London. During the winter months
the long waves give nearly continuous service over this period. Under
summer conditions considerable difficulty is frequently experienced
in maintaining the long waves during the afternoon period in New
York, corresponding to the evening period in London. At these
times, however, the short waves are usually effective.
The projected transatlantic telephone cable will use new magnetic
loading materials and new insulating compounds for submarine cables
recently developed by the Bell Telephone Laboratories. It will have
at least one intermediate repeater point at Newfoundland. A circuit
of this kind, differing radically from radio circuit in its characteristics
will add both to the message capacity and to the reliability of the
transatlantic service.
Connections to South America
Figure 53 indicates a short-wave radiotelephone circuit from New
York to South America which, it is expected, will be in service early
in 1930. The South American transmitting and receiving stations,
which will be in the vicinity of Buenos Aires, will be owned and oper-
ated by the companies who operate the local telephone service in
Buenos Aires and the wire lines extending to other points in South
America.
Special Services
Telegraph Circuits
While the Bell System handles practically no commercial telegraph
message business, it plays an important part in meeting the communi-
cation needs of the United States by furnishing a large mileage of
telegraph circuits for the private use of individuals and institutions,
and for the use of governmental departments. Over two million
kilometers of such circuits are now in use. One-third of this amount
is used by newspapers and press associations. The greater part of
TELEPHONE SYSTEM OF THE UNITED STATES
69
70
BELL SYSTEM TECHNICAL JOURNAL
TELEPHONE SYSTEM OF THE UNITED STATES 71
the remainder is used by commercial, financial and other organiza-
tions. Between New York and Chicago, a distance of approximately
1,400 kilometers, there are slightly over 300 such circuits now in
operation.
Figure 54 shows the system of telegraph circuits furnished by the
Bell Companies to one of the press associations. An indication of
the importance of private communication systems to commercial and
financial institutions is given in Fig. 55 which shows the telegraph
circuits furnished by the Bell Companies to a single brokerage com-
pany.
The greater part of such telegraph circuits have in the past been
operated by hand-speed Morse telegraph. At the present time, how-
ever, nearly a third of the mileage is operated with telegraph printers
and this method of operation is rapidly increasing. Two speeds of
service employing printers are offered, one operating at 40 words
per minute and the other operating at 60 words per minute. At the
present time, in view of the use to which this service is put, no demand
has arisen for multiplex operation, but this method of operation is
possible and will be used if it should become desirable.
The telegraph circuits were originally all obtained as a by-product
of the telephone business by compositing or otherwise superposing
them on the telephone wires, using direct current for the telegraph
circuits. At the present time approximately two-thirds are obtained
in this way. The remaining third are obtained by " carrier current "
methods. The carrier current system of open-wire lines uses fre-
quencies above the voice range and provides ten duplex telegraph
circuits on each pair of wires. The carrier current system used on
cable circuits employs frequencies within the voice range, the currents
being transmitted over an ordinary telephone four-wire cable circuit.
This system gives twelve duplex telegraph circuits on each such
circuit.
Telephone Circuits Provided for Private Use
In addition to the usual telephone message business, the Bell
Companies furnish telephone circuits for the private use of individuals
and organizations.
So-called " special contract " telephone circuits are set up between
particular parties for their private use at definite times specified in
the contract. Approximately 2,000,000 circuit km. hours of such
facilities are now in use during each complete business day. This is
the sum of the figures obtained by multiplying the length of each such
special circuit by the number of hours per day it is continued in use.
72 BELL SYSTEM TECHNICAL JOURNAL
About three quarters of this total is accounted for by circuits where
the contract calls for 12 hours operation per day, nearly all the re-
mainder is accounted for by circuits which remain in vService 24 hours
per day. A remaining small fraction is made up of shorter period
contracts which are permitted to be as short as 30 minutes per night
one night per week, or 10 minutes per day five days a week.
As an illustration of the extent of use of this service, there are at
present 158 full-time special contract circuits between New York and
Philadelphia and 89 of such circuits between New York and Boston.
Foreign Exchange Service
Closely related to the above is the furnishing of what is called
foreign exchange service. This consists of an arrangement whereby
a customer in one exchange area is provided with a circuit for his
exclusive use to another exchange area, this circuit being associated
with a telephone number in a distant exchange so that other telephone
stations in that exchange can be connected to the special line without
toll charge. By this means, a business office in Boston, for example,
can be given a New York telephone number, all New York calls for
that number being treated as local calls but being actually completed
over the special line to Boston.
This type of service has a considerable popularity, there being over
1,000 such lines in service at the present time. Most of them are
for relatively short distances, but some are for material distances,
the longest being between Cleveland and New York, a distance of
about 900 kilometers.
Telephone Networks for Program Transmission to Radio Broadcasting
Stations
Radio broadcasting has resulted in the development of networks
of telephone circuits for transmitting programs from studios or other
places at which they are picked up to the radio station or system of
stations from which they are broadcast. By such telephone wire
systems the ceremonies of the Presidential Inauguration on March 4,
1929, were simultaneously transmitted to 118 radio stations located
all over the United States. A statement regarding these interesting
telephone networks, the requirements which they must meet and their
importance in program broadcasting in the United States is given
in a separate paper presented to this Congress (see paper on Wire
Systems for National Broadcasting by A. B. Clark).
TELEPHONE SYSTEM OF THE UNITED STATES
73
74 BELL SYSTEM TECHNICAL JOURNAL
Electrical Transmission of Pictures
A commercial service for the electrical transmission of pictures
between the cities of New York, Chicago and San Francisco was
inaugurated in April, 1925. The eight cities now connected to this
service and the routes of the lines used in connecting them are shown
in Fig. 56. In addition, a portable transmitter is provided which
may be moved to any desired point. At present this is located in
the city of Washington, D. C.
The pictures as transmitted are of about twelve centimeters by
seventeen centimeters. Any size picture, of course, can be photo-
graphed to come within these dimensions. The detail of each picture
corresponds to 39.4 lines per centimeter in each direction, that is,
each picture is composed effectively of about v300,000 independent
elementary areas. The line time of transmission with the present
commercial system is about 7 minutes.
Pictures may be sent from any of the cities shown to one or to
more of the other cities which are reached by this service. Newspapers
use this service for the transmission of pictures of events of national
importance or where matters arise in any part of the country of large
news interest. For example, pictures of the inauguration of President
Hoover were sent in this way to the newspapers in San Francisco.
In view of the three hours difference in time between Washington and
San Francisco the pictures were published in newspapers sold at a
time of day earlier than that at which the event took place.
The majority of the pictures transmitted are for business or social
purposes including pictures of legal documents, advertising material
to be simultaneously released at a number of separate points, pictures
showing new styles in ladies' wearing apparel, personal greetings in
the handwriting of the sender and finger-prints of criminals.
The Western Union and Postal Telegraph Companies now have a
service in which they will accept telegraph messages for " facsimile "
transmission over this picture system between those cities which the
system reaches. This service has not yet been offered long enough
to show how much it will be used.
Telephony in Connection with Aircraft Operation
Telephony promises to play a very important part in the practice
of commercial aviation. The Bell Telephone Laboratories are carry-
ing out a large amount of development work on all phases of telephony
for this purpose. One-way receiving sets have been developed per-
mitting an airplane pilot to receive weather reports and to determine
the direction of radio beacons. Experimental radio sets suitable for
TELEPHONE SYSTEM OF THE UNITED STATES 75
two-way conversations between a moving plane and the wire telephone
system have been developed and demonstrated.
Safety of airplane travel depends a great deal on the rapid accu-
mulation and dissemination of meteorological data. An experiment
on a promising method of handling such data is being carried out on
an airplane route between San Francisco and Los Angeles in the State
of California. At each terminal landing field and at two intermediate
fields meteorologists are located. At six periods during the day each
of these is rapidly connected in succession by telephone to outlying
weather observation points varying in number at the different points
from three to sixteen. The information thus accumulated and co-
ordinated at each of the four landing fields is rapidly transmitted to
the other three fields by means of printer telegraph circuits connect-
ing them. This constant rapid observation of weather conditions
along the airplane route and over a considerable territory around it
permits very accurate prediction of the weather conditions which
any plane will meet in its travel over the route. Such weather pre-
dictions may be communicated to the airplanes before starting or by
radio during their flight.
Printer telegraph circuits appear to be a particularly convenient
means of interchanging information among important landing fields
along airplane routes.
Ship-to-Shore Telephony
The Bell System development work on ship-to-shore telephony was
originally started with wave-lengths in the neighborhood of 400 meters,
which were later taken into the broadcasting range. In 1920 shore
transmitting and receiving stations in northern New Jersey were
equipped to operate simultaneously three separate telephone channels
in this range. Through these radio stations any telephone subscriber
could be connected experimentally to the steamships " Gloucester"
and " Ontario " which were engaged in coastwise shipping from Boston
southward. In October, 1920, a talk to one of these ships furnished
an interesting part of a demonstration at a banquet in New York
City tendered to the delegates to the " Preliminary International
Communication Conference" which was meeting in Washington at
that time.
Development of ship-to-shore telephony has been delayed because
of uncertainties regarding the commercial situation and wave-length
assignments. At the present time the work is again being actively
pushed using wave-lengths under 100 meters. A transmitting and a
receiving station will shortly be in course of construction near the
76 BELL SYSTEM TECHNICAL JOURNAL
seacoast of northern New Jersey and a radio-telephone set is about to
be installed on the steamship "Leviathan" to operate with these shore
stations. As this ship approaches or leaves New York it is expected
to be possible to talk from it to any telephone in the Bell System.
This is intended not only as a demonstration of the technical features
of such a service but to afford an indication of the extent to which
such a service will be used under commercial conditions.
Radiotelephony is being used from shore stations to coastal boats
in a number of cases in the United States, but not connected to the
commercial telephone system. These include particularly certain
boats of the U. S. Coast Guard Service. A careful study, including
tests, has been made of telephone service to tugboats operating in
New York harbor for the purpose of controlling and thus making
more efficient the operation of such craft. So far, it is not clear that
this service will be commercially justified.
Telephony to Other Mobile Stations
Consideration has been given to telephone connections for types
of mobile stations other than ships and airplanes. Communication
with moving trains can technically be carried out with facilities now
available. Active studies are under way to determine the practica-
bility of providing such service at a cost which would be attractive
commercially and with apparatus which can be limited to a reason-
able space on the train.
Telephone Services of Railroads and Other Public Utilities
The operation of railway systems requires a large amount of com-
munication service. The dispatching of trains was, until recent years,
carried out largely by the use of telegraph. This has been rapidly
changed until at the present time on over 60 per cent of the total
railway mileage the train dispatching is by telephone. The railroads'
telephone service to stations in the Bell System is through P.B.X.'s
leased to them by the telephone companies. In addition to this,
the railroads frequently own private telephone circuits extending
along their rights of way which connect to and are switched through
these same P.B.X.'s.
Similar arrangements are provided for meeting the special com-
munication needs of electric power companies, oil pipe-line companies,
and other utilities.
Telephone Public Address Systems
Experience in many cases has shown that with the public address
system used by the Bell Companies it is possible to amplify speech
TELEPHONE SYSTEM OF THE UNITED STATES 77
or other sounds so that they can be heard by an audience of prac-
tically unlimited size. Such public address systems as they are called
are used very extensively in large auditoriums and at large public
gatherings. For example, the ceremonies of inauguration of President
Hoover held on the steps of the Capitol in Washington were amplified
by the public address system so as to be heard by a gathering esti-
mated at a hundred thousand persons, gathered within a radius of
about 300 meters.
Furthermore, by using the public address system with suitable
long distance telephone circuits, it is possible to convey the proceed-
ings of such occasions simultaneously to audiences in all parts of the
country. The local distribution of such proceedings is, however, now
done largely by radio broadcast rather than by use of the public
address system.
A use of the public address system which so far has been taken
advantage of only on a few special occasions is by providing two-way
operation to interconnect two or more meetings held simultaneously
in different places. A notable example of this usage is the joint
meeting of the American Institute of Electrical Engineers and the
Institution of Electrical Engineers in London on February 16, 1928,
interconnected by the transatlantic telephone circuit. In this meet-
ing, addresses were heard by both audiences and a resolution made
in London and seconded in New York was jointly and unanimously
carried. It is possible that this may foreshadow a future important
use of a public address system.
Television
The possibility of transmitting pictures of a scene over electrical
circuits at so high a speed that the effect is given of seeing at a dis-
tance has naturally interested telephone people for a considerable
while. However, the large amount of detail which is taken in by
the human eye and the resulting broad band of frequencies required
to transmit this detail as well as the necessary complexity of the
terminal apparatus has, so far, prevented the development of a prac-
tical service of this kind.
In 1927 the Bell engineers demonstrated to a large number of
interested people a television circuit which extended from New York
to Washington, a distance of about 440 kilometers. The television
pictures so demonstrated had a detail corresponding to 50 lines in
each direction, that is 2,500 elementary areas and 18 such pictures
were shown each second. Two circuits especially corrected for volume
and phase distortion over a band width of about 20,000 cycles were
7H BELL SYSTEM TECHNICAL JOURNAL
employed between the two cities. These circuits were, for the most
part, in open wire although approximately 13 kilometers of specially
loaded cable were necessary at the ends in entering the cities. By
means of a separate talking circuit a person at one end of the system
could talk to, as well as see, a person at the other end. Systems of
approximately twice the detail and also systems adapted to the view-
ing of larger scenes such as athletes in action have since been developed
and demonstrated.
Time Service
Arrangements have been made in many parts of the country to
furnish subscribers who desire it, accurate information as to the time
of day. A subscriber wishing the information asks for or dials a
particular number assigned for this purpose and is connected either
to an operator who advises him individually as to the time or is
switched across a bus-bar to which is connected the amplified speech
of an operator repeating at fifteen second intervals the exact time
of day. In the present development of this service it is the practice
to localize in one place the time service for an entire exchange area.
By-Products
Certain interesting and important by-products of the telephone
development work justify a brief mention. Three arts separate from
the telephone art have been radically changed by such by-product
developments. These include submarine telegraphy, phonographs
and motion pictures.
The changes in submarine telegraphy have resulted from develop-
ment by the Bell Laboratories of the materials known as " permalloy "
and " perminvar " which have unusual magnetic properties at low flux
densities. Submarine cables so loaded can transmit approximately
10 times as many words per minute in one direction as compared to
cables of the same weight as previously constructed. As such loaded
cables are not duplexed the effective increase in speed of transmission
is approximately five times.
Development work in connection with the faithful recording and
reproduction of sound has greatly improved phonographs and their
records. The "Orthophonic Victrola" is an example of such devel-
opment.
An extension of this work led to the development of the "talking "
motion picture. The systems known under the names " Vitaphone "
and "Movietone" followed from this work. Great interest has been
aroused in such systems in the amusement field in the United States,
TELEPHONE SYSTEM OF THE UNITED STATES 79
Moving picture houses in the important cities and towns are already
equipped to show pictures of this type and it appears destined to
revolutionize the motion picture art.
A study of speech and hearing in connection with telephone service
has led to the development of various devices of value to those having
abnormal hearing or speech. This work has been carried out in close
cooperation with interested members of the medical profession. One
of these devices, the "audiometer," is useful in determining the con-
dition of hearing of individuals by determining the smallest volume
of sound at a considerable number of different frequencies which
the individual can hear. This device, in rapidly testing large groups
of people such as in the public schools, is believed to be of consider-
able importance. Sound amplifying devices are provided for those
hard of hearing.
Another interesting by-product is an artificial larynx for those who
have lost their natural larynx as a result of pathological conditions.
Apparatus has also been constructed to permit the totally deaf to
understand speech sounds by holding their fingers against a moving
diaphragm. In one form the individual fingers and thumb are held
against separate vibrating bodies and the important range of speech
sounds is divided by electrical filters and one part of it applied to
each of these five vibrating bodies. This partial electrical analysis
of sound appears to be of considerable help in this tactual apprecia-
tion of sound.
Other tools of interest to the medical profession include electrical
stethescopes and electro-cardiographs. The first of these permits
any desired number to listen to chest or other sounds in medical
patients. Electrical filters may be interposed in such arrangements
to exaggerate or subordinate certain part of the sound. The electro-
cardiograph, by permitting the amplification and recording of slight
differences of electrical potential between selected points of the skin
of a patient give an indication of the condition of his heart beat.
Conclusion
In the above discussion, while emphasis has been placed upon engi-
neering matters, it has naturally been impossible in the discussion
of results to separate engineering considerations from many other
important phases of the telephone communication problem. While
engineering is essential to the results that have been obtained, they
are due also to these other factors, commercial and general in their
character, and to the policies as regards service and operations which
guide the Bell Telephone System. Furthermore, the solution worked
80 BELL SYSTEM TECHNICAL JOURNAL
out has been designed specifically to meet conditions in the United
States, conditions which in many respects are different in the different
countries.
It is, of course, not possible in a paper of such broad scope to give
technical details of the engineering problems involved. These have,
however, been quite fully set forth in numerous articles in the tech-
nical press of the United States. For the convenience of those who
may wish to refer further to these matters, a bibliography containing
a selected list of some of the more impoitant articles is attached to
this paper.
In looking forward, there seems to be no doubt that the develop-
ment of telephone communication in the United States, commercially
and technically, will be more rapid than in the past, not less rapid.
There are strong indications that in the future very much larger
amounts of telephone service, both exchange and toll, will be de-
manded than at the present time, and in fact that for a number of
years at least the rate of growth will continue to increase. The type
and extent of services supplied will be modified to meet the broaden-
ing and multiplying demands of the changing business and social
structure of the country. Finally, it is evident that the rapid advance
of science will continue to bring forward new possibilities by means of
which new and improved forms of communication systems, apparatus,
and materials, can be developed.
These facts all indicate that the engineering work for the telephone
communication system of the United States is not complete nor de-
creasing in magnitude or importance, but on the contrary it is increas-
ing in volume and complexity and in the importance of the problems
to be undertaken and solved.
Authors' Note
The authors wish to acknowledge their indebtedness to a large
number of members of the organization for their assistance in the
preparation of this paper. It is impracticable to mention all who have
been of assistance but they wish to express their appreciation par-
ticularly to Messrs. O. B. Blackwell, W. E. Farnham, W. H. Harden,
H. S. Osborne, and W. A. Stevens.
TELEPHONE SYSTEM OF THE UNITED STATES
81
Partial Bibliography of Papers Relating to the Bell Com-
munication System
General
Ideals of the Telephone Service.
Bell Telephone Quarterly, Vol. 1, Oct. 1922, pages 1-11.
Science, Vol. 57, Feb. 23, 1923, pages 219-224, Annual
Report of Smithsonian Institution, 1922, pages 533-
540.
Semi-Centennial of the Telephone.
Bell Telephone Quarterly', Vol. 5, Jan. 1926, pages 1-11.
Telegraph and Telephone Age, Vol. 44, March 1, 1926,
pages 98-101.
Fifty Years of Telephone Progress, 1876-1926.
Telegraph and Telephone Age, Vol. 44, Feb. 1, 1926,
pages 51-53.
Building for Service.
Bell Telephone Quarterly, Vol. 7, April 1928, pages
69-81.
General Engineering Problems of the Bell System.
Electrical Communication, Vol. 4, Oct. 1925, pages 111-
125.
Bell System Technical Journal, Vol. 4, Oct. 1925, pages
515-541.
Bell System Research Laboratories.
Electrical Communication, Vol. 2, Jan. 1924, pages
153-163.
Development and Research in the Bell System.
Bell Telephone Quarterly, Vol. 4, Oct. 1925, pages 266-
280.
The Budget Plan of the Bell System.
Bell Telephone Quarterly, Jan. 1923, pages 32-42.
Electrical Communication, April 1923, pages 64-68.
Service in the Making.
Bell Telephone Quarterly, Vol. l,Oct. 1922, pages 26-33.
Functions and Management Problems of the Traffic Depart-
ment.
Bell Telephone Quarterly, Vol. 5, Oct. 1926, pages 203-
218.
Standardization in the Bell System.
Bell Telephone Quarterly, Vol. 8, Jan. 1929, pages 9-24,
and April 1929, pages 132-152.
J. J. Carty
J. J. Carty
J. J. Carty
H. P. Charlesworth
H. P. Charlesworth
E. B. Craft
E. B. Craft
C. A. Heiss
K. W. Waterson
K. W. Wateison
H. S. Osborne
Local Service
General
Selection of Central Office Names.
Bell Telephone Quarterly, Vol. 6, Oct. 1927, pages 231-
237.
The Planning of Telephone Exchange Plants.
American Institute of Electrical Engineers, Transac-
tions, July 1928, pages 809-817.
Cable Plant
Development of Cables Used in the Bell System.
Bell Telephone Quarterly, Vol. 2, Apr. 1923, pages 94-
106.
1800-Pair Cable Becomes a Bell System Standard.
Bell Telephone Qu arterly, Vol. 8, Jan. 1929, pages 25-29.
A. E. \'an Hagan
W. B. Stephenson
F. L. Rhodes
F. L. Rhodes
82
BELL SYSTEM TECHNICAL JOURNAL
Switching Systems
Machine Switching Telephone System for Large Metropolitan E. B. Craft
Areas. L. F. Morehouse
Bell System Technical Journal, Vol. 2, Apr. 1923, pages H. P. Charlesworth
53-89.
Ameiican Institute of Electrical Engineers, Transac-
tions, Vol. 42, Feb. 1923, pages 187-201.
Machine Switching Private Branch Exchanges and Their \V. H. Harrison
Application to Railroad Service.
In American Railway Association, Telegraph and Tele-
phone Section, Papers, 1924, pages 418-440.
Panel Type Machine Switching System in the United States. H. P. Clausen
Electrical Communication, Vol. 4, Oct. 1925, pages
91-97.
Telephone Switchboard — ^Fifty Years of History.
Bell Telephone Quarterly, Vol. 7, July 1928, pages
149-165.
Buildings
Housing the Bell System.
Bell Telephone Quarterly, \^ol. 5, July 1926, pages
131-139.
Post Office Electrical Engineers' Journal, \'ol. 19, Jan.
1927, pages 325-334.
F. B. Jewett
H. P. Charlesworth
Toll Service
Short Distance Toll Service
Tandem System of Handling Short-Haul Toll Calls.
American Institute of Electrical Engineers, Transac-
tions, Jan. 1928, pages 9-20.
Long Distance Service
General
Engineering the Long Lines.
Bell Telephone Quarterly, Vol. 2, Jan. 1923, pages 18-31.
Advance Planning of the Telephone Toll Plant.
Ameiican Institute of Electrical Engineers, Transac-
tions, Vol. 47, Jan. 1928, pages 1-8.
Telephone Toll Lines
Telephone Transmission Over Long Distances.
Electrical Communication, Vol. 2, Oct. 1923, pages
81-94.
American Institute of Electrical Engineers, Transac-
tions, Vol. 42, Oct. 1923, pages 984-995.
Some Very Long Telephone Circuits of the Bell System.
Bell System Technical Journal, Vol. 3, July 1924, pages
495-507.
Transmission Features of Transcontinental Telephony.
American Institute of Electrical Engineers, Transac-
tions, Vol. 45, Sept. 1926, pages 1159-1167.
Open Wire and Carrier Circuits
Carrier Current Telephony and Telegraphy.
American Institute of Electrical Engineers, Transac-
tions, Vol. 40, Feb. 1921, pages 205-300.
Electrician, Vol. 36, May 6, 1921, pages 551-554.
Practical Application of Carrier Telephone and Telegraph in
the Bell System.
Bell System Technical Journal, Vol. 2, Apr. 1923, pages
41-52.
F. O. Wheelock
E. Jacobsen
J. J. Pilliod
J. N. Chamberlin
H. S. Osborne
H. H. Nance
H. H. Nance
E. H. Colpitts
O. B. Blackwell
A. F. Rose
TELEPHONE SYSTEM OF THE UNITED STATES
83
Making the Most of the Line.
Electrical Communication, Vol. 3, July 1924, oages 8-21.
Carriei Systems on Long Distance Telephone Lines.
Bell System Technical Journal, Vol. 7, July 1928, pages
564-629.
American Institute of Electrical Engineers, Transac-
tions, Vol. 47, Oct. 1928, pages 1360-1386.
Carrier Telephone System for Short Toll Circuits.
American Institute of Electrical Engineers, Transac-
tions, Vol. 48, Jan. 1929, pages 117-139.
Toll Cables
Boston to Chicago Telephone Cable — Section of Largest and
Longest Cable Line in the World Being Completed
to Pittsburgh, Pa., by A. T. & T. Co.
Telephony, Vol. 81, Dec. 31, 1921, pages 15-18.
Philadelphia-Pittsburgh Section of the New York-Chicago
Cable.
Bell System Technical Journal, Vol. 1, July 1922, pages
60-87.
American Institute of Electrical Engineers, Transac-
tions, Vol. 41, June 1922, pages 446-456.
Development of Cables Used in the Bell System.
Bell Telephone Quarterly, Vol. 2, Apr. 1923, pages 94-
106.
F. B. Jewett
H. A. Affel
C. S. Demarest
C. W. Green
H. S. Black
M. L. Almquist
L. M. Ilgenfritz
R. W. King
J. J. PiUiod
F. L. Rhodes
Bancroft Gherardi
William Fondiller
Thomas Shaw
William Fondiller
Toll Cables — Loading
Commercial Loading of Telephone Circuits in the Bell System.
American Institute of Electrical Engineers, Transac-
tions, Vol. 30, pt. 3, June 1911, pages 1743-1764.
Commercial Loading of Telephone Cable.
Electrical Communication, Vol. 4, July 1925, pages
24-39.
Development and Application of Loading for Telephone Cir-
cuits.
Bell System Technical Journal, Vol. 5, 1926, pages
221-281.
American Institute of Electrical Engineers, Transac-
tions, Vol. 45, Feb. 1926, pages 268-292.
Electrical Communication, Vol. 4, April 1926, pages
258-276.
Permalloy; the Latest Step in the Evolution of the Loading F. L. Rhodes
Coil.
Bell Telephone Quarterly, Vol. 6, Oct. 1927, pages
239-246.
Toll Cables — Transmission
Telephone Transmission Over Long Cable Circuits. A. B. Clark
Ameiican Institute of Electrical Engineers, Transac-
tions, Vol. 42, Feb. 1923, pages 86-97.
Electrical Communication, Feb. 1923, pages 26-40.
Bell System Technical Journal, Vol. 2, Jan. 1923, pages
67-94.
Building-Up of Sinusoidal Currents in Long Periodically J. R. Carson
Loaded Lines.
Bell System Technical Journal, Vol. 3, Oct. 1924, pages
558-566.
Distortion Correction in Electrical Circuits with Constant O. J. Zobel
Resistance Recurrent Networks.
Bell System Technical Journal, Vol. 7, July 1928, pages
438-534.
84
BELL SYSTEM TECHNICAL JOURNAL
Toll Circuit Equipment
Telephone Repeaters.
American Institute of Electrical Engineers, Transac-
tions, Vol. 38, Oct. 1919, pages 1287-1345.
Practical Application of the Telephone Repeater.
The Western Society of Engineers Journal, Vol. 27,
May 1922, pages 129-142.
Telephone Repeaters.
Electrical Communication, Vol. 1, Aug. 1922, pages 6-
10; Nov. 1922, pages 27-36.
Telephone Equipment for Long Cable Circuits.
American Institute of Electrical Engineers, Transac-
tions, Vol. 42, June 1923, pages 742-752.
Echo Suppressors for Long Telephone Circuits.
Ameiican Institute of Electrical Engineers, Transac-
tions, Vol. 44, Apr. 1925, pages 481-490.
Electrical Communication, Vol. 4, July 1925, pages
40-50.
Bancroft Gherardi
F. B. Jewett
H. S. Osborne
Bancroft Gherardi
C. S. Demarest
A. B. Clark
R. C. Mathes
Toll Line Construction
Poles.
Bell Telephone Quarterly, Vol. 1, Oct. 1922, pages 34-44.
Bell System Sleet Storm Map.
Bell System Technical Journal, Vol. 2, Jan. 1923, pages
114-121.
Specializing Transportation Equipment in Order to Adapt It
Most Economically to Telephone Construction and
Maintenance Work.
Electrical Communication, Vol. 1, Feb. 1923, pages
50-59.
Bell System Technical Journal, Vol. 2, Jan. 1923, pages
47-66.
Open Tank Cieosoting Plants for Treating Chestnut Poles.
Bell System Technical Journal, Vol. 4, Apr. 1925, pages
235-264.
Bell Telephone Quarterly, Vol. 4, Jan. 1925, pages 132-
142.
Recent Toll Cable Construction and Its Problems.
Telephone Engineer, Vol. 32, Sept. 1928, pages 31-33.
Switching of Toll Circuits
Toll Switchboard No. 3.
Bell System Technical Journal, Vol. 6, Jan. 1927, pages
18-26.
Electrical Communication, Vol. 5, Apr. 1927, pages
255-259.
Maintenance of Toll Circuits
Measuring Methods for Maintaining the Transmission Effi-
ciency of Telephone Circuits.
American Institute of Electrical Engineers, Transac-
tions, Vol. 43, Feb. 1924, pages 423-433.
Electrical Tests and Their Applications in the Maintenance
of Telephone Transmission.
Bell System Technical Journal, Vol. 3, July 1924, pages
353-392.
Practices in Telephone Transmission Maintenance Work.
American Institute of Electrical Engineers, Transac-
tions, Vol. 43, 1924, pages 1320-1330.
Bell System Technical Journal, Vol. 4, Jan. 1925, pages
26-51.
F. L. Rhodes
J. N. Kirk
J. N. Kirk
T. C. Smith
H. S. Percival
John Davidson, Jr.
F. H. Best
W. H. Harden
W. H. Harden
TELEPHONE SYSTEM OF THE UNITED STATES
85
International Connections
Connections in North America
Key West-Havana Submarine Telephone Cable System.
American Institute of Electrical Engineers, Transac-
tions, Vol. 41, Feb. 1922, pages 1-19.
Comiections to Europe
Telephoning to England.
Radio Broadcast, Vol. 2, March 1923, pages 425-426.
Transatlantic Radio Telephony.
Bell System Technical Journal, Vol. 2, Oct. 1923, pages
116-144.
American Institute of Electrical Engineers, Transac-
tions, Vol. 42, June 1923, pages 718-729.
Transatlantic Radio Telephone Transmission.
Bell System Technical Journal, Vol. 4, July 1925, pages
459-507.
Institute of Radio Engineers, Proceedings, Vol. 14, Feb.
1926, pages 7-56.
Radio Telephone Developments of the Bell System.
Bell Telephone Quarterly, Vol. 5, Oct. 1926, pages 219-
237.
New York-London Telephone Circuit.
Bell System Technical Journal, Vol. 6, Oct. 1927, pages
736-749.
\'oices Across the Sea.
North American Review, Vol. 224, Dec. 1927, pages
654-661.
Transatlantic Telephony — The Technical Problem.
American Institute of Electrical Engineers, Journal,
Vol. 47, May 1928, pages 369-373.
Bell System Technical Journal, Vol. 7, Apr. 1928, pages
161-167.
Transatlantic Telephone Service — Service and Operating Fea-
tures.
American Institute of Electrical Engineers, Journal,
Vol. 47, Apr. 1928, pages 270-273.
Bell System Technical Journal, Vol. 7, Apr. 1928, pages
187-194.
W. H. Martin
G. A. Anderegg
B. W. Kendall
R. W. King
H. D. Arnold
Lloyd Espenschied
Lloyd Espenschied
C. N. Anderson
Austin Bailey
J. O. Periine
S. B. Wright
H. C. Silent
Bancroft Gherardi
O. B. Blackwell
K. W. Waterson
Special Services
Telegraph Circuits
Metallic Polar-duplex Telegraph System for Long Small-gauge
Cables.
American Institute of Electrical Engineers, Transac-
tions, Vol. 44, Feb. 1925, pages 316-325.
\'oice-Frequency Carrier Telegraph System for Cables.
Electrical Communication, Vol. 3, Apr. 1925, pages 288-
294.
Telephone Networks for Program Transmission to Radio Broad-
casting Stations
Telephone Circuits Used as an Adjunct to Radio Broadcasting.
Electrical Communication, Vol. 3, Jan. 1925, pages
194-202.
Telephoning Radio Programs to the Nation.
Bell Telephone Quarterly, Vol. 7, Jan. 1928, pages 5-16.
How Chain Broadcasting is Accomplished.
Radio Broadcast, Vol. 12, June 1928, pages 65-67.
J. H. Bell
R. B. Shanck
D. E. Branson
B. P. Hamilton
H. Nyquist
M. B. Long
W. A. Phelps
H. S. Poland
A. F. Rose
L. N. Stoskopf
C. E. Dean
86
BELL SYSTEM TECHNICAL JOURNAL
Electrical Transmission of Pictures
Transmission of Pictures Over Telephone Lines.
Bell System Technical Journal, Vol. 4, Apr.
187-214.
1925,
pages
Telephone in Connection with Aircraft Operation
Airways Communication Service.
Bell System Technical Journal, Vol. 7, Oct. 1928, pages
797-807.
Aviation, Vol. 25, Oct. 6, 1928, pages 1090-1091, 1136,
1138, 1140, 1142, 1144, 1146.
Ship-to-Shore Telephony
Radio Extension of the Telephone System to Ships at Sea.
Institute of Radio Engineers, Proceedings, Vol. 11,
June 1923, pages 193-239.
Telephone Services of Railroads and Other Public Utilities
Telephone Equipment for Train Dispatching Circuits: A dis-
cussion of the Requirements, Development and
Design of Latest Types of Equipment for High
Grade Train Dispatching Systems Including Vac-
uum Tube Amplifiers and Loud Speakers.
Electrical Communication, Vol. 2, Oct. 1923, pages
111-140.
Recent Developments in Telephone Train Dispatching Cir-
cuits.
Railway Signaling, Vol. 17, Feb. 1924, pages 73-75;
May 1924, pages 208-211; June 1924, pages 253-
256; Aug. 1924, pages 320-322.
Telephone Public Address Systems
Use of Public Address System with Telephone Lines.
Bell System Technical Journal, Vol. 2, Apr. 1923, pages
143-161.
Electrical Communication, Vol. 1, Apr. 1923, pages
46-56.
American Institute of Electrical Engineers, Transac-
tions, Vol. 42, Feb. 1923, pages 75-85.
High Quality Transmission and Reproduction of Speech and
Music.
Electrical Communication, Vol. 2, Apt. 1924, pages
238-249.
American Institute of Electrical Engineers, Transac-
tions, Vol. 43, Feb. 1924, pages 384-392.
Television
Television.
American Institute of Electrical Engineers, Transac-
tions, Vol. 46, June 1927, pages 913-917.
Production and Utilization of Television Signals.
American Institute of Electrical Engineers, Transac-
tions, Vol. 46, June 1927, pages 918-939.
Synchronization of Tele\asion.
American Institute of Electrical Engineers, Transac-
tions, Vol. 46, June 1927, pages 940-945.
Wire Transmission System for Television.
American Institute of Electrical Engineers, Transac-
tions, Vol. 46, June 1927, pages 946-953.
Radio Transmission System for Television.
American Institute of Electrical Engineers, Transac-
tions, Vol. 46, June 1927, pages 954-962.
H. E. Ives
J. W. Horton
R. D. Parker
A. B. Clark
E. B. Craft
H. \V. Nichols
L. Espenschied
W. H. Capen
W. H. Capen
W. H. Martin
W. H. Martin
Harvey Fletcher
H. E. Ives
Frank Gray
R. C. Mathes
H. M. StoUer
E. R. Horton
D. K. Gannet
E. I. Green
E. L. Nelson
TELEPHONE SYSTEM OF THE UNITED STATES
87
R. W. King
O. E. Buckley
By-Products
By-Products of Telephone Research.
Bell Telephone Quarterly, Vol. 7, Oct. 1928, pages
304-312.
Loaded Submarine Telegraph Cable.
Bell System Technical Journal, Vol. 4, July 1925, pages
355-374.
Electrical Communication, Vol. 4, July 1925, pages
60-70.
American Institute of Electrical Engineers, Transac-
tions, Vol. 44, June 1925, pages 882-890.
Telegraph and Telephone Age, Vol. 43, Nov. 16, 1925,
pages 524-525.
Permalloy Loaded Cable.
Electrical Communication, Vol. 2, Apr. 1924, pages
232-234.
Man-made Ears for the Deaf; Why Many Deaf People Hear
Normally in Noisy Places and Over the Telephone.
Scientific American, Vol. 8, Nov. 1925, pages 320-321.
Recent Advances in Wax Recording.
Bell System Technical Journal, Vol. 8, Jan. 1929, pages
159-172.
Sound Recording with the Light Valve.
Bell System Technical Journal, Vol. 8, Jan. 1929, pages
173-183.
Synchronization and Speed Control of Synchronized Sound
Pictures.
Bell System Technical Journal, Vol. 8, Jan. 1929, pages
184-195.
A Sound Projector System for Use in Motion Picture Theatres.
Bell System Technical Journal, Vol. 8, Jan. 1929, pages
196-208.
Miscellaneous
Telephone Transmission.
Sibley Journal of Engineering, Vol. 31, Apr. 1917, pages
_ 177-180.
Transmission Unit and Telephone Transmission Reference
Systems.
Bell System Technical Journal, Vol. 3, July 1924, pages
400-408.
American Institute of Electrical Engineers, Transac-
tions, Vol. 43, June 1924, pages 797-801.
Statistics of the Telephone Industry of the United States
Figure 57. Number of Telephones in the United States.
Figure 58. Telephone Development in the United States.
Figure 59. Percentage Distribution of Bell Stations in Fifteen Large Cities in the
United States.
Figure 60. Telephone Conversations — Average Number Daily in Millions in the
United States.
Figure 61. Average Daily Number of Toll Messages in the United States.
Figure 62. Yearly Telephone Messages per Capital in the United States.
Figure 63. Kilometers ot Telephone Wire in the United States.
Figure 64. Kilometers of Exchange and Toll Wire in the United States.
Figure 65. Telephone Wire in the Bell System.
Figure 66. Growth of Various Classes of Physical Property in the Bell System.
Figure 67. Bell System Revenues.
Figure 68. Table Showing Initial Period Toll Rates.
Figure 69. Map of the Bell System Showing Territories of the Associated Com-
panies.
Figure 70. Telephone Employees in the United States.
Figure 71. Table Showing Population and Telephones which may be Connected
by Transatlantic Telephone Service.
F. B. Jewett
Harvey Fletcher
H. A. Frederick
D. MacKenzie
H. M. Stoller
E. O. Scriven
Bancroft Gherardi
W. H. Martin
88
BELL SYSTEM TECHNICAL JOURNAL
8.000.000
16.000.000
14.000.000
12,000.000
0,000.000
8,000.000
6,000.000
4.000,000
2,000,000
OfVJ^<D<00(\J^<X)<DOCM<J'<OOOOrvj'if(0<00(\J^>000 ^
eOeOoO<0<00><J)0>0)0>00000 — — — — — c\jfVJ<\lf\jf\j
(Ocoo<o<D<ooOcoaOooo>o>0)0>o>a)o>o>a>o>o>a>a)a>o>
Fig. 57 — 'Number of telephones in the United States.
TELEPHONE SYSTEM OF THE UNITED STATES
89
z
o
CL
o
CL
o
o
oc
u
Q.
Ul
U
z
o
r
Q.
U
_l
u
I-
Jjnujry lit o^ eich yeir
Fig. 58 — Telephone development in the United States.
90
BELL SYSTEM TECHNICAL JOURNAL
<
•A
<
tn
W
H
<
a
w
H
w
a
CO
U
a
o
<
H
O
H
<!
H
C/)
>-l
w
b
O
iz;
O
H
H
tn
o
<:
H
u
u
Oh
C3
O
CI
BQ
re <^' ^' -^ O r^' O O 00 00 O 00 rc r<l tN
oo' o' 0>' oc' oc' l~~' C~ t^ C'" C^ OC C" O OS o
CI
g C rt
n ^ u
OvoO'-iTt^OCC'l'^rrj^O"^-*-^-^'*
•^' sd rf -- ■* "S ■^' >-< O "-^ oo' oo' r^ O 0\
f*^ CN r^i i>i c^i CN ro CN t^i ^, — ^ --H r^ '-I
3
'■3
00^'O'*'*"^ONv0Orc-+'O'^'*00
roOiOCO'+-rfCir;C^t^C>r^'-iOOO
■^ ir> ir> o O O- Sv O >-< J>- O "-^ fO O O
d
O O
^ O)
C'oj
01
09
Osirj'O'-iON'^^'^ioior^-H-^ooo
oo-H"ooo<^Ooo-*CNCsr^-^-*oO'*
00Ot^t^'*"0'-'O'-iCNOO0000iO
(^4'cN~lO T}<"'-rtC-o"ro'>o"c-ro Tf '-H \0 ■^
q ca
? flj
o
cu
OOOOOOCOOOOOOOO
ooooooooooooooo
•^ O CN "* CN O 00_^lO^O_lO 0_"T_0_0_0_
0~ O" CT '-H 00~^-^ LO fO^ ■^ ■—< >0 lO On tN lO
•.-iio'*icNt^mrr>ossOLOt^cN-HOv-^
r0CNO00O<~0'-iOCNt^OiO"^'^t^
VO~ fO CN ^~ -4" T^" -^ •rt
.2
a
rjl
TELEPHONE SYSTEM OF THE UNITED STATES
91
80
70
60
50
■40
30
20
10
O
1898 1908 1918 1928
Fig. 60 — Telephone conversations — Average number daily in millionsin United States.
00
CO
0>
o
00
0>
00
5>
00
if)
u
o
<
to
(0
UJ
li.
o
(O
z
o
_J
_l
Fig. 61 — -Average daily number of toll messages in the United States.
92
BELL SYSTEM TECHNICAL JOURNAL
o
in
o
o
0>
o>
a\
in
o
in
00
(M
f\J
t\j
0^
0^
a>
200
150
100
50
Fig. 62 — Yearly telephone messages per capita in the United States.
TELEPHONE SYSTEM OF THE UNITED STATES
93
120
110
100
90
in
CL
80
UJ
t-
u
2
O
70
_l
^
u.
60
O
eo
z
50
o
-J
_l
40
i
r
1
^^
^^
n
«
^
^
OTHE
R COMP/i
^NIES
1
w
■
VWWYVV
■
■
■
30
20
10
1920 1921 1922 1923 1924 1925 1926 1927 1928
Fig. 63— Kilometers of telephone wire in the United States.
94
BELL SYSTEM TECHNICAL JOURNAL
120
to
UJ
\-
u
Q
U-
O
V)
z
o
1920 1921 1922 1923 1924 1925 1926 1927 1928
Fig. 64 — Kilometers of exchange and toll wire in the United States.
TELEPHONE SYSTEM OF THE UNITED STATES
95
100
in
z
o
J
1920 1921 1922 1923 1924 1925 1926 1927 1928
Fig. 65 — Telephone wire in the Bell System in millions of kilometers.
96
BELL SYSTEM TECHNICAL JOURNAL
4000
4000
— — — — — — — — — — CMt\JCy(\JC\Jt\Jf\IC\J(\JC\Jfr)
Fig. 66 — Growth of various classes of physical property of the Bell System.
o — f\in^«'><oi~-(0<j)o — (\iro^>n(£)h-<oo>o— (\j(0^ inioi^oo
OOOOOOOOOO— — — —— — — — — — (M(\|(\j(\J(\|(\j(\)(\j(\|
Fig. 67 — -Bell System revenues in millions of dollars.
TELEPHONE SYSTEM OF THE UNITED STATES
97
Report
Charge
O O lO o >o O
LO lO lO NO NO t^
•o o o to o o
t^ GO On ON O O
o o c c o o
ooc c o o
oo c oc o
O' C O C O C'
o ooooo
O O C' O 3 O
oo oooo
oo oooo
«^
Person-to-
Person
All Hours
ID lO LO O lO O
-;f lO t^ O <>1 to
to o o to o to
t^ C to r^ O f^l
lo o >o o o to
t^ O '^1 to O i>l
o to to o to o
to t^ r^i to r— C
O >o o to lo o
to r^ C '^l i~~ C
to o o to o to
^J to o r^i to l^
CNl CN fN ro ro rr)
ro Tt -rh Tj- to to
to NO NO vO J^ t^
r^ t^ 00 00 00 C^.
O^. On C O O -H
^H -H CM C^l <rv) (M
c
o
a
g
Si
1 ■
o >oo o o o
•r-i ■— 1 CNi ro Ttl LO
to to O O to O
NO t^ On O r^ to
o "^ c o o to
to r^ O O O fN
'>^ oi ro ro fO ro
C C to O O to
to to t^ o o o)
CO ro" ro -^ >* •*
o o to o o to
to to t^ O O <M
■* Tl^ -^ to to to
o o >o to o to
to to t^ f^ o oi
to to lO lO o o"
m
1 ■
8^
lO lO o to o to
NO «^ 00 On -H tv)
to to o o to c
'* no 00 o <>i to
lO O to to to o
t^ o '>) ^j rv) to
lo o to o o to
r- O CN) to to t^
O to O to to c
o <~Ni to t^ t-- o
to o o to o to
(vj to to r^ O oi
■^ •■-H ^ ■« tN CN)
*>! r^ f^l ro '^ ro
r^ 't ^ "* -f •*
'Tf to to to to to
O NO O nO nO r^
t^ t^ t^ t^ 00 GO
1 .
lO) to to to to to
On O ■" ro to t^
O to O to C to
O tN to t-^ O fN
o to o to o to
to t^ O '^1 to t^
O to o to O to
O f^ to t^ O f^l
o to o to o to
to t~~ O ^1 to t^
o to o to o to
O ^1 to t-. O <^J
•■— 1 tN r^ CN CNi r^i
^ ro ro ro -* -^
Tf" •<* to to to to
NO O nO nO f^ t^
t— t^ GO 00 00 00
On On ON ONO O
1-1 i-H
Air Line
Kilometers
(Fractions
Omitted)
711- 763
763- 814
814- 866
866- 969
969-1072
072-1175
175-1303
303-1432
[432-1561
561-1689
689-1818
818-1947
NO •* (^ J^ tN O
r^ O ^ On NO On
O tN r^ "^ "* to
r^ r-j r^i o] r-1 CN
1 1 1 1 1 1
t^ O ^ ^O t^ Ol
Tj< t^ O 'T) On NO
On O O) f^ "^ ^
2590-2719
2719-2848
2848-2977
2977-3105
3105-3234
3234-3363
3363-3492
3492-3620
3620-3749
3749-3878
3878-4006
4006-4135
4135-4264
4264-4393
4393-4521
4521-4650
4650-4779
4779-4907
C, CO
tj Si
Pi'O
cooooo
o o to to to to
O O O O to to
CN) fN tN CN C-) rNI
lo too o o o
(M <M oo ro fO "n
lo to to to o o
ro ro ro ro ^ ^
o o to to to to
TT '^ ^^ ^* "^ '^
m^
Person-to-
Pcrson
AH Hours
C to O to o o
Ol 1^ ro r<-> ^ to
lO »o o *o o to
lO NO r— t-~ 00 00
o to o to to o
On ON O O -^ -^I
lo o o to o to
"^ rO -f -* to to
to C to O O LO
O J^ I— 00 On ON
o to to o to o
O O -^ i^l '^1 ^
€«?^
rs (>4 cNi fvi tN r^i
^3
6
c
'-t~t
B
8:30 P.M.-
4:30 A.M.
o "o o 'o o to
•--< .r-i CN C^l (^ fO
to to to to >o to
ro r^ rO ro rO rO
loco to o o
rO '^^ •* -^ to to
to to c c to o
to to O NO O t^
o to o o to o
r^ t-~. 00 00 00 On
O to o O lo to
ON Ov OO o o
&^
^H T^ 1— 1 ^H
7:00 P.M.-
8:30 P.M.
o >o o 'o o to
T-^ ■^ r>i CN (-O f^
to to o O 'o c
ro rr> ^ ^ 'Tf iri
lO to O to o to
to to no o t^ t^
o <o >o o c o
00 00 oo ON o c
to O to O 'O C
O -- — r^i '^1 oo
lo o to o to o
-O -f -f to lO o
4^
o to o "o o to
— I— 1 (TNI r-j ro fO
o to c to o to
-:f -+ to to O nC
O to O to O VO
t-~. t^ GC 3C ON O^
O to O to O "O
O O ■^ >— 1 c^l (M
o to o to o to
-O oo 'f ^ to to
o to o to o to
O O t— t^ 00 00
€r%
11
c
js.uonieters
(Fractions
Omitted)
On On On 00 00 00
"-I r^I ro -f to NO
O On On On 00 00
-H "-l '^, ^ to
68- 77
77- 90
90-103
103-116
116-129
129-145
^ t^ rO >0 — 1 —
nO t^ O^ i-^ ^ o
^H T-i ■^ 1^1 rs) (>i
1 1 1 1 1 1
lO ^H f^ ro to ^H
"* NO J— O^ CNl -Tf
-H ..-H ^ .rt tV| CVI
O On On l-~ t^ NO
00 On -^ to I-^ On
ITS 1^1 ro ro "^ oo
1 1 1 1 1 1
'-^ O O^ On «^ J^
O oc On >— ' to I^
"^ CN CN ro ro ro
396-415
415-454
454-473
473-492
492-531
531-550
o GO '^ o to — 1
r^ O oo O 00 ^
in ^ ^ ^ ^ t^
1 1 1 1 1 1
O O 00 rt" O to
to t^ o 'o o 00
lo to o o o o
CO
NO
o
98
BELL SYSTEM TECHNICAL JOURNAL
X
w
E
^1 i - ' -
'" s' 4 ie \\ n
'.■Si V t* t'
CO
'S
a
o
cj
c
4)
a
o
•a
o
o
CO
m
(U
01
o
be
c
o
CO
>.
C/3
<u
c
o
I
bb
TELEPHONE SYSTEM OF THE UNITED STATES
99
1898 1908 19 18 1928
Fig. 70 — -Thousands of telephone emplo^'ees in the United States.
100
BELL SYSTEM TECHNICAL JOURNAL
Transatlantic Telephone Service
List of places which may be connected with the transatlantic telephone service at the
present time. The figures shown both for population and telephones are estimates
for January 1, .1929.
Total
Population
Total
Telephones
Number Served by Trans-
atlantic Connection
Population
Telephones
England, Scotland, Wales,
and Noithern Ireland
Germany
Belgium
Holland
Switzerland
France
Denmark
Norway
Sweden
Danzig
Spain
Austria
Hungary
Czechoslovakia
Gibraltar
45,830,000
64,860,000
8,000,000
7,750,000
3,990,000
41,370,000
3,530,000
2,820,000
6,150,000
400,000
22,600,000
6,950,000
8,620,000
14,600,000
17,000
280,000
1,780,000
3,000,000
225,000
250,000
236,000
1,000,000
340,000
180.000
484,000
17,300
156,000
175,000
134,000
140,000
500
9,000
45,830,000
64,860,000
8,000,000
7,750,000
3,990,000
37,900,000
782,000
250,000
760,000
400,000
22,600,000
1,970,000
1,000,000
725,000
17,000
50,000
1,780,000
3,000,000
225,000
250,000
236,000
935,000
130,000
44,500
169,000
17,300
156,000
116,000
50,000
35,000
500
Luxemburg
3,000
Total Europe
Spanish Morocco
237,767,000
1.000,000
1,000,000
118,500,000
9,800,000
15,500,000
3,650,000
8,126,800
600
196,884,000
37,000
7,147,300
300
Total Africa
United States
Canada
600
19,197,000
1,330,000
70,000
80,000
20,677,000
28,804,400
32,800,000
37,000
118,500,000
9,800,000
1,500,000
3,650,000
300
19,197,000
1,330,000
31,600
80,000
Mexico
Cuba
Total North America
Grand Total
World Total
147,450,000
386,217,000
1,930,000,000
133,450,000
330,371,000
20,638,600
27,786,200
Percentage of Number
Served to World Total
1 7Cr
t ' /o
85%
Fig. 71
Structure and Nature of Troostite ^
By FRANCIS F. LUCAS
In this paper the structure and nature of the constituent troostite (found
in hardened steels) is discussed. High power metallography was first ap-
plied to this problem about six years ago and the early results were presented
in an address before the Franklin Institute.
Since that time many improvements in technique have been developed
which have resulted in better resolution and definition. The subject has
been reviewed in the past two years and with the aid of the improvements
in technique, hardened steels are found to be largely mixtures of the things
which metallographers call martensite and troostite.
In small specimens of 0.90 per cent carbon tool steel hardened to C-65 on
the Rockwell scale, innumerable particles of troostite are found. When
these particles of troostite are examined by present high power methods the
structure is clearly resolved into laminated pearlite. In certain stages of
development of a troostitic nodule its structure borders on the verge of
present methods of resolution.
Nodular troostite develops under favorable conditions as a globular mass.
At the center is a nucleus about which the growth occurred. Radial, fan-
shaped grains extend outward from the nucleus and these grains show orien-
tation phenomena when revolved about the optical axis of the microscope.
It is believed that when martensite forms, the structure develops on the
old austenitic crystallographic planes. Troostite appears not to follow the
old austenitic system but seems to be a reorientation of the freshly trans-
formed alpha iron about a nucleus which usually is an inclusion, a void, a
sharp corner in a grain boundary or some other detail of structure.
The structure of troostite in various stages of its formation is illustrated by
means of high power photomicrographs, many of which are shown at this
Congress for the first time.
The following conclusions were reached:
Nodular troostite appears to be an aggregate of ferrite and carbide and in
the very early stages of formation its structure is on the border of present
methods of resolution. The condition of the ferrite and carbide in relation
to each other is not stable — they tend to stratify, forming pearlite.
Troostitic nodules grow about a nucleus which may be an inclusion, a
void, a corner in a grain boundary or some other detail of structure.
The nodules contain fan-shaped radial grains.
The development of troostite results in a reorientation of the ferrite —
seemingly without particular reference to the old austenitic crystallographic
planes. Martensite does follow the old system of austenitic planes.
The small fan-shaped grains in nodular troostite may persist as small
grains or they may undergo grain growth by union. It is a matter seem-
ingly dependent upon the thermal treatment of the specimen.
IN a paper^ presented before the Franklin Institute in the year 1924
some observations on the structure and probable nature of the
constituent troostite were given. Two types of troostite were shown
to occur in hardened steels depending on the mode of heat treatment.
If a bar of 0.50 per cent carbon steel is given a taper heat treatment,
1 Presented by the author before World Engineering Congress, Tokio, Japan,
October 30, 1929.
^ Lucas, "High Power Metallography — Some Recent Developments in Photo-
micrography and Metallurgical Research," Journal of the Franklin Institute, Vol. 201,
February 1926.
101
102 BELL SYSTEM TECHNICAL JOURNAL
troostite occurs which was described as flocculent border type for lack
of a better designation. It seemed to be largely ferrite and appeared
to be the means by which the excess constituent (in this case ferrite)
appeared at the grain boundaries.
If a small specimen of the same steel is heated to a high tempera-
ture and quenched in oil or water depending on the circumstances of
the experiment, a structure results which may be largely martensite
needles with scattered particles of troostite. Sometimes relatively
large areas on the prepared surface of the specimen may be almost
entirely of the constituent troostite. As is well known, this condition
is controlled by the rate of cooling in the quenching operation. The
type of troostite found in uniformly heated and quenched specimens
was defined as nodular troostite. This paper deals further with this
particular constituent of hardened steel.
Since these early experiments in which high power metallography
was first applied with success to the structures of hardened steel, there
have been many improvements in technique. These improvements
have resulted in a much higher order of resolution and it is the object
of this paper to review the past work in the light of the improved meth-
ods now available and to present some new results.
To quote from the Franklin Institute paper:
"... These nodules develop from innumerable nuclei throughout
the austenite and martensite matrix. . . . The nuclei increase in
number and the developing nodules become larger and larger. Irregu-
larities in growth due to interference of nodules occur as the growing
particles increase in number and size until finally the whole mass seems
to be composed of nodules, some spherical in shape but many deformed
due to mutual interference and to irregularities in growth. A selec-
tivity or preference in crystal habit probably prevails for crystallo-
graphic planes since spines, branches, and interconnected crystallites
may be found occasionally. In reality these are poorly formed nodules,
growth in some one or more directions having been arrested."
" It is quite evident that if the entire mass of the metal passes through
the nodular troostitic stage, this constituent must contain carbide or
carbon in some form."
"When one of the globular-shaped crystal masses which has developed
under favorable conditions of growth is sectioned in such a way as to
divide the mass along a plane passing through the center; the nucleus
is found at the center and fan-shaped grains extending from the center
toward the outside. When freshly formed and under the highest pow-
ers of the microscope these radial grains have all of the appearance of a
solid solution. The nodule must contain carbon in some form, as
stratification soon takes place."
STRUCTURE AND NATURE OF TROOSTITE 103
Moreover, it was shown that each of the fan-shaped grains is a
separate crystalline unit for if a nodule of troostite is revolved about the
optical axis of the microscope, these very small fan-shaped grains dis-
play orientation phenomena in exactly the same way as a system of
polyhedral grains in a pure metal will do if revolved about the optical
axis of a microscope while being kept under observation at 100 or 200
diameters magnification. The only difference lies in the fact that in
nodular troostite the grains are fan-shaped and quite small, making it
desirable to carry out the observations with an oil immersion lens
which will yield high magnifications.
Fig. 1 is reproduced from the Franklin Institute paper and illus-
trates a typical section on a plane passing through the center of a single
nodule. Fig. 2, also from the same paper, is a diagrammatic represen-
tation of how the fan-shaped grains develop along axes of crystalliza-
tion A, B, C, etc.
Fig. 1 not only shows crystallization about a nucleus but it also
supplies evidence for the conclusion at that time that nodular troostite
is either a solid solution of iron carbide in iron or it is a very fine aggre-
gate of iron and iron carbide — the carbide so finely dispersed as to lose
its identity under the microscope. Failure at that time to resolve the
structure of troostite into its ultimate constituents compelled one to
recognize the existence of the two possibilities as to structure.
The improvements in technique previously mentioned have thrown
some new light on the structures found in hardened steels and these
have been discussed in later papers.^' '*■ '^
Certain it is that the structures found in hardened steel are largely
mixtures of the things which metallographers call martensite and
troostite, the name martensite in this case meaning a needle-like
structure. Troostite, generally, is regarded as a lower order of decom-
position than martensite. This, however, is not believed to be substan-
tiated by the evidence.
It has been shown ^ that martensite is a decomposition of the austen-
ite along the octahedral crystallographic planes. That is, martensite
is a structure superimposed by decomposition of the austenite on the
old crystallographic system of the austenite. Two changes are in
' Lucas, "A Resume of the Development and Application of High Power Metallog-
raphy and the Ultra Violet Microscope," Vol. I, Proceedings International Congress
for Testing Materials, Amsterdam, September 1927.
* Lucas, "Photomicrography and Its Application to Mechanical Engineering,"
Mechanical Engineering, Vol. 50, March 1928.
* Lucas, "Further Obseivations on the Microscructure of Martensite," Trans.
American Society for Steel Treating, Vol. XV, February 1929.
'Lucas, "The Micro-Structure of Austenite and Martensite," Trans. American
Society for Steel Treating, Vol. VI, No. 6, December 1924.
104 BELL SYSTEM TECHNICAL JOURNAL
volved: an allotropic change of the iron from gamma to alpha and a
precipitation of the carbide FesC. This matter is more fully discussed
in a recent paper ^ to which those interested are referred.
Troostite is not like martensite in respect to habit of formation.
It does not assume fully the old austenitic crystalline symmetry. It
seems to have a new crystalline orientation of its own.
Troostite develops along grain boundaries and within the grains.
It may develop as a spine or branch along a crystallographic plane. The
nodules may be roughly spherical masses; they may be semi-spherical
masses, the flat side being bounded by a crystallographic plane or a
grain boundary, or they may be rounded but irregular shaped masses.
In any event, whether in ball-shaped masses or some constricted form,
the small fan-shaped grains are found radiating from a nucleus of
growth. This nucleus in most cases can be identified as an inclusion,
a void, or a sharp corner in a grain boundary. Thus the tendency is
for reorientation of the iron to occur when nodular troostite develops.
It is well known that a slow rate of cooling promotes more troostite.
Rapid cooling results in more needles or the constituent we call marten-
site. Evidently when the rate of cooling is favorable the freshly trans-
formed alpha iron is given time to reorient itself and does do so by
growing about some convenient inclusion or other body. Thus nodular
troostite develops. Whether the carbide is held in solid solution in the
freshly transformed alpha iron seems to be a matter of speculation.
A small specimen of steel weighing less than ten grams, heated to
1000° C. in a vacuum for a suitable length of time, and quenched in ice
and brine will contain almost innumerable troostitic bodies, many of
them very small, and some quite large. The larger ones perhaps are
a few ten-thousandths of an inch in diameter and from this dimension
the troostitic particles decrease in size to the vanishing point of present
microscopic resolution which is around 200 atom diameters.
The specimen itself will have a hardness on the Rockwell scale of
about C-65. Nevertheless the troostitic bodies have been clearly re-
solved to show the presence of fully laminated pearlite. So that in a
small specimen of steel quenched from a high temperature in a very
effective cooling bath, one finds not only the needle constituent mar-
tensite but nodules of troostite containing fan-shaped grains of fully
stratified pearlite.
The question naturally arises as to whether the steel in its transition
from austenite to pearlite first develops a needle structure (martensitic)
and then this in turn is replaced by a nodular (troostitic) one.
Some light ^ was thrown on this angle of the problem by a high power
examination of an iron carbon allov. The carbon content was 2.65
STRUCTURE AND NATURE OF TROOSTITE 105
per cent and by quenching small pieces from very high temperatures,
polyhedral grains of austenite containing martensitic needles and
troostitic nodules were found to occur. Both constituents were found
to occur in the same grain and both seemed to be entirely surrounded
by austenite. Had the needles formed first and the nodules developed
from the needles, one might expect to find some nodules with untrans-
formed needles sticking out around the boundaries of the nodules.
This was found not to be the case. The boundaries of the troostitic
nodules are always sharply defined.
In some specimens of commercial plain carbon steels in which some
tempering had taken place troostitic nodules were found in which it
appeared that the nodule had grown at the expense of some martensitic
needles. The needles seemed to be dimly visible in outline in the back-
ground of the nodule. Cases of this kind appear very infrequently.
Microscopic evidence does not support the conclusion that one type of
structure replaces the other.
If a specimen of commercial tool steel heat treated to produce some
troostite in a martensitic matrix is tempered, one might expect the
troostite nodules to grow in size if the nodular form of structure re-
places the needle structure. As a matter of fact the nodules remain the
same size and the carbide which they contain tends to coalesce into
small globular particles, marking not only the border outline of the
nodule but also the outlines of the fan-shaped grains.
The needle and nodular patterns are structures which result from
quenching and not from tempering. The excess constituent in the
case of hypo- or hyper-eutectoid steels appears to be eliminated or
cleared by means of the constituent troostite. The constituent mar-
tensite (needles) appears not to be involved in this phenomena in
quenched specimens when both troostite and martensite are present.
If one examines a normalized specimen of plain carbon tool steel of
about 0.90 per cent carbon content he will find a large polyhedral struc-
ture marked by a carbide network, but within these grains will be found
a great many smaller grains of pearlite, usually fan-shaped. In many
cases the outlines of the old troostitic nodules can be traced without
difiiculty. From the configuration of the pattern it seems likely that
these small grains within the larger (old austenitic) grain must differ in
their inner crystalline symmetry, i.e., it is probable that the ferrite is not
everywhere oriented the same throughout the old austenitic grain.
Under some circumstances controlled by heat treatment, it appears that
grain growth does occur among these small fan-shaped grains and the
old austenitic grain may be uniformly oriented ferrite containing sphe-
roidized particles of cementite which by their positions mark the old
structure and tell the history of the transformations.
106 BELL SYSTEM TECHNICAL JOURNAL
Professor Honda believes that martensite forms first and troostite
develops secondly, replacing the martensite. '^ In his discussion of the
subject he appears to deal with the ultimate nature and composition of
the constituents and not with their outward form.
A number of typical illustrations are included to show in detail the
structure of troostite. For these experiments a high grade tool steel
of about 0.90 per cent carbon was used. Small specimens weighing
about 10 grams were suitably heated in a vacuum furnace to a high
temperature and quenched in ice and brine solution. The hard-
ness of the specimens was quite uniform and averaged C-65 on the
Rockwell scale. From a study of the photographs it is quite apparent
that in a specimen of the kind, we may have not only the constituents
martensite and troostite, but in the troostite also the constituent pearl-
ite in the form of fan-shaped grains. From the work of Mathews,^
Bain,^ Enlund,'" and others, it is also apparent that some retained
austenite may be present. Hardened steel, therefore, is a complex
structural aggregate at best.
Conclusions
Nodular troostite appears to be an aggregate of ferrite and carbide,
and in the very early stages of formation its structure is on the border
of present methods of resolution. The condition of the ferrite and car-
bide in relation to each other is not stable; they tend to stratify form-
ing pearlite.
Troostitic nodules grow about a nucleus which may be an inclusion,
a void, a corner in a grain boundary or some other detail of structure.
The nodules contain fan-shaped radial grains.
The development of troostite results in a reorientation of the ferrite,
seemingly without particular reference to the old austenitic crystal-
lographic planes. Martensite does follow the old system of austenitic
planes.
The small fan-shaped grains in nodular troostite may persist as small
grains or they may undergo grain growth by union. It is a matter
seemingly dependent upon the thermal treatment of the specimen.
^ Honda, "Is the Direct Change from Austenite to Troostite Theoretically Possi-
ble? " Journal British Iron and Steel Institute, 1926, Vol. GXIY, No. 2.
* Mathews, "Austenite and Austenitic Steels," Trans. American histitute of Mining
and Metallurgical Engineers, Vol. LXXI, 1925; "Retained Austenite," British Iron
and Steel Institute, 1925, No. 11, Vol. CXII.
'Bain, "The Persistence of Austenite at Elevated Temperatures," Tratis.
American Society for Steel Treating, \'ol. VIII, 1925.
1° Enlund, "On the Structure of Quenched Carbon Steels," Journal British Iron
and Steel Institute, 1925, Vol. CXI, No. 1.
STRUCTURE AND NATURE OF TROOSTITE
107
Fig. 1 — Mag. 3230X. Fig. 1 is reproduced from the Journal of the Frank-
lin Institute and shows a typical troostitic nodule sectioned on a plane passing through
the center. The nodule has developed as a globular mass about a nucleus. Radial
grains have developed. These grains change from light to dark when the nodule is
revolved about the optical axis of the microscope. Therefoie it is clear that the
small fan-shaped grains are differently oriented. Where nodular troostite forms
regranulation must occur. Where changes in orientation take place, grain boundaries
must result. The structure of the troostite has not been fully resolved in this photo-
graph. Compare with others which follow.
108
BELL SYSTEM TECHNICAL JOURNAL
Fig. 2 — A Diagram. Fig. 2, also from the Journal of the Franklin Institute,
illustrates diagrammatically the mode of crystalline growth in a troostitic nodule.
STRUCTURE AND NATURE OF TROOSTITE
109
t
#
Fig. 3 — Mag. 3500X. Fig. 3 sliows the early stages in the formation of troos-
titic nodules. The background will be recognized as martensite. The dark particles
are troostite. The field is on the border of an area containing large well developed
nodules. This position in the specimen is one in which thermal conditions promoted
tne development of the needle structure but did not fully inhibit the development of
nodules. The very small dark particles are about five-millionths of an inch in diam-
eter. The larger ones are from about ten to twenty times larger.
no
BELL SYSTEM TECHNICAL JOURNAL
Fig. 4 — Mag. 3500X. Fig. 4 shows a somewhat later period in development of
troostite. The troostite appears to have formed along grain boundaries. The excess
constituent is clearly seen, and here and there a laminated structure, pearlite. Evi-
dently whatever the state of the carbide with reference to the iron in troostite —
wnethei contained in solid solution as first formed or whether disposed as a fine aggre-
gate with the iron — the condition must be very unstable, otherwise evidences of finely
laminated pearlite would be lacking.
STRUCTURE AND NATURE OF TROOSTITE
111
Fig. 5 — Mag. 3500X. Fig. 5 is of a small but well developed nodule showing
four fan-shaped grains about a nucleus of growth. Some excess constituent has
appeared but only the very early stages in the process of stratification to pearlite
are visible. It is apparent, however, that the nodule is not composed of a solid so-
lution. The nodule is about eight ten-thousandths of an inch in diameter.
112
BELL SYSTEM TECHNICAL JOURNAL
Fig. 6 — Mag. 3500X. Fig. 6 is of a larger nodule than illustrated in Fig. 5
and shows obviously a little later stage in the process of stratification. This nodule is
about the same as the one illustrated in Fig. 1 except that the structure has been
resolved.
STRUCTURE AND NATURE OF T ROOST ITE
113
Fig. 7— Mag. 3500X.
than that shown in Fig.
throughout.
Fig. 7 illustrates a condition somewhat further advanced
6. Stratification is well advanced and is plainly shown
114
BELL SYSTEM TECHNICAL JOURNAL
Fig. 8 — Mag. 3500X. Fig. 8 shows a troostitic development which had formed
along an old austenitic grain boundary. The excess constituent is starting to clear
at the grain boundary. Well developed pearlite is revealed. The large light-
colored grain covering the center of the field is just starting to break up into two con-
stituents. Formerly grains of this kind, becauseof lack of resolution, were thought to
be in all probability solid solution grains. These grains must represent a state very
nearly that of freshly formed troostite.
STRUCTURE AND NATURE OF TROOSTITE
115
Figs. 9 and 10— Mag. 3500X. Figs, 9 and 10 show two typical nodules along
grain boundaries or crystallographic planes. The excess constituent, the radial
grains, some practically irresolvable and others fully resolved, and the center of
growth (in Fig. 9) are clearly revealed.
116
BELL SYSTEM TECHNICAL JOURNAL
Fig. 10— Mag. 3500X.
STRUCTURE AND NATURE OF TROOSTITE
117
Fig. 11 — Mag. 3500X. Fig. 11 illustrates the wide range in structure to be
found in hardened steel. The background is martensite whicn contains a troostitic
nodule. One grain of the nodule is fully laminated pearlite. The other grains are
in all stages of stratification.
118
BELL SYSTEM TECHNICAL JOURNAL
Fig. 12 — Mag. 3500X. Fig. 12 illustrates the condition which prevails when
growing nodules interfere and the whole area is troostite. Small fan-shaped grains
are found in the different stages of stratification.
STRUCTURE AND NATURE OF TROOSTITE
119
Fig. 13 — Mag. 3500X. Fig. 13 is of a field similar to Fig. 12 except that a
more advanced stage in stratification is present. This photograph is reproduced from
the Proceedings of the Inlertiational Congress for Testing Materials.
120
BELL SYSTEM TECHNICAL JOURNAL
v^ 'Ciia^Si'^tj^lJ'lL ♦'"'V
\,P7' i/%»»
:/V^
■ \\ *"%• * * *
.*&'
t .
••♦♦■
v^
% ^
Fig. 14 — Mag. 3500X. Fig. 14 is of a specimen which was quenched and then
drawn for ten minutes at 650° C. The outHne of a troostitic notkile is clearly marked
by globular carbide particles. The hardness of the specimen was C-28 after temper-
ing.
Radio Broadcasting Transmitters and Related Trans-
mission Phenomena ^
By EDWARD L. NELSON
This paper is a brief discussion of recent developments in American
practice concerning radio broadcasting transmitters. Descriptive material
and photographs pertaining to several new commercial transmitting equip-
ments are included. Reference is also made to the more important aspects
of the related transmission problem. On account of the scope of the
subject, the treatment is necessarily superficial, but it may serve to indi-
cate the present status of the transmitter art and its relative position with
respect to the industry as a whole. A short bibliography containing some
of the more important recent contributions to the subject is attached as
an appendix, to which reference may be had for more detailed information.
Radio Transmitters
THE radio transmitter is essentially a focal point in the present-
day broadcasting system, since upon it the program circuits
converge and from it the radio distribution network emanates. For
this reason, the requirements which have been imposed on trans-
mitting apparatus are extremely rigorous, and all phases of trans-
mitter performance have been subjected to the most careful scrutiny.
Under these stimulating influences, the last few years have brought
about some very noteworthy advances in this portion of the broad-
casting field.
As long as music and entertainment continue to hold a prominent
place on broadcasting programs, fidelity of transmission will probably
remain the most sought-for characteristic, not only for the radio
transmitter itself, but for all of the apparatus units in the system.
A very high standard of performance has now been attained in this
respect. Fig. 1, below, shows the overall frequency-response charac-
teristic of a new type 50-kw. equipment, the first of which has gone
into service at one of the leading American broadcasting stations
within the past few months. It will be noted that this characteristic
is substantially flat between 30 and 10,000 cycles. The greatest
departure from the horizontal line which is the ideal characteristic
is less than 1 db. The frequency discrimination which this represents
is of such a low order that it probably could not be detected in ordi-
nary listening tests, even by a skilled musician.
Another recognized prerequisite to a high degree of fidelity is exact
proportionality between audio input and sideband output. Increased
\Read before the World Engineering Congress, Tokio, Japan, October, 1929;
Proceedings of Institute of Radio Engineers, November, 1929.
121
122
BELL SYSTEM TECHNICAL JOURNAL
emphasis on accurate reproduction has recently led to the introduc-
tion of improved technique for checking this important characteristic
under dynamic conditions. The method employed consists of im-
pressing a pure sine-wave input on the transmitter at various fre-
quencies throughout the audio range and subjecting the output of
-si \ I LU I I I I Mill
30
100 1000
FREQUENCY IN CYCLE.S PER SECOND
10,000
Fig. 1 — -Frequency-response characteristic of Western Electric 7-A (50-kw) radio
transmitter.
a Straight-line rectifier to harmonic analysis. One type of harmonic
analyzer which has been used with excellent results is that due to
Wegel and Moore. ^ This device produces a photographic record, an
example of which is shown in Fig. 2. Measurements of this type are
2ND HARMONIC -10.5%
3RD HARM0NIC= 9.0%
75 100 125
FREQUENCY IN CYCLES PER SECOND
200
Fig. 2-
-Harmonic analyzer graph indicating overloading (2nd harmonic, 10.5 per
cent; 3rd harmonic, 9 per cent).
of particular importance under present conditions since current Ameri-
can practice is tending toward the extensive use of transmitters in
which modulation is accomplished at relatively low power levels and
the required power output is obtained by means of subsequent stages
amplifying modulated radio-frequency power. Such amplifying stages
2 R. L. Wegel and C. R. Moore, "An Electrical Frequency Analyzer," Bell Syst.
Tech. Jour., p. 299-323, April, 1924.
RADIO BROADCASTING TRANSMITTERS 123
are susceptible of serious amplitude distortion unless the conditions
under which the tubes operate (direct plate and grid voltages and
impedance of the connected load) are carefully predetermined. For
this purpose, the harmonic analyzer has proved to be invaluable.
Through its use, commercial transmitters are now available in which,
at the working upper limit of modulation, the harmonics generated
are not greater than 5 per cent.
The attainment of such high standards for fidelity leaves little
opportunity for progress, and it is improbable that significant ad-
vances in this direction will be made in the near future. Accordingly,
in continuing their efforts toward further improvements in broad-
casting service, transmitter engineers have been led to divert their
attention to the problem of rendering less conspicuous and objection-
able the background of noise and interference which, in the past, has
so seriously impaired the artistic effect of programs except in the
immediate vicinity of transmitting stations. This is the principal
justification for the present movement toward higher power outputs
for broadcasting stations. It has also resulted in increased emphasis
on the maintenance of a high average degree of modulation, a develop-
ment which is rapidly bringing about a very perceptible improvement
in general broadcasting conditions.
The degree of modulation of the carrier in a radio telephone trans-
mitter is a somewhat intangible factor which necessarily varies rapidly
through wide limits during the rendition of a program. With every
transmitter, however, there is a definite modulation limit which is a
characteristic of the design and which cannot be exceeded without
bringing about serious distortion. This limit is an important per-
formance index which, for lack of a better name, has been called
"modulation capability." The modulation capability of a trans-
mitter may be defined as the maximum degree of modulation (expressed
as a percentage) that is possible without appreciable distortion, em-
ploying a single-frequency sine-wave input and using a straight-line
rectifier coupled to the antenna in conjunction with an oscillograph
or harmonic analyzer to indicate the character of the output.
For a number of reasons, some technical and some economic, many
of the broadcasting transmitters in use have been so constructed that
overloading of the audio power stage with consequent distortion
occurs whenever the degree of modulation exceeds approximately
50 per cent. The usual practice in placing broadcasting transmitters
into service consists of determining, by means of a suitable vacuum-
tube voltmeter or other "volume indicator," the audio level at the
input of the set for which distortion becomes evident. The average
124 BELL SYSTEM TECHNICAL JOURNAL
operating level is then established at a suitably lower value, frequently
6-10 db. Recently, by modulating at low power levels, transmitters
have been produced which are capable of 100 per cent modulation
without noteworthy distortion. It is obvious that, if a transmitter
of this latter type is employed and the same margin is observed in
determining the average audio input level, the resulting sidebands will
be twice the amplitude of those produced by a transmitter whose
modulation capability is only 50 per cent. To produce equivalent
sidebands with a transmitter capable of but 50 per cent modulation
requires that the carrier amplitude be doubled or the power output
multiplied by four. In other words, insofar as signal-to-noise ratio is
concerned, which is the factor that usually determines the coverage
of a broadcasting station, the increase in modulation capability men-
tioned results in an improvement that in the older type of apparatus
could only be had by quadrupling the rated output of the transmitter.
From the coverage standpoint, the night range of a given station can
be approximately doubled in this manner. Since this is accomplished
without increase in the carrier power, the outlying zone in which the
station may produce serious beatnote interference with others as-
signed to the same channel will not be extended. Accordingly, the
use of transmitters capable of a high degree of modulation is a notable
contribution toward the more effective utilization of the medium,
which is the outstanding technical problem in American broadcasting
today.
. Another important factor, from the standpoint of intensive devel-
opment of the available frequency band, is frequency maintenance.
In a system involving so many stations as are now operating in the
United States, accurate maintenance of the assigned frequencies pre-
sents a very difficult problem. The maximum deviation permitted
by the existing government regulations (± 500 cycles) is somewhat
beyond the capabilities of the ordinary wavemeter and difficulty has
been experienced in obtaining a satisfactory substitute. In the ab-
sence of adequate frequency control apparatus, very serious beatnote
interference has been of frequent occurrence. During the past year,
however, considerable improvement has been brought about by the
extensive adoption of piezo-electric reference oscillators and automatic
piezo-electric control. Equipment for the latter purpose capable of
a relatively high standard of performance is now being offered com-
mercially and it is probable that apparatus of this type will be installed
in the near future by the majority of stations. Its use is expected
to avoid entirely heterodyne interference on the "cleared" channels,
where the beatnotes are those produced between the carriers of sta-
RADIO BROADCASTING TRANSMITTERS 125
tions having adjoining frequency assignments. There is also reason
to beheve that the general adoption of such apparatus will materially
improve conditions on the "shared" channels, each of which is occu-
pied by several stations located at suitable distances, provided the
assigned frequencies can be maintained with sufficient accuracy to
preclude the reproduction of audible beats or other objectionable
interference effects.
This problem of "synchronization," or preferably "common fre-
quency operation," is beginning to receive considerable attention from
all factors in the broadcasting industry. It promises important
contributions in at least two directions:
(1) Improvements in the coverage of a common service area by two
or more stations all broadcasting the same program;
(2) The attainment of minimum geographical spacings between sta-
tions operating on the same nominal frequency and broad-
casting different programs.
The degree of frequency maintenance required for these two cases
is apparently quite different. For case (1), the evidence indicates
that very rigorous requirements must prevail. The most successful
operations of this type have employed wire lines connecting the sta-
tions for the transmission of a base frequency from which the carriers
were derived by means of harmonic generators. For case (2), how-
ever, there is reason to believe that comparatively wide limits will
suffice.
Expeiience has shown that if the entertainment value of a pro-
gram is not to be seriously impaired by interference, the ratio of
wanted to unwanted carrier at the receiving point, in terms of field
intensity, must be at least 100 : 1. From a relative interference
standpoint, the significant factors are the wanted sidebands, the
unwanted sidebands and the unwanted carrier, each of which produces
a component in the detector output by interaction with the wanted
carrier. With equal modulation at both stations, which is one of
the conditions assumed, the ratio of the audio components due to
the sidebands will, in general, be approximately the same as that
between the carriers, or 100 : 1, representing a difference in level of
40 db. Due to the frequency difference between carriers, demodu-
lation of one of the unwanted sidebands will result in the original
signal with each of its elements shifted upward in pitch by an amount
corresponding to this difference, while the other sideband will produce
a signal which is similarly displaced in the reverse direction. The
interfering signal mav be badlv garbled, therefore, but its disturbing
9
126 BELL SYSTEM TECHNICAL JOURNAL
efifects insofar as enjoyment of the program is concerned will be sub-
stantially unaffected. The beatnote, which results from the inter-
action of the unwanted and wanted carriers, will be 6-10 db above
this sideband interference level if average practice, as previously
described, is followed. From this analysis, it appears that if the
beatnote can be held to a value below the lowest frequency which it
is desired to transmit and if one of the circuit elements of the repro-
ducing system can be designed to provide some 10 db discrimination
against the beat frequency, interference due to the latter can be so
subordinated that the service areas of the stations involved will be
defined by the limiting condition assumed for sideband interference,
or a 100 : 1 ratio between carrier field intensities. Under these cir-
cumstances, no beatnote interference will be experienced in those
areas where reasonably good service can be given. In adjoining
regions, where the carrier ratio is less than 100 : 1, beatnote inter-
ference may continue to be observed but is of no importance since
satisfactory reception in such areas is precluded by the sideband
interference.
To meet the requirements outlined, it is probable that ultimately
frequencies will have to be maintained to approximately 10 cycles,
which would result in a maximum beatnote near the lower limit of
aural frequency response. Such precision seems hardly necessary,
however, under the conditions existing at the present moment. Al-
most all loud speakers now commercially available discriminate not-
ably against frequencies below 100 cycles. A material improvement
in beatnote conditions could probably be brought about, therefore,
by the adoption of automatic control apparatus capable of main-
taining the assigned frequencies to ± 50 cycles. Such performance
is within the capabilities of the piezo-electric apparatus now available.
Under the circumstances it is expected that considerable progress
will be made in this direction during the coming year.
The foregoing considerations lead to the formulation of an impor-
tant system requirement affecting receiving apparatus, which in this
case includes both the radio receiver proper and the loud speaker.
In a system involving a relatively large number of stations assigned to
cleared and shared channels at 10-kc. intervals, such as exists in the
United States, beatnote interference in the form of components at
approximately zero cycles and at 10 kc. is an inherent characteristic.
If a maximum frequency deviation of ± 10 cycles is accepted as the
ultimate limit, in order to avoid such interference the receiving appa-
ratus must be so designed that at frequencies below 20 cycles and
above 9,980 cycles there will be introduced sufficient attenuation to
RADIO BROADCASTING TRANSMITTERS 127
suppress effectively the beatnotes likely to be encountered under any
practical operating condition. Developed in this manner the propo-
sition is more or less self-evident, but due to the rapidity with which
the audio spectrum of broadcasting apparatus is being extended,
some emphasis on the matter seems desirable.
Still another factor of importance from a system standpoint is
control of radio harmonics. Spurious radiation of all types is inimical
to intensive development and must be avoided. The harmonic prob-
lem presents unusual difficulties since efficiency requires that the
tubes in the final power-amplifier stage be used in such a manner that
relatively large harmonic voltages are impressed on the output circuit,
yet the harmonic power radiated must be held to an extremely small
amount. A measure of the purity of wave form required may be
gained from the fact that a 5-kw. transmitter operating on a good
antenna is capable of establishing an electromagnetic field of approxi-
mately 0.5 V per meter at a distance of one mile. Under the circum-
stances, a harmonic of 0.1 per cent represents a field intensity of
500 ^v per meter at the same distance. Acceptable service in many
areas is being obtained with field intensities of this order of magnitude.
To bring the interfering field down to the static level would probably
require reduction of harmonics to 0.01 per cent or less. From an
apparatus standpoint, such performance represents a very difficult
problem and it is questionable if it can be justified at the present
time. Practice on this point is still in a state of flux, but there is
reason to believe that some intermediate value, such as 0.05 per cent,
will prove to be the proper solution, and will be applied to all broad-
casting stations in the near future.
One circumstance that has undoubtedly contributed to the delay
in formulating definite requirements concerning the control of har-
monics has been the difficulty of obtaining suitable apparatus for
the evaluation of such components in quantitative terms. Field
strength measuring sets have recently been made commercially avail-
able, however, which are capable of covering the necessary range in
frequency and intensity. A photograph of one of these sets is shown
in Fig. 3. It consists essentially of a sensitive, stable superheterodyne
receiver incorporating a calibrated attenuator at the input of the
intermediate-frequency amplifier and a supplementary radio-frequency
oscillator from which a voltage of the frequency of the station under
measurement can be introduced in the antenna circuit. The oper-
ating characteristics of such an instrument have been described by
Friis and Bruce.^ By means of a series of removable loops and coils,
*H. T. Friis and E. Bruce, "A Radio Field-Strength Measuring System for
Frequencies up to Forty Megacycles," Proc. I. R. E., 14, 507-519; August, 1926.
128
BELL SYSTEM TECHNICAL JOURNAL
the set shown is capable of measuring field strengths ranging from
approximately 0.01 to 7,000 mv per meter throughout the hand 250
to 6,000 kc. Apparatus of this type is now in use by the radio inspec-
tion division of the Department of Commerce.
Fig. 3 — -Commercial field-strength measuring set. Range: 250-6,000 kc, 0.01-7,000
mv' per meter.
In the light of this discussion of present trends in transmitter devel-
opment, a brief description of some recent transmitting equipments
may be of interest. A particularly noteworthy example of current
practice is the 50-kw. Western Electric transmitter, one of which
has been placed in service within the past few months by the Crosley
Radio Corporation at Mason, Ohio. Views of this equipment are
shown in Figs. 4, 5, 6, 7, and 8. The transmitter proper is shown
RADIO BROADCASTING TRANSMITTERS
129
in Fig. 4. As will be seen, it consists of seven panel units with a
screen enclosure in the rear. The first unit on the left is the oscillator-
modulator. This is essentially a low-power transmitter capable of an
output of 50 watts and 100 per cent modulation. It is followed by
three push-pull stages amplifying modulated radio-frequency power.
The first power-amplifier stage, which employs two 250-watt tubes,
occupies the second unit. The tubes for the second power stage,
which are water cooled, and the associated tuned output circuit are
contained in the third and fourth units, respectively. The final power
Fig. 4 — Western Electric 7-A (50-kw) radio transmitter.
Oscillator-amplifier assembly.
stage, incorporating six water-cooled tubes each capable of a peak
output of approximately 40 kw., occupies the fifth unit. The last
two panels constitute the front of an electrically screened enclosure
housing the output circuits for this latter stage. All of the panels
are aluminum covered with several coats of black lacquer grained by
rubbing with abrasive paper. A full complement of meters is pro-
vided, the cases of which are either grounded or mounted behind
glass for the protection of the operating personnel. In designing the
equipment, special consideration has been given to safety. Access
to the apparatus in the rear of the panels can be had only through
the door on the left which is held closed by a bolt operated by the
hand wheel shown. The rotation of this wheel opens the transmitter
control circuits putting the equipment out of operation. It then
grounds the high-voltage supply busses and finally withdraws the
bolt. As an additional precaution a manually operated disconnect
switch for the main power supply is provided just inside the gate
which can be opened on entering. Access to some of the tubes is
had by opening the glass windows in the various panels, but these are
130
BELL SYSTEM TECHNICAL JOURNAL
secured by electrically operated latches unless the wheel is in the
grounded position. Door switches are provided in the control circuits
which prevent the transmitter from being placed in operation unless
all doors and windows are closed.
The power panel and rectifier assembly is shown in Fig. 5. The
general arrangement corresponds to that of the transmitter proper
and similar safety features are provided. In the power panel, which
is on the left, are centralized the necessary power distribution and
control facilities. The equipment requires a 3-phase input of approxi-
Fig. 5 — -Power panel and rectifier assembly for SO-kw radio transmitter.
mately 250 kw. at 440 volts. The control arrangement is such that
the transmitter can be started and stopped by means of a single set
of push buttons, the various circuits being energized in proper sequence
by means of suitable relays and contactors. The central unit is a
three-phase half- wave rectifier supplying power at 1,600 volts to the
plates of the air-cooled tubes. The six-tube rectifier on the right
supplies plate power at 17,000 volts for the water-cooled tubes. The
filament and plate transformers and smoothing filter for the latter
are located in the power room on the floor below. The filter consists
of two units, one for each side of the push-pull circuit, employing a
6-/if condenser and a 12-henry inductance. Two 24-volt, 550-ampere
direct generators (one a spare) supply power to the filament circuits.
RADIO BROADCASTING TRANSMITTERS
131
These machines are slot wound and employ composition brushes, a
filter consisting of a 1-mh. inductance and four 1,000-juf electrolytic
condensers being used to suppress commutator and slot ripples. Grid
bias voltages are obtained from a 2-kw., 300-volt unit, which is also
installed in duplicate. The only other rotating apparatus is that
associated with the water-cooling system. The tubes are cooled by
means of distilled water which is conducted to the anodes of the
amplifier tubes through insulating hose coils. The total heat trans-
Fig. 6 — Antenna coupling and tuning unit for SO-kvv radio transmitter.
ferred by the cooling water is approximately 175 kw. A flow of 75
gallons per minute is maintained. Four 56-in. by 58-in. radiator units
are employed, each consisting of a bank of copper tubes with spiral
fins over which air is blown by a 37-in. fan. Ample radiator capacity
is provided to maintain the water below 180 deg. F. under all atmos-
pheric conditions.
To promote antenna efficiency and to reduce the intensity of the
electric field in the station building, the equipment is arranged to
deliver its output to the antenna through a radio-frequency trans-
mission line approximately 500 ft. long. The line is balanced to
ground and is designed for a characteristic impedance of 600 ohms.
The antenna coupling and tuning unit is shown in Fig. 6. It is in-
132
BELL SYSTEM TECHNICAL JOURNAL
tended for installation in a small building with a grounded copper
roof located at the base of the antenna downlead. It consists of two
tuned circuits, each housed in separate shielded compartments. In
the photograph the doors and two of the screen panels have been
removed to show the interior arrangement. The line is terminated
by the parallel tuned circuit on the left which is inductively coupled
to the antenna circuit to preserve an approximate balance to ground.
The antenna is tuned by means of the series condenser and coil shown
on the right. Accurate adjustment of the inductance of the coil is
Fig. 7 — Artificial antenna for 50-kw radio transmitter.
provided for by means of a short-circuited single-turn secondary
which is located within the coil and arranged so that it can be rotated
through approximately 90 deg. by the motor mounted on the floor
beneath. The latter may be controlled from the operating room by
a reversing switch placed on the right-hand panel of the transmitter
assembly. A polyphase position indicator is provided to indicate the
angle and movement of the secondary. The direct-current circuit of
the thermal ammeter in the antenna circuit is also carried back to a
bracket-mounted instrument on the end of the transmitter. These
facilities permit the antenna tuning to be checked and adjustments
RADIO BROADCASTING TRANSMITTERS
133
made to compensate for minor variations in antenna conditions with-
out leaving the operating room.
Another feature of interest is the artificial antenna shown in Fig. 7.
This unit is essentially a 600-ohm non-inductive resistance capable of
dissipating approximately 75 kw. which can be connected to the
output circuit of the final power amplifier stage in place of the trans-
mission line. The heat dissipating elements consist of a series of
woven wire grids mounted in the units at the top of the framework.
The resistance of these grids is substantially independent of frequency,
but the combination presents a slight inductive reactance which is
compensated for by means of the condenser and coil combination
shown. These elements are inserted into the circuit symmetrically
Fig. 8 — Piezo-electric crystal mounting and temperature control apparatus.
in order to maintain an approximate balance to ground. The struc-
ture is completely shielded and is fitted with safety door and ground-
ing switches similar to those already described.
The piezo-electric crystal mounting and temperature-control appa-
ratus which is a part of the oscillator-modulator unit is shown in
Fig. 8, dismantled to facilitate inspection. The quartz plates em-
ployed are approximately one and a quarter inches square and are
cut parallel to one of the faces of the natural rock crystal. This plate
is mounted between two lapped metal plates and covered with a
porcelain cap carrying a terminal to which the upper electrode is
connected by means of a short section of metal foil. The mounted
crystal is supported by a brass block, through the center of which
extends a spiral bimetallic thermostat. The top of the block is also
lapped and the crystal mounting is secured to it by means of the four
134
BELL SYSTEM TECHNICAL JOURNAL
springs shown. The heating element consists of a winding of re-
sistance wire inserted in the block concentric with the thermostat.
The assembly is mounted in a thermally insulated box, shown on its
side in the photograph. Two of these units are provided, one located
on each side of the oscillator-modulator unit directly below the win-
dow. A detachable handle for adjusting the contacts of the thermo-
stat and a suitable thermometer extend through the box to the front
of the panel. The brass mounting block is provided with a groove
to receive the bulb of the thermometer. The thermostat does not
operate directly in the heater circuit but controls the grid bias of a
Fig. 9 — Simplified circuit schematic of 7-A (50-kw.) radio transmitter.
vacuum tube in the plate circuit of which a suitable relay is placed.
The quartz plates are ground to oscillate at the assigned frequency
at approximately 50 deg. C, and the final adjustment is made by
varying the operating temperature. The temperature coefficient of
the plates varies from 30 to 100 parts in a million per deg. C. The
degree of constancy attained necessarily depends on the diligence of
the operating personnel. With proper maintenance the maximum
deviation has been held to ± 30 cycles for long periods of time.
A simplified circuit schematic is shown in Fig. 9. Features of the
electrical design are the modulation system, the push-pull amplifier
stages with cross neutralization, the capacity coupling arrangement
used to facilitate control of parasitic oscillations, and the provisions
for the suppression of harmonics. The modulating amplifier is a
50-watt tube operating at 750 volts. The audio power stage employs
a 250-watt tube at 1 ,500 volts. In this manner, ample audio-frequency
voltage and power are provided to effect complete modulation without
RADIO BROADCASTING TRANSMITTERS
135
distortion in the audio tube. With so powerful an equipment, the
suppression of radio-frequency harmonics to a satisfactory degree
becomes a difficult problem. The push-pull circuits, capacity coup-
Fig. 10 — -Panel assembly for Western Electric 5-C (5-k;w.) radio transmitter.
ling, three tuned circuits in cascade, shielding of all coils, and the
two tuned shunts adjusted to the second harmonic which are con-
nected between each side of the transmission line and ground all
/fe_il ./L
Fig. 11 — Rear view of panels in 5-C radio transmitter.
contribute to superior performance in this respect. The amplitude of
the harmonics radiated, as determined by field strength measure-
ments, is less than 0.03 per cent.
A 5-kw. equipment of similar general design is shown in Figs. 10
136
BELL SYSTEM TECHNICAL JOURNAL
and 11. It consists of six units: A power panel, a 10,000-volt rectifier
for the water-cooled tubes, a piezo-electric oscillator unit, an inter-
mediate amplifier unit, a power amplifier unit employing two 10-kw.
tubes, and an output unit. An air-cooled transformer for the rectifier,
the associated filter, and an artificial antenna are assembled in a
Fig. 12 — -Western Electric 6-B (l-k\v.) radio transmitter.
screened enclosure in the rear of the panels. Three motor-generator
sets are provided to supply filament power, grid bias, and plate power
for the air-cooled tubes. A 3-phase power input of 30 kw. at 220
volts is required. The equipment is capable of fidelity in trans-
mission comparable with that of the 50-kw. unit. The amplitude of
the harmonics radiated is held to approximately 0.2 per cent.
A 1-kw. equipment of the same type is shown in Figs. 12 and 13.
It involves only two panels, a piezo-electric oscillator unit and an
amplifier unit. The final power stage employs a 4-kw. water-cooled
tube. Two motor generators are used, one supplying 24 volts and
RADIO BROADCASTING TRANSMITTERS
137
250 volts for filaments and grid bias, the other 2,000 volts and 4,000
volts for the plates of the air-cooled and water-cooled tubes, respec-
tively. A power input of 10 kw. is required.
Fig. 13 — Rear view of 6-B radio transmitter.
Radio Transmission Phenomena
Radio transmission phenomena in the broadcasting band have
been given considerable study, and the general nature of the efTects
likely to be encountered are fairly well understood. Important con-
tributions have been made by Bown and Gillett, by Bown, Martin,
and Potter, by Goldsmith, and by Espenschied.'' The second paper
referred to is particularly noteworthy on account of the insight which
it afTords into the complexities of the process of transmission and the
evidence which it presents concerning the injurious effects of fre-
quency modulation. The latter has not yet fully received the atten-
tion which it deserves; many otherwise well designed transmitters
* See attached list of references.
138 BELL SYSTEM TECHNICAL JOURNAL
are still in operation that are subject to frequency changes of the
order of ± 1,000 cycles during modulation. This condition is not
only conducive to impaired fidelity at moderately distant receiving
points, but it increases interference and precludes successful common
frequency operation. Fortunately, the use of automatic frequency
control apparatus in its present form is effective in minimizing this
efifect as well as in limiting frequency variations of much longer period.
It is probable, therefore, that with the more general use of automatic
piezo-electric control, this matter will rapidly cease to be a problem.
As might be expected, the attention being given to intensive devel-
opment has materially stimulated interest in transmission. There is
a very evident need for much information of a more quantitative
nature than is now available. Data concerning attenuation over
city and rural areas as a function of frequency, suitable separations
between stations of various powers operating on a common carrier
frequency, allowable distances between transmitting stations and
nearby populous communities, relative day and night ranges, relative
summer and winter ranges, time of the day and season of the year
at which the transition occurs, and other questions of a similar nature
have become of great practical importance. The problem is rendered
particularly difiicult by the range in climatic, topographic, and cul-
tural conditions which exist in the United States. Under the cir-
cumstances, there are excellent opportunities for important work in
this field.
A significant tendency disclosed by recent measurement work in
a number of city areas is public acceptance of and demand for field
intensities which a few years ago would have been considered objec-
tionably high. For some time it has been more or less generally
agreed that a field intensity of 10 mv. per meter would afford a satis-
factory high-grade broadcasting service. Recently, however, in spite
of increased eftectiveness due to higher degrees of modulation and in
spite of continued improvement in the sensitivity of commercial
receiving sets, stations establishing field strengths of 10-15 mv. per
meter have been greatly handicapped in competing with others capable
of producing 30-50 mv. per meter in the same areas. In several
densely populated districts measurements have disclosed field inten-
sities of 300-500 mv. per meter without any noteworthy number of
complaints provided the programs were of a high character. There
is little to indicate whether this tendency is the result of a decreased
interest in distant stations, a desire for higher standards in reproduc-
tion involving lower noise levels, or a combination of these factors
with others, but it is evidently a matter which must be given careful
consideration in engineering future installations.
RADIO BROADCASTING TRANSMITTERS 139
It is interesting to contrast this situation with that existing in
some of the large rural districts as exemplified by the recent survey of
conditions in the Middle West by Jansky.^ Here over large areas
acceptable service is being obtained with field strengths of 50 and
100 /iv per meter. Giving due consideration to the difference in noise
levels, which is undoubtedly a factor of great significance, such a dis-
crepancy can only be reconciled on the basis of a vast difference in
service standards. That such conditions will be allowed to continue
for any considerable period of time is very doubtful. This is further
evidence indicating that the movement toward more powerful stations
is technically sound.
One phase of the transmission problem which deserves increased
attention is antenna performance and design. It is an interesting
circumstance that while the accurate rating of broadcasting stations
is a matter of great practical concern to the industry, to date con-
sideration has been confined to the power delivered to the antenna.
Variations in the efficiency of the latter have been almost entirely
neglected in spite of the fact that, due to this cause, the power actually
radiated can be shown to vary through a range of four to one, or
greater. There is little doubt that stations should be rated, either
directly or indirectly, in terms of field intensity measurements. That
such a system of rating has not already been put into effect is probably
due to the lack of suitable measuring apparatus. With such equip-
ment now available, rapid progress in this direction is expected.
An interesting feature of current American practice with respect
to broadcasting antennas is a definite tendency toward the use of
higher supporting structures. For the past few years, most of the
towers erected have been from 150 to 225 ft. in height. Several of
the more recent stations are employing 300-ft. towers, and it is not
improbable that some 400-ft. structures will be put up in the near
future. Since the natural frequency of grounded steel towers of these
dimensions falls in the broadcasting band and may approximate the
assigned operating frequency, low-capacity porcelain insulators are
inserted at the base. The latter effect a considerable increase in the
natural frequency of the towers and preclude serious distortion in
the field intensity pattern due to heavy induced currents in the steel.
The antennas themselves are of such dimensions that the current
antinode is positioned well up on the vertical section. The effect is
to concentrate the radiated power along the ground plane and to
increase materially the field intensity in the local service area. Such
antenna systems promise a better economic balance between the in-
* See attached list of references.
140 BELL SYSTEM TECHNICAL JOURNAL
vestment for generating modulated radio-frequency power and that
for radiating it.
References
Ralph Bown, Carl R. Englund, and H. T. Frus. Radio Transmission Measure-
ments. Proc. I. R. E., 11, 115; April, 1923.
D. G. Little. KDKA Telephone Broadcasting Station of the Westinghouse Elec-
tric and Manufacturing Co., East Pittsburgh, Penna. Proc. I. R. E., 12,
255; June, 1924.
Ralph Bown .\nd G. D. Gillett. Distribution of Radio Waves from Broadcasting
Stations over City Districts. Proc. I. R. E., 12, 395; August, 1924.
Edward L. Nelson. Transmitting Equipment for Radio Telephone Broadcasting.
Proc. I. R. E., 12, 553; October, 1924.
Julius Weinberger. Broadcast Transmitting Stations of the Radio Corporation
of America. Proc. I. R. E., 12, 745; December, 1924.
Ralph Bown, DeLoss K. Martin, and R.\lph K. Potter. Some Studies in Radio
Broadcast Transmission. Proc. I. R. E., 14, 57; February, 1926.
Alfred N. Goldsmith. Reduction of Interference in Broadcast Reception. Proc.
I. R. E., 14, 575; October, 1926.
Lloyd Espenschied. Radio Broadcast Coverage in City Areas. Bell Syst. Tech.
Jour., VI, 117; January, 1927.
D. K. Martin, G. D. Gillett, and I. S. Bemis. Some Possibilities and Limita-
tions in Common Frequency Broadcasting. Proc. I. R. E., 15, 213; March,
1927.
Knox McIlw.vin .\nd W. S. Thompson. A Radio Field Strength Survey of Phila-
delphia. Proc. I. R. E., 16, 181; February, 1928.
I. F. Byrnes. Recent Developments of Low Power and Broadcasting Trans-
mitters. Proc. I. R. E., 16, 614; May, 1928.
P. P. EcKERSLEY. The Design and Distribution of Wireless Broadcasting Stations
for a National Service. Proc. Wireless Section, I. E. E., 3, 108; June, 1928.
H. M. O'Neill. Characteristics of Certain Broadcasting Antennas at the South
Schenectady Development Station. Proc. I. R. E., 16, 872; July, 1928.
S. W. Edwards and J. E. Brown. The Use of Radio Field Intensities as a Means
of Rating the Outputs of Radio Transmitters. Proc. I. R. E., 16, 1173;
September, 1928.
C. M. Jansky, Jr. Some Studies of Radio Broadcast Coverage in the Middle
West. Proc. I. R. E., 16, 1356; October, 1928.
Wire Line Systems for National Broadcasting ^
By A. B. CLARK
The interconnecting of radio broadcasting stations by special telephone
lines for the simultaneous broadcasting of radio programs began on a
commercial basis in 1923. Today well over 30,000 miles of program
transmission circuits are in use in the United States and transcontinental
broadcasts by means of such wire lines are a daily occurrence.
The paper first states the radio limitations which make wire lines neces-
sary for broadcast coverage of large nations. A map and data are given
showing the present broadcasting chains in the United States and indi-
cating the extent of their use. An explanation is given of why program
transmission circuits must have transmission characteristics materially
different from message telephone circuits and a brief discussion of some
of the important transmission characteristics of such circuits, including
particularly " frequency range " and " volume range." The present chains
in the United States which are made up almost entirely of open-wire cir-
cuits on a voice-frequency basis are briefly described. The manner in
which these chains are tested and the way control is exercised are also
indicated. To exercise this control requires an elaborate network of tele-
graph wires now aggregating over 40,000 miles and a corps of special men
over 300 in number.
WHAT we are here considering, as an important factor in pro-
moting national solidarity, is the tying together of a whole
nation so that a single broadcast will instantly reach even the most
remote points. Radio broadcasting stations (employing the more
generally used frequencies) are essentially local distribution centers
serving effectively points up to 50 miles (80 kilometers) or, in favor-
able cases, 100 miles (160 kilometers) or more from the radio trans-
mitter. For the larger nations it is evidently necessary to make
division into areas, locating a radio transmitter in each area for its
coverage, and then to provide a network of circuits connecting the
transmitters in the various areas with the point at which the broad-
cast originates. At the present time wire telephone systems are
employed almost exclusively for this national distribution of broad-
casts. It is the purpose of this paper to discuss the wire networks
which are now being provided in the United States by the Bell Tele-
phone System.
In the United States at the present time (January 15, 1929) pro-
grams are being regularly distributed over extensive wire networks
or " chains " as indicated on the map of Fig. 1.- The various chains
^ Presented before the World Engineering Congress at Tokio, Japan, October,
1929, Proc. of the I. R. E., November, 1929.
- This map has been revised to show the network chains as of September 1, 1929.
141
10
142 BELL SYSTEM TECHNICAL JOURNAL
are usually referred to by colors and are so designated on the map.
As a regular procedure most of these chains operate about six hours
each day. Following are the numbers of radio stations served by
each chain together with the lengths of telephone circuit involved.
(An additional chain which operates only one hour each week is not
included.)
Radio Telephone
Stations Circuit Miles
Red network 5 41 10,500 16,600 kilometers
Purple network 41 8,450 13,600
Blue network 12 3,650 5,900
Green network 8 3,600 5,800
Orange network 5 1,700 2,700
Brown network 3 450 700
Total 110 28,150 45,300
' See table on Fig. 1 for revised data as of September 1.
On occasions when events of particular importance take place,
several of the regular chains may be merged together and additional
circuits added so as to pick up programs from various parts of the
country. For example, on November 5, 1928, the evening before
the United States presidential election, the networks shown in Fig. 2
were in operation, about 85 radio stations being mcluded. At various
times during this evening, five separate programs were broadcast from
several different points in New York City; Palo Alto, California;
Little Rock, Arkansas; and Pittsburgh, Pennsylvania. The United
States was thus virtually one great auditorium, with listeners esti-
mated as no less than fifty million.
From the technical standpoint, program transmission circuits are,
of course, very different from message telephone circuits. In the
first place, message telephone circuits must be arranged so that to
and fro conversations can take place practically instantaneously.
Program transmission circuits on the contrary are single-direction
transmission circuits. They are, therefore, not complicated by prob-
lems of electrical echo, singing and the like, which are ever present
with long message telephone circuits. However, although free from
the problems of two-way working, the design and operation problems
of program transmission circuits are by no means easy as compared
with those of message telephone circuits. On the contrary, in many
respects, these problems are considerably more difficult, the reason
being that the requirement as to approach to absolute fidelity of
reproduction is much more severe than for message telephone circuits.
A frequency band width of 2,500 cycles furnishes, if properly util-
ized, a telephone circuit over which speech is transmitted very clearly
so that conversations may be easily carried on. This band is not
t^*
10500 1&900
11,100 n;8oo
:^20 ^900
1,700 ZTOO
4S0 700
27,300 44,000 _.,"
OTSHOWN.-THIS INVOLVES
p) WHICH Goes UP l/2 HOUR A WEEK ANO INVOLVES
40W N0RK4AL DIRECTION OF TRANSMISSION
ARE CONNECTED-
27^0 <4,0C»
ONE CHAIN( P PA J WHICH GCCS UP ONLY ONC HOUR * WtEK NOT SHOWN -THIS INVOLVES
30 STATIONS AND 5000 CIRCUIT MILES. ALSO ONC CHAlNlGOLDJWhICH GOCS UP t/f HOUR A WETK AJJO INVOLVES
35TATI0MS AND ItOO CIRCUIT MILES IS NOT SHOWN ARROWS SHOW NORMAL DlftTCriON or TRANSMISSION
•-INDICATES POINTS WHERE ONE OR MORE RADIO STATIONS AAE CONNECTefr
Fig. 1 — Routes of Bell System Program Network Chains in the United States as of September 1, 1929.
WIRE LINE SYSTEMS FOR NATIONAL BROADCASTING 143
■5fc ,«S>i $ a S 9 »
C
O
o
lU
3
c
Oh
00
OS
O
J3
tn
oi
c
o
bfl
c
a
to
O
a;
O
s
-4-J
CO
bfl
WIRE LINE SYSTEMS FOR NATIONAL BROADCASTING 143
c
o
c
'tn
00
O
M
Z
c
o
bo
c
a
CO
J*!
O
(U
s
O
Ui
Oh
s
CO
144 BELL SYSTEM TECHNICAL JOURNAL
adequate, however, for program transmission because of the different
character of the transmitted material. The bulk of present-day
broadcast programs consists of musical selections, including a fair
amount of high-grade material. To reproduce music, and particu-
larly high-grade music, in a pleasing manner calls for a materially
widened band. This wider band also gives a high degree of natural-
ness to speech which is particularly desirable when loudspeakers are
used for reception.
At the present time in the United States the frequency band which
is transmitted over the long distance program chains extends from
about 100 cycles to about 5,000 cycles. It is, of course, possible to
transmit an even wider band than this, although the cost of the
circuits will, of course, increase as the band is widened. In consider-
ing how wide the band should be, the complete system, including
pickup apparatus, wire transmission line, radio transmitters, radio
transmission paths through the ether, radio receiving apparatus and
loud speakers must be considered. It seems probable that as the art
progresses a band wider than the above will be found desirable. On
the wire line systems, development work is going forward looking
toward the possibility that such wider bands may be found desirable
in the future. At the lower frequencies, where most people consider
that improvement is particularly desirable, consideration is being
given to the possible extension of the band down to 50 cycles and
possibly lower. Consideration is also being given to the possible
addition of two or three thousand cycles to the top of the band.
In addition to this broad band transmission requirement, program
transmission circuits must be designed to handle wide ranges of vol-
ume, particularly for the transmission of musical programs. Much
of the enjoyment in listening to good music appears to come from
the ranges of volume, so that in order to deliver such musical programs
properly these ranges of volume must be preserved in large part at
least. At the present time the volume ranges are "compressed"
somewhat by adjustment of amplification under control of an operator
at the pickup point. This tends to make easier the radio trans-
mission problem as well as the wire transmission problem. The
range of volume which is now delivered, as read by a "volume indi-
cator" (a meter which roughly indicates the peaks), is of the order
of 30 decibels (3.4 nepers), which means that during the fortissimo
parts of programs the power which is transmitted is about 1,000 times
as great as it is during the pianissimo portions.
The designer of the wire circuits must be concerned lest during
those periods when the program power is strong, the program circuits
WIRE LINE SYSTEMS FOR NATIONAL BROADCASTING 145
produce an undue amount of disturbance in neighboring circuits
which may be transmitting other programs or telephone messages.
The designer is also concerned lest when the program power is weak
the programs be unduly interfered with by noise or crosstalk from
other circuits. He must particularly consider the noise and cross-
talk which may be heard during pauses in programs. During such
pauses it is very annoying to the listeners to hear a background of
noises of various sorts and it is essential that the listeners be unable
during such pauses to pick up intelligible speech from telephone
message circuits crosstalking into the program circuit.
At the present time generally satisfactory results are being obtained
in transmitting the volume range of about 30 decibels (3.4 nepers).
Considerably more must be done both in the radio and in the wire
systems, however, before there can be transmitted volume ranges
comparable with those put out by symphony orchestras, high-grade
artists, and the like.
Having indicated in a general way the requirements of program
transmission circuits, there will next be described the wire systems
which are now in use in the United States.
The present-day program transmission circuits in the United States
are "on a voice-frequency basis," which means that the waves trans-
mitted over the circuits are essentially copies of the sound waves
impinging on the microphones. Most of the circuits now being pro-
vided are carried by the familiar open wires, usually copper wires
0.165 inch (4 mm.) in diameter spaced about 1 foot (30 cm.) apart
on the crossarms. The transmission properties of an open-wire pair
without loading are well suited for program transmission purposes
since the distortion is comparatively small although it is far from
negligible. Spaced at intervals on these circuits, averaging roughly
150 miles (240 kilometers) apart, are one-way repeaters or amplify-
ing devices. Along with these amplifiers are other electrical devices
for counteracting the distortion introduced by the open-wire circuits,
incidental cables involved, etc. Other one-way repeaters are pro-
vided at the terminals of the circuit. Considerable technical refine-
ment is, of course, involved in the design of these amplifiers and of
the auxiliary apparatus associated therewith which cannot be gone
into here.
In setting up the program transmission circuits, an important part
of the work consists in making measurements at different single
frequencies within the band which it is desired to transmit over the
circuit. Before making such overall measurements, the amplifiers
and auxiliary apparatus are so adjusted locally as to compensate for
146
BELL SYSTEM TECHNICAL JOURNAL
the amount of distortion which theory and experience indicate should
be expected. Then, final adjustments are made by certain specially
provided adjustable parts in accordance with the overall measure-
ments. Such overall tests and adjustments are, in general, made
daily.
In setting up these circuits, another important consideration is that
each amplifier carry its proper load or, in telephone parlance, each
amplifier deliver to its associated line the proper output level. To
insure this, diagrams are prepared in advance, showing the desired
transmission levels at each repeater, a typical diagram being shown
in Fig. 3. In setting up the circuits, the repeater gains are first set
^ TO LOCAL
.BROADCASTING
STATION MONITOR'S
i __ _ AMPLIFIER
-Sdb LEVEL
3D LLVLL I
FROM LOCAiTI
CIRCUIT ~-f.l_,
KEY TO SYMBOLS
LONG DISTANCE CIRCUIT [r] FILTER
LOCAL CIRCUIT
^ AMPLIFIER
^ REPEATING COIL
^ LINE EQUALIZER
P TRANSMITTING INPUT EQUALIZER
[|] 600<-J ARTIFICIAL LINE
[E] LOCAL EQUALIZER
VOLUME INDICATOR
?
Fig. 3 — -Typical Circuit Layout and Transmission Level Diagram of Program
Network Circuits.
to values which theory and experience indicate should result in con-
ditions as shown in the prescribed transmission level diagram. Test-
ing current is then applied to the sending end of the circuit and sensi-
tive measuring devices are applied at the output of each repeater.
If the results of these measurements do not accord with the trans-
mission level diagram, suitable adjustments are then made.
In building up the large chains which tie together a considerable
number of radio transmitters, wire distributing centers are provided
at strategic points. Figure 4 shows the circuit layout of the various
chains which have been referred to and indicates in a general way how
noMO
5tx; TLe.WAjn
naw
tugene
^ ^
f^Odeburq
npo
SAN
2AN FKMC nCO,
CAUF
Y Y Y
SAN LWIS
oetspo
V^-
KFl
5] Y 7 BA«CR3
f RESMC
Molly wo
1> 0€
LOS ANGCLES. CALIF
ra
ff?
'*m
^^ ^oi la — 1 ~t^ — gcs=
WISNb lyVVThU
nOUlTQN .TEJAS
Fig. 4— Layout of Permanent and Recurring Program Network Circuits with Associated Repeater Equipment Stiown on the Basis of Normal Direction of TranBausaiun as of January 15, 1921>.
KtV
• IVmonent Circuit
■ Recurnnq Circuit
Repeater and dsaociaced equipmenc
1 Radio Scatlon
Oni Cham (PPA) is not shown. This ch«sin is
usedonlyonehuwi* awecK and consists of
3460 ctrcvit miles involving 19 ridio stations.
WIRE LINE SYSTEMS FOR NATIONAL BROADCASTING 147
the various chains are interconnected and arranged for switching at
certain distributing centers.
In the United States the largest distributing center is, naturally,
in New York City, since the bulk of the program material originates
at that point. At such a distributing center a special collection of
various forms of equipment is provided consisting of one-way ampli-
fiers, loud speakers, multifrequency oscillators, various forms of trans-
mission measuring devices and miscellaneous apparatus. The photo-
graph of Fig. 5 shows a portion of the program layout in the New
Fig. 5 — Portion of Program Apparatus Layout in New York Long Distance
Telephone Office as of January 15, 1929.
York long distance telephone office as of January 15, 1929. The
various bays at the left carry the line apparatus associated with
branches of various chains. In the rear are located the transmission
measuring apparatus and multifrequency oscillators. In the fore-
ground are the terminals of various telegraph order wires.
In transmitting programs over a wire network, as has been pointed
out above, it is important that the volume range be held within
WIRE LINE SYSTEMS FOR NATIONAL BROADCASTING 147
the various chains are interconnected and arranged for switching at
certain distributing centers.
In the United States the largest distributing center is, naturally,
in New York City, since the bulk of the program material originates
at that point. At such a distributing center a special collection of
various forms of equipment is provided consisting of one-way ampli-
fiers, loud speakers, multifrequency oscillators, various forms of trans-
mission measuring devices and miscellaneous apparatus. The photo-
graph of Fig. 5 shows a portion of the program layout in the New
Fig. 5 — Portion of Program Apparatus Layout in New York Long Distance
Telephone Office as of January 15, 1929.
York long distance telephone office as of Januar}^ 15, 1929. The
various bays at the left carry the line apparatus associated with
branches of various chains. In the rear are located the transmission
measuring apparatus and multifrequency oscillators. In the fore-
ground are the terminals of various telegraph order wires.
In transmitting programs over a wire network, as has been pointed
out above, it is important that the volume range be held within
148 BELL SYSTEM TECHNICAL JOURNAL
proper limits. It is one of the obligations of the one who " picks up "
the program to hold his range of volume between proper limits. At
the central distributing point those in charge of the wire circuits
usually find it desirable to make checks from time to time to insure
that the proper range of volume is maintained. This checkup is
made by means of a device known as a "volume indicator" similar
to the one which the program supplier uses for purposes of regulating
his volume range. Other volume indicators are provided at various
strategic points in the wire network in order to insure that the proper
range of volume is reaching these points. In addition to regularly
making these observations by means of volume indicators, loud-
speaker monitoring observations are continually made at practically
all repeater points.
The results of these observations are transmitted back to the con-
trol points periodically by means of telegraph order wires so that the
control operator knows at all times the condition of transmission at
every point in his territory.
With the network chains grown to such vast proportions as indi-
cated in Figs. 1 and 4, it is essential that the system for controlling
the networks be such that all points involved be in instant commu-
nication with certain designated control points. To accomplish this,
the United States has been divided into four areas, each area of which
is under the control of a distributing center or control station. The
four control stations in the United States at present (January 15,
1929) are. New York covering the eastern section, Chicago the western
section, Cincinnati the southern section, and San Francisco the Pacific
Coast section. Each of these control points is connected to ever>-
repeater point in its area by means of telegraph order wires and in
addition is connected to every radio station in the area served by
the networks under its control. The various control points are also
connected together by means of order wires and arrangements are
provided so that New York can be placed in communication with any
of the radio stations in the United States which are served by the
chains. The total telegraph wire mileage employed for this service
is now approximately 43,000 miles (70,000 kilometers).
A large corps of specially trained telephone men is needed to properly
supervise the transmission performance of the chains as well as to
take care of the switching and general coordination work involved.
At present, about 300 men are employed in the United States for
this service, these men, of course, being in addition to those who care
for the regular wire and equipment maintenance.
WIRE LINE SYSTEMS FOR NATIONAL BROADCASTING 149
Acknowledgment
Acknowledgment is made to Mr. H. S. Hamilton for considerable
assistance in connection with the preparation of the text and par-
ticularly of the drawings, and to Mr. G. S. Bibbins and Mr. H. C.
Read for furnishing most of the statistical data.
Notes on the Heaviside Operational Calculus
By JOHN R. CARSON
This paper briefly discusses the following topics: (1) the asymptotic solu-
tion of operational equations; (2) Bromwich's formulation of the Heaviside
problem, and its relation to the classical Fourier integral; and (3) the
existence of solutions of the operational equation. The paper closes with
some general remarks on the interpretation of the operator and the opera-
tional equation, emphasizing the purely symbolic character of the latter.
THE large amount of work done in the past thirteen years, start-
ing with important papers by Bromwich ^ and K. W. Wagner,^
has served to remove whatever mystery may have surrounded the
Heaviside operator, and has placed his operational calculus on a quite
secure and logical foundation. However, certain phases of the prob-
lem still do not appear to the writer to have as clear or adequate
treatment as perhaps might be desired ; these it is the object of the
present paper to discuss. The topics dealt with are (1) the asymp-
totic solution of operational equations; (2) Bromwich's very important
formula and its relation to the classical Fouiier integral; and (3) the
existence of solutions of the operational equation.
In the following it will be assumed that the reader has a general
acquaintance with the Heaviside operational calculus as well as the
Fourier integral, but a brief sketch of the former may not be out of
place. It will be recalled that the Heaviside processes were originally
developed in connection with the solution of electrical problems:'
more precisely, the determination of the oscillations of a linearly
connected system specified by a set of linear differential equations
with constant coefficients or a partial differential equation of the type
of the wave equation. This system is supposed to be in a state of
equilibrium at reference time / = 0, when it is suddenly acted upon
by a 'unit' force (zero before, unity after time / = 0) ; the subsequent
behavior of the system is required. In the solution of this problem,
Heaviside's first step was the purely formal and symbolic one of
replacing the differential operator d/dt by the symbol p, thereby
1" Normal Coordinates in Dynamical Systems," Proc. Lond. Math. Soc. (2),
15, 1916.
2"Uber eine Formel von Heaviside zur Berechnung von Einschaltvorgange,"
Archiv. Elektrotechnik, Vol. 4, 1916.
^ Since this paper is addressed largely to physicists and engineers, we shall employ
to some extent the language of circuit theory rather than pure mathematics; no loss
of essential generality is involved.
150
NOTES ON THE HEAVISIDE OPERATIONAL CALCULUS 151
reducing the differential equations to an algebraic form, the formal
solution of which we shall write
Here h = hit) is the variable with whose determination we are con-
cerned and H{p) is the Heaviside function, derived as stated from
the differential equations of the problem. This equation is as yet
purely symbolic, and its conversion into an explicit solution for h,
as a function of t, constitutes the Heaviside problem.
Bromwich ^ formulates the problem as the infinite integral
hit) =^. TWrr^^P- (2)
The writer's formulation of the problem is, that h is uniquely
determined by the integral equation ^
^W^"'* = WW) <'^
This equation is valid for all values of p, for which its real part is
greater than some finite constant c; c must be at least large enough
to make the infinite integral converge. In the majority of physical
problems this constant may be taken as 0; in some, however, the
equation is valid only when c is greater than some finite constant.
The equivalence of (2) and (3) is very easily established in a num-
ber of ways; perhaps the simplest is to show, following March,^ that
(2) is the formal solution of (3). Either can be deduced from the
other. The Bromwich solution can, of course, be derived directly
from the Heaviside problem, as shown below.
I
One of the most interesting and perhaps the least generally under-
stood of Heaviside's methods of solving the operational equation is
the process whereby he derives a series solution, usually divergent
and asymptotic, in inverse fractional powers of t. What I have termed
the Heaviside Rule ^ for deriving this type of solution may be formu-
lated as follows:
* "The Heaviside Operational Calculus," B. S. T. J., 1922; Bulletin Amer. Math.
Soc, 1926.
*"The Heaviside Operational Calculus," Bulletin Amer. Math. Soc, 1927.
* In terming this process the Heaviside Rule I do not in any sense imply that
Heaviside himself would have applied it incorrectly. In fact in one case he adds
an extra term which contributes to numerical accuracy although the series itself is
152 BELL SYSTEM TECHNICAL JOURNAL
If the operational equation h — \/II{p) admits of formal series
expansion in the form
h = aa -\- aiVp + a^p + a^p^p + a^p- + . . . , (4)
a solution, usually divergent and asymptotic, results from discarding the
/— ^" 1
terms in integral powers of p, and replacing />"\/> by -7— — -^ , whence
;.'^ao + |ax + a3|^ + a.|^,+ ---}-i=- (5)
As stated in a forthcoming paper, this divergent series is a true
asymptotic expansion, as defined by Poincare, if and only if, the
singularities in 1/H(p) all lie to the left of the imaginary axis in the
complex plane. Otherwise the series may require the addition of
an extra term or factor, or even be quite meaningless.
An excellent illustration of the preceding principle is furnished by
the operational equation,
V/> + X
For convenience and without loss of essential generality we take
|X| = 1 and X = e*^; that is, the parameter X may lie anywhere on a
circle of unit radius in the complex plane.
Now the solution of (6) is easily derived by well known processes
of the operational calculus: it is
h{t)=- \ ' dr (7)
^ Jo Vr-V/ - r
^ Jo ^T^r^
= dT. (8)
T
The solution is also known to be '^
h{t) = e-O^^l'U, (^], (9)
where /o(X) is the Bessel function Jo(ix).
a true asymptotic expansion. On the other hand Heaviside in his frequent appli-
cations of the Rule gives no hint or indication of the restrictions imposed on its
applicability. Fortunately in most applications of the operational calculus to physi-
cal problems, the Rule leads to correct results.
^ See formula (p) of the table of integrals in Chap. IV, "Electric Circuit Theory
and Operational Calculus."
NOTES ON THE HEAVISIDE OPERATIONAL CALCULUS 153
Now return to the operational equation (6), and expand as follows,
without reference to convergence,
h =
1 /, , p\-''-' r
-P 1 +Y V^
=
\2l
(^)(fr-
■■■■)*
Application
of the Heaviside Rule
; now
gives the divergent solution
hit)
~ S{\t).
p. 32
2!
(iif-
] VttX/
(:
(10)
We have now to distinguish three cases:
1. Xfi > 0. (Real part of X > 0.)
In this case it can be shown from (7) that ^
;z(/) ~ 5(X/) (11)
and that the Heaviside Rule leads to a true asymptotic expansion,
as defined by Poincare. When X = 1, by the known expansion of
the right hand function in equation (9) we find that the error com-
mitted by stopping with any term in the divergent series is less than
that term. This property, however, does not characterize the series
for all complex values of X for which the real part is positive.
2. X« < 0, X = - M, Mfl > 0.
In this case, comparison of (8) with (7), gives by aid of (11),
h{t) ~ e'"5(M/), (12)
which again is a true asymptotic expansion. The expansion differs,
however, from that given by the Heaviside Rule, by the factor g''^
and the alternation in sign of the odd terms of the series.
3. Xfl = 0, X = iw.
In this case it is easily shown that ^
A(/) = ^-"'-'/^>/o(f ), (13)
where /o is the Bessel function of order zero. From the known
asymptotic expansion of this function, we find that
hit) ~e-"'-'/^T^^*"'''^-5(^'c«^0]Rea.Part (14)
with an error less than the last term included.
* L.c. by the process described in Chap. V.
^ L.c. formula (w) of table of integrals. Chap. IV.
154 BELL SYSTEM TECHNICAL JOURNAL
Perhaps the simplest way of establishing the Heaviside Rule for
the asymptotic solution of the operational equation h = l/H(p) and
the conditions under which it is valid, is as follows: We start with the
integral equation
r h(t)e~p'dt = l/pH(p) (15)
Jo
and specify that the singularities of llpH(p) and its derivatives are
all confined to the left hand side of the complex plane, except at the
point ^ = 0, in the neighborhood of which
ao
- + ai + a•2^1p + a^p + a^p^fp + • • • . (16)
pH{p) ^jp
In other words, l/pH{p) admits of expansion in powers of V/?
Now since
•M p-pt 1
we have from (15)
r h^dt= ^ (17)
Jo
r(
"-^)'-"-pm-w <'«^
By virtue of the restrictions imposed on l/pH{p), equation (18) is
valid at ^ = 0, whence by (16)
r("-s)*="- ''"'
Now differentiate (18) with respect to ^; we get
Now add H ^ ^ dt to the left of (20) and its value a2/2-yJp to
Jo 2 ^j^^t
the right hand side; we have
NOTES ON THE HEAVISIDE OPERATIONAL CALCULUS 155
Now set ^ = 0; from (16) we have
r(
"-^+14'^'=-- ''''
a formula which again is valid by reason of the restrictions imposed
on l/pH(p).
Proceeding in this manner we get the formula
Jo
(h - Sn)-t-dt = (- l)"w!a2n+i, (23)
'0
where
+ (- 1)"1.3 ••• (2« - 1) '^'"
(20"
= first {n -\- \) terms of the divergent Heaviside series. (24)
Also since
S.» = 5. + (- !).+■ '■'•;3//.: + '^ ^ (25)
we have from (23) by changing w to (« + 1),
f
Jo
h-s.- (- i)"+^ ^-^"(J^!: + ^^ ^)/"+w/
= (- iy+'{7i + l)!a2„+3. (26)
Equations (23) and (26) establish the fact that (h — 5„) converges,
for indefinitely great values of t, at least as rapidly as l/t"+^-y[t, since
otherwise the integrand of (26) would diverge; stated in mathematical
notation
h - Sn= 0(1/ 1"+^'^). (27)
Consequently the series S when divergent is a true asymptotic ex-
pansion, as defined by Poincare, of the function h.
The foregoing says nothing, it will be noted, regarding the error
committed when 5„ is employed to compute the function h. Nothing,
in general, can be said about this question, which requires an inde-
pendent investigation in every specific problem. In some cases the
error will be less than the magnitude of the last term of Sn, but this
is the exception rather than the rule. In other exceptional cases the
series may even be absolutely convergent.
156 BELL SYSTEM TECHNICAL JOURNAL
The foregoing results can undoubtedly be derived by integration
of the Bromwich integral (2) along the contour suggested by March
{I.e.). Wiener in his paper on "The Operational Calculus" {Math.
Ajinalen, Bd. 95, 1925) gives an entirely different treatment of the
problem. The operational calculus he deals with, however, differs
under some circumstances from that of Heaviside, as Wiener himself
remarks. A paper by Tibor v. Stacho on "Operatoren Kakiil von
Heaviside und Laplaceshe Transformation" (publication 1927 VI 15
by the Hungarian University, Francis Joseph) may also be consulted.
n
Subject to certain well known restrictions a function f{t) can be
expressed as the Fourier integral
/W = T"; HP)e^''dp. (28)
the path of integration being along the imaginary axis. We assume
for the moment that this equation is valid.
Now suppose that f{t) represents a force applied to an electrical or
dynamic system whose "steady state" or forced response to an applied
force F{p)eP^ is
H{p)' •
Then the forced response g{t) of the system to the applied force /(/)
is given by
However, in applying the foregoing to the Heaviside problem we
encounter an initial difficulty. This is that if /(/) is taken as the unit
function (zero before unity after, / = 0) it does not admit of formu-
lation as the Fourier integral (28). The unit function, however, when
multiplied by e~'' when c is a positive real constant, does admit of such
formulation, and it is easy to show that the unit function itself is
given by
c-fioo ,
— dp c > 0. (30)
c-i« P
Consequently, if the unit function is the force impressed on the sys-
tem, the forced response is
iTTiJc-io, pn{p)
NOTES ON THE HEAVISIDE OPERATIONAL CALCULUS 157
If now all the singularities of the integrand lie to the left of the imag-
inary axis, then k{t) = /?(/) and (31) is the formulation of the
Heaviside problem. Suppose, however, that the electrical or dynamic
system specified by II{p) is "unstable"; that is, it contains some
internal source of energy which makes its transient oscillations in-
crease with time / instead of dying away. In such a case H{p)
will have zeros to the right of the imaginary axis, and in order that
(31) shall be the solution of the Heaviside problem, c must be taken
so large that all the singularities of the integrand lie to the left of
the path of integration. Consequently
h{t) -^ I -^TTTT^P'
J_ r+'" g^' ^^ (2)
iiriX-i^ pn{p)
pro\ided c is so chosen that all the singularities lie to the left of the
path of integration in the complex plane. This is Bromwich's formu-
lation of the Heaviside problem. ^°
From the foregoing it follows that the Fourier integral
-f
is, in general, the formulation of the Heaviside problem if and only
if, all the singularities of the integrand lie to the left of the imaginary
axis. If there are singularities on the imaginary axis, the integral
is ambiguous, while if there are singularities to the right of the im-
aginary axis, the integral gives an incorrect solution of the Heaviside
problem. ^^
As a simple example consider the operational equation
h = \/H{p) = ^
P-is'
where the real part /Sr of /3 is positive. The correct solution as given
by either (2) or (3) is
// = / <
= e^' t > 0,
^° The appropriate mathematical methods of solving the infinite integral (2) are
dealt with in great detail by Jeffreys in his "Operational Methods in Mathematical
Physics" (Cambridge University Tracts).
" To prevent misunderstanding it should be stated that the application, when
permissible, of the classical Fourier integral (2a) to the Heaviside problem, was
known long prior to the work of Bromwich. Bromwich's essential and important
contribution lay in showing that the path of integration must be shifted to the
right of all the singularities, together with a verification of an important form of
solution, first given by Heaviside, of the operational equation.
11
158 BELL SYSTEM TECHNICAL JOURNAL
whereas the Fourier integral (2a) gives
h = - e^' t < 0.
= / > 0.
There is another reason why care must be exercised in applying
the classical Fourier integral to the Heaviside problem. This is that
in solving the operational equation, h = \/H{p), the appropriate
expansion of \/H{p) may introduce singularities on or to the right of
the imaginary axis in the component terms. This offers no difficulty
if either (2) or (3) is employed, but renders the Fourier integral (2a)
inapplicable. As an example consider the equation
V^+ 1
One form of solution is gotten by multiplying numerator and denomi-
nator by V^ — 1 , whence
Ji =
_ 4p
p - 1 p - 1
and each term has a singularity at ^ = 1.
A physical interpretation of the foregoing may not be without
interest. Suppose that an elementary force F(p)eP^dp, where p =
c + io), is applied at an indefinitely remote past (negative) time to a
system specified by H{p) . The response of the system is then
^''^^ e^'dp-^ T^{t)dp,
H{p)
where Tp{t)dp is the concomitant transient or characteristic oscillation
of the system. If c is chosen sufficiently large then at least for t >
the transient term can be made as small as we please compared with
the first term. Finally if the impressed force is the unit function
(zero before, unity after, time t = 0) and it is written as
1_ r+*"£!!^
5W Jc-i 00 P
the total response and therefore h{t) is given by
J_ f
<^+<«' .tp
e'
pllip)
dp,
NOTES ON THE HE AVI SIDE OPERATIONAL CALCULUS 159
provided c is sufficiently large to make the transient term
/^C-|-lQO
T„{t)dp
^ C—iaa
negligibly small. Analytically this requires that c be so large that
the zeros of pllip) shall all lie to the left of the axis pji = c.
Ill
The foregoing discussion tacitly assumes the existence of an unique
solution of the operational equation. On the part of the physicist
this assumption is entirely proper because if the operational equation
is the symbolic formulation of a correctly set physical problem an
unique solution must and does exist. When approached from the
purely mathematical standpoint, however, the case is different and
there is no assurance of the existence of a solution. As an example
consider the operational equation
h = e"
The corresponding integral equation
f = r
P Jo
h{t)e-p'dt pR >
has no solution, while Bromwich's formula
c-ioo P
1 r'+"^ ov
gives h = t < — 1
- 1 / > - 1
which is obviously incorrect. As a matter of fact the operational
equation itself has no solution.
To formulate the necessary and sufficient conditions for the exis-
tence of a solution we may proceed as follows: If a solution exists
it is given by either of the equations
h{t) = ^ r j\P)e"'dp, (2)
KP) = f
00
hit)e-p'dt pR > c, (3)
.
160 BELL SYSTEM TECHNICAL JOURNAL
where f{p) denotes \/pH{p). Substitution of the value of //(/), as
given by (2), in (3), gives the transform
f(p) = -L e-,i(U J\z)e''dz. (32)
IlTl Jo Jc-lx
In addition, since //(/) = for / < 0, we must have
• C+ioo
-^ I f{p)e'Pdp = when / < 0.
(33)
Equations (32) and (33) formulate the necessary and sufficient
restrictions on f(p) for the existence of a solution of the operational
equation
h = pfip) = l/H{p).
To correlate the transform (32) more closely with the classical
Fourier transform, write p — n -\- ioo and
f{ii + 7co) = 0(w) i( and co real.
Then the transform (32) becomes
0(co) =— c'""(It (l>{x)e''Hx (34)
ZrJo J -00
for all values of u. > c. Also since //(/) = 0, for / < 0, the lower
limit of integration with respect to / in {33) may be replaced by — oo ,
whence
0(
w
= ^ e-^'hll I <p{x)e"'dx, (35:
27rJ_oo J_oo
which is the classical Fourier transform.
The foregoing naturally suggests a few remarks regarding the mode
of approach to the operational calculus. If we regard, as Heaviside
certainly did, the operational equation as the symbolic formulation
of a definite physical problem, it is not permissible to define the sig-
nificance of the operator p a priori. The meaning of the operator p
and methods of solution of the equation must be so determined as
to give the correct solution of the original physical problem. Hea\i-
side's procedure here was purely heuristic and "experimental"; equa-
tions (2) and (3), however, provide a sound logical basis for the de-
velopment of the operational calculus. On the other hand, from the
purely mathematical standpoint it is possible to develop an opera-
NOTES ON THE HEAVISIDE OPERATIONAL CALCULUS 161
tional calculus on the basis of certain mutually consistent definitions
and conventions adopted at the outset, just as it is possible to develop
different geometries and algebras. An operational calculus so devel-
oped, however, may or may not agree with that of Heaviside and
may or may not give the correct solution of the Heaviside problem.
In a number of recent papers on the Heaviside operator this procedure
has been adopted. To the writer this appears both illogical and
doubtful, and is certainly not the method of Heaviside himself, as is
sometimes implied.
In the interpretation of the operational equation // = l/II(p) it is,
in the writer's opinion, extremely important to recognize the fact that
it is not a true equation and has no literal significance of itself, but is
simply and solely the symbolic or shorthand way of writing down
equation (2) or its equivalent (3). If this fact is kept clearly in mind
the 'operator' p loses the mysterious character it seems to possess for
so many students and all real danger of misinterpretation and incorrect
solution is eliminated. In the writer's opinion, Heaviside's achieve-
ment in the development of his operational calculus does not consist in
inventing a novel and mysterious kind of mathematics, but in formu-
lating a body of rules and processes whereby recourse to the actual
equations of the problem is rendered unnecessary.
There is another fact which it is also important to clearly recognize.
In the original differential equations from which the operational equa-
tion is derived, the symbol p" denotes d^'/dt'^ and its reciprocal ^~",
corresponding multiple integration, and the index « is always integral.
If, as in the case in important electrotechnical problems, non-integral
or fractional powers of the symbol p occur in the operational equation,
it is due to algebraic manipulations and operations, which in essence
rob p of its original significance. That is to say, in such cases it is not
permissible nor indeed possible to assign to the operator p its original
significance. For example the operational equation
// = \'p
does not mean
/dV'
hit) = (— j -1 (1 = unit function)
which is itself meaningless, but simply
1 r'C+ioo „tp
hit) =-^ -^dt c >
162 BELL SYSTEM TECHNICAL JOURNAL
or
1 /•«
}i{t)e-p^dt pit >
_L- r*
V^ Jo
More broadly stated, the operational equation is the shorthand state-
ment of true equations in which p has lost its original significance and
is simply the complex argument of functions which obey all the laws of
algebra and analysis.
Failure to recognize these simple principles is responsible for a large
amount of confusion, loose reasoning and profitless discussion of so
called 'fractional differentiation,' a term which, to the writer at least,
is quite meaningless. On the other hand, their recognition should go
far towards removing whatever mystery may have surrounded the
Heaviside operator and the Heaviside processes.
Contemporary Advances in Physics, XIX.
Fusion of Wave and Corpuscle Theories.
By KARL K. DARROW.
In this article certain of the simple and familiar phenomena of optics and
of electronics — for instance, refraction at a boimdary between two media,
and diffraction by a grating — ^are interpreted by both of the theories, undu-
latory and corpuscular, which have so often been condemned as incom-
patible with one another; the attitude being, that the theories may be
brought into concordance by modifying one at least in ways which, extra-
ordinary as they seem, do not quite destroy its character.
NOT quite five years ago I published in this journal an article
entitled Waves and Quanta, expounding there the data which
invited a corpuscular theory of light, regardless of the great array of
classical phenomena of optics which demanded with no less insistence
the long-triumphant undulatory theory. Today, not only are those
data still extant and undeniable; they have been reinforced by obser-
vations on electron-streams which have compelled a wave-theory of
free negative electricity, despite the very abundant evidence for free
corpuscular electrons. Most physicists expect that not only light and
negative electricity, but whatever other fundamentals there may be —
meaning, probably, positive electricity and nothing else — will be
found to conform in some ways to simple wave-theory, and in some to
simple particle-theory. Most physicists, I think, would concede that
the two ideas must be forced into one scheme, whatever violence it
may entail to others of our preconceptions, inborn or inbred. We
must stretch the theories and our minds, so that corpuscles and
waves shall appear no longer as alternatives of which election must
be made, but as complementary aspects of one reality.
To make a beginning with this process of stretching, I propose to
treat some of the very simplest and most familiar of the phenomena,
which up to lately have been interpreted by ofie only of the theories:
phenomena such as the refraction of light in passing from air to water,
the bending of the paths of electrons in passing from vacuum into
metal, the diffraction of light and electrons from a ruled ditifraction-
grating. (None of these examples, incidentally, involves a theory of
the structure of the atom.) Each of them shall be interpreted by the
other theory — not in order to substitute the other for the one, but in
order to practice the art of using both theories in alliance.
163
104 BELL SYSTEM TECHNICAL JOURNAL
Refraction of Waves and Refraction of Corpuscles.
I presume that every textbook of optics and every history of physics
informs its readers that anciently there was a controversy between a
wave-theory of Hght (attributed to Huyghens) and a corpuscular
theory (accredited to Newton) which was totally decided in 1850 by
an experiment of Foucault. Light is refracted toward the normal in
passing from air to water, and should therefore move more rapidly in
water than in air if it consists of particles, but not so rapidly if it
consists of waves — so runs the argument. Foucault and Fizeau
discovered that light does move less rapidly in water than in air.^
Let us analyze the argument more closely before deciding what was
proved.
The reasoning from the "wave-theory" is usually made in graphic
fashion by showing "Huyghens' construction" (Fig. 1) which should
remind many a reader of his high school days ! This is a very crude
form of wave-theory, much too primitive to account for most of the
phenomena which the physicist has in mind when he says that light
(or electricity, or matter) is of the nature of waves; but for the present
purpose it will do.
In Fig. L A A' is the trace, on the plane of the paper, of a wavefront
moving through air (say) in the direction LM toward the boundary
between air and water. It is the trace of the wavefront at a par-
ticular moment, say /; at a later moment, say /', the front has moved
on to another position BB' . Denote by v the speed of the wave-
front in air; then the perpendicular distance between BB' and A A'
is equal to v{t' — t). While the wave is advancing through this
distance, its intersection with the boundary of the water sweeps over
the distance AB, which we will denote by D. Designate by d the
angle between wavefront and boundary, the "angle of incidence."
From the diagram one sees immediately:
sin 6 = v{t' - t)ID. (1)
Now in Huyghens' view, whenever the oncoming wavefront passed
over an atom in the boundary-surface it incited that atom to emit a
"wavelet." The circles drawn around various points on the line AB
are the traces on the plane of the paper, of halves of those spherical
wavelets — the halves expanding downwards into the water. Accord-
1 Foucault usually gels all the credit, but Fizeau and Breguet were working at
the same time, incited by the same suggestion of Arago, and using the same method
with differences in detail; and they announced their result only six weeks later.
Indeed, at the meeting of the Academic des Sciences (May 6, 1850) at which Foucault
reported his success. Fizeau said that if the sun had shone that day or the day
before, they too would have had data to present.
CONTEMPORARY ADVANCES IN PHYSICS
165
ing to "Huyghens' Principle" the ongoing wavefront in the water is
the envelope of these spheres. In Fig. 1 they and the ongoing wave-
front are represented for the moment t' when the wave in the air
reaches B. The radius ^C of the wavelet expanding from A is then
the distance which light traverses in water during time (/' — /), for
that wavelet started when the wave in the air reached A. Denote by
v' the speed of light in water and by 6' the angle between the new
WATER
Fig. 1.
wavefront and the boundary, the "angle of refraction " ; then from the
diagram :
sin d' = v\t' - t)ID (2)
and from (1) and (2) together, we obtain:
sin ^/sin d' = vjv'.
(3)
From this familiar equation it follows in general, that the ratio
(sin 0/sin d') is independent of the angle of incidence. (It is called
the index of refraction of the second medium with respect to the first ;
I denote it hereafter by N.) Also it follows in particular, that when
light is refracted towards the normal the wavefronts must move more
slowly in the second medium than in the first, which is what Foucault
verified, or rather, thought he had verified.
Now try it by the corpuscle-theory. In Fig. 1, I have the line LAIN
redrawn as a heavy line, and the lines at right angles to it left out; for
the line LAIN, one of the "rays" of light, is now to be interpreted as the
path of a corpuscle, and there are no wavefronts.
So long as the corpuscle is too far from the boundary-surface to feel
any force from the water, it moves in a straight line with unchanging
momentum; for the forces exerted on it by the air, being equally
applied in all directions, balance one another out. In the region near
166
BELL SYSTEM TECHNICAL JOURNAL
the boundary, this remains the truth for the components of force
parallel to the surface; but the components along the normal, applied
respectively from the direction towards the air and the direction to-
wards the water, need not be perfectly equal. After the corpuscle
has gone through the transition region and reached the depths of the
water, it continues in a straight line with a momentum of which the
component parallel to the boundary — the "tangential" component,
say — is still the same as it was in the air, while the normal component
is changed. Denote by pt and pn these two components of the original
momentum of the particle through the air, by p the magnitude of their
resultant which is the original momentum ; by p/, p„' and p' the corre-
Fig. 2.
spending quantities for the final flight of the corpuscle through the
water. From Fig. 2 we see :
sin d = Ptfylpt'' + pn" = pt.lp, sin 0' = p/jp,
and since pt = p/ :
sin djsin 0' = p'/p.
(4)
(5)
The corpuscle-theory therefore leads to the statement that the sines
of the angles of incidence and refraction stand to one another as the
momenta of the corpuscle in the first medium and in the second; and
when light is refracted towards the normal, the corpuscles must move
with a greater momentum in the second medium than in the first.
CONTEMPORARY ADVANCES IN PHYSICS 167
Comparing the equations (5) and (3) to which the two conceptions
lead, one sees that far from contradicting one another, they are both
acceptable, provided that :
PIP' = v'lv. (6)
We may hold both the theories simultaneously, we may interchange
the two at will, provided we assume that the momentum of the cor-
puscles varies inversely as the speed of the wavefronts. In spite of
the outcome of Foucault's experiment, we may adopt either the wave-
theory or the corpuscle-theory or both at once to describe refraction,
provided we assume that when a beam of light is refracted toward the
normal, the speed of the wavefronts diminishes but the momentum of
the corpuscles grows greater.
Why then did everyone concede that the corpuscular theory of light
was killed by the experiment of Foucault? Because everyone was
making two assumptions which seemed so obvious as to be hardly worth
the stating, and so certain that it would have been regarded as absurd
to call either into question:
(A) It was being assumed, that the momentum of a corpuscle must
always be strictly proportional to its velocity; in other words, that the
mass of a corpuscle must be invariant.
(B) It was being taken for granted that in a wave-theory of light
the speed of the waves, and in a corpuscle-theory of light the speed of
the corpuscles, must be identified with the actual speed of light as
measured in any actual experiment.
When these assumptions are made, equation (5) goes over into the
form,
sin e/sln d' = p'/p = v'/v, (7)
which is contradictory to equation (3) and disproved by the experiment
of Foucault.
But it no longer seems radical to change the first of these assump-
tions, for it is known from observation that there are particles — elec-
trons, for example — of which the mass is not invariant, but depends
upon the speed. For such a particle the momentum is not exactly
proportional to the velocity. It is then not quite so revolutionary to
go further, and suppose that the corpuscle of light is of so strange a
nature that its velocity and its momentum are in magnitude inversely
proportional to one another. If one made this supposition then one
could accept the second assumption, and still explain the refraction
of light by the corpuscle-theory.
Even the second assumption, however, is not sacred. It may seem
absurd to set up a wave-theory of light, and then say that the speed of
168 BELL SYSTEM TECHNICAL JOURNAL
the wavefronts is not to be identified with the measured speed of light.
It does seem absurd to set up a corpuscle-theory, and then say that the
speed of the corpuscles is not necessarily the same as that of light.
Yet it may turn out in the end that a theory of either kind is strength-
ened, and made more competent to account for a variety of facts, by
abandoning that easy and natural identification. I will try to prove
by actual examples that it does so turn out. Meanwhile I summarize
this section in a sentence:
If we wish to interpret light, or electricity, or matter, by both a corpuscle-
theory and a ivave-theory, the momentum of the corpuscles must be supposed
to vary inversely as the speed of the ivaves.
I have omitted the special reference to refraction, for any more
general theory must include that particular case, or fall down com-
pletely; I have added allusions to electricity and matter, for the test
of any alteration of the two classical assumptions will depend chiefly
on whether it helps in understanding the wavelike properties of these
two, and not of light alone.
We now carry the wave-theory a great step beyond the primitive
form in which Huyghens left it, by introducing the ideas oi frequency
and wave-length.
Wave-length of Waves and Momentum of Corpuscles
Instead of the single "wavefront" of Fig. 1, suppose a train of sine-
waves of frequency v, period T{— 1/f), wave-length X and wave
number /x(= 1/X) travelling through air along the course LMN. For
definiteness, think of sound-waves. The condensation ^ of the air
conforms to the equation:
p = Po sin 2ir {vt — p.s -{- a), (8)
wherein 5 stands for distance measured from some arbitrary plane
perpendicular to LM, and a for some constant. I write the equation
down because one like it (or more than one) occurs in every wave-
theory. In that of light there are six such equations, with components
of electric and magnetic field strength replacing p; but it will be
sufificient to think of one. In the wave-theory of matter there is one,
with a quantity of very abstract meaning replacing p.
Now when the wave train passes through into the water, its fre-
quency remains the same. With sound-waves, or any mechanical
vibrations of matter, this is obvious; two pieces of matter in con-
tinuous contact must vibrate in unison, or not at all. We generalize
this statement to cover light-waves, and waves of other varieties later
^ The excess of the density over the normal vahie, dulded by the normal value.
CONTEMPORARY ADVANCES IN PHYSICS 169
to be considered. Using primes to designate the values which things
have in the second medium, we put:
/ = p. (9)
The speed of the waves is the product of their wave-length by their
frequency :
V = v\ v' = v'\'\ (10)
consequently:
v'jv = y/\. (11)
The wave-lengths of the wave train on the two sides of the boundary
vary directly as the speeds.
Return now to the last section, and introduce this result into equa-
tion (6) ; one gets:
P'/P = V^' (12)
which means: we can interpret refraction of light (or of electricity, or of
matter) by both the wave-theory and the corpuscle-theory, provided
that we make the momentum of the corpuscle vary inversely as the
wave-length of the waves.
Write accordingly,
p\ = constant. (13)
Now there are several remarkable experiments which show that this
relation actually holds, and moreover that the constant which appears
in it is the universal constant h of Planck:
p = h/X. (14)
For instance, one may pour a stream of X-rays — that is to say,
high-frequency light — into a gas, after having measured its wave-
length in the known and reliable way depending on one of the phenom-
ena in which X-rays behave as waves. A certain portion of the rays
is scattered; it is scattered as though it consisted of corpuscles, each of
which strikes an individual free electron and bounces off, the electron
meanwhile recoiling from the blow.- Further analysis of the data
shows that there is conservation of momentum — that the momentum
which the electron gains is equal to that which the corpuscle of light
has lost, provided that the momentum of this latter is equal to the quotient
of h by the wave-length of the rays. For the wave-length of the scattered
X-rays, measured in the same way as that of the primary rays was
measured, is not the same as theirs; and the difference between the
values of /;/X, before and after scattering, is equal to the momentum
which the electron received.
- The Compton effect (cf. the seventh article of this series).
170 BELL SYSTEM TECHNICAL JOURNAL
Again, one may pour a stream of electrons against a crystal or an
optical ruled grating, after having measured the speed of the electrons
in one of the well-known ways depending ultimately on the deflection
of such a beam in known electric and magnetic fields.^ The mass
of the electrons being known, one knows also their momentum. Now
the crystal or the grating, whichever it may be, forms from the primary
beam a diffraction-pattern of new beams. Well! the formation of a
diffraction-pattern is the primary reason for saying that light is wave-
like, and it gives the primary way of measuring wave-length of light.
One is equally obliged to admit that a stream of free negative elec-
tricity is wavelike, and to accept the value for its wave-length which
the diffraction-pattern gives. Again it turns out that the wave-length
is equal to the quotient of h by the momentum of the electrons.
It may be objected that in all of those experiments, the corpuscles
were observed in a vacuum. Compton measured X-rays before and
after scattering, but during the measurements they were in vacuum
or at any rate in air. Davisson and Germer, Thomson and Rupp,
observed electrons returning through the same evacuated space as
they had crossed on their way to the diffracting lattice. One might
emphasize that all these savants compared momenta and wave-lengths
for different beams in the same medium instead of comparing them for
the same beam in different media. The distinction is certainly worth
noticing; but happily there are experiments which bear directly on
refraction. Davisson and Germer measured, not precisely the
refraction of an electron-stream passing from vacuum into nickel, but
a minor perturbation of the diffraction-pattern which is due to that
refraction. We will analyze their result, for nothing shows more
clearly the relations — or lack of relation, the reader may think —
between speed of waves, speed of corpuscles and measured speed of
stream.
Davisson and Germer came to values of the index of refraction
(sin 0/sin d') which were greater than unity — which corresponded
therefore to a bending of the stream towards the normal, as it passed
from vacuum into nickel — which therefore signified that the speed of
the waves is not so great in nickel as in air.
On the other hand, it is known that when an individual electron
passes from vacuum into a metal, its kinetic energy and its velocity
increase as it goes through the surface. We have in fact the situation
described in the corpuscle-theory picture of refraction, a few pages
back. Return to equations (4) and (5), and consider a corpuscle for
^ The experiments of Davisson and Germer, of G. P. Thomson, and of Rupp (cf.
the eighteenth article of this series).
CONTEMPORARY ADVANCES IN PHYSICS 17 1
which the momentum p, the velocity u, the kinetic energy K, the
mass m are related to one another as in Newtonian mechanics —
properties which are practically those of electrons except when these
are moving much more rapidly than any involved in these experiments:
p = mil, K = y^mu^. (15)
Use Ui and Un to denote tangential and normal components of speed;
use primes to designate the values which things have in the second
medium (nickel). Starting from equation (5), we continue:
sin djsm 9' ^ N = M'jM = u'/u;
iV2 _ 1 = (m'2 _ ^^2)1^^2 = (^z _ x)/K. (16)
The quantity {K' — K) is the gain in kinetic energy which the electron
wins on passing into the nickel ; and this gain, as I have said, is positive;
hence by equation (16) the index of refraction must be greater than
unity. This is in agreement with the result of Davisson and Germer;
the agreement, in fact, appears to be quantitative.'*
It is always pleasant to get an agreement; but note how we got this
one. We got it by dropping the assumption that the speed of the
corpuscles and the speed of the waves must be the same. Or rather,
by not making that assumption. For though the fact of experience
is always the same — the swerving of the electron-stream toward the
normal as it enters the nickel — it is interpreted by the two theories in
opposite ways; the waves are slowed down, but the corpuscles are
speeded up, in passing from the vacuum to the metal. Even if wave-
speed and corpuscle-speed were the same in empty space, they could
not be the same in any other medium.
This is more serious than it may appear at first. It amounts in
effect to saying that a beam of free negative electricity has two dif-
ferent speeds; one when we visualize it as a jet of particles, another
quite different when we visualize it as a train of waves.
But is not one of these "the right one" and the other "a wrong one,"
and can we not settle between them by measuring the actual time
which the electricity takes to pass a measured distance? Let us
examine this possibility. We shall find that after all it is not so easy
to evade the ambiguity in such a fashion.
Phase-Speed and Group-Speed
Suppose an endless train of perfect monochromatic sine-waves
marching along through space. For definiteness, think again of sound-
^ There is a remarkably interesting correlation between these results and the new
statistical theory of the electron-gas inside the metal (cf. my article in the October
1929 number of this Journal, pp. 710-716).
172 BELL SYSTEM TECHNICAL JOURNAL
waves. It might seem as if we could measure their speed by picking
out one crest, as A of Fig. 3, and checking off with a stop-watch the
moments when it passes two fixed markers placed a known distance
apart. Not so; for we cannot see or hear or in any way perceive the
individual crests. The wave train produces a perfectly uniform tone
in the ear which it strikes. If two listeners are stationed at different
points along the path of the sound, neither can recognize the moment
at which any particular crest glided by. All they can recognize, all
they can compare, is the moment of passage of a perhirbation of the
wave train ; a sudden beginning, a sudden ending, a transient swelling
of the sound. Most measurements of the speed of sound, in fact, are
measures of the speeds of something violent — the crack of a pistol or an
electric spark, the roar of an explosion — something very unlike a uni-
form train of sine-waves.^
Now a sine-wave with a perturbation is in effect a sum of two or
more sine-waves each of endless extent and constant amplitude, but
having different wave-lengths and different amplitudes. This state-
ment is the content of Fourier's principle from which the method of
Fourier analysis is derived. One might represent even the sudden and
violent pulsation of air due to an explosion, or the electrical spasm due
to an outburst of static, by a summation of properly-chosen endless
monochromatic sine-wave trains. I take however the simplest con-
ceivable case: the wave train composed of only two sine-waves of dif-
ferent wave-lengths.
The reader will probably recall that when the difference between the
wave-lengths is only a small fraction of either, this composite wave
train resembles a sine-wave with regular fluctuations of amplitude —
that is to say, with "beats" (Fig. 3). The maximum or centre of a
beat occurs where a crest of one sine-wave coincides with a crest of the
other — the minimum between beats, where crest falls together with
trough. Denote the two wave-lengths by X and X -|- AX. One
sees by inspection that a wave-length is the same fraction of the dis-
tance D between two consecutive beat -maxima, as the discrepancy AX
is of the wave-length -.^
D/\ = X/AX. (17)
Of course this statement is exactly true only in the limit of vanishingly
small AX. We shall always stay close to this limit, though some of the
following statements would be valid even otherwise.
^ I except so-called measurements of the velocity of sound which are really measures
of frequency and wave-length in stationary' wave-patterns, these being then multi-
plied together.
^ The principle of the vernier.
CONTEMPORARY ADVANCES IN PHYSICS
173
Fi^. 3.
12
174 BELL SYSTEM TECHNICAL JOURNAL
Now if the two component waves advance with equal speed, the
beats are simply carried along with a speed equal to theirs. But if the
velocities of the two component waves are not the same, then the
velocity of the beats is not the same as either, nor the mean thereof.
It is in fact something totally different.
To see this, imagine that you are moving along with one of the sine-
waves; for definiteness, that you are riding on the crest B of the train
with the shorter waves (Fig. 3). At a certain moment, say / = 0,
it coincides with a crest A of the other sine-wave, and you are at the
top of the beat. Meanwhile the other train is moving relatively to
the first; for definiteness suppose that the longer waves move faster,
so that relatively to the shorter they are gliding upward. After a cer-
tain time they have gained on the shorter waves by a distance AX, the
difference between the two wave-lengths. But when this time has
elapsed, the top of the beat is no longer where you are, but where the
crest B' of the first train coincides with the crest A' of the second.
It has dropped back through the distance X, while the second wave
train was getting ahead by the distance AX. Perhaps it will be easier
to realize that while the second wave train is gaining on the first by X,
the beat is dropping back by the distance D between consecutive beats;
by equation (17) this comes to the same thing.
Therefore when the longer waves travel faster than the shorter, the
beats travel more slowly than either. If the longer waves were the
slower, the beats would travel more rapidly; but this case is never
realized in nature, not at least with light-waves '' and waves of elec-
tricity^ and matter.
We now deduce the formula for the actual value of the speed of the
beats. Denote by v and v -\- f^v the speeds of the two sine-waves of
which the wave-lengths are X and A -f- AX, respectively; by g the
speed of the beats. It is sufficient to put into notation what has just
been said in words. Relatively to the former wave train, the velocity
of the latter wave train is Av, that of the beats is {g — v). Relatively
to the former wave train, the latter moves a distance AX while the
beats are moving a distance X in the opposite sense, therefore with a
minus sign. Hence:
{g - v)/M' = -X/AX (18)
" The exception to this statement — the case of light having wave-lengths lying
within a region of anomalous dispersion of t he transmitting substance — has been an-
alyzed by Sommerfeld and L. Brillouin {A7in. d. Phys 44, pp. 177-202,203-240; 1^14)
who find that in this case the group-speed defined by (20) loses its physical im-
portance, and a segment of a wave train is transmitted with a speed never exceed-
ing the speed of light in vacuum. This appears to be related to the absorption
which always gees with anomalous dispersion.
CONTEMPORARY ADVANCES IN PHYSICS 175
and solving for g,
.?=^'-X^, (19)
or going over to the differentia! notation, which will not only look more
natural but will signify that the result which we have just attained is
strictly valid in the limit for infinitesimal differences of wave-length:
g^ V - X(dvi'd\). (20)
This is the formula for the group-speed; for the term "group-speed"
is the usual one for what I have been calling "speed of beats." Like-
wise phase-speed is commonly used to denote the speed of the individual
sine-wave trains.
The term "group-speed" is in one respect unfortunate; for it implies
that any "group," that is to say any sequence of uneven and irregular
wave-crests and troughs, is propagated with a perfectly definite speed.
However this is true only for the simplified group which we have been
considering, the beat formed of no more than two wave trains; and
even for this it is exactly true only in the limit, where the wave-length-
difference between the trains approaches zero. All other groups
change in form as they advance. Now there is always something
arbitrary in defining "speed" for something which changes as it goes,
like a puff of smoke or a cloud. The arbitrariness is nil in only the
limiting case which I have just been formulating. However, it must
not be exaggerated. A bunch of irregular crests and troughs may
retain enough of its form and compactness, as it travels over a distance
many times as great as its width, to justify the statement that it has a
speed of its own. And if such a group turns out, on being analyzed in
Fourier's way, to consist mainly of sine-waves clustered in a small
range of wave-lengths, then its speed will not be far from the value of g
computed by equation (20) for a wave-length in that range.
Now these deductions explain a very remarkable experiment by
Michelson, which otherwise might have disproved — indeed I do not
see how it could have been interpreted otherwise than as destroying —
both the wave and the corpuscle theory of light. I will preface the
account of this experiment by saying that for light in empty space the
speed of all wave-lengths is the same,** so that there never is any dif-
* The chief evidence for this statement is astronomical. If light of one color
traveled faster than light of another, a luminous star emerging from behind a dark
one would be seen first in the faster-travelling hue; in fact there would be a .se-
quence of colors, the same for every emergence of every such star, and spread out over
a time-interval proportional to the distance of the stars. Nothing of the sort has
ever been observed, although there are plent>- of luminous stars revolving around
dark ones which regularly occult them.
176 BELL SYSTEM TECHNICAL JUURXAL
ference between velocity of groups and velocity of wave-crests; they
both have the same universal constant value c. However this cannot
be true for light in transparent material media such as glass, water,
or carbon bisulphide; for the refractive index of all these media varies
from one wave-length to another — they are said to be dispersive.
Now Michelson measured the time taken by a flash of light to cover
a measured distance, first through air (very nearly the same as vacuum)
then partly through air and partly through carbon bisulphide. The
source of light shines continuously, and an incessant beam falls on a
revolving mirror and is reflected in a continuously-changing direction;
a second, stationary mirror receives this reflected beam during a very
small fraction of each complete revolution and sends it back, so that
the twice-reflected beam is a series of segments cut from the primary
beam. It was the time taken by the segments to travel a known dis-
tance which Michelson measured.^ Reasoning back from the data,
he computed that they took (1.76 ± 0.02) times as long to go a given
distance in carbon bisulphide as in air. But the refractive index of
carbon bisulphide, in the range of the spectrum where Michelson's
source of light was brightest, is about 1.63; so that the primitive wave-
front-theory gives 1.63 for the ratio of the speeds in air and CS^,
and the corpuscle-theory gives (1.63)~^
Foucault and Fizeau, be it remembered, had done the experiment
with water. It happens that for water the derivative dvld\ is much
smaller, and the group-speed therefore much closer to the wave-speed,
than for carbon bisulphide. Also their experiments, though performed
by the same method as Michelson was later to adopt and adapt, were
less accurate than his. But if they had performed the Michelson
experiment in 1850, the result would have been astounding. For
Arago had asked, in effect: is it the speed of the wave-fronts in the
wave-theory, or the speed of the corpuscles in the corpuscular theory,
which agrees with the measured speed of a piece of light? Arago had
said: "These experiments . . . will permit no further hesitation as
between the rival theories. They will settle mathematically (I employ
this word on purpose) they will settle mathematically oneof the greatest
and most disputed questions of natural philosophy." He had proposed
a question to Nature, and had written down two and only two answers.
Everyone thought that Nature must reply by ratifying one of the
9 When the segments returned from the second to the first mirror they found that
the latter had revolved a little further beyond the oiientation which it had when
they left it, so that it reflected them onward not quite along the path to the source of
light, but along another path inclined to that one at an angle twice as great as that
through which it had revolved. Michelson measured the angle, and knowing the
rate of revolution of the revolving mirror he then knew how long the light had taken
to go from it to the stationary mirror and back.
CONTEMPORARY ADVANCES IN PHYSICS 177
answers. Foucault and Fizeau reported that she had repHed: the
former. But they had not heard distinctly ; for her actual response was :
tieither.
Michelson's experiment however came after the idea of group-
velocity as distinp;uished from wave-velocity had been invented and
established. The refractive index of carbon bisulphide varies with
wave-length. On determining the wave-speed or phase-speed v from
the refractive index (by the equation N = c/v) and then the derivative
dv/d\, it is found '" that in the region of the visible spectrum, the term
\{dvld\) amounts to about seven per cent of the term v, on the right-
hand side of equation (20) — that is, the group-speed should be some
seven per cent lower than the wave-speed in carbon bisulphide. In
air, however, group-speed and phase-speed are sensibly the same.
The ratio of the group-speeds in air and CSo falls close to Michelson's
value."
Coming as it did, therefore, the Michelson experiment merely showed
that those who had subtilized the Huyghens' theory by introducing
sine-waves had incidentally invented something able to move with the
measured speed of a light-flash, though nothing of the sort had been
available in the original form. Had it come earlier — well, there is no
way of knowing what would have been inferred; but people might have
come to think that after all a wavefront-theory or a corpuscle-theory
of light may have some use and value, even though the speeds assigned
to the waves or the corpuscles do not agree with those actually meas-
ured. Such an attitude of mind would be rather advantageous,
today. As a corollary for the present I submit: in picturing a jet of
free negative electricity as a beam of waves or a stream of corpuscles,
we should not be too confident that either the speed of the waves or
the speed of the corpuscles is the speed with which a segment dissected
from the jet would move from place to place, until someone succeeds in
making actual measurement of this last. Fundamental theory has
something to say on this point, which we will presently consider.
'" I take all the numerical values in this section from a re\'ie\v of Michelson's
work by J. Willard Gibbs {Am. Jour. Sci. 31, pp. 62-64; 1886) which so far as I know
is the latest critical discussion of the data.
'1 The problem is more complex than I have intimated, not only because Michelson
observed light covering a very wide range of wave-lengths so that i' and dvjdX both
extend over wide ranges of values, but also because different parts of a wave-front
are reflected from different parts of the mirror at different moments, and therefore
from differentl y-incl ined parts. Quite a controversy went on during the eighteen-
eightiesin the pages of " Nature" as to what it was that Foucault had really measured.
Rayleigh at first {Nature 24, p. 382; 1882) thought it was g; then changed his mind,
(25, p. 52; 1882) and decided it was t'-/g; then was convinced by Schuster {3i, pp.
439-440; 1886) that it was really i'V2(t' - g). J. W. Gibbs then took a hand {Zi,
p. 582; 1886) and contended that after all it was really g. The contro\"ersy seems to
have rested there. It may be added that Michelson's data eliminate v'-jg, but do not
quite discriminate between g and Schuster's expression.
17cS BULL SYSTEM TECIISICAL JOURNAL
r'.ROrP-Sl'EED AXn rORI'lSCI.K-SPEED
Thus far I have said thai if we wish to use wave-theory and corpuscle-
theory alternatively, we must make the momentum of the corpuscle
equal to the quotient of the constant h by the wave-length of the waves;
but I have said nothing about the energy of the corpuscle.
Let us adopt the universal assumption — based on a multitude of
experiments, for instance those on the photoelectric effect — that the
energy £ of a corpuscle of light is equal to the product of its frequency
V b>- the same universal constant //; and let us extend it to the other
kinds of corpuscles which we may associate with other kinds of waves,
and vice versa.
Then the complete description of the particles associated with waves
of wave-length X is as follows:
p = ///X, E= hv^ hv/X. (21)
Here, as before, v stands for the phase-speed of the waves (not the
particles).
Returning to the formula (20) for the group-speed, we now can write
it thus:
r^= V - \{dv:d\) = v\ - \dip\)ld\ ,
= -XHdp/dX) = - i\'lh)idE/d\). ^ '
Suppose next that the energy and the momentum of the corpuscles
in question are related to each other and to their speed in the well-
known fashion of ponderable bodies, to which it is known that electrons
conform. Thus for sufficiently low speeds, the relations are practically
those of the "classical" mechanics:
p = Wo/^ E = i^Wo«-, whence E = p-,2n!Q. (23)
Here nio stands for the constant mass, u for the speed of the corpuscles
(not the waves).
The energy of the corpuscles is a function of the momentum only,
and continuing to develop the formula (22) for the group-speed, we
find:
o = {- }^jh)(d E;dp){dp;d\) = dE;dp .^^.
— p^i — I3~lm() = II.
The group-speed of the waves is equal to the speed of the corpuscles.
The same conclusion follows if we use the relativistic definitions for
the energy and the momentum of a particle,
E = WocVVl - (3-, p = Wn^f/Vl - f3- (J3 = II 'c),
E = c^niifc- -\- p-
as the reader mav test for himself.
CONTEMPORARY ADVANCES IN PHYSICS 179
Summarizing: if the corpuscles associated with the waves have the
properties of ordinary material bodies — if, let us say, for short, the
corpuscles are material particles, their speed is equal to the group-speed of
the waves.
This is a ver\' happy and agreeable result. It compensates very
largely for our having been forced to concede that if we want both
waves and corpuscles, the wave-speed and the corpuscle-speed must
be different. The wave-theory has supplied another velocity which is
equal to that of the corpuscles. Moreover it is precisely the \'elocity
with which we should expect an isolated segment of a wave train to
move from place to place. If someone were to cut a piece out of an
electron-jet and measure the time it took to traverse a known distance,
the speed which he would deduce from his data would probably agree
both with the corpuscle-speed and with the group-speed, and disagree
with the wave-speed. It would be interesting to try this out.
In the equations ^23) I have taken account only of the kinetic
energy of the corpuscles; in the equations (25), only of their kinetic
energy and of the 'rest" energy associated with their mass. But the
explanations of refraction by the two theories will no longer be con-
cordant, unless the potential energy also is admitted. Let us denote
the potential energy of a corpuscle by U; and, since as yet these theories
have been verified only for negative electricity, let us immediately
write eV for U, e standing for the charge of an electron and V for the
electrostatic potential in the region where it is. For the total energy
of the corpuscle, then, we have instead of (25) the relati\'istic
expression,
E = nl,c^i^l - ff' -\- U = WocVVF^^^ + eV, (26)
which for small values of the corpuscle-speed u (= 3c) reduces to the
classical expression,
E = hnuu- -\- U = hnuii" + eV. (27)
In an earlier section we interpreted the refraction of an electron beam
passing from vacuum into metal by thinking of the metal and the
vacuum as being two regions in which different values of electrostatic
potential prevail, the potential thus changing sharply from one value
to the other at the surface which bounds the solid. Xow when the
beam considered as a stream of corpuscular electrons passes across
such a surface, the energy of each electron as expressed by (26) or (27)
remains the same, though the proportion which is kinetic energy is
changed; and therefore the frequency Ejh of the equivalent wave-train
remains the same. If then we keep the assumption that the wave-
ISO BULL SYSTEM TECIIXICAL JOCRXAL
length of the waves is equal to // divided by the monientuin of the
particles, we have the following value for the ratio between the
wave-speeds v' and v on the two sides of the surface:
„y_x;.v.(|^)/(|*).,V,, (28)
and the speed of the waves varies inversely as the momentum of the
corpuscles, which is just what is required in order that we may hold
both the theories simultaneously.
But how about the theorem that corpuscle-speed is equal to group-
speed? Returning to the equations (25), we see that the introduction
of the potential energy has altered the relation between energy and
momentum; we now have:
E = c^Jnioc' -f p- + eV. (29)
But so long as we are comparing different electron-streams in the
same medium (vacuum, for instance), the potential energy is the
same for all and does not depend on the momentum; and differentiating
E with respect to p to obtain the value of the group-speed g, we get:
.? - dE/dp = ^^ = c^m.nHX_-^ ^
E - eV nioc-l^l - (3-
and thus group-speed and corpuscle-speed are equal, as before.
I will write down the expression of the phase-speed, although for
the physicist it is of minor importance, not being measurable — a fact
which exempts us, temporarily at least, from pondering over the
curious feature that it depends on the value of the potential energy
of the corpuscles, and therefore (for electrons) on the value accepted
for the electrostatic potential of the region where the wave-train is,
even though in practice it is generally assumed that electrostatic
potential may be measured from an arbitrary zero. The formula is
v = E/p = -^^-h'^^i±U
Wo«/Vl - /3- (30)
= c-ju + U/p,
and if we put the potential energy of the corpuscles equal to zero, we
find the phase-speed varying inversely as the corpuscle-speed,^'- and
greater than the speed of light.
1- There is a paradox here which, as I can testify from personal experience, is a
dangerous source of confusion. The formula v = c'-jii sounds like an approximation
to tlie formula v = const' p which I have gi\'en as the retpiisite relation lictwecii
CONTEMPORARY ADVANCES IN PHYSICS 181
Stationary Waves and Oscillating Particles
We have tried out, separately and in tandem, two alternative ways
of interpreting a beam of radiation advancing through space; first as a
stream of corpuscles, then as a train of waves. We will now try out
two alternative ways of interpreting radiation enclosed in a box; first
as a system of stationary waves, then as a quantity of corpuscles rush-
ing to and fro and bouncing from the walls. To simplify the case as
much as possible, think only of motions parallel to one side of the box;
or to make the pictures more graphic, think of a tube or pipe like those
often used in experiments on sound, in which the waves travel along the
axis.
Now it is well known that when a train of sound-waves is sent
through a tube, or generated by vibrations somewhere in the tube, it is
partially reflected from the far end, then again partially reflected from
the near end, and so on over and over again; the overlapping wave
trains passing to and fro interfere with one another; and when the
wave-length is related in a certain way to the length of the tube, the
overlapping wave trains form a stationary ivave-pattern of alternating
loops and nodes — the tube is said to be in resonance. If the two ends
of the tube are alike (both open, or both closed) so that reflection takes
place in the same way as both, the waves which admit of resonance
are those of which the half-wave-length or an integer number of half-
wave-lengths fits exactly into the tube; denoting by d the length of the
tube, these wave-lengths are given by the formula-:
wQj= d, n= 1,2,3, . . . (41)
This equation defines what may be called the characteristic wave-
lengths of the tube. The tube distinguishes these, or the wave trains
possessing these wave-lengths, from all the others.
Suppose on the other hand we had particles rushing back and forth
along the axis of the tube, and rebounding without loss of energy
whenever they struck either wall. Denote by ii the speed of a particle ;
it takes a time-interval Id/u to describe a complete round-trip with
two rebounds, and one might say crudely that it has a frequency u/ld.
I say "crudely" because the corpuscle is not moving with a sinusoidal
motion, like a pendulum-bob; its speed does not vary as a sine-function
wave-speed and momentum. However the two relate to entirely different situations.
The first is a comparison between wave-speeds and corpuscle-speeds for different
beams in the same medium. The second is a comparison between wave-speeds and
corpuscle-momenta for the same beam in different media. The resemblance between
the two is accidental and misleadin;^.
I am ind(-l)ted to Professors C. 11. lukarl and E. C. Kemblc for elucidation of liiis
point.
182 BELL SYSTEM TECHNICAL JOURNAL
of time, but retains the same value throughout except for the change
of direction; if we were to apply a Fourier analysis to this motion, we
should find not only the frequency ///2c?, l)ut all of its overtones. Let
us think however only of this fundamental frequency. Now it
is evident that there is nothing, in our ordinary conceptions of particles
rushing back and forth and rebounding from walls, to distinguish any
value of speed or frequency above any others. The phenomenon of
resonance sets certain wave-lengths apart from others, but there is
nothing to correspond to resonance in this latter case, and set certain
speeds apart from others.
But instead of sound, think of some kind of radiation which we have
interpreting both as corpuscles and as waves — light, for example.
Light enclosed between parallel reflecting walls forms stationary
waves,''' provided that its wave-length is related to the distance d
between the walls by the equation (41), which I rewrite:
X = Id/u, ;/ = 1, 2, 3 . . . (42)
The parallel reflecting walls, or the limitation which they set upon the
space accessible to the light, thus single out certain characteristic
wave-lengths and distinguish them from all others. How interpret
this fact by corpuscle-theory?
Well, we have been associating waves of wave-length X with cor-
puscles of momentum p = h/\; let us continue to do so. The reflecting
walls, then, single out certain characteristic values of momentum given
by this equation, derived straight from (42):
p = nh/ld, (43)
which I proceed to rewrite thus,
2d-p = uh ;/ = 1, 2, 3 . . . (44)
These values of momentum, I have said, are set apart from all the
rest. If waves and corpuscles are interchangeable as bases for a
theory of light, then the feature of wave-motion known for short as
"resonance" obliges us to make that supposition. But in what way,
and to what extent, are they set apart? According to modern quan-
tum-theory, they are actually the only possible values. A particle
describing a cyclic motion of this character, in which it moves a fixed
distance with a fixed momentum and then moves the same distance
backward with the same momentum reversed and so forth ad infinitum,
is constrained by something in the order of nature to have one or
'^Interference f)atleriis are essentially of this type, thoiii'li iisiiali\- they are
formed between mirrors ol)liciiie to one another.
CONTEMPORARY ADVANCES IN PHYSICS 183
another of the "permitted" momenta defined by equations (43) and
(44).
Examining equation (44), one sees how this definition of the per-
mitted momenta may be stated. The quantity on the left of (44) is
the product of the momentum of the particle, by the distance which it
traverses each time it performs its cycle. '^ This product must be
equal to an integer multiple of the Planck constant //.
Now the quantum-theory of the atom developed fifteen years ago
by Bohr, Sommerfeld and W. Wilson — the first and greatest of the
forward steps in the contemporary conquest of the problem of atomic
structure — was based on the assumption that an electron perfomiing a
cyclic motion must perform it in such a way, that its momentum
conforms to a condition of which equation (44) is but a special case.
This is the condition alwavs written thus:
/
pdq — 7ih, w = 1, 2, 3 . . . (45)
If the electron is oscillating to and fro in a straight line through a
position of equilibrium, q stands for its distance from that position and
p for its momentum, and the integral is taken once around a complete
oscillation. It is evident that (44) is the special form of this equation
for the case in which the force acting on the electron is vanishingly
small until it hits the wall and then suddenly becomes enormous. If
the electron is revolving in an orbit in two or three dimensions, there
are two or three equations like (45) all postulated at once; but I shall
not take up such more complicated cases.
Summarizing the outcome of this section in a phrase: if we associate
waves of wave-len»th X with corpuscles of momentum hi\, and stationary
waves in an enclosure with corpuscles flying hack and forth between its
walls, then the condition that the waves must fulfil to form a stationary
system is equivalent to the quantum -condition imposed upon the corpuscles.
This is an illustration of wave-mechanics. How extraordinarily
fruitful and valuable such comparisons have proved in the hands of
Louis de Broglie, of Schroedinger, Rose, Fermi and Sommerfeld — to
name only a few — I have shown in part, in earlier issues of this journal.
Here it must suffice to say that Schroedinger developed the principle
into a form suitable for predicting the stationary states of atoms; Bose
constructed out of it a competent theory of radiation in themial equili-
'^ It travels a distance d in the forward sense with a momentum p, and then an
equal distance in the backward or negative sense with a momentum of equal amount
but reversed sign, so that the total product of distance by momentum is
t>d + (- p)(-d) = 2dp.
1S4 BULL SYSTL.M TliCIINICAL JOURNAL
liriiim, considered as a gas of which the atoms are corpuscles of hght;
while Fermi. Dirac and Sommerfeld between them used it to make a
powerful theory of the free negative electricity in metals, conceiving
this alternatively as a gas of which the atoms are electrons, and a
system of stationary waves enclosed within the surface of the metal as
in a box with reflecting walls.
Diffraction' of Waves axd Diffraction' of Corpuscles
The effect of a diffraction-grating upon a beam of light projected
against it has always been considered the most striking evidence that
light is of the nature of waves and not of corpuscles. Indeed it is
considered to suffice in itself to prove the corpuscle-theory untenable.
With any common understanding of the term corpuscle-theory,
this statement is correct; but we had better put it in the softer form,
that the effect of a diffraction-grating on a beam of light proves that if
we adopt a corpuscular theory we must endow the corpuscles with some
very strange property which nobody ever thought that particles could
possess, and which may even seem to be in contradiction with their
nature. W'e had better put the statement in this milder way, because
it now is known that in spite of all the evidence for individual electrons,
a beam of negative electricity is affected by a grating in much the same
way as a beam of light.
Take then almost the simplest conceivable case of diffraction; a
plane-parallel beam of light falling perpendicularly on a wall containing
many equally-spaced parallel slits, and a part of the light passing
through the slits to a screen infinitely far away. On this infinitely-
distant screen — which may in practice be brought up to a convenient
nearness, b}- means of a lens — one sees a peculiar pattern of light and
shade. I single out one particular feature of this pattern : the fact that
there are maxima of illumination along certain lines parallel to the
slits. One of these, for instance, is straight ahead from the slits, along
the direction of the incident beam prolonged; another is oft' to one side,
in a direction making a certain angle (say </>) with that of the incident
beam; another is equally far off to the other side. These two last-
named, the Jirst-order maxinia, are those we shall consider; it will be
enough to speak of one.
By the wave-theory, a first-order maximum is explained as follows.
Each of the slits is the source of a secondary wave train of spherical
wave crests, stimulated by the primary wave train, and having the same
frequency and wave-length. Consider any two adjacent slits.
Secondary wave crests start from the two at the same moment. At
any point equally distant from the slits, they arrive simultaneously, and
CONTEMPORARY ADVANCES IN PHYSICS
185
reinforce each other; this is the explanation of the central bright fringe.
At any point not quite equally distant from the slits, they do not
arrive quite simultaneously, and the reinforcement is impaired. But
at a point which is further from one slit than from the other by just
the wave-length X, the wave crest arriving from the latter meets the
next previous crest from the former, and the reinforcement is re-
stored. The first-order maximum is located at these points.
Fig. 4.
From Fig. 4 one sees ''^ that when the screen is very far away, the
points distant from the slit 6\ by one wave-length more than they are
distant from S-i are situated in the direction inclined at to the straight-
ahead direction, the angle being given approximately by the formula
sin 4> = ^,<3,
(46)
where a stands for the distance between the slits. When the screen is
infinitely far away, the formula is exact. (I must admit that it is
somewhat disingenuous to simplify the problem by solving only the
special case in which the screen is infiniteh' far away, for the general
case opposes much more serious difiiculties to the corpuscle- theory-;
but this is the special case of greatest physical importance, and one has
to make a beginning somewhere.)
We have now explained the presence of a first-order maximum in
the pattern of light and shade on the screen, though it cannot be said
that we have "verified" formula (46), for that formula serves as the
practical definition of wave-length : wave-lengths are measured by
'■' From the figure wq see that for di and d>, the distances from .S'l and 6- to the
point P on the screen, we have:
d{- = D' + X-, dy = D' + (.V - a)\ d, = D sec <^, .v = D tan <^
and hence
(d, - d-i)(d, +<f,) = lax - a\
When D, x, di and </•: all become infinite together, the second factor on the left becomes
equal to 2D stc ^ and the second term on the right may he neglected.
1<S6 BI'.LI. SYSTEM TECHNICAL JOURNAL
measuring the angle 4> and using equation (46). Let us now try the
corpuscle-theory on the problem.
Putting as heretofore the value h/X for the momentum of the cor-
puscles, translate (46) into the language of the alternative theory; one
gets:
sin (/) = h/ap. (47)
In words: a corpuscle of momentum p, passing through any slit, is
particularly likely to bend around through an angle (/> of which the sine
depends in a certain way on its momentum and on the distance to the
next slit.
Which is to say: the likelihood that a corpuscle entering a slit will
bend its course through a certain angle depends on the presence of
other slits in the same wall, and on the distance between these slits.
But the reader will inquire: how does the corpuscle entering one of
the slits know that the other slits are there? If all the other slits were
suddenly stopped up, the first-order maximum would vanish, the
likelihood that the corpuscle would turn in the direction given by (47)
would fall to zero; but how could it know that they had been stopped
up?
Well! this is precisely the strange and extravagant property with
which we are forced to endow the corpuscles, if we want to use the
particle-theory to explain diffraction. It must be supposed that when
passing through a slit, a particle of light knows whether there are other
slits and, if so, how they are spaced. It must be supposed that an
X-ray particle striking an atom in a crystal knows that there are other
atoms in a regular array, and knows moreover just the pattern and the
scale of that array. It must be supposed that electrons enjoy a like
omniscience. Or to express it in more technical language; the prob-
ability that a corpuscle of light, of electricity or of matter shall be
deflected through a given angle when it strikes an atom or passes
through a slit must be supposed to depend on the arrangement of the
other atoms or the other slits in the vicinity. This idea is not easy to
accept; but it must be accepted, if one is to build up a complete cor-
puscular theory of any of these entities.
But if one accepts it, one finds that the stipulation (47) turns out to
be another example of the general quantum-condition of which, in (44),
we have already met one instance. For write it thus:
ap sin (j) = apt = nh, « = 1, 2, 3 . . ., (48)
the factor n being now introduced to take account of the maxima of
second, third, and higher order which also occur on the screen, though
CONTEMPORARY ADVANCES IN PHYSICS 187
I refrained from mentioning them earlier. I have used the symbol pt
for the quantity p sin 0, for this, as one sees immediately, is the tan-
gential component of momentum which the corpuscle acquires at the
deflection, not having had any before. The wall containing the slits,
or the row of atoms if we consider instead the difi^raction of X-rays by a
crystal, receives an equal momentum in the opposite sense. We may
therefore say that diffraction occurs in such a way, that the regularly-
spaced series of slits or atoms receives a momentum pi given by the
formula :
apt = nh. (49)
But now what is the product apt? It is the product of the mo-
mentum of the row of atoms or slits by the distance a between any
adjacent two; it is therefore the integral J'pdqoi the general principle
(45), evaluated for the range of integration a. Now the general
principle is supposed to apply when the range of integration covers a
complete cycle of a periodic motion. There is nothing obviously
periodic about a steady sidewise sliding of a row of atoms with a
constant momentum. But in a sense, there is after all something
periodic. For if the row of equally-spaced atoms (or slits) extends to
infinity in both directions, then when it has moved sidewise through the
distance a each atom lies exactly in the former place of another atom,
and the original arrangement is to all appearances restored. The
steady onward motion of the regular array is also a cyclic departure and
return to a periodically-restored arrangement; and the maxima of the
difl^raction-pattern are determined by applying the quantum-condition
to this cyclic motion.
The reader may ask: how about the component of momentum in
the direction at right angles to the grating? Without precisely answer-
ing that question, I will end the article by applying the corpuscular
theory to a case in which all the components of momentum are duly
taken into account: diffraction of X-rays or of electrons by a three-
dimensional crystal.
Suppose an "ideal" crystal extending infinitely far in all directions.
It is composed of similar and similarly-oriented "atom-groups" — I
will use the language and the symbols of the eighteenth article of this
series — arranged upon a "space-lattice," of which the three spacings
shall be denoted by a, a', a". If we start with one atom-group A,
then along one direction from it there is an infinite sequence of such
groups at distances a. la, 3a, . . . and also at distances (— a), (— 2a),
(— 3a), ... in the opposite sense. Call that the .v-direction. Then
along another direction through .1, say the y-direction, there is an
188 BELL SYSTEM TECHNICAL JOCRXAL
infinite sequence of groups at distances a', 2a', 3a', . . . and (— a'),
(— 2a'), etc.; and along a third or 2-direction through A, there is an
infinite sequence of atom-groups spaced at intervals a".
Now think of the atom-groups as hard particles, and the corpuscle
of light or of electricity (the "X-ray quantum" or the electron) as a
hard particle which rushes into the lattice, hits one of the atom-
groups — A, say — and bounces oft". Denote by 0, </>', 0" the angles
which its original direction of motion makes with the x, y, z directions
respectively; Ijy 0, C, 6" the angles which its final direction of motion
makes with these three. Before the defiection, the corpuscle has a
momentum of magnitude p, parallel to its original direction of flight;
afterward it has a momentum of the same magnitude, but parallel to
its final direction of flight. At the deflection, then, it loses — that is,
it communicates to the lattice — a momentum of which the three
components along .v, y, z have the values:
/?(cos d - COS (p); picos d' - cos 0'); picos d" - cos (/>").
Now if, following the foregoing procedure, we equate the first of these
to some integer multiple of h/a, the second to some integer multiple
of h/a', and the third to some integer multiple of h/a", and then
translate momentum of corpuscles into wave-length of waves by the
now-familiar formula p = h/X, we get:
a(cos d — cos (f)} = ii\,
a'(cos d' - cos 4>') = ii'X, (50)
a"(cos d" - cos 0") = ;/"X,
where ;/, ;/', ii" stand for any three integers. Now these are the
equations (numbered 3, 4, 5 in the eighteenth article) to which conform
the " Laue beams," which is to say, the directions in which electrons
and light are actually diff^racted by crystals.
Perhaps I should close with two or three admonitions. To make the
wave-theory and the corpuscle-theory equivalent for a few simple cases
is of course not at all the same as making them equivalent universally.
Also, the examples in this article are not always so elementary as they
may seem. The first involved two distinct media with a sharp bound-
ary between; and discontinuity is always less agreeable than continuity
to the mathematician. The last but one involved a non-sinusoidal
vibration, which is much more complex than a sinusoidal one. More-
over, the concepts of light-waves and quanta are not nearly so beauti-
fully welded together as those of electricity-waves and electrons.
Nevertheless these illustrations may help to weaken the idea that there
is no way out of the present situation but to abandon either waves or
corpuscles; for decidedly, there is a way.
Wave Propagation Over Continuously Loaded Fine Wires
By M. K. ZINN
The paper contains the resuhs of a theoretical investigation of wave prop-
agation along a pair of wires that are "loaded" by enclosing each wire
in a continuous sheath of magnetic material. The results of greatest
practical interest are certain approximate formulas that are sufficiently
simple to be adapted to engineering design studies, while having a high
degree of precision for all practical dimensions and frequencies.
THE purpose of this investigation is to define the character of
wave transmission along a pair of wires each of which is loaded
with a continuous sheath of magnetic material. Exact expressions
for the propagation constants are developed from the general theory
that applies to such a system. Also, simple approximate formulas
are given for the sizes of wires that are generally used in paper-insu-
lated cables.
Wave Propagation Along a Pair of Wires with Magnetic
Sheaths
For the benefit of those who are not interested in following the
theoretical work in detail, a general sketch of the method and a sum-
mary of the mathematical results will be given first, together with
a discussion of some numerical examples. Details of the theoretical
work have been placed in the Appendices.
The analysis here given follows closely the methods developed by
John R. Carson ^ in a solution of the transmission of periodic currents
along a system of coaxial cylinders. The analysis for the case where
the outgoing and return conductors are coaxial is applied, with only
small modifications, to the case where the two conductors are parallel
and not coaxial. This application of the theory ignores the "proximity
effect." - That is to say, it assumes that the electric and magnetic
forces within each conductor are functions only of the distance from
its axis and of the coordinate in the direction of propagation, which
is strictly true where the cylindrical conductors are coaxial.
1 "Transmission Characteristics of the Submarine Cable," John R. Carson and
J. J. Gilbert, Jour>ial of the Franklin Institute, December 1921.
- This is the usual method of dealing with problems involving balanced parallel
conductors. The alternating-current resistance of the s\-stem may be expressed as
the product of the a.c. resistance, assuming a concentric return, and a "proximity
effect correction factor," which takes into account the influence of the parallel return
conductor. The "proximity effect" is in general negligible at voice frequencies for
conductors of sufficiently small cross-section, such as those of paper insulated cables.
References: "Wave Propagation over Parallel Wires: The Proximity Effect." John R.
Carson, Phil. Mag., April 1921, and "Wave Propagation over Parallel Tubular
Conductors," Sallie Pero Mead. Bell System Technical Journal, April 1925.
13 189
IQO
BELL SYSTEM TECHNICAL JOURNAL
The physical system contemplated is shown in Fig. 1. The out-
going and return systems of conductors, each comprising a cyhndrical
wire with insulated cylindrical sheath, are assumed to be identical
in all respects. For the sake of generality, it is assumed that the
magnetic sheaths may be insulated from the wires, as shown. The
interesting practical case where wire and sheath are contiguous, form-
ing a bi-metallic conductor, then appears as the limiting case of
infinitesimally thin insulation.
WIRE
•E';
DIRECTION
OF PROPAGATION I
MAGNETIC SHEATH
3L| bg cl2
INSULATION (ADMITTANCE Yg)
MAGNETIC SHEATH (X2,|J.2)
^\ E" *i XV, + 4^ dz \— INSULATION (ADMITTANCE Y,)
— ^ » 7 — 1-
iHl'+^'dz (I) ^Ii Y WIRE(X,.|I,)
* oz ^
Fig. 1 — Illustrating various quantities involved in the analysis.
The problem consists in finding a solution for the propagation con-
stant of the system from Maxwell's equations. If the magnetic
sheaths are in contact with the wires, the propagation constant is
given in the usual form, V = \ Y^Z, where Y2 is the admittance across
the insulation between the sheaths and Z is the series impedance
of the system. The admittance is, in general, either a known, or
an experimentally determined, quantity; so that for this case the
theoretical problem resolves itself into that of finding the series im-
pedance.
An important part of the investigation is, however, to determine
WA VE PROP A GA TION VER CONTIN UO USL Y LOA DED WIRES 1 9 1
what the effect would be of introducing insulation between the wire
and its sheath. In this more general system, shown by the sketch,
the solution for the propagation constant has two values, because
two layers of insulation are involved, and cannot be expressed in
the usual form. It is found, however, that it can be expressed in
terms of the propagation constant for the elementary case where wire
and sheath are in contact by introducing two other known propagation
constants that determine transmission along the separate pairs of
conductors in the system. The expression for the propagation con-
stant, when given in this form, shows directly the effect of insulating
the wires from their sheaths.
It is necessary first to define certain impedances. Let /] be the
total current in one of the wires and lo the total current in its sheath.
The tangential electric forces in the surfaces of wire and sheath are
denoted by Ei" , £«' and E^" , as shown in Fig. 1. These electric
forces are linear functions of the currents, as follows:
-C,2 — -^21 1 1 I ■'^22 -'2)
-C,2 ^^ •^21 J 1 ~r ^^22 ■*2)
E," = Zn"h.
(1)
The impedances which appear in these equations as the coefihcients
of the currents are functions of the electrical constants and dimensions
of the wires and sheaths. Their values are given in Appendix A.
Now let
7 = propagation constant determining transmission along the loaded
wires if the wires and their sheaths were in contact = VKoZ.
7i2 = propagation constant determining transmission along one wire
with its sheath as the return, when the sheath is insulated
from the wire = VFiZio.
722 = propagation constant determining transmission along the two
sheaths if the wires were removed = V ^2^22.
Then, from (1)
£2' - Ey"
Z12 —
Z22 =
Z =
I Ai — Zii — Z21 ~l~ Z22 -(- Ai, 1-2 = — /i
2E2"
h
IE."
+ A2 = 2Z22" + X:
In these equations,
+ X2 = 2
Z22
7 'i
Z]i — Z21 ~r Z22
/i = y (2)
+ X2
192 BELL SVSTE.U TECHNICAL JOURNAL
Xi = iwLio = reactance arising from the magnetic field between the
outer surface of the wire and the inner surface of
the sheath.
A'o = iccLoc, = reactance arising from the magnetic field between the
two sheaths.
The terms in brackets in the equation for Z give the "internal
impedance " of one of the loaded wires for the elementary case where
wire and sheath are in contact, and Xo is the additional reactance
that arises from the magnetic field outside the wires.
With the elementary propagation constants, 7, 712 and 722 so defined,
it is found that the propagation constant, F, of the general system
can be expressed as follows:
2r~ = 712- + 722- ± \(7i2- + 722-)^ - 47-712-. {^)
It is convenient also to express the two solutions for T in the form
of series :
-r -1 ^ 2 T12'- I J 7i2^ , r, r, 712*^ ,
712" + 722- (712- + 722"/ (712- + 722-)
To- = 7i2- + 722" — Ti'-'. (4)
lation between the wire and sheath. For, if
is small
The solutions in the series form show the effect of introducing insu-
47-7i2"-^
(719- + 722")-
compared to unity, as it would be in a continuously loaded wire with
a thin magnetic sheath of high resistance, then, to a first order of
approximation, the principal propagation constant Fi is less than 7,
the propagation constant that determines transmission when wire and
sheath are in contact, by the factor
The other propagation constant, Fo, is, in this case, very large com-
pared to Fi and plays no appreciable part in defining the character
of transmission except at points very near to the terminals of the
system. For practical purposes, the system may be considered to
have only one significant mode of propagation.
Case of a Wire with Contiguous Sheath
The Internal Impedance
The practical case where the magnetic sheath and the wire are
contiguous, forming a bi-metallic conductor, is of special interest.
WAVE PROPAGATION OVER CONTINUOUSLY LOADED ]VIRES 193
In this case, the propagation constant is uniquely determined from
a knowledge of the admittance between the loaded wires and of their
series impedance. The "internal impedance" of the loaded wires
comprises the larger part of this impedance. For the purpose of
engineering design work, it is convenient to have at hand approximate
formulas for the "internal impedance."
The exact expression for the impedance is given by the last of
equations (2). \A'hen the magnetic sheath is thin, as compared to
the radius of the copper wire, certain approximations can be made.
These are explained in Appendix A. The result is the following
formula for the "internal impedance":
where F = wjULob
Z i _\ — orF + ioiCi
-^ + lull
^7rX,X,M-.>/'^ + Mi&(^^-^log^
(5)
II = 2ir''-\,n.af \\t{\, - Xi) + Xi^l +2^/
/? /?
R = p ,'p = d.-c. resistance of one of the pair of bi-metallic
conductors,
Ri = , ,., = d.-c. resistance of the inner part of the conductor
TrKib- .
(the wire),
i?2 = ^ , ., 7T- = d.-c. resistance of the outer part of the
irXMi- — 0~) , / , , 1 X
conductor (the sheath),
Lo = 2^2 log T = low-frequency inductance contributed by the
sheath,
b = radius of the wire,
a = outside radius of the sheath,
t = a — b = thickness of the sheath,
^1, Ml = conductivity and permeability of the wire,
X2, M2 = conductivity and permeability of the sheath,
CO = Itt times the frequency,
I'M
BELL SYSTEM TECHNICAL JOURNAL
The total series "loop" impedance of the pair of loaded conductors
per centimeter is Z = Zi -\- Xo-^
For the purpose of indicating the degree of precision of the approxi-
mate formula, data are given in Fig. 2 on the internal resistance and
inductance of various copper wires coated with loading material to
300
250
tf)
I
200
a.
us
a.
LU
o
z
<
in
IS)
LU
a.
(0 0.0248
>
a.
z
UJ 0.0246
uj 0.0244
_1
cc
LU
a
ai
o
z
<
I-
o
D
a
z
0.0242
0.0240
0.0238
O0236
'^^s^
o
(
) (
1
26 GAUGE
^
"V.
:^
16
3
4 6 8
FREQUENCY — KILOCYCLES
10
12
pi^_ 2 — -Internal impedance of wires of various sizes with continuous loading ol
approximately 25 millihenrys per wire mile (for very small currents,
i.e., hysteresis losses not included).
3 All quantities are expressed in the electromagnetic c.g.s. system of units. To
obtain the result in ohms per loop mile, multiply by 160,934 (10~»j. In the case
of cable circuits, A'2 ( = zcoL22) is an experimentally determined quantity, L22 having
a value of about .001 henry per mile.
WAVE PROPAGATION OVER CONTINUOUSLY LOADED WIRES 195
such a depth as to give an internal inductance of about .025 henry
per wire mile. The magnetic material in the sheath has been assumed
to have a permeability of 3,000 and a conductivity of .77 x 10~^ in
e.m.u. (resistivity 13 microhm-centimeters, in practical units). The
data shown by the solid lines are exact while the points give the
results obtained by means of the approximate formula (5). A com-
parison of results is tabulated below for the largest wire (16 B. & S.
gauge), where the errors of the approximate formula are greatest.
Internal Resistance and Inductance (of One Wire)
Ohms and Millihenrys per Mile
Frequency —
Kilocycles
Exact
Approximate
Errors
Res.
Ind.
Res.
Ind.
Res.
Ind.
2
5
8
10
21.065
31.674
86.795
186.65
276.04
24.77
24.56
24.37
24.05
23.75
21.065
31.63
86.18
183.6
269.4
24.77
24.63
24.41
24.01
23.66
- .14%
- .71
-1.63
-2.41
+.29%
+ .16
-.17
-.39
The errors are roughly proportional to the quantity, WajM2X2.
For a loading material having, say, one-quarter the permeability and
the same conductivity, the errors would be about twice as large,
therefore, if the inductance and the wire size remain the same.
Hysteresis Loss
The real part of the internal impedance given by (2) or (5) is the
effective internal resistance of the bi-metallic wire, taking into account
the heat losses that arise from the electric current, namely, d.-c.
resistance, eddy current loss and "skin effect loss." The formulas
do not take into account hysteresis loss, which is a magnetic phe-
nomenon as distinguished from these electric phenomena. The de-
termination of hysteresis loss rests upon experimental data. If the
energy loss due to hysteresis in the magnetic material per unit volume
per cj'cle is Ji (ergs), then the resistance increment due to hysteresis is
R,
hrdr.
(6)
For the low values of magnetic force that obtain in telephony, it is
found that // = ■qB'^, where r? is the hysteresis coefficient and B the
induction density. Therefore
Rk
1' X
IPrdr.
(7)
1<)6
BELL SYSTEM TECHNICAL JOURNAL
Since the majjnetic coating is thin, and the "demagnetizing," or
"screening," effect of eddy currents small, it may be assumed that
// = 2//r. (It will not exceed that value, at least.) Using this
approximate value for //, the resistance increment due to hysteresis is
Ri, = 8r?co/x-'/
a
ab
= 2r]fJ.O)BaL2,
(8)
0,8
0.6
Z
UJ
o
<
I-
o
H
UJ
I
0.4
Z
u
CC
a.
O
0.2 I
I-
<
a:
4 6
FREQUENCY — KILOCYCLES
Pig 3 — .Illustrating the fractions of the total current that are carried by the copper
wire and by the magnetic sheath (19 gauge (B. & S.) wire with
continuous loading of 25 millihenrys per mile).
where Ba is the induction density at the outside boundary' of the
sheath.
The Distribution of the Current in Wire and Sheath
It is a matter of interest to know how much of the current is carried
by the magnetic sheath and hf)W the current is distributed over the
cross-section of the wire and sheath at various frequencies. The
WAVE PROPAGATION OVER CONTINUOUSLY LOADED WIRES 197
solution of this problem is not an essential part of the investigation,
but it helps in understanding what takes place in the bi-metalhc
conductor.
The ratio of the currents in wire and sheath to the total current, as
computed from (1), is plotted in Fig. 3. It will be noted that the
fraction of the total current carried by the sheath becomes greater
800
5
o
a
s
If)
Q.
Z
<
z
u
Q
UJ
CC
u
600
400
200
FOR A TOTAL CURRENT
OF I AMPERE IN THE
WIRE AND SHEATH
• DIRECT CURRENT
^ Uir\C.^ I v_-'-'r\r\ui"< I
*M0 KILOCYCLES
10 KC.
2KC,
UJ <
O
UJ
o
UJ
< o
? <
Q- J
90
r^
60
10 KC
IjzKC.
JU
n
_^-^_^_-
L
1 2KC.^
*^IOKC.
"
-30
0.01
0.02 0.03
RADIUS OF COPPER WIRE-
CENTIMETERS
0.04 0.045580 0.046758
HSHEATHk-
THICKNESS
/ SCALE \
I ENLAROEDJ
Fig 4— Illustrating the current density throughout the cross-section of a Avue loaded
with a continuous magnetic sheath— for direct current and 2 and 10 kilocycle
alternating currents. (Same 19 B. & S. gauge wire as that of Fig. 3.)
as the frequency increases. But the fraction carried by the copper
nevertheless remains very nearly unity at all frequencies. This
behavior is explained In^ the curves representing the phase angles
involved. These show, of course, that at very low frequencies the
198 BELL SYSTEM TECHNICAL JOURNAL
copper current and the sheath current are nearly in phase, but with
increasing frequency, the copper current lags behind the sheath current,
until at high fretjuencies the two currents approach a queidrature
jihase relation.
It may be said that at high frequencies the current in the loading
material is practically all "wattless" current, in the sense that it
contributes very little to the energy delivered to any receiving device
connected to the line, but it dissipates energy, of course. At 10
kilocycles, for the 19-gauge loaded wire, the current carried by the
magnetic sheath contributes only 2 per cent of the useful current
(see Fig. 3) ; yet 75 per cent of the energy loss occurs in the sheath
(see Fig. 2).
The difference in phase between the component currents in wire
and sheath is explained by the consideration that the reactance of a
given filament of current is proportional to the magnetic flux external
to it. In the copper, therefore, the elementary current paths have a
small resistance, but a large reactance, due to the fact that nearly all
the magnetic flux is in the loading material. Near the outer surface
of the loading material, on the other hand, the current paths have
less internal reactance, but the resistance is large.
This brings the discussion to Fig. 4, which shows how the amplitude
and phase of the current varies over the cross-section of the bi-metallic
conductor for direct current and for 2 and 10 kilocycle alternating
currents. For the 19-gauge loaded wire, illustrated, the "skin effect"
in the copper is seen to be very small, the alternating current dis-
tribution being practically uniform, as for direct current. At the
boundary between the copper and the magnetic material, the current
amplitude suffers a discontinuity, but the phase is continuous. The
discontinuity in the current amplitude conforms to the law that the
component of electric force along the conductor must be continuous
at a boundary, which requires that the ratio between the current
amplitudes on the two sides of the boundary must equal the ratio of
the conductivities of the two materials. The current density dis-
tribution over the cross-section of the magnetic sheath is uniform for
direct current, of course, but for alternating currents, the density
increases and the phase advances abruptly toward the outer surface
of the sheath.
APPENDIX A
The impedances ' which appear in equation (1) in the bod\- of the
paper as the coefficients of the currents are given b\-:
^ See abo\'e noted paper (reference i) tor ihc development of tliese forniulas.
WAVE PROPAGATION OVER CONTINUOUSLY LOADED WIRES 199
7 " —
2io}iJ.2 Z/a — 1
X2 U,' '
7 " —
2zCOyU2
X'2
U-2
u-r
Z2/ =
2/CO/i2
.V2
f// •
Z.J =
2iwfJL2
X2
1
z,/' =
llCO/Jil
1 2/aJAi]
/oGvi)
(y)
where
^'' ^ .V: L^' .vi /o'(.ri)
(Note that Zoo' = Zoo" - Zoi"),
t^y = - 3'y[/o(xy)i^o'(>'y) - /o'Cv/)A'o(xy)],
V, = - ylMyi)K,{xj) - Mxj)Ko(yn,
U/ = - ylJo'{xj)Ko'(yi) - Jo'{y^)Ko'(xjn
V/ = - ylMyi)Ko'(xj) - Jo'(.r,)i^o(v/)].
(10)
Jo and Ko are Bessel functions of zero order of the first and second
kind, respectively, and Jo' and Ko' are their derivatives with respect
to the arguments, which are given by
X: = ad-^A:Tri(j)iXjKj,
(11)
3'; = hji->^AiriwiXj\j,
where w = It times the frequency, i — V— 1, a; and bj are the outer
and inner radii respectively of conductor j, and ixj and Xy are its
permeability and conductivity. Quantities wath the subscript 1 refer
to the wire and those with the subscript 2 refer to the sheath. All
quantities are expressed in the electromagnetic c.g.s. system of units.
Writing Maxwell's Law, curl E = — -7- , around the contours indi-
cated by the dotted rectangles in Fig. 1 gives
where I'l, V2 are the potential differences between the surfaces of
200 Bh.LL SVSrF.M TECHNICAL JOURNAL
the conductors, as shown, and $), $2 are the normal values of the
magnetic flux that cuts the surfaces bounded by the contours. The
term — 2E->" results from the symmetry of the system, which imposes
the condition that the electric and magnetic forces at corresponding
points in the outgoing and return conductors are equal and oppositely
directed. Also, it is unnecessary to write a third equation for the
field between the other wire and its sheath, because this equation
would be the same as (12). Therefore, the transmission is charac-
terized by only two modes of propagation.
Since all the variables are propagated at the same rate, and since
sinusoidal currents are being considered, djdz may be replaced by — F
and didt by iw. Then
£,' - E," + TV, = XJu (14)
- lEo" + TFs = X.iU + /2), (15)
where V is the propagation constant and
A'l = i(joLi2 = reactance arising from the magnetic field between the
outer surface of the wire and the inner surface of
the sheath.
Xo = ioiLoi = reactance arising from the magnetic field between the
two sheaths.
The potential dilTerences Fi, Fo can be expressed in terms of the
currents by writing Maxwell's Law, curl II = 4tI, around contours in
the outside surfaces of wire and sheath. (Such a contour for the
wire is indicated by dotted lines in the sketch.) This gives
2T:a^-^= - 47rFiFi, (16)
dz
27ra.^^= - 47rF2Fo. (17)
az
where
Y\ = admittance across the insulation between wire and sheath.
F2 = admittance across the insulation between the two sheaths.
Smce ill = — and //■> = >
a 1 a 2
r/i = \\Yu (18)
r(/i + Li) = F2F2. (19)
WAVE PROPAGATION OVER CONTINUOUSLY LOADED WIRES 201
Substituting (18) and (19) in (14) and (15), respectively, gives
(20)
(21)
Xl — yT ] IX = £0' — El",
X, - Y^ ) (/, + /2) = - 2E,",
and substituting (1) in (20) and (21) gives the two equations of the
currents. In order that they shall be consistent, the determinant
of the coefficients must vanish. Therefore
r2
X\ — ^ — Z'ji -\- Zii
J 1
Xi — — + 2Zoi
■^ 2
A'o
7 '
+ 2Z.J
= 0.
(22)
The roots of this equation give the required solutions for the propaga-
tion constant. First, however, it is convenient to introduce two
known propagation constants. Let
712 = propagation constant determining tran smissio n along one
wire with its sheath as the return = -ylYiZn-
722 = propagation constant determining transmiss ion a long the
two sheaths if the wires were removed = VFoZoa-
Then, from (1), (20) and (21),
Z12 — Zii — Z21 ~r Z22 r A 1)
Z22 ^^^ 2/^22 I -^2>
substituting (23) in (22) and rearranging,
712- - r2
(23)
Fi
- 2Z..0'
— /v 22
T22" ^ "
Fo
= 0.
(24)
Expanding
r - P(722- + 7l2-) + T12-722- - 2Z22''FlF2 = 0.
(25)
The remaining impedance can be eliminated by introducing 7, the
propagation constant that would characterize transmission if the
202 BELL SYSTEM TECHNICAL JOURNAL
wires were in contact with the sheaths. (In order not to disturb the
dimensions, it may be imagined that the insulation between wire and
sheath be replaced by an infinitely conducting material, which, how-
ever, is assumed to conduct no current axially. Then E^' — Ei"
— Xili.) To find 7, make Yi infinite and solve (25). Then
r = Y.Z,
and
Z = Z,, - ^^ • (26)
Zi2
Therefore
2Z22''J^ll^2 = T22^7]2' ~ 7"7l2"- (27)
Finally, substituting (27) in (25) and solving the resulting equation
gives the two solutions for the propagation constant,
2P = 712- + 722' ± V(7i2'' + 722'/' - 47-712-. (28)
The arbitrary constants remain to be determined. The currents
are, in general,
7i = ^iie-f-- + ^126-1- + ^iie^- + ^126^-%
(29
The condition of principal practical interest is that of a long cable
with connection made to the two wires and with the sheaths left free
at the sending end. For this case, the conditions are
(1) At s = 0, /i = Jo and h = 0,
(2) At s = 00 , /i = and A = 0,
where Jo is the current delivered to the cable pair at the sending end.
From the second condition,
Bii = B12 = B21 — B22 — 0. (30)
From the first condition,
Au +-4 12 = /o,
(31)
A21 + ^22 = 0.
But these constants must satisfy, for each of the two values of F,
the equations of the currents, whose coefficients are given in the
WAVE PROPAGATION OVER CONTINUOUSLY LOADED WIRES 203
determinant (22). Therefore
w
here
A21 = KiAn,
A 22 ^^ -'^2-'4i2i
(32)
^12 — ^2*2 — T r~ ^22 — —■^22 —
p 2 p 2
ZT~ ^22 — —^22 — ^,
Z22' 7 , r,^
- Z.2 + 3r
F - F.-
Z7 ' ~ 7 7' -
{2^2>)
Zii ^ . r
2
-Z22+Y:
Substituting (32) in (31) and solving
A - T ^^ A - - T ^'
-^11 — -' ^? T^ } ^12 — -f
-A.2 — -A-i 7V2 — iV 1
^21 — -'O'i? ^ — ~ ^22-
A2 — Ai
(34)
Finally, the currents are given by
/n
(35)
This completes the analysis for the more general system where
the magnetic sheaths are insulated from the wires. For the special
case where wire and sheath are contiguous, 712" is infinite and (28)
shows that Fi = 7 and r2 = «> . The transmission is, therefore,
defined by only one mode of propagation. The series impedance of
the system is, from (23) and (26),
Z - 2 |^Z22" - z,- - Zo/ + Z22' J + ^''
(36)
where the terms in brackets give the internal impedance of one of
the loaded wires, and X2 is the reactance that arises from the magnetic
field between them. The internal impedance can be obtained also by
2£./'
finding -j — f-f- directlv from the last two of equations (1), of course.
l\ ~T Li
204 BELL SYSTEM TECHNICAL JOURNAL
The constant K-2. becomes — 1 and the total current, / = /i + I2,
is propagated in accordance with / = Iq€~'^^, where 7 — -sZY-y.
The constant Ki, which is the ratio of the current in the sheath to
tliat in the wire, is of interest. It becomes
Jo T, /^ ]•> — i^oo ^]] — ioo] /2*"^
T-=-f^i = ^—f = ^—f (^0
1 I Z/22 ■'-'■22
The approximate formulas for the case where wire and sheath are
contiguous are derived as follows: The arguments, xo and ,V2, of the
Bessel functions differ by only a small amount when the magnetic
sheath is thin. This situation is favorable to an advantageous use
of Taylor's series. Joixn), for example, can be expressed in terms of
Joiy^), its derivatives and the difference of the arguments in a Taylor
series as follows:
, /o(.r) = My) + rJo'iy) + ^j /o"(v) + fi M"(y) + • • • > (38)
where r = .v — _v (x-i, y2 being written simply, x, v, here, for con-
venience). Furthermore, Bessel functions are subject to recurrence
formulas,^ which enable us to express each of the derivatives occurring
in the series in terms of the function of zero order, its first derivative
and the argument. Therefore, by applying the recurrence formulas
to the Taylor series, we find functions U and V (see Appendix B)
such that
Mx) = UMy) + VJo'(y), (39)
Ko{x) = UKoiy) + VKo'(y) (40)
{U2, V2 being also written now, U, V). Differentiating (39) and (40)
with respect to t,
Jo'(x) = U'My) + V'Miy), (41)
K,'(x) = U'Ko(y) + V'Ko'(y), (42)
3t7 , dV
where U = — — > V = -z— •
OT OT
''The two recurrence formulas required are:
J,/{Z) = ^' /(=) - /„ + ,(3),
J,/{z) = /„_,(=) - "/„(=).
l*lie Bessel Functions ut" the second kind satisfy the same furmuius.
WAVE PROPAGATIUX Ol'ER CONTINUOUSLY LOADED WIRES 205
If (39) to (42) be solved for U, V, U', V, it can be verified that the
solutions are the definitions of these functions already given in equa-
tions (10).*^
The exact formula for the internal impedance of a wire with con-
tiguous sheath has been given in (36). In terms of the functions U
and V, this formula becomes
By using the series for these functions and discarding all terms
of degree higher than w", the approximation given in the body of the
paper (equation 5) may be obtained.
APPENDIX B
\^ hen the recurrence formulas are applied to the Taylor series, it is
found that
T- r'' rW 3 \ t'" / 2 1 2 \
^='+2+6T. + 24('-/)-T2o(,v-7)
^ 2v 6V V-/ 24Vv v^
, rW . 7 , 24\ T« /3 X^ , 120\ , ,,.,
These series converge for
< 1, which condition is satisfied by
the sheath dimensions of any practical continuously loaded conductor.
A considerable number of the terms in the series for U and F are
^ A relation that can be used to advantage at times is
V h
U'V - uv = -■- = --.
X a
This relation corresponds to the similar one for the Bessel functions themselves,
namely:
J„'{z)KJz-) - J„{z)K„'(z) =1.
z
14
206 BELL SYSTEM TECHNICAL JOURNAL
parts of well-known series that define certain elementary functions.
It can be verified readily that
U = cos T -\- '—
F=sin.+,[log(l+^)-^]+^
(46)
+ TT7fl+ •••. (47)
i2oy 2403; ' 24oy
U' = — sin T + ^ 1- -— (above remainder of (3)), (48)
T
V d
V = COS T 1- T- (above remainder of (4)). (49)
1+- ^'
y
The series (46) to (49) possess a certain advantage for computing in
that the quantities in brackets are real numbers. ( Note that
- = ^^-7 • \ They have been used also in obtaining the approxi-
mate formulas given in the body of the paper.
The quantities discussed above all pertain to the sheath. For
finding Z//, involved in the last of the formulas (9), the series are not
valid, of course. For this we have the well-knowii series,
1 /o(.rO 1
U,' Jo'{xO .Vi
^4 ^96 1536 ^ 23040
(50)
— see, e.g.. Gray, Mathews and McRoberts, " Bessel Functions,'
2d edition, p. 170.
Theory of Vibration of the Larynx^
By R. L. WEGEL
The vibration in the larynx is caused by an automatic modulation by
the vocal cords of the air stream from the lungs. Analytically the mechan-
ism is the same, and physically, closely analogous to that of the vacuum
tube oscillator. It depends principally on the resonance of the vocal
cords, the modulation of air friction in the glottis by their motion and
the attraction due to constriction of the air stream between them. When
these forces exist in certain relative proportions and phases, sustained
oscillation as in singing takes place. The whole mechanism may be rep-
resented analytically b>' force equations, from which conditions for accre-
tion or subsidence of the vibration or for sustained oscillation may be
easily deduced. These equations also show the analogy with other types
of oscillating systems.
IT is customary in treating the theory of the voice to assume the
glottis or space between the vocal cords to be a source of a steady
stream of air with superimposed periodic impulses caused by the
vibration of the vocal cords. The harmonic content of these impulses
is modified by the "resonating" vocal cavities before being radiated
into free air. It is the nature of this modification which receives most
attention. The mechanism by which the vibration of the vocal cords
is maintained has not been carefully studied.
The vocal cords are maintained in a state of sustained vibration
by the proper balance between the various mechanical constants of
the complete system, which thus act as a transformer of a part of
the non-vibratory power derived from the air stream from the lungs
into the vibratory power resulting in sound. It is a simple theory
of this mechanism which is considered here.
The method used is to obtain the force equations, which describe
the vibrations of the complete mechanical system, by means of the
Lagrange equations, from expressions of the total instantaneous
kinetic and potential energies, the instantaneous forces acting and
rate of dissipation of energy. The resulting simultaneous equations
relating to the displacements and velocities of the various parts are
then studied to find the frequencies of free vibration and the relations
which must obtain between the various mechanical parameters of
the system in order that one of these frequencies be sustained. The
method is an application of the theory of H. W. Nichols, published in
Physical Revinv, August, 1917.
The theory is reduced to easily workable form by the introduction
of simplifying approximations which will be described in the progress
1 Presented before Acoustical Society of America, May 11, 1929.
207
208
BELL SYSTEM TECHNICAL JOCK SAL
of the discussion. The principal one of these is the neglecting of
all reactions of second or higher order, thus leaving a set of linear
differential equations.
Structure of the Vocal Tract
The vocal tract consists of three principal parts, the lungs and
associated respiratory muscles for maintaining a flow of air, the
EPIGLOTTIS
FALSE VOCAL
CORDS
GLOTTIS
VOCAL CORDS
TRACHEA
Fig. 1 — Anterior-Posterior Section of the Larynx.
larynx (see Fig. 1) for producing the periodic modulation and the
upper vocal cavities, pharynx, mouth and nose for varying the rela-
tive harmonic content of sound originating in the larynx.
THEORY OF VIBRATION OF THE LARYNX 209
The capacity of the lungs in an adult man is capable of being varied
from about two to five liters. The av'erage in quiet breathing is
about 2.6 liters. The average expiration of air in quiet breathing is
about .5 liter. The rate of expiration of air in medium loud singing
varies from 40 to 200 cm. ^ sec, the lower values obtaining for trained
singers.
The larynx (see Fig. 1) consists of an irregularly shaped cartilaginous
box at the top end of a tube, the trachea, about 12 cm. long by 2 cm.
in diameter, leading from the lungs. The larynx contains the vocal
cords, a pair of fibrous lips which in vibrating vary the width of the
slit called the glottis, between them. The length of the glottis in
the adult male averages about 1.8 cm. and in the female 1.2 cm.
The width of the glottis varies widely with differing sounds. A few
tenths of a millimeter may be considered representative. The tension
and separation of the vocal cords are controlled by muscles.
The principal upper vocal cavities are the pharynx, a space just
over the larynx, the mouth and the nasal cavities. The first and
second may be varied in size and shape at will, but the effect on the
last is controlled only by varying the communicating aperture be-
tween it and the pharynx.
Equations of Motiox of the Larynx
Fig. 2 shows a cross-section of a model which illustrates the
essential details of the larynx in so far as it is necessary for this treat-
ment, ^o represents the area of the opening to the trachea. The
vocal cords are represented by elastically hinged members of com-
bined effective area S-^. By effective area is meant the area of aper-
ture which displaces the same Aolume of air as the vocal cords when
it moves the distance 50 of the tips of the cords. This area is less
than that of the vocal cords. The tips of the vocal cords are separated
to form a gap, the glottis, of area Su A positive or up and outward
displacement q-i of the vocal cords increases S^. It will be assumed
that the air is not appreciably compressed in the neighborhood of
the glottis, that is, any tendency to compression is relieved by flow
into the trachea or pharynx.
The pressure in the lungs forces a steady current of air through
the glottis. Let the velocity in the trachea of this steady flow be /o
and in the glottis /]. Small vibrations of the vocal cords superimpose
additional small velocities, Zo and /i, in the trachea and glottis re-
spectively. If the instantaneous velocity of the vocal cords be u
and it be assumed that they are constrained to move in synchronism
(/o + H)S, = {h + h)S, -f HS2. (1)
210
BELL SYSTEM TECHNICAL JOURNAL
The above material is a description of a simple model of two degrees
of freedom which simulates the principal characteristics from the
standpoint of performance of the more complex larynx which has
many degrees of freedom. It is this idealized model which will be
considered in the subsequent treatment. Such points of performance
of the actual larynx which may be due to the action of ignored and
Fig. 2 — Schematic Larynx Model.
presumably subsidiary modes of motion will, of course, not be pre-
dicted by the theory. These are assumed to be of minor importance.
The possible independence of motion of the two vocal cords will be
considered later, however.
The contraction of the air stream at the glottis introduces a rela-
tively large concentrated kinetic energy in the air stream at this point
similar to that at the mouth of a Helmholtz resonator. The inertia
of a small plug of air between the vocal cords may then to a first
approximation be treated as a mass L\. A concentration of frictional
resistance also occurs at this point due to viscosity and to turbulence.
A positive displacement q-z (outward) of the vocal cords causes an
increase in the mass of the plug of air in the glottis and a change in
the effective resistance, R, encountered by it. The inertia Z-/ and
resistance R of the glottis are therefore both functions of q-i, the dis-
placement of the vocal cords from a mean position, and of the width
of the glottis. If further Q^ represent the average displacement of
THEORY OF VIBRATION OF THE LARYNX 211
the vocal cords from an appropriately chosen reference position, L^
their inertia coefficient and K^ their effective stiffness, all measured
at the tips, the total kinetic energy, T, and potential energy, V, of
the larynx are
T = hL,'(A + hy + H2^2^ (2)
V = ^K,(Q2 + 92)-. (3)
The Lagrange equation of forces for the nth coordinate of any
system is
_ddT_dT dV
dtdtn dqn dqn,
in which Fn is a reaction due to friction. The force equations for
the glottis and vocal cords therefore become
17 P I ^ ^
dt dti
hU'ih + hY
(5)
^ = ^^^ + ^^ + ^^^^^ + ^^^ 2 ^ • ^^^
Nature of the "Constants" of the System
It is quite safe to conclude that none of the coefficients (inertia,
dissipation and stiffness) of the larynx are sensibly constant over the
range of operation of the coordinates. Direct measurements are
evidently impossible. It is conceivable that they may be arrived at
indirectly by means of a comparison of experimental data, especially
taken for the purpose, on voice curves and the results of dynamic
analysis of the kind described here. The problem may also be studied
by means of models. In order to solve equations 5 and 6 it is, how-
ever, necessary to evaluate the space and velocity derivatives.
A few simple experiments were performed on models for the sole
purpose of determining the qualitative nature of variation of resistance
of the glottis with displacement of the vocal cords. A diagram of
the model used in the measurements is shown in Fig. 3. This con-
sists of a brass tube, a, ^i" in diameter, beveled off on the top at an
angle of 45° with the axis, and two 3 s" brass plates, h, fitted on these
beveled surfaces so as to leave a slit, S, which was made adjustable
in width. A cross-section of this model is shown in c. The bottom
of the tube was attached to a large air chamber in which the pressure
and velocity of air flow could be regulated and measured.
Three shapes of "glottis" were measured. The first had square
corners, as shown on Fig. 3f. The second, M, was the same as ic,
212
BELL SYSTEM TECHNICAL JOURNAL
except that the corners of the Hps were rounded. The third, Fig. 3>e,
had square corners as before, but the sUt was about .1 mm. wider
in the middle than at the ends.
Fig. 3— Glottis Models.
The resistance R is given as the ratio of the product of pressure
and slit area to the linear velocity of flow. Measurements were made
in each case through a range of pressures such as to give fluxes through
the slit through a range of 50 to 200 cm.^/sec. (Stanley and Sheldon
values, see Sci. Am., Dec. 1924) and through a range of slit width W
of .01 to .10 cm. The data can be represented approximately in
this range for the three slits by the following formula?:
i? = 3.6 PW'-' X 10-«,
i? = 6.1 PW- X 10-«,
R = 800 /-n^--3 X io-«.
In these expressions / is the velocity of flow of air through the slit.
More careful data taken through a wider range of / and W would
undoubtedly have given i? in a power series.
These formula^ are taken to indicate that the resistance of the
actual glottis increases faster than a linear function of / and W due
to turbulence and may be represented as a single valued function of
either displacement of the vocal cords q-i (or glottis width) or of air
velocity as expressed by a Taylor's series as follows:
R = R,
, d,R
dq-i
, d,R .
+ 1
d^R ., , 2d,fR .
dqi" dqodti
+
do'R
~d7r
ir + etc. (7)
THEORY OF VIBRATION OF THE LARYNX 213
In this expression Rq is the resistance measured in the reference
position at which point the derivatives are taken, where ii and q-
are zero. The experiment mentioned above determines the signs ot
the coefficients of q-i and i\ as positive. If the flow were purely lami-
nar, i.e. due to viscosity only, the first would be negative and the
second zero.
The Reaction F^
By definition, £o = RJi, where £o is the force of the lung pressure
on the glottis and /i a corresponding linear velocity of fiow of air.
If a force Fi slightly greater than Eq act on the glottis and result in
a velocity I = h -\- i],
zh, = ^- («'
A combination of (7) and (8) constitutes an evaluation of Fi for sub-
stitution in the force equation (5). To a first order approximation
then :
/^, =J?,/, + (i?. + /,^)/, + /,^^,.. (9)
The coefficient of g-^ is dimensionally a stiffness and that of /'i a re-
sistance. In what follows they will be denoted by
F, = RJ, + Rrii + K„q.. (10)
Glottis Mass (L/) Reactions
The kinetic energy of the air stream being proportional to the volume
integral of the square of the velocity is largely concentrated in the
glottis on account of the relatively high velocity at this point. On
account of the irregularity in shape and turbulence in the stream it
is impracticable to attempt an integration. If the velocity were so
small that the turbulence were absent an approximate value of the
air mass would be obtained by taking the mass of a cylinder of air
having the length of the slit and a diameter equal to its width. This
would make the mass L/ proportional to W~, or since the width is
proportional to displacement of the vocal cords, to q-i'. Owing, how-
ever, to turbulence and other non-linearities, the mass is probably
more nearly described as a tongue of air issuing from the glottis,
the inertia L/ of which varies as some power function of the width
and also of the velocity. -
- It has been found since experimentally that the mass reaction is very nearly
that of a cylinder as described but reduced somewhat in diameter due to viscous
or turbulent drag at the tips of tiie "socal cords.
21-1 BELL SYSTEM TECHNICAL JOURNAL
It might be seen by carrying through an expression for this glottis
mass involving a function of velocity similar to that for R of equa-
tion (7) that only a quantitative change in effective mass would
result in the final equations and that no new type of reaction would
be introduced. This demonstration is not included here. In order
to save space in this qualitative treatment it is ignored. For small
displacements go from a reference position at which the velocity of
the air is /o, the glottis inertia may be represented by the direct
function:
Ly = Li +-3— ^2 + ^^-^^2- + etc., (11)
in which the coefficient of 92 is obviously positive. The second term
of the second member of (5) may now be evaluated by performing
the differentiations as indicated. Neglecting second and higher order
terms and denoting dq^/dt by /n the reaction in question becomes
The glottis mass of air, therefore, introduces two kinds of reactions:
a simple inertia and a reaction proportional to the velocity of the vocal
cords. For simplicity of notation (12) will be written
Uf^ + Gi,. (13)
This completes the evaluation of the terms (5), the force equation
of the glottis, which may now be written
£0 == Roh + RJ, + K,.q, + ^ + Gu. (14)
Force Equation' of the \'ocal Cords
The force equation (6) of the vocal cords contains four terms. The
first is the inertia reactance of the vibrating lips. The mass Lo is
the effective vibrating mass which, if multiplied by one-half the square
of the velocity at the cord tip, gives the kinetic energy of their motion.
If the distribution of the velocity in the vocal cords were known
this might be found by integration. The second term Fo in equa-
tion (6) represents the internal dissipation and is assumed propor-
tional to the small velocity u. The third term is the elastic reaction
which is proportional to displacement.
The fourth term is a " gyrostatic " term. This term ma>- be written
as follows:
THEORY OF VIBRATION OF THE LARYNX 215
- T- = - HA + h)- -^ — h -^-^ g^ + etc. • (15)
dq2 ^ dq-i dq-r )
Again by neglecting second and higher order effects this reaction
becomes
It will be seen that the first term of this expression represents a
static force tending, since it is negative, to draw the vocal cords
together. This is the BernoulU effect utilized in a venturi meter.
This steady force is counterbalanced by an elastic reaction of the vocal
cords with which it combines to determine an equilibrium position
which obtains when the cords are not vibrating. This term may,
therefore, be dropped from the fin'al equations representing only
superimposed motions.
The coefficient of i] is identical, except for a sign, with G of (13).
It represents a force on the vocal cords due to a superimposed part
of the Bernoulli effect caused by the small superimposed velocity ix
in the glottis. The coefficient of qo is dimensionally a stiffness. This
apparent stiffness is due to the nature of the air flow and is inde-
pendent of any elastic members. It is negative if the second differ-
ential of glottis mass with respect to cord displacement is negative,
positive when this coefficient is positive and vanishes when this
coefficient is zero. It simply adds or subtracts in effect from the
stiffness K-i of the vocal cords. The first possibility is the more likely.^
These terms may then be written for simplicity
-^= - F - Gh - K^q,. (17)
oq-i
Force Equations of the Larynx
The force equations of the glottis and vocal cords with constants
thus evaluated are
E, = Lr^ + R,H + Gk + R,U + /v„?o, (18)
= Lo^ + R^u + A>, + K.Q. - F - Gi, - K^q.,. (19)
As explained before, Eq = i^o^i and F = KoQn; so these cancel and
are of no interest here. In the following it will be seen that the field
^ This coefficient has since been found to be negative.
216 BELL SYSTEM TECHNICAL JOURNAL
stiffness Kj. is included in K^ to simplif\' notation. This leaves (18)
and (19) finally:
= L, ^ + R,H + Gu + K.,q,, (20)
= Lo^ + Roi. + K.q. - Gi\. (21)
It should be noted that these equations represent all first order
internal reactions of the idealized model of the larynx. The series
expansions have been carried out, to show to what approximations
these equations hold. It should also be pointed out that the effects
of mechanical hysteresis of the parts, which make the relative posi-
tions of the parts dependent on the previous history of their motion,
have not been considered. A consideration of hysteresis complicates
the theory considerably and is ignored for the same reason and with
the same justification and limitations that it is ignored in the ele-
mentary treatment of electrical circuits containing coils with magnetic
material and condensers with electrostatic hysteresis.
External Reactions of the Trachea and \'ocal Cavith^s on
THE Larynx
So far the modifying effect of the trachea and lungs, as well as the
upper vocal cavities, on the motion have not been considered. Before
using the equations it is necessary to evaluate these reactions and
add them in their proper places.
Imagine a weightless piston fitted into the trachea just below the
vocal cords such that the volume of air thus enclosed in the larynx
is so small in comparison to that of the trachea and lungs that its
compressibility may be neglected. If the vocal cords are held rigid
and the plug or piston of air in the glottis is forced inward, a reaction
in addition to the resistance and inertia of the glottis will be encoun-
tered due to the impeding effect of the trachea piston, which impe-
dance is determined by the constants of the lower chambers. If a
small force /o act on the trachea, causing a small velocity, to, and
we assume linearity of response /o = Zoio where Zo is a constant which
may, due to a positive inertia reactance or a stiffness, contributed
by air compression in the lungs, involve either a time derivative or
integral of displacement. For the present consider it to be a gener-
alized impedance operator. Due to the relative incompressibility of
the air in the larynx, the volume displaced by the trachea piston is
i^So = iiS]. Since also the instantaneous pressure inside the larynx
THEORY OF VIBRATION OF THE LARYNX 217
is constant on all its walls, including the surface of the trachea piston
f^/So = f.'Si. We then have
f\ = zMn. (22)
This reaction due to the trachea must be added to those of the glottis
given in (20). In like manner if the effective area of the vocal cords
is 52 a reaction h must be added to their force equation
h = zMi,. (23)
•Jo"
Due to the steady component of air flow there is a static component
of pressure tending to force the cords outward. This is counter to
the static Bernoulli term and again, if second order effects of small
quantities be neglected, serves only to alter the equilibrium position
and may therefore be disregarded here.
When the glottis plug of air is displaced inward a force is exerted
on the vocal cords tending to move them outward which is relieved
to a certain extent by a yield of the trachea piston. This force on
the vocal cords may be shown by reasoning similar to that above
to be
Zo%i-^\. (24)
Since this part of the system is linear, the reaction between glottis
and vocal cords through this channel is reciprocal so a force is exerted
on the glottis when the vocal cords are displaced of
Zo%|i2. (25)
It will be noticed that ^i is a variable because of the variation in
width of the glottis while vibrating. The effect of this variation in
these terms is obviously second order since ii is small and will therefore
be neglected.
The reactions of the upper cavities might be similarly added, but
they are apparently relatively small and since they are at present not
quantitatively known, are disregarded in the general equations be-
cause of the increased complexity. Generally, however, Zo may be
thought of as representing the additive effects of both upper and
lower chambers.
The complete force equations of the voice for small vibrations,
2 IS
BELL SYSTEM TECHNICAL JOi'RNAL
taking into account all major external as well as internal reactions,
mav then be written:
dix
'dt
A^ + R,H + Gh + K,^q,+ ^"'^'" -• ' ""^^
^'o^
ZqSiSo
^1 H rr-TT— ^2
6'o^
di
■'^n^j'i" • , JunOiOt .
i) = Lo-f-\- R2i2 + A'og. - Gi\ + -^~- k +
t,.
(26)
(27)
These equations may be put in a somewhat simpler form by virtue
of the fact that they are linear differential equations with constant
coefficients. In such a case the time differential may be replaced
by an algebraic operator p such that i = pq, di/dt = p^q, where p is
of the dimensions and nature of a frequency
= ifU + pR: + pZof,]qx +(pG + pZ,^+K„^q,, (28)
= (- pG + pZo^ jq,
+ ( p-U + pR-i + K. + />Zo|^;
The determinant of this system is (calling pZ^ = Fo)
(29)
D
p'L, + pR, + Yo-^^pG + Yo^ + A'„
- pG+ Yn^)U-'L. + PR, + Ao + F„|^n
(.SO)
This determinant represents the complete reactions of the larynx and
the external effects of communicating air chambers.
If the effects of the air chambers be disregarded the system is
represented by placing Fo == 0, giving the simple form
D
(p'U + pRi)
- pG
ipG + K„)
ip-Lo + pR.2 + K.)
(31)
Nature of This System
The voice system represented by determinant (3) is very closely
analogous to other vibrators, such as the microphone oscillator or
door buzzer and the vacuum tube oscillator. The literature on the
latter subject is now so extensive that the pointing out of the analogy
should make the method of solution for sustained oscillation, as in
singing, or for subsidence or accretion of the oscillation, as in speak-
ing, clear to any one familiar with it.
THEORY OF VIBRATION OF THE LARYNX
219
Fig. 4a is a schematic diagram of a three-element vacuum tube
oscillator circuit known as the "tuned grid" circuit. This is one of
many kinds. The transformer coupling between the plate and grid
circuit is represented by an auto-transformer. Fig. 4b represents
the same circuit schematically but with circuit elements only. In this
R2 represents that part of the resistance of the coil which belongs to
the grid circuit and any other associated dissipation, L2 is the in-
ductance of the coil as seen from the grid mesh and Ko the reciprocal
of the combined tuning capacity across the grid and that of the grid-
filament. It is the electrical stiffness or elasticity of the grid mesh,
K,
^
:^2
AAAAM-
nmr^
Fig. 4—" Tuned Grid " Oscillator.
in other words. Ri is a plate-filament resistance ("a.c") and Lj
that part of the coil in the plate circuit. M is the mutual inductance
of the transformer which is not part of the mesh impedance of either
plate or grid. The element K,, is the " uni-lateral mutual impedance "
(G. A. Campbell, 1914) between the plate and grid meshes and is
numerically equal to (jlK^, where ^ is the amplification constant of
the tube. Other internal tube impedances are as usual neglected.
The impedance determinant may be written directly from the circuit
diagram Fig. 3b.
{p'-L, + pR,) {^M + K„)
pHI ip-Lo + pR, + K.2)
D
(32)
The quantities on the principal diagonal of this determinant, that is
the first and last elements, are as usual in a circuit determinant the
mesh impedances while the others are the mutuals. The principal
features of the analogy may be seen by comparison of determinants
(31) and (32). Except for the thus far undefined external or trachea
impedance the mesh impedances are the same, from which it appears
that the glottis is analogous with the plate-filament path in the vac-
uum tube and the vocal cords with the grid-filament path. The air
220
BELL SYSTEM TECHNICAL JUl'RXAL
velocity in the glottis 7i corresponds to the plate current. In the
vacuum tube this plate current is modulated by varying charge, q-i,
on the grid. In the larynx the glottis air velocity is modulated by
varying displacement, q2, of the vocal cords. The charge on the
plate (again neglecting internal capacities except the grid-filament)
causes no effect on the grid mesh and in the larynx the position of
any element of glottis air has no effect on the vocal cords. The uni-
lateral mutual impedance, K,,, is the same in both.
The analogy breaks down at the point where the " feed back " part
of the mechanisms is compared. The " feed back" is the bilateral
part of the mutual impedance between the two meshes. In the
vacuum tube circuit this is p-M, the mutual of the transformer, while
in the larynx it is pG, the " gyrostatic " mutual. The latter is a type
of element which does not occur in electrical circuits, arising as it
^im(y^
mw
^^AAJW — I
^im^
■^ww
a b
Fig. 5 — Tuned Grid and Wind Reed Circuits.
does from a variation of a mass or inductance with a displacement.
Inductance, being a function purely of the geometry of a circuit,
can only vary with mechanical displacement and not with electrical
displacement or charge. The gyrostatic mutual is common in the
mechanics of rotating bodies whence it derives its name. It is also
the mutual in an electromagnetic telephone receiver or relay be-
tween the electrical circuit and the armature or diaphragm.
In order to fix the rather useful concept of the analogy in mind.
Fig. 5 is added showing the schematic circuit of the vacuum tube
(5a) and a circuit diagram (5b), which represents determinant (31)
the characteristic formulation of the dynamics of the larynx. Fig. 5b
is represented by the conventions of an electrical circuit, except for
the element G for which a different convention is necessary. The
one taken here is that of a resistance enclosed in a rectangle. From
(31) it will be seen to be similar to a resistance in its association with
frequency p but different from resistance in that it occurs non-sym-
metrically in sign in the determinant. It does not involve dissipation.
THEORY OF VIBRATION OF THE LARYNX
221
Its occurrence here is the simplest possible for when there are appre-
ciable concealed or ignored modes of motion it may have the form
of a generalized impedance containing at least one element of resist-
ance, but will always be non-symmetrical as a whole in sign in the
determinant.
The use of the circuit for representing the mechanical system is an
extension of an old but recently popularized method of studying
mechanical or electrical vibrating systems by the help of analogy,
one with the other. The extension consists in the explicit representa-
tion by diagram of the gyrostatic mutual which makes the deter-
minant unsymmetrical in sign and of the unilateral mutual which
makes the determinant unsymmetrical in magnitude. Fig. 6 is a
K.
■m^5^
AAAAA
^m^
AAAAA-
Fig. 6 — General Wind Reed Circuit.
diagrammatic representation of the more general system of deter-
minant (30). This includes the external Yq reactions as well as the
internal.
Having thus described the extended method of analogy the follow-
ing study of the larynx with the help of the circuit diagram of its
determinant should be clear.
Sustained Vibration of the Simple Larynx
In vibrating, the vocal cords do not receive excitation of the fre-
quency at which they vibrate. The source of power is in the air
stream /i which enters the equations in iv",,, the unilateral mutual
impedance. Since this is treated as a constant circuit or dynamical
element this air stream may be ignored as a drive and the resulting
15
222 BELL SYSTEM TECHNICAL JOURNAL
vibration considered as the free oscillation of the system. The de-
terminant (31) (or 30) is then used to determine the free frequencies
and decrements of the system. The method is as usual to solve for p
in the equation
D = {). {32,)
To simplify the demonstration the simple larynx without the load of
the air chambers will be considered. Taking D of (31) then and
expanding:
^^L,Lo + p\L,R, + L.Ri) + p\UK. + R,R. + G^)
+ p{R,K, + GK^,) = 0. (34)
If this be divided by Z1L2 and the uncoupled decrements and natural
frequency defined :
-^1; ;77- = Ao; -7- = coo-, (35)
= 0. (36)
2.Li\ 2.L11 Li
then
p"- + p\l^, + 2A2) ^ pi coo- + 4A,A2 + -^ )
One of the roots of this equation is zero and another is negative
real since all coefficients are positive. This root is therefore the decre-
ment of a mode of non-vibratory motion. The remaining two roots
may be real, imaginary or generally complex, of the form
Aija,. (37)
If it is found that A = 0, then an oscillation once started will be
sustained. If A be negative then any existing oscillation must sub-
side or if A be found positive then an impulse will start an oscillation
which of itself increases in amplitude to a point where its violence
modifies the constants to such an extent as to make A vanish, leaving
a sustained oscillation, or negative leaving the oscillation to subside
to a lower amplitude or completely if sufficient permanent changes
have been made.
If now (36) be written
Ap-' + Bp~ + Cp + D ^ Q (38)
and the tirst root (37) be substituted, two equations result, one from
the real and the other from the imaginary terms, as follows:
^A(A- - 3co-) + 5(A- - CO-) + CA + /^ = 0, (39)
^(3A- - CO-) + B{2^) + C = 0. (40)
THEORY OF VIBRATION OF THE LARYNX 223
Now the condition for sustained oscillation is that A = and if the
value of o) when this obtains be wo then
D C
coo- = -g and '^°^""J' (^^)
or the condition for sustained vibration in terms of the constants is
AD = BC. (42)
In addition to this if use be made of the fact that in an algebraic
equation such as (36) the coefficient of p- is the negative sum of all
the roots then this coefficient is the real root. Let this be Ao and
then (36) may be written
p^ + p'-^, + ;^coo- + Aocoo- = 0. (43)
The coefficient of p in (36) is therefore the square of radial frequency
at which sustained oscillation will take place and this is seen to be
higher than the natural frequency wa of the vocal cords, the difference
being increased when the damping of either mesh is greater or when
the coupling mutual G is greater.
It might be noted in passing that (43) is the free oscillation equation
for any system which may be represented by a cubic equation and
is not confined to the simple larynx. Such an equation always results
when there is only one kind of reactive element in one of the meshes.
It holds also for the tuned grid circuit.
The condition for sustained oscillation to be fulfilled for the con-
stants may from (42) be reduced to:
R,R, = G'[^- ij- (44)
It is rather difficult to place a simple physical interpretation on
this formula. The qualitative import of it may however be seen by
substituting the values of G and Ku from (13) and (10):
R\R'i = Li' I]'
doR/dqo doLi/dqo
doLi/dqo ,..
The first term in brackets is in the nature of a resistance modula-
tion constant, a fractional change in glottis resistance per unit dis-
placement of the cords, to be designated by r and the second term
similarly a glottis mass modulation constant, /. The quantity Lil^
is the momentum of the air in the glottis. This equation is then
i?ii?2 = {UU)\r - /)/. (46)
224 BELL SYSTEM TECHNICAL JOURNAL
Thus it appears that the resistance modulation must always be greater
than the mass modulation and when the difference is small the air
momentum must be increased to compensate. Owing to the physical
limitation in accuracy of continuous maintenance of adjustment in
the larynx, if a large momentum is depended upon to compensate for
a small modulation difference, an unsteadiness or instability is likely
to result. It is common experience that it is impossible to produce a
sound with the voice with less than a certain minimum intensity.
This corresponds, with the most favorable adjustment of the modu-
lation constants which are physically possible, to a minimum momen-
tum of air from the lungs which satisfies (46). It will be evident
that this interpretation must not be taken too seriously quanti-
tatively.
Subsidence and Accretion of Vibration of the Simple Larynx
Oscillograms made of the speaking voice show that, among other
things, the amplitude of the oscillation and the pitch are in a con-
tinuous state of change. This is also true in singing but not nearly
to the same extent. It seems therefore that in singing the adjust-
ment of the voice system for sustained oscillation as described in
(44) above is of major importance, while in speaking conditions for
variation are of most importance.
The principle of the investigation of variation is simple enough
but in all but the most elementary systems the algebra involved is
impracticably awkward. If by solving {ii) directly for the roots of p,
it be found that A is positive, then any existing vibration will tend to
increase while if A is negative, then vibration will tend to subside.
The algebraic difficulties arise in the general solution but these are
largely obviated by making the assumption, which is most likely
usually fulfilled in practice, that the real parts of the roots may be
treated as small quantities when compared with the imaginary parts.
A common frequency for a man's voice is 150 cycles per second for
which coo is 1000 in round numbers. The decrement of a telephone
receiver is ordinarily 100 to 200 in open air. The decrement of a
tuning fork is represented by a fraction. Judging from variations
in amplitude in an oscillogram (from which of course decrements
may not be read directly) it would seem reasonable to assume that A
is small compared with wo- The study of variation thus becomes an
investigation of small departures from a condition of sustained oscil-
lation, the reference condition being that critical adjustment for which
the roots of interest of {ii) are pure imaginary.
THEORY OF VIBRATION OF THE LARYNX 225
Suppose in (38) that A = \; then without loss in generahty:
pz ^ Bp"' + Cp + D = 0. (47)
In such an equation the roots are continuous functions of the co-
efficients. The same is true of their derivatives except at the one
point where transition occurs from pure real to complex values. The
values of the roots of interest in this connection are in their complex
region at the point where the real part of the root passes through a
zero value. This is the point at which free oscillation of the oscil-
lating mode occurs, the values of the roots of this mode being as shown
before, ± jwo.
If it now be supposed that one cause produces small variations,
directly or indirectly on each of the coefficients and that the magni-
tude of this cause be .r, then :
(3,= + 2., + C)g + .= f + .f + f = 0. (48)
The problem then is to determine dp resulting from any assigned
cause dx when p = jooo- From (43) we have at this point B = Aq,
C = coo" and D = Aqcoo^.
.Ao\ dp ,dB , . dC , dD
This is the frequency (complex) variation equation taken in the
neighborhood of free oscillation.
When any readjustment of the larynx takes place all of the "con-
stants" entering the coefficients undergo change, in particular those
of the glottis Ku, Ri, G. Suppose for simplicity that one only varies,
then this variation dKu, dRi, or dG may be taken as the magnitude
of the cause dx. In particular if Ku vary,
dB = = dC and dD/dx = G/L^L^,
.Ao\ dp G
coo / dKu IwrfLiLo
If in addition Ao be small compared with coo,
. GdKu /, , .Ao\
dp = -> or J- 1 +J— • (.-^1)
This shows that if a condition of sustained oscillation is departed
from by slightly increasing Ku, an increase in the amplitude of vibra-
tion begins which is proportional to the logarithm, since {p -\- dp)
226 BELL SYSTEM TECHNICAL JOURNAL
is the exponent, of the increment dKu and the frequency (imaginary
part) of vibration increases sHghtly in proportion. If K,, were the
only varying element the vibration would continue indefinitely to
increase.
If on the other hand K„ be assumed constant, the variation being
in Ri, then it may be similarly shown that
dp =1— T,
- ( 4A, + -^ + Ao-^ ) + /""-^'^
Li\Li2 I COq
(52)
whence it appears that a small increase in glottis resistance dR\ (or
(/Ai) introduces a subsidence of vibration but an increase in frequency
of oscillation as before. A decrease — dR^ of course produces the
opposite efifect.
If the change be in G, it turns out that
dG
dp =
ZcOo'LiyLiO
A^-2GAo ) + /•( ^^^ + 2couG
Wo
(53)
Here it appears that an increase in the gyrostatic mutual, G, may
introduce either a subsidence or an accretion in amplitude but like
the others makes for an increase in frequency of oscillation.
Variation in other elements produces similar conflicting tendencies
not only in damping but in frequency.
The physical picture to be drawn from this is that in speaking the
voice modulates from one amplitude and frequency to another by
proper relative variations in adjustments in its constants, being con-
stantly in a state of changing subsidence or accretion. It would
seem also that the principal cause of change in frequency is in the
vocal cords and that of amplitude variation in the glottis. Speaking
is, in this respect, a more intricate process than singing.
Other Types of " Feed Back"
The detailed study of the larynx has so far been limited to the
assumption that the " feed back" is entirely gyrostatic. This is of
course actually not the case. How much influence is exerted by the
general Yq is difficult to estimate.
If the trachea were a long tube but still shorter than a quarter
wave-length of sound at the frequency of oscillation and rather smaller
in diameter, and substantially open at the end the mass of the air
in it would then be appreciable and Fo in (30) would be written p''Lo.
If in addition the gyrostatic term were negligible the system would
then be exactly analogous with the tuned grid system and (32) rather
than (31) should be the subject of detailed study.
THEORY OF VIBRATION OF THE LARYNX 111
If on the other hand the lungs acted substantially as a solid walled
chamber of comparatively small size, the elasticity of the contained
air would be represented by taking K^ for Fo. The surface area in
the lungs is very large compared with a regular chamber of equal
volume so considerable dissipation must be encountered by vibration.
If this were the most important reaction Fo should have been replaced
by pR^.
Unquestionably all three types of reaction enter. A more general
treatment to include them is plainly not a subject for a short paper.
It is interesting however to note that in the dynamical system of
brass horns these latter Fo reactions exert controlling influences. In
this case the lips of the player perform the same function as do the
vocal cords of the voice while the external load, the horn, corresponds
to the pharynx, the reaction of which is the same dynamically as
the trachea. In this case the frequency of the horn is that of sustained
oscillation and not that of the lips. The same is true of the wood-
wind, in which case the reed or reeds replace the lips or vocal cords.
In these cases Fo is proportional inversely to the hyperbolic tangent
of the frequency or may be approximately represented by the im-
pedance of an anti-resonant element.
Abstracts of Technical Articles From Bell System Sources
Notes on the Effect of Solar Disturbances on Transatla?itic Radio
Transmission.^ Clifford N. Anderson. In 1923 when the relation
between abnormal long-wave radio transmission and solar disturbances
was first noted, the outstanding abnormality was the great decrease in
night time signal field strength accompanying storms in the earth's
magnetic field. There was a slight increase in daylight signal field but
this was distinctly secondary to the efi^ect upon night field. Previous
to 1927, data on signal fields were limited to one set of measurements
a week, and although daylight signal field strengths were higher during
periods of increased magnetic activity, it was somewhat difficult to
determine the efi"ect of individual storms. The present notes show the
efl^ects of individual storms of 60-kc transatlantic radio transmission
and also give some indication as to their efi'ect on short-wave radio
transmission.
The Mutual Impedance Between Adjacent Antennas} Carl R.
Englund and Arthur B. Crawford. The simple theory for the
computation of reflecting or multibranch antenna systems is sketched.
If the points at which observations of electrical quantities are to be
made are definitely specified, a knowledge of the self and mutual
impedances (properly defined) between antennas is sufficient to make
the computations determinate. Of the circuit constants, the most use-
ful and accessible is the antenna current ratio
and in the work here reported has been measured in the range 0.33 X
to 1 X. Experiment has shown that in this range is that theoretically
calculable for a Hertzian doublet. Actually this range is equivalent to
X/3 to 00 . The discussion of experimental procedure is purposely
thorough.
An Experimental Method for the Determination of the Ballistic De-
magnetization Factor.''^ Donald Foster. A method is described for
experimentally determining the ballistic demagnetization factor. By
means of a double search coil ot novel design the magnetization and
1 Proceedings of the Institute of Radio Engineers, September, 1929.
- Proceedings of the Institute of Radio Engineers, August, 1929.
^Philosophical Magazine, September, 1929.
228
ABSTRACTS OF TECHNICAL ARTICLES 229
the magnetic field intensity are determined from ballistic galvanometer
deflections. While the discussion refers mainly to circular cylinders,
the scheme is adaptable to specimens of other shapes. It is particularly
designed to obtain accurate measurements of field intensity in cylinders
of small diameter.
Details of a special design are given.
Curves are given which illustrate the variation of the demagnetiza-
tion factor with the magnetization, as well as the dependence of this
relation on the material and on the dimensional ratio.
The Use of Continued Fractions in the Design of Electrical Netivorks.^
Thornton C. Fry. In U. S. Patent No. 1,570,215 and in several
technical papers by Bartlett and Cauer it has been shown that con-
tinued fractions can often be used in designing networks with pre-
assigned impedances. The chief difficulty of the method has been that
it frequently required the structures to contain negative resistances,
inductances or capacities and therefore the results, though correct in
theory, were often worthless in practice because the networks could
not be constructed.
The present paper removes this difficulty in virtually all cases w^here
the analytic character of the desired impedance is known, that is,
where it can be represented by a formula and not merely by a graph.
In such cases the choice of a type of structure, as well as the assignment
of values to the elements, becomes almost a matter of routine with the
definite assurance in advance that no negative elements will be
required.
A Voltage Regulator for Gas Discharge X-Ray Tubes} F. E.
Ha WORTH. This note describes a device used in connection with a gas
discharge x-ray tube, to regulate the voltage across it by automatically
adjusting a mercury valve between the tube and the pumps, thus con-
trolling the pressure of the gas. It has been used with tubes of the
Hadding and Shearer types and has operated satisfactorily for more
than a year. It was designed to replace the regulator described by
Bozorth, which is similar in principle but has certain disadvantages,
for example the moving parts have high inertia and adjustment is
required when the atmospheric pressure changes.
The Significance of the Hydrogen Content of Charcoals/' H. H.
LowRY. Most studies of the thermal decomposition of hydrocarbons
^ Am. Math. Soc. Bull., July-August, 1929.
'" Journal of the Optical Society of America, August, 1929.
s Journal of Physical Chemistry, September, 1929.
230 BELL SYSTEM TECHNICAL JOURNAL
are confined to an examination of the composition of the Hquid and
t^aseous products. Among exceptions to this generalization may be
mentioned the interest in coke, carbon black, and charcoal. Even in
these cases the physical properties rather than the chemical composi-
tion are regarded as the factors which determine their suitability for
specific uses. However, in an earlier paper it was pointed out that
certain physical properties of a group of charcoals were rather simply
related to the per cent hydrogen which was contained in them as de-
termined by ultimate analysis. This group of charcoals was prepared
in a gas-fired furnace from a single, specially-selected lot of anthracite
coal. As stated in this earlier paper, careful consideration of the
commercial records taken at the time of preparation indicated that
the hydrogen content was probably determined by the maximum
temperature to which the samples were heated during their preparation.
The hydrogen contents ranged from 0.21 to 0.53%, while the probable
range of maximum temperature was 900° to 1200°. The presence of
hydrogen in these charcoals was shown to be consistent with a point
of view that so-called "amorphous" carbons are hydrocarbons of low
hydrogen content built up of polymerized residues from the thermal
decomposition of hydrocarbons of greater hydrogen content. Since
the significance of the hydrogen content of charcoals has been generally
overlooked, the present study was undertaken in order to evaluate the
factors which may ordinarily be varied in the preparation of charcoals
for various purposes. The factors which were independently varied in
this study were the maximum temperature, the time of heating, the
atmosphere surrounding the sample during heating and the raw
material. To a limited extent the effect of previous heat treatment
was also determined. A later paper will give the results of the study of
the correlation of hydrogen content and some adsorptive properties of
charcoals prepared under carefully controlled conditions.
Btginnings of TelepJiovyJ Frederick Leland Rhodes, Outside
Plant Development Engineer, Department of Development and Re-
search, American Telephone and Telegraph Company.
It is only within the past decade or so that science and business have
become subjects for literature. Somehow these great phases of human
endeavor have been sadly neglected in the literary world until very
recently, and now it seems as though, conscious of the lack of good
literature in these fields, engineers, scientists and business executives
are making up for lost time. Frederick Leland Rhodes has written a
new book which undoubtedly will be of great assistance to those in the
' Harper & Brothers, New York and London, 1929.
ABSTRACTS OF TECHNICAL ARTICLES 231
telephone industry, for it supplies them with an accurate picture of the
technical background of a great industry. It is greatly to the ad-
vantage of an individual to know the history of his own business, and
Mr. Rhodes has supplied it in an interesting form, thoroughly accurate
and readable. No effort has been made to set down the more recent
achievements in the world of telephony, but only to carry each chapter
to what might be termed the "middle period" in development. There
are many phases of the telephonic art which have not been touched
upon in the volume, but at the same time, one is not conscious of any
lack in this respect as one reads through its interesting pages.
Any volume is the better off for illustrations, and Mr. Rhodes' book-
is generous in that it carries fifty-four illustrations scattered through
260 pages.
The first portion of the book naturally deals with Alexander Graham
Bell and occupies three chapters. Following this we have two chapters
called "The Bell Patents." As General John J. Carty, Vice President
of the American Telephone and Telegraph Company, says: "Never
before had the claims of an inventor been subjected to such exhaustive
litigation and judicious scrutiny, and never before d'd an inventor
receive such a complete and dramatic vindication." The remainder
of the fourteen chapters deals with the truly romantic progress of
telephone plant, its improvements and expansion over a term of years
when telephony was young and the road was fraught with difficulty.
Of special interest are the numerous references to original and authentic
sources, and in this regard the author has unquestionably used great
care and much labor in order to give his reader the most accurate
information possible, thus more truly gaining his end of supplying a
concrete picture of the younger days of a great industry.
Mr. Rhodes' volume is a great contribution, not only to the literature
of telephony, but also to that rapidly growing library which contains
in its pages the romance of business in America. As a library reference
book it will be valuable to the technical student. Any member of the
Bell System would do well to familiarize himself with this work, not
only because it will help him in his job, but because he will find it a
really interesting story.
Further Note on the Ionization in the Upper Atmosphere.^ J. C.
ScHELLENG. In this paper Mr. Schelleng records certain considera-
tions that were omitted from a previous paper, which omission resulted
in some difficulty.
* Proceedings of the Inslilute of Radio Engineers, August, 1929.
232 BELL SYSTEM TECHNICAL JOURNAL
Transmission Networks and Wave Filters.-' T. E. Shea. In this
book is summarized the research and experience of the Bell System in
the application of electric wave filters, equalizers, balancing networks
and similar electrical systems. The preface discusses the nature of the
signals transmitted over communication systems and a statement f)f
the principal ways in which selective networks are used to modify sig-
nal transmission. A detailed example of the application of selective
networks to an actual long distance telephone circuit gives specific en-
gineering requirements and limitations.
The next portion of the book deals with some of the more general
principles governing network analysis. The engineering terms used to
evaluate network performance are described and a number of general
theorems and equivalences which simplify the analytic treatment of
networks are demonstrated. A considerable discussion is also given of
the characteristics of the elementary two-terminal networks most used
as constituents of larger structures.
With this background the author is now ready to consider the proper-
ties of wave filters. Conditions for free transmission and attenuation
in ladder networks are set up and the particular networks of chief
practical importance are described in detail. The various structures
revealed by this listing differ widely among themselves as regards
propagative and impedance characteristics even when they transmit
the same frequency bands. Since the ideal network characteristics
seldom correspond exactly to any one of these structures, filter re-
quirements are usually met most efficiently by composite networks,
containing sections of several different types. The author describes
the conditions which must be satisfied before different sections are
joined together and gives several examples of methods of computing
the performance of such composite structures.
This treatment of networks deals only with their response to steady
single-frequency electrical impulses. It cannot be applied directly to
communication systems, since signals are of more complicated wave
forms and are transient in character. In the last portion of the book
therefore, the author discusses the use of Fourier analysis in relating the
characteristic of the network computed on a steady-state basis to its
response to a transient impulse of arbitrary character.
Some Principles of Broadcast Frequency Allocation}'^ L. E. Whitte-
MORE. This paper discusses some of the technical factors which must
be considered in the allocation of frequencies to broadcasting stations
8 D. Van Nostrand Company, New York.
^^Proceedings, Institute of Radio Engineers, August, 1929.
ABSTRACTS OF TECHNICAL ARTICLES 233
in such a way as to provide the best possible coverage of a given country
or continental area.
A given frequency or channel can be used for either of two kinds of
service; (1) by one station, exclusively, to give high grade service to
the immediate locality and opportunity for service over broad rural
areas when transmission conditions are good, and (2) by two or more
stations simultaneously, to give local service to a number of separate
regions, each of rather restricted area. The problem, therefore, in-
volves a determination of (1) the proper balance between the two kinds
of service, rural and urban, and (2) the proper basis for the apportion-
ment of the assignments.
Reference is made to the basis of apportionment of radio broad-
casting assignments laid down in the U. S. Radio Act of 1927, and to
certain suggestions which have been made for the apportionment of
broadcasting frequency assignments among the countries of Europe.
A brief discussion is given of the relation between field intensity, or
signal strength, and distance of transmission at broadcast frequencies.
The paper also discusses briefly the effects produced in the case of (1)
a single station operating exclusively on a "clear" channel, and (2)
two or more stations operating simultaneously on the same channel.
It is suggested that the distribution of assignments on "clear"
channels, in a given continental area be made proportional to the
population of each of several large geographical units or zones and that
the distribution of assignments on "multiple assignment" channels
be made to comparatively small geographical units in proportion to
their areas.
Contributors to this Issue
John R. Carsox, B.S., Princeton, 1907; E.E., 1909; M.S., 1912;
American Telephone and Telegraph Company, 1914-. Mr. Carson
is well known through his theoretical transmission studies and has
published extensively on electric circuit theory and electric wave
propagation.
A. B. Clark, B.E.E., University of Michigan, 1911; American
Telephone and Telegraph Company, 191 1-. Toll Transmission De-
velopment Engineer, 1928-. Mr. Clark's work has been largely con-
cerned with toll telephone and telegraph systems.
Karl K. Darrow, B.S., University of Chicago, 1911; University
of Paris, 1911-12; University of BerUn, 1912; Ph.D., University of
Chicago, 1917; Western Electric Company, 1917-25; Bell Telephone
Laboratories, 1925-. Dr. Darrow has been engaged largely in writing
on various fields of physics and the allied sciences. Some of his earlier
articles on Contemporary Physics form the nucleus of a recently
published book entitled "Introduction to Contemporary Physics"
(D. V^an Nostrand Company). A recent article has been translated
and published in Germany under the title "Einleitung in die Wellen-
mechanik."
Bancroft Gherardi, B.Sc, Polytechnic Institute, Brooklyn, N. Y.,
1891; M.E., Cornell University, 1893; M.M.E., Cornell University,
1894. New York Telephone Company, Engineering Assistant, 1895-
99; Traffic Engineer, 1899-1900. New York and New Jersey Tele-
phone Company. Chief Engineer, 1900-06. New York Telephone
Company, and New York and New Jersey Telephone Company,
Assistant Chief Engineer, 1906-07. American Telephone and Tele-
graph Company, Equipment Engineer, 1907-09; Engineer of Plant,
1909-18; Acting Chief Engineer, 1918-19; Chief Engineer, 1919-20;
Vice President and Chief Engineer, 1920-. Mr. Gherardi is a Past
President of the American Institute of Electrical Engineers.
Frank B. Jewett, A.B., California Institute of Technology, 1898;
Ph.D., University of Chicago, 1902. American Telephone and Tele-
graph Company, Transmission and Protection Engineer, 1904-12.
Western Electric Company, Assistant Chief Engineer, 1912-16; Chief
Engineer, 1916-21; Vice President and Chief Engineer, 1921-22;
Vice President, 1922-25. International Western Electric Company,
234
CONTRIBUTORS TO THIS ISSUE 235
Vice President, 1922-25. Manufacturers Junction Railway, Vice
President, 1922-25. American Telephone and Telegraph Company,
Vice President, and Bell Telephone Laboratories, President, 1925-.
Dr. Jewett is a Past President of the American Institute of Electrical
Engineers.
Francis F. Lucas, Associated Bell Telephone Companies, 1902-10;
Western Electric Company, 1910-25; Bell Telephone Laboratories,
1925-. Mr. Lucas has specialized in the development and appli-
cation of microscopy. He has received international recognition
and awards for the development of high power metallography and
ultra-violet microscopy and for numerous scientific papers which he
has contributed on the subjects of metallurgical and biological re-
search. For several years he has been Consulting Technical Expert
for the War Department, U. S. A., Watertown Arsenal.
Edward L. Nelson, B.S. in E.E., Armour Institute of Technology,
1914; Western Electric Company, 1917-25; Bell Telephone Laborato-
ries, 1925-. As Radio Development Engineer of Bell Telephone Lab-
oratories, Mr. Nelson is responsible for the development and design of
commercial radio apparatus, which includes radio broadcasting equip-
ment.
R. L. Wegel, A.B., Ripon College, 1910; Assistant in Physics,
University of Wisconsin, M. A., 1910-12; Western Electric Company,
1914-25; Bell Telephone Laboratories, 1925-. Mr. Wegel has written
several papers on theory of telephone receivers and on the theory of
hearing. The article appearing in this issue is taken from lecture
notes on Mechanics of Vibrating Systems by the author. It is planned
to publish these notes in future issues of the Bell System Technical
Journal.
M.K. ZiNN, B.S. in E.E., Purdue University, 1918; American Tele-
phone and Telegraph Company, 1919-. Mr. Zinn's work has been
related particularly to the design of loading for telephone circuits.
The Bell System Technical Journal
April, 1930
Developments in Communication Materials^
By WILLIAM FONDILLER
The subject of engineering materials is one of increasing importance, as is
evidenced by the expenditure of over a half bilhon dollars annually in new
construction by the Bell System. This has led to the concentration of the
research and engineering work on materials in a group devoted particularly
to this field of activity. Studies of the chemical and physical properties of
materials must be combined by the materials engineer with a knowledge of
the operating requirements of telephone apparatus.
The paper covers broadly the materials used in communication engineer-
ing and gives instances in which the needs of the telephone plant imposed
requirements which were not satisfied by commercially available materials.
Some of the instances cited are phenol fiber having improved resistance to
arcing for use in sequence switches; a composite molded plastic for use in
terminal strips; textile materials for central office wiring treated to improve
their electrical insulating quality and non-ferrous metals of more uniform
rViprarf prictiVc Prnhlfime inirnlA/incr tVip nisp n( duralumin fnr radin hrnarl-
CORRECTION SLIP FOR ISSUE OF JANUARY, 1930
Page 153: Equation (10) should read
//(/) ~
{^-h(D-^rii:r--}^'
//(/) ^
S(\t) (10)
of the signal, sound or scene to distant points, or their recording.
Up to about ten years ago the average manufacturer left to his
designing engineer the problem of selecting and testing the materials
which were to be embodied in a design, and he in turn was dependent
on the manufacturers of raw materials as to the variety and quality
of the materials available. Without depreciating the ability or ini-
tiative of manufacturers of engineering materials, it will be evident that
the special needs of a particular industry would, in general, not be as
fully appreciated by an outside manufacturer as by an engineer working
1 Presented before A. I. E. E. on November 13, 1929.
237
16
The Bell System Technical Journal
April, 1930
Developments in Communication Materials^
By "WILLIAM FONDILLER
The subject of engineering materials is one of increasing importance, as is
evidenced by the expenditure of over a half billion dollars annually in new
construction by the Bell System. This has led to the concentration of the
research and engineering work on materials in a group devoted particularly
to this field of activity. Studies of the chemical and physical properties of
materials must be combined by the materials engineer with a knowledge of
the operating requirements of telephone apparatus.
The paper covers broadly the materials used in communication engineer-
ing and gives instances in which the needs of the telephone plant imposed
requirements which were not satisfied by commercially available materials.
Some of the instances cited are phenol fiber having improved resistance to
arcing for use in sequence switches; a composite molded plastic for use in
terminal strips; textile materials for central ofilice wiring treated to improve
their electrical insulating quality and non-ferrous metals of more uniform
characteristics. Problems involving the use of duralumin for radio broad-
casting transmitters and the light valve used in sound pictures are also de-
scribed. Particular emphasis is laid on the benefits resulting from the con-
tinuous research in magnetic materials which have produced successively —
powdered electrolytic iron cores for loading coils, permalloy, and recently
perminvar.
Summing up, the work on materials has resulted in benefits along two
general lines:
1. Improvement in quality of commercial materials.
2. Discovery or development of valuable new materials.
THE subject of this paper, "Developments in Communication
Materials," perhaps needs some definition with the rapid addition
of new fields to the pioneer arts of telegraphy and telephony. Today
we must include high frequency wire telegraphy and telephony by
means of carrier currents, radio, telephotography, television and, in a
sense, sound pictures. All of these modes of communication of intelli-
gence are characterized by the use of electrical means for the transfer
of the signal, sound or scene to distant points, or their recording.
Up to about ten years ago the average manufacturer left to his
designing engineer the problem of selecting and testing the materials
which were to be embodied in a design, and he in turn was dependent
on the manufacturers of raw materials as to the variety and quality
of the materials available. Without depreciating the ability or ini-
tiative of manufacturers of engineering materials, it will be evident that
the special needs of a particular industry would, in general, not be as
fully appreciated by an outside manufacturer as by an engineer working
1 Presented before A. I. E. E. on November 13, 1929.
237
16
238 BELL SYSTEM TECHNICAL JOURNAL
on these problems. Thus it has come about in the Bell System, as with
other large consumers of materials, that the investigation of materials
has been organized as a distinct branch of research and engineering
activity. Studies of the chemical, physical and metallurgical proper-
ties of materials are embraced in this work. In general the materials
engineer should not only be well versed in materials, but should also
have a good knowledge of the operating characteristics of the apparatus
to be designed. Thus he can discuss the materials side of the problem
with the designing engineer on equal terms and make his contribution
to the best advantage. The importance of a thorough knowledge
of materials in the telephone business will be appreciated from the fact
that, during 1929, it is estimated that about $590,000,000 will be spent
for additions to the Bell System plant.
In telephony the general introduction of the dial system has imposed
more severe requirements than heretofore because of the need for the
utmost in reliability of performance of the large number of switches,
relays, etc., which are required to operate automatically with a mini-
mum of maintenance. In the central ofihce small size of apparatus con-
stitutes a very important consideration, not only because of building
space required, but the mass and travel of the automatic switches have
an important effect on the speed with which connections can be es-
tablished and hence on economy of operations. Thus, close control
of the quality of materials and the need for small, compact apparatus
are important design considerations.
In a brief survey of progress in the development of materials, it will
be necessary to select a few typical items of interest. The items
selected deal primarily with the telephone problem as this is, at the
present time at least, the largest single factor in the communications
group. The subject may be divided broadly into insulating materials
and metallic materials.
. Insulating Materials
Phenol Fiber
Considering first sheet insulating material, we have been using the
term "phenol fiber" to cover such materials as bakelite-dilecto, mi-
carta, formica and similar fibers made by various manufacturers.
Phenol fiber is used extensively in telephone apparatus. One of its
applications is in the sequence switch which has insulators alternating
with conducting segments, as shown in Fig. 1. The sequence switch,
which is used in the dial system, draws out an arc when in operation
which sometimes causes carbonization of the insulators. In some cases
a hole was burned through the insulator and in other cases the arc was
DEVELOPMENTS IN COMMUNICATION MATERIALS 239
Fig. 1 — Sequence switch, used in dial system.
Fig. 2 — Detail of apparatus for arcing test of phenol fiber.
240
BELL SYSTEM TECHNICAL JOURNAL
sustained over the insulation to such an extent that the circuit was not
broken at the proper moment. An examination of the various grades
of phenol fiber commercially supplied indicated that they varied widely
as to their resistance to arcing. Fig. 2 shows testing apparatus de-
signed to evaluate this characteristic.
The sample under test was made into a sequence switch cam and
rotated on the fixture at a speed of 10 r.p.m. The set is wired to give a
circuit condition comparable with that causing failure in service, except
that slower speed and higher voltage are used to accelerate the test.
The position of the rear brush is so adjusted that after the material
has become carbonized through an arc of 15 degrees or a hole has been
burned through the insulation, the machine would be stopped by means
of a circuit breaker, shown in Fig. 3. This instrument makes the
Fig. 3 — Assembly of apparatus for arcing test
failure value independent of the operator's judgment, and has proven
so satisfactory that it has been employed for specification purposes.
Fig. 4 shows insulators tested by this instrument; those at the top
having been rejected, and those at the bottom being satisfactory. An
improvement of 20 to 1 in arcing characteristics was obtained. This
was brought about by close cooperation with the Bakelite Research
Laboratories, which developed a special grade of resin to be used in the
manufacture of this material. In this case the materials engineer
developed a method of test for evaluating the particular quality de-
sired which enabled the supplier to improve his product in the desired
respect.
Even though resistant to moisture in the ordinary sense, phenol
fiber absorbs a certain amount of moisture depending on the quality
of the material furnished. As this moisture is given up, the material
DEVELOPMENTS IN COMMUNICATION MATERIALS 241
Fig. 4 — Insulators subjected to arcing test.
Top — F"ailure value, 20 rev.
Bottom — Failure value, 1200 rev.
%
Fig. 5 — Telephone relay showing phenol fiber insulators between contact springs.
242
BELL SYSTEM TECHNICAL JOURNAL
tends to shrink. If the fiber is not sufficiently hard as manufactured,
it will also flow under pressure.
In telephone relays of a commonly used type, illustrated by Fig. 5,
the contact springs are insulated from each other by thin sheets of
phenol fiber, and any material change in dimensions of these insulators,
due to moisture absorption or cold flow, will alter the spacing of the
contacts, thus throwing the relay out of adjustment. To measure
these tendencies on materials used in spring pile-ups, we use the method
illustrated by Fig. 6. It will be seen that a Brinell machine, usually
Fig. 6 — Modified Brinell machine for flow-test of insulator laminations.
employed for metals testing, has been modified to use a flat-ended
plunger resting on a pile of insulating material. The test material
is first cut into pieces 'jA" square and then subjected to atmospheric
conditions which would cause it to take up an amount of moisture
comparable to that expected under manufacturing conditions. The
pieces are then stacked and a pressure of 2,000 pounds per square inch
applied. The testing apparatus is installed in a heat insulated box
DEVELOPMENTS IN COMMUNICATION MATERIALS
243
Fig. 7 — Flow-test apparatus of Fig. 6 enclosed for temperature control.
7
6
5 5
^^
^AP^
^^
'^■——-^
Z
1-
O
lU
i:^ 3
/
^
/
f
PRESSURE 2000 LB
PER SQUARE INCH
TEMPERATURE 120° F
UJ
o
2
/
/
P
HENOL
. FIBRE
1
U
^
f
24
48 72
TIME IN HOURS
96
120
Fig. 8 — Flow-test results for hard rubber and phenol fiber.
244
BELL SYSTEM TECHNICAL JOURNAL
shown in Fig. 7, so that the temperature throughout the 24 hour test
may be maintained at 120° F. corresponding to the maximum Hkely to
be experienced in service. The amount of shrinkage or flow is meas-
ured on the dial previously shown. Fig. 8 shows the relative per-
formance of hard rubber and phenol fiber under the conditions of this
test.
Molded Plastics
In recent years there has been great activity on the part of manu-
facturers of molded plastics to develop improved molding compounds,
and we have endeavored to keep informed of new developments by
examining new compounds as they became available. An interesting
problem presented itself in the application of suitable molding com-
pounds to a device known as a test strip, shown in Fig. 9. It will be
Fig. 9 — 100 point test strip used in switchboards.
seen that it consists of a number of metal terminals mounted flush on
the face of the strip and projecting at the back to provide soldering lugs
for the central office wiring. In operation it is necessary to touch a
metal contact plug to the appropriate test strip contact which will pro-
duce an audible signal in the operator's receiver. In passing the plug
over "live" terminals an arc is drawn out, which is accentuated by a
habit of some operators of running their pencils along the grooves
leaving a conducting path. Such arcs caused permanent conducting
paths in the surface of the bakelite, despite the adoption of strenuous
cleaning routines. The need for a better insulating material for this
use became even more urgent with a demand for a test strip having 200
terminals Instead of 100 in the same space.
Studies of compounds having such base materials as cellulose-nitrate,
shellac, hard rubber, casein, and cellulose-acetate showed the last
mentioned to give desirable arcing resistance. Foreign conducting
material on the surface was burned off by the arc; the products of
combustion of the small amount of cellulose acetate actually burned
by such an arc are largely volatile, and the residue is non-conducting.
The compound used was found not to be sufficiently heat resistant
to be satisfactory for the body of the test strip. The problem was
solved by using it as a veneer on the test face of the bakelite strip.
DEVELOPMENTS IN COMMUNICATION MATERIALS
245
This face is farthest from the heated ends of the terminals, is free from
mechanical strain and is therefore not damaged by soldering operations.
Since it was the practice to mold this test strip using several partially-
cured preforms, the veneer construction was introduced with only a
slight increase in cost. The cellulose acetate has nearly the same
molding temperature as the phenol plastic, so that the composite test
strip could be molded in one operation. Fig. 10 shows the appearance
■TERMINALS IN HOLES IN PREFORMS
PREFORM OF PHENOL PLASTIC
PREFORM OF PHENOL PLASTIC
PREFORM OF PHENOL PLASTIC
PREFORM OF CELLULOSE ACETATE
Fig. 10 — Method of molding composite 200 point test strip,
of the modified test strip and the method of molding.
Textile Insulation
Another development was in the improvement of textile insulation
which was recently described before the Institute.^- ^ It is mentioned
here only in passing, because of its great commercial importance.
As a result of several years of study in the laboratory, it was found
that the insulating quality of textiles depended on (1) the kind of
fiber; (2) impurities present in the fiber; (3) moisture. The salts of
sodium and potassium were found to be highly detrimental from an
insulation standpoint. A very great improvement was effected by a
washing treatment of the textile. Thus it has been possible to make
^ "The Predominating Influence of Moisture and Electrolytic Material Upon
Textiles as Insulators," R. R. Williams and E. J. Murphy, Trans. A. I. E. E., Vol.
48, 1929.
^ "Purified Textile Insulation," H. H. Glenn and E. B. Wood, Trans. A. I. E. E.,
Vol. 48, 1929.
246 BELL SYSTEM TECHNICAL JOURNAL
cotton an acceptable substitute for silk as wire insulation, as well as to
improve greatly the insulating properties of silk. In one instance,
central office distributing frame wire, of which the Bell System uses
about five hundred million conductor feet annually, it was found pos-
sible to use double silk insulated conductor of treated thread where
formerly triple silk insulation was required. An actual improvement
in insulation was effected at the same time that a considerable economy
resulted.
Metallic Materials
N on- Ferrous Metals
Telephone apparatus uses about 30,000,000 lbs. yearly of brass,
bronze and nickel silver as structural members, springs and bearings.
Because of space limitation the parts are necessarily small, many are
formed into irregular shapes; spring parts must maintain accurate
adjustment and have long fatigue life; certain other parts must resist
wear. Experience with commercial grades of brass indicated wide
variations under existing specifications and unsatisfactory means of
testing the quality. At first blush there may not appear to be any
connection between the temper of a metal spring and the grade of
telephone service furnished, but looking at the matter broadly we were
convinced that the stakes were large enough to warrant our launching
an investigation of non-ferrous metals with the object of arriving at a
better purchasing specification. Accordingly the Bell Telephone
Laboratories initiated a joint study with the Western Electric Company
and the American Brass Company which has extended over a period
of several years. The results of this work have been described in
considerable detail in appropriate papers before the American Society
for Testing Materials.'' •"
This has resulted —
1. In a more accurate knowledge of the physical properties of brass,
phosphor bronze and nickel silver.
2. Development of improved methods of test.
3. Preparation of better purchasing specifications with resulting
improved control of the quality of the materials.
As an instance of the benefits derived, the work on hardness testing
may be cited. For many years the scleroscope had been used as a rapid
means of controlling the quality of sheet metal but trouble was fre-
quently encountered because results could not be readily duplicated on
* "Physical Properties and Methods of Test for Sheet Brass," H. N. Van Deusen,
L. I. Shaw and C. H. Davis, Proc. Amer. Soc.for Testing Materials, 1927.
*" Physical Properties and Method of Test for Sheet Non-Ferrous Metals,"
J. R. Townsend, W. A. Straw and C. H. Davis, Proc. A. S. T. M., 1929.
DEVELOPMENTS IN COMMUNICATION MATERIALS
247
different instruments and it was necessary to allow rather wide limits
on each temper resulting in considerable overlapping of the temper
tolerances. While tensile strength is usually considered the reference
test for cold worked metal, it is necessary to have a test which can be
used for more rapid inspection. As a result of our study we were able
100
40,000 60,000 aO,000 100,000 120,000
TENSILE STRENGTH fPSl)
Fig. 11 — Relation between tensile strength and Rockwell hardness — sheet brass.
to develop means for maintaining the Rockwell hardness tester to an
accuracy within 2 points compared with 8 or 10 points on the sclero-
scope. Fig. 11 shows the relation between tensile strength and Rock-
well hardness for a rolling series made up by the American Brass Com-
pany under carefully controlled manufacturing conditions. This
rolling series covered all ranges of hardness and thickness of sheet
metal generally used in telephone apparatus. The tension test is used
248
BELL SYSTEM TECHNICAL JOURNAL
as a reference test and is resorted to only when the Rockwell test
indicates the material to be close to the limiting values specified.
Work has been completed, resulting in the preparation of improved
specifications for leaded brass, annealed brass, nickel silver and
phosphor bronze and a similar investigation of rod stock in all grades
of these metals is now under way. It is interesting to note in passing
\^
9*
r
/s
%4
H
Q>
^
Fig. 12 — Rotary selector used in dial system.
that in the course of our investigations we determined that the endur-
ance limit of non-ferrous metals is only half that established for ferrous
metals, averaging approximately }i of the ultimate strength.^
For one of the rotary selectors used in the panel dial system we
developed a leaded phosphor bronze sheet containing approximately
3 per cent of lead which proved very valuable in terms of increased
6 "Fatigue Studies of Non-Ferrous Sheet Metals," J. R. Townsend and C. H.
Greenall, Proc. A. S. T. M., 1929.
DEVELOPMENTS IN COMMUNICATION MATERIALS 249
life of the switch. This selector consists of an arrangement of closely
spaced terminals referred to as the " bank " and a set of rotating brushes
contacting with the bank terminals as shown in Fig. 12. Experience
in the field indicated that under severe service conditions these selectors
have a comparatively short life. As a result of our studies we replaced
the brass brushes in the rotor with phosphor bronze, and the brass
terminals of the bank with leaded phosphor bronze, a combination which
has given approximately four times the life obtainable with brass parts,
with corresponding maintenance savings. The reduced wear seems to
be due in part at least to a lubricating effect of the lead constitutent in
the bank terminals.
Aluminum alloys have had considerable application to telephone
apparatus not only in die castings but in sheet form as diaphragms in
certain of the new developments in telephone transmitters and re-
ceivers. One of the most interesting of the aluminum alloys is dur-
alumin, an alloy of aluminum, copper, silicon and magnesium. This
material has about one-third the specific gravity of steel and like steel
can be increased in strength by heat treatment in the manufactured
form. Our first application of duralumin was as a stretched diaphragm
in radio broadcasting transmitters. Here it was necessary to obtain
material with as small a mass as possible and with the necessary
strength to allow stretching to give a high natural period essential for
good quality transmission. The material used in this case was 1.7
mils thick and had a tensile strength between 70,000 and 80,000 pounds
per square inch.
Probably one of the most difficult applications of sheet duralumin is
to the light valve used in the film method of sound picture recording.
The light valve is an electromechanical device actuated by amplified
speech currents, and consists of a loop of duralumin tape supported in a
plane at right angles to a magnetic field. A view of the light valve is
given in Fig. 13 which shows the tape held by two wind-lasses, AA"^, at
one end, and wrapped over a spring-supported pulley B at the other.
This places the tape under considerable tension. The tape is 6 mils
wide and .5 mil thick. The central portion of the loop is supported on
insulating bridges just above the face of the pole piece which constitutes
the armature of an electromagnet.
Viewed against the light, the valve appears as a slit 2 mils wide by 256
mils long. In operation the amplified speech current is passed through
the duralumin tape which, reacting with the magnetic field of the
electromagnet causes variations in the width of the slit controlled by
the variations in the speech current. The light beam directed toward
the film is thus modulated by the slit in accordance with the variations
250
BELL SYSTEM TECHNICAL JOURNAL
of the speech current. In order to avoid distortion, severe require-
ments were imposed on the straightness of the edges of the tape, and
on the strength, in order to permit stretching to give a natural period in
excess of 7,000 cycles per second. To obtain these properties special
heat treatments and methods of rolling the material had to be de-
veloped.
Fig. 13 — Light valve used in recording of sound pictures.
AA'^ — Wind-lasses.
B — Pulley supported by spring.
CC^ — Insulating bridges.
Ferrous Metals
Some interesting problems have been encountered in the use of
ferrous metals in telephone apparatus, particularly in the operator's
calling dial. Considerable trouble was encountered from slippage of
the dial governor resulting from premature wear or breakage of the
tips of the pawls or the teeth of the pinion. These parts had been made
out of low carbon steel which had been found satisfactory for the
subscriber's dial. The operator's dial, however, being used for a
greater number of times, presented a more severe condition and case
hardening was applied to obtain better wear resisting properties.
This treatment was found to be unsatisfactory because the parts have
thin sections and the combined weight of the two parts amounts
to only 2 grams. Case hardening either produced too deep a case
giving brittleness or too shallow a case which soon wore through.
A nickel-chrome steel, originall}- developed for the automotive industry
was finally adopted for the pawl and pinion combined with a special
heat treatment. It was thus possible to obtain a useful life of 8 million
operations as compared with an average of }4 million operations for
the steel formerly used. This is another instance in which an increase
DEVELOPMENTS IN COMMUNICATION MATERIALS
251
in first cost resulted in appreciable savings in annual cost of the device,
considered from the operating companies' standpoint.
Ferro- Magnetic Metals
Up to about 15 years ago, telephone engineers used the magnetic
materials in their designs which had been originally developed for the
power industry, viz., magnetic iron and silicon steel. An exception was
the use of 4. mil hard drawn steel wire for loading coil cores where
extremely low permeability was desired.
The increasingly severe requirements imposed by compositing and
phantoming of telephone circuits and the introduction of vacuum tube
Fig. 14 — Loading coils showing core rings of liighly compressed powdered iron.
repeaters, made necessary the development of materials which would
more adequately meet the new requirements. It was in 1915 that the
Western Electric Company first produced compressed powdered
electrolytic iron cores for loading coils. The construction of such
powdered iron core coils is illustrated by Fig. 14. Electrolytically
deposited iron is ground to a fine powder; the particles are covered with
an insulating film and then compressed at a pressure of 200,000 lbs.
per square inch to form rings as shown in the figure. This material
was sensational in the improvements which it afforded over the core
materials theretofore available as it combined with extremely high
resistivity, high stability of A.C. permeability under conditions of
powerful superposed or residual D.C. magnetization. The change in
252
BELL SYSTEM TECHNICAL JOURNAL
A.C. permeability resulting from the temporary application of large
magnetizing forces did not exceed 2 per cent as compared with changes
of the order of 30 to 40 per cent commonly found in previously available
materials.
The next important step was the discovery of permalloy, a nickel-
iron alloy having extremely high permeability which had its first
application in the loading of submarine telegraph cables. This mate-
rial with its extremely low hysteresis loss and high induction for feeble
magnetizing forces, has since been applied extensively in the design of
transformers, relays, receivers, and other telephone apparatus. Fig. 15
6000
4000
2000-
-2000-
-4000
-6000
1.^
»UCON
STEE
/
A
,>-^
/
/
>■
o
-I
_l
«»•
/
/
/
2
a
UJ
a.
y
/
/
CO
K
/
y
'
■^^
-0.10 -0.8 -0.6 -0.4 -0.2 O 0.2 0.4 0.6 0.8
H
Fig. 15 — Hysteresis loops of silicon steel and permalloy.
0.10
shows comparative hysteresis loops for permalloy and silicon steel.
The much smaller hysteresis loss of permalloy, approximately one-
seventh of that of the silicon steel sample is indicative of its greatly
reduced tendency to remain magnetized after the removal of a mag-
netizing force, a property which is of great importance in the operation
of quick release types of relays. In transformers and in continuously
loaded cable, the very high permeability at small magnetizing forces
of this material, strikingly shown in Fig. 16, is of great value. It is the
high permeability of permalloy that made it possible to load telegraph
cables successfully and thereby attain a threefold increase in telegraph
speed. In transformers such as those used in vacuum tube amplifiers,
the high permeability permits the designer either to achieve equivalent
quality with a much smaller apparatus volume or, in the same space,
to furnish equipment of better quality. The latter result is shown by
the curves of Fig. 17 which indicate how transformer performance at
DEVELOPMENTS IN COMMUNICATION MATERIALS
253
100,000
80,000
^ 60,000
ffl
<
ui
S
a.
UJ
a
i. 40,000
20,000
4 000 8,000 12,000
B
Fig. 16 — Permeability curves of soft iron and permalloy.
16,000
<
o
u. 3
_l
Q.
<2
u
o
<
I-
>
-
1
r
^'
.^-
^
^
^
i
A = SILICON STEEL CORE
B = PERMALLOY CORE Tsl °/o
10
100 1,000
CYCLES PER SECOND
10.000
Fig. 17 — Showing improvement in quality of voice frequency amplifier due to permal-
loy core transformers.
17
254 BELL SYSTEM TECHNICAL JOURNAL
very low frequencies is improved by the use of permalloy. Sheet
permalloy has been followed by compressed powdered permalloy^ and
this by perminvar,^ the newest member of the magnetic alloy family.
Compressed powdered permalloy has replaced the powdered iron as
it has all of the desirable properties of the latter and to an even greater
degree. By virtue of higher permeability combined with lower hys-
teresis loss, it has made possible the design of smaller coils of superior
performance characteristics. As an illustration the two loading coils
of Fig. 18 are shown, the smaller of these being the electrical equivalent
Fig. 18 — Relative size of powdered iron (left) and powdered permalloy (right)
loading coils.
of the larger in all respects and in some its superior. In general the
reduction in coil size made possible by the use of powdered permalloy
in place of powdered iron amounts to about 75 per cent giving very
substantial savings in manufacturing costs, handling problems and
installation space required.
Perminvar is remarkable in an entirely unique respect. Its permea-
ability is not exceptionally high, being of the same order as that of
ordinary soft iron at moderately low magnetizing forces, but it is
exceptionally constant with respect to magnetizing forces. This is
shown in Fig. 19 from which it will be noted that there is substantially
no change in permeability up to a force of about 2 gausses whereas over
this same range, the permeability of soft iron undergoes a change of
more than 2,000 per cent. Up to somewhat smaller magnetizing forces,
perminvar has a vanishingly small hysteresis loss. Fig. 20 depicts this
loss for perminvar. It is to a material of constant permeability and
low hysteresis loss that the transformer designer turns when he has a
difficult requirement as to low modulation to meet. Unfortunately,
while perminvar has these properties over a limited range of magnetiza-
' "Compressed Powdered Permalloy, its Manufacture and Magnetic Properties,"
W. J. Shackelton & I. G. Barber, Trans. A. I. E. E., Vol. 17, 1928.
* "Magnetic Properties of Perminvar," G. W. Elmen, Jotir. of Franklin Institute,
Vol. 206, 1928.
DEVELOPMENTS IN COMMUNICATION MATERIALS 255
3600r
3200
2600
2400-
2000-
:i.
1600-
1200-
800-
400
1300
I
i---. AIR
QUENCHED
/
-ANNEALED
/
J
/
baked!
y
y
y
/
y
Fig. 19 — Permeability curves for perminvar.
1200
1000
800
CD 600
400
200
1200
1000
800
OQ600
400
200
A
A
*'
/
h
y
4,
/
1
y
/
J
y
Ofl 1.6 2.4 32
H
6000
5000
4000
3000
2000
1000
(0
-1000
-2000
-3000
-4000
-5000
-6000
0.8 1.6 2.4 3.2 -5-4-3-2-1 I 2 3 4 5
H H
A
1
^
y
/
y
1^
•y
/
h
^
K
/
1
^
[
}
<y
/\
^
^A
1
y
B
/
1
^
/
1
J
,y
/
/
^
/
r
'
-
^
/'
/
y
1
f
/
L
i
/
I
J>
/
y
1
y
\
y
y
1
/
'
I
/
1
Fig. 20 — Hysteresis loops for perminvar.
A — Air quenched.
5— Baked.
256 BELL SYSTEM TECHNICAL JOURNAL
tion, if this range is exceeded they are lost and so it is necessary that
it be used within suitable limits.
These materials have been described in technical papers before
various scientific societies and are not, therefore, discussed in detail
here.
As indicating the wide scope of magnetic performance that is de-
manded of materials for use in communication apparatus, some of the
necessary properties are listed below :
High permeability — at very feeble and at high inductions.
High saturation value of induction.
Low residual magnetization.
Low hysteresis loss at feeble and moderate magnetizations.
Low eddy-current losses over the frequency range from to 80,000
cycles.
High constancy of permeability over a wide range of magnetization.
Small effect on A.C. permeability at feeble currents with superposed
or residual D.C. magnetization.
Certain of these requirements are imposed from the simultaneous
transmission of D.C. telegraph currents, speech currents and carrier
frequency telephone or telegraph currents through the transformers,
loading coil or other iron-core apparatus in the circuit. Interference
between channels, due to magnetic modulation in the cores, must be
kept at an extremely low value for satisfactory quality of transmission.
Summing up our work on materials, the results have been along two
general lines: (1) improvement in quality of commercial materials and
(2) development of new materials. As regards the first, we have
worked in close cooperation with material suppliers whose progressive
attitude has made possible certain of the advances described. The
more striking advances have been due to the discovery of new or im-
proved materials in our laboratories, the savings from which have
amply justified the program of continuous research which has been the
Bell System policy for a number of years. To take a single instance,
the field of magnetic alloys — probably the first to which we applied
intensive effort, — a single invention, the powdered electrolytic iron
core resulted in savings of such magnitude as to far overshadow the
cost of the investigational work. As already noted, this material has
since been superseded by the powdered permalloy core which represents
an equally great advance.
There is one point which should be emphasized and that is, that the
most economical material is not necessarily the cheapest one. Treated
textiles cost more per pound than ordinary textiles; permalloy costs
more per pound than silicon steel. In these particular instances so
DEVELOPMENTS IN COMMUNICATION MATERIALS 257
much less material is required to obtain the desired result that there is a
net saving in cost of manufacture. The true criterion of relative
economy, however, takes into account not alone the cost of manufac-
ture, but the serviceability of the device throughout its operating life.
Hence the designer, if he be free to decide on purely engineering
grounds, will make his decision as to the best materials to use on the
basis of the lowest annual charge over a period of years, thus taking
into account the important item of maintenance cost.
Transoceanic Telephone Service — Short-Wave
Transmission
By RALPH BOWN i
The discussion relates to the transmission problems involved in short-wave
radiotelephony over long distances and the transmission bases for design of
the systems used in commercial transatlantic service. Choice of operating
frequencies, amounts of transmitter power, directive transmitting and re-
ceiving antennas, automatic gain controls in receivers, and voice-operated
switching devices are all factors which may be invoked to aid in solving these
problems. The way in which they have been applied in the transatlantic
systems and the results which have been obtained are set forth briefly.
TRUNK circuits between London and New York which furnish
telephone service between these two cities and also permit suc-
cessful conversation by means of toll wire extensions between the
United States and Europe more generally are being carried over both
long waves and short waves. It is the purpose of this paper to consider
the transmission side of the new short-wave circuits which the American
Telephone and Telegraph Company and the British General Post Office
have made available for this service. In doing this we shall proceed
from the more general considerations, relating to wave-lengths and
communication channels, through a discussion of the principles govern-
ing the general design of the system, into a brief summary of practical
performance results.
The frequency range so far developed for commercial radio use is
roughly 20 to 30 million cycles wide, extending from about 10 kilocycles
to perhaps 25,000 kilocycles per second. There are two parts of this
whole spectrum suitable for transoceanic radiotelephony — the long-
wave range which is relatively narrow, extending roughly from 40
kilocycles to 100 kilocycles, and the short-wave range which in its
entirety is much broader, extending from about 6000 kilocycles to
25,000 kilocycles.
It is evident that the long-wave region, including perhaps only 50
kilocycles, offers opportunity for development of relatively few tele-
phone channels, particularly in view of the fact that it is in use by a
number of telegraph stations. Also it must be borne in mind that for
telephony these waves are suitable for only moderate distances of the
order of 3000 miles and for routes in the temperate zones where static
1 Presented at the Winter Convention of the A. I. E. E., New York, N. Y., Jan.
1930.
258
TRANSOCEANIC TELEPHONE SERVICE 259
interference is moderate. The first transatlantic radiotelephone cir-
cuit opened in 1927 was a long- wave circuit (58.5-61.5 kilocycles).
In providing the next few channels for the initial growth of the service
the opportunity to determine the utility of short waves was embraced.
The short-wave range is vastly wider in kilocycles but, nevertheless,
has its limitations as to the number of communication facilities it
afifords. For a given route of a few thousand miles a single frequency
gives good transmission for only a part of the day. For example,
from the United States to Europe a frequency of about 18,000 to 21,000
kilocycles (17 to 14 meters) is good during daylight on the Atlantic.
But in the dawn and dusk period a frequency of about 14,000 kilocycles
(22 meters) is better. For the dark hours something like 9000 kilo-
cycles (33 meters) gives best transmission and for midnight in winter an
even lower frequency near 6000 kilocycles (50 meters) is advantageous.
Thus, in considering the short-wave range in terms of communication
circuits, we must shrink its apparent width materially to take account
of the several frequencies required for continuous service.
At the present time the frequency spaces between channels are
much greater than the bands of frequencies actually occupied by useful
transmission. This elbow room is to allow for the tendency of many
stations not to stay accurately on their nominal frequencies but to
wander about somewhat. But in spite of this allowance, cases of
interference are common and one of the activities which must be carried
on in connection with a commercial system is the monitoring of inter-
fering stations and the accurate measurement of transmitting fre-
quencies to determine the cause of the conflict. To permit intensive
development of the frequency space offered by Nature the greatest
possible constancy and accuracy of frequency maintenance in trans-
mitting sets will be required.
The fact that channels have been assigned (within wide bands set
aside for a particular service) with little regard to the geographical
location of stations may result in neighboring channels having much
stronger signals than those in the channel being received. When
this is so, a severe requirement is placed on the selectivity of the re-
ceiver to prevent interference.
Interconnecting with Wire Circuit Extensions
The skeleton of a radiotelephone circuit is in its essentials very
simple. It consists merely of a transmitter and a receiver at each end
of the route and two oppositely directed, one-way radio channels
between them. These two independent channels must be arranged
at the terminals to connect with two-wire telephone circuits in which
260
BELL SYSTEM TECHNICAL JOURNAL
messages in opposite directions travel on the same wire path. The
familiar hybrid coil arrangement so common in telephone repeaters
and four-wire cable circuits might appear to solve this problem, were
there not difficulties peculiar to the radio channels. In the short-wave
case large variations in attenuation occur in the radio paths within
short intervals of time. These would tend to cause re-transmission of
received signals at such amplitudes that severe echoes and even singing
around the two ends of the circuit would occur unless means were
provided to prevent this.
To overcome these fundamental transmission difficulties, an auto-
matic system of switches operated by the voice currents of the speakers
has been developed.^ These devices cut off the radio path in one
UNITED
STATES
„ BOARD
RECEIVING
AMPLIFIER
"LI — '
RECEIVING
DELAY
TOLL awn
SWITCH-
kSiXl
TSW^
hw^ —
TELEPHONE
OPERATOR
HYBRID COIL
TESQ-
RADIO
RECEIVER
ETCOr
J-
FROM
ENGLAND
a
RECEIVING
DETECTOR
w.
TRANS-
MITTING
DETECTOR
TRANS-
MITTING
DELAY
LAWRENCEVILLE
CO
2
RADIO
-► TRANS-
MITTER
J-
TO
ENGLAND
Fig. 1^ — Circuit diagram illustrating operation of voice-operated switching device.
direction while speech is traveling in the reverse direction and also keep
one direction blocked when no speech is being transmitted. The
operation is so rapid that it is unnoticed by the telephone users. Since
this system prevents the existence of singing and echo paths, it permits
the amplification to be varied at several points almost without regard
to changes in other parts of the system, and it is possible by manual
adjustment to maintain the volumes passing into the radio link at
relatively constant values, irrespective of the lengths of the connected
wire circuits and the talking habits of the subscribers.
Fig. 1 gives a schematic diagram of the United States end of one of
the short-wave circuits showing the essential features of a voice-
operated device which has been used. This kind of apparatus is
^ For detailed description of this system see "The New York-London Telephone
Circuit" by S. B. Wright and H. C. Silent, Bell System Tech. JL, Vol. VI, October,
1927, pp. 736-749.
TRANSOCEANIC TELEPHONE SERVICE 261
capable of taking many forms and is, of course, subject to change as
improvements are developed. The diagram illustrates how one of
these forms might be set up. This form employs electro-mechanical
relays. The functioning of the apparatus illustrated is briefly as
follows: the relay TES is normally open so that received signals pass
through to the subscriber. The relay SS is normally closed to short
circuit the transmitting line. When the United States subscriber
speaks his voice currents go into both the Transmitting Detector and
the Transmitting Delay circuit. The Transmitting Detector is a
device which amplifies and rectifies the voice currents to produce
currents suitable for operating the relays TES and SS which thereupon
short circuit the receiving line and clear the short circuit from the
transmitting line, respectively. The delay circuit is an artificial line
through which the voice currents require a few hundredths of a second
to pass so that when they emerge the path ahead of them has been
cleared by the relay SS. When the subscriber has ceased speaking the
relays drop back to normal.
The function of the Receiving Delay circuit, the Receiving Detector,
and the relay RES is to protect the Transmitting Detector and relays
against operation by echoes of received speech currents. Such echoes
arise at irregularities in the two-wire portion of the connection and are
reflected back to the input of the Transmitting Detector, where they
are blocked by the relay RES which has closed and which hangs on for a
brief interval to allow for echoes which may be considerably delayed.
The gain control potentiometers shown just preceding the transmitting
and receiving amplifiers are provided for the purpose of adjusting the
amplification applied to outgoing and incoming signals.
The relief from severe requirements on stability of radio transmission
and from varying speech load on the radio transmitters which this
system provides permits much greater freedom in the design of the
two radio channels than would otherwise be possible.
The Radio Channels
One of the first questions which comes up in considering the design
of a radio system is the power which can be sent out by the transmitter.
The word "can" is used advisedly, rather than "should," since in the
present art the desideratum usually is the greatest amount of power
that is technically possible and economically justifiable. There are
few radio systems so dependable that increased power would not
improve transmission results. At very high frequencies the generation
of large powers is attended by many technical difficulties but fortu-
nately the radiation of power can be carried out with much greater
262 BELL SYSTEM TECHNICAL JOURNAL
efficiency than is feasible at lower frequencies. At 18,000 kilocycles
(about 16 meters) a single half- wave radiator or doublet is only about
25 ft. long and it is possible to combine a number of them, driven in
phase by a common transmitter, into an antenna array which concen-
trates the radiated power in one geographical sector. In that direction
the effectiveness may be intensified 50 fold or more (17 db) and waste
radiation in other directions reduced materially. Thus, one of the
transmitters at Lawrenceville, New Jersey, used in the short-wave
transatlantic circuits when supplying 15 kw. radiates in the direction
of its corresponding receiving station as effectively as would a non-
directive system of about 750 kw.
The transmitting antennas also give some directivity in the vertical
plane, increasing the radiation sent toward the horizon and decreasing
that sent at higher angles. It is not yet certain that vertical directivity
is always advantageous and this effect has not been carried very far.
At the receiving station the radiated power has dwindled to a small
remnant which must be separated from the static as far as possible
and amplified to a volume suitable for use in the wire telephone plant.
Here again directive antenna arrays are of value. A receiving antenna
system sensitive only in a narrow geographical sector, and that lying
in the direction from which the signal arrives, excludes radio noise from
other directions and thereby scores a gain of perhaps 40 fold (16 db)
in the power to which the signal can be amplified without bringing
noise above a given value. It also scores against noise which arises in
the tubes and circuits used for amplification, since the combined action
of the several antennas of the array delivers more signal to the initial
amplifier stage where such noises originate.
Thus, it is evident that transmitter power, transmitting directivity,
receiving directivity, and quiet receiving amplifiers are of aid in pro-
viding signal transmission held as far as possible above the radio
noise. In a well designed system the relative extents to which these
aids are invoked will depend upon economic considerations as well as
upon the technical possibilities of the art.
There is one other type of noise than that provided by Nature which
is of particular importance at short waves,- — electrical noise from the
devices of man. One of the worst offenders is the ignition system of the
automobile. The short-wave transoceanic receiving station at Net-
cong. New Jersey, is so located that automobile roads are at some
distance, particularly in the direction from which reception occurs.
Service automobiles which produce interference cannot be allowed near
the antenna systems unless their ignition systems have been shielded.
Also, electrical switching and control systems incidental to the power,
TRANSOCEANIC TELEPHONE SERVICE 263
telegraph, and telephone wire systems at the station are shielded or
segregated.
At both the transmitting and receiving stations at least three antenna
systems are supplied for each circuit, one antenna for each of the three
frequencies normally employed. The design and arrangement of
these are dictated by the requirements flowing from their uses. The
purpose of the transmitting antenna is to concentrate as much power
as possible in one direction. The purpose of the receiving antenna is to
increase reception from the desired direction and to cut down reception
at all other angles. In the former the forward-looking portion of the
characteristic is of greatest importance, while in the latter the rearward
characteristics need greatest refinement.
Transmission Performance
In short-wave telephone systems the width of the sidebands is so
small a percentage of the frequency of transmission that tuning charac-
teristics of the antennas and high-frequency circuits are relatively
broad and impose little constriction on the transmission-frequency
characteristic. A flat speech band is easy to obtain over the range
of approximately 250 to 3000 cycles employed for these commercial
circuits. This relieves the short-wave circuits from many of the
problems of obtaining sufficient band width which are troublesome in
designing long-wave systems.
Short-wave transmission is subject to one frailty which particularly
hampers its use for telephony. This is fading. Where fading is of
the ordinary type, consisting of waxing and waning of the entire trans-
mitted band of frequencies, automatic gain control at the receiving
station is of value and is employed in the transoceanic circuits under
discussion. The amplification in the receiver is controlled by the
strength of the incoming carrier and is varied inversely with this
strength so as to result in substantially constant signal output. Ob-
viously this control can be effective only to the extent that the signal
seldom falls low enough to be overwhelmed by radio noise.
When fading is of the selective type, that is, the different frequencies
in the transmitted band do not fade simultaneously, the automatic gain
control system is handicapped by the fact that the carrier or control
signal is no longer representative of the entire signal band.
Selective fading is believed to result from the existence of more than
one radio path or route by which signals travel from transmitter to
receiver. These paths are of different lengths and thus have different
times of transmission. Wave interference between the components
arriving over the various paths may cause fading when the path lengths
change even slightly.
264 BELL SYSTEM TECHNICAL JOURNAL
If the path lengths differ by any considerable amount, for example, a
few hundred miles, the wave interference is of such a character as to
affect the frequencies across a band consecutively rather than simul-
taneously.
With the presence of selective fading there comes into being the
necessity of guarding against rapid even though small variations in the
transmitted frequency, since if such variations are present a peculiar
kind of quality distortion of the telephone signal results.
The varying load which speech modulation places on the transmitter
circuits tends to cause slight variations in the instantaneous equivalent
frequency which are known as "frequency modulation" or "phase
modulation" depending on their character. To prevent this effect
the control oscillator must be carefully guarded against reaction by
shielding and balancing of circuits and the design must be such as to
preclude variable phase shifts due to modulation in subsequent circuits
of the transmitter.
It is apparent that if there are two paths of different lengths, two
components which arrive simultaneously at the receiver may have left
the transmitter several thousandths of a second apart. If the transmit-
ter frequency has changed materially during this brief interval trouble
may be expected. The trouble actually takes the form of a distortion
of the speech as demodulated by the receiving detector.^
Defects in short-wave transmission due to radio noise, minor varia-
tions in attenuation, fading, and distortion are nearly always present to
some extent and, when any or all are severe, cause a certain amount of
lost service time. These interruptions are of relatively short duration
and, furthermore, there is enough overlap in the normal times of useful-
ness of the several frequencies available, so that shifting to another
frequency may give relief. There is, in addition, a kind of interruption
which from the standpoint of continuity of service is more serious. At
times of disturbance of the earth's magnetic field, known as " magnetic
storms," short-wave radio transmission is generally subject to such
high attenuation that signals become too weak to use and sometimes
too weak to be distinguishable. These periods affect all the wave-
lengths in use and may last from a few hours to possibly as much as two
or three days in extreme cases. They are followed by a recovery period
of one to several days in which transmission may be subnormal.
Severe static may cause interruption to both long- and short-wave
services at the same time but the short waves are relatively less affected
by it and are usually able to carry on under static conditions which
3 For a discussion of this phenomenon see "Some Studies in Radio Broadcast
Transmission" by Bown, Martin, and Potter, 7. R. E. Proc, Vol. 14, No. 1, p. 57.
TRANSOCEANIC TELEPHONE SERVICE
265
prevent satisfactory long-wave operation. On the other hand severe
fading or the poor transmission accompanying a magnetic disturbance
may interrupt short-wave service without affecting the long waves
adversely, — in fact magnetic disturbances often improve long-wave
transmission in the daytime. The service interruptions on the two
types of circuits are thus nearly unrelated to each other and have no
definite tendency to occur simultaneously. This is the principal
reason why both long-wave circuits and short-wave circuits appear
essential to reliable radiotelephone service.
On routes which are very long or which cross tropical areas which
result in static sources facing the directive receiving antennas, long
+ 10
+ 5
.0
in
X
o
z
If) -10
o
y -15
>-20
I-
<
Ml
a. -25
-30
-36
r\K\ IQ ayi
Kkt- t
1& kJlC-r(rDC\ nCTAI M
1 Xr\ KJF\A/
o- o
•- ■•
SOUTHGATE, ENGLAND.
TRANSMISSION ON60KC (5000 METERS) ROCKY POINT, L.I. TO
CUPAR, SCOTLAND.
HORIZONTAL COMPONENT OF EARTH'S MAGNETIC FIELD.
k
i
L
r
^^
1
, J
^,
c
</
N
k
/
/
/
t
*-<
"■—^
Y"
r
>
/ A
A
1
!
/ V
>-— — <
^
/
y
\ \
\ 1
^
V
\
\
I
A
y
r-
\
L \
\ I
\\
Y
y
/
/
/
1 — <
_,-'
<'
•
/
/
f
/ /
/ /
/ /
/ /
r /
1
r
18,740 -5^
<
■2.
2
18,730 <
sl£-
Q
18,720 H
18.710 p
UJ
z
o
18,700 <
18,690
"i
I-
<
5 4 3 2
HAYS BEFORE
I
STORM
3 4 5 6
DAYS AFTER
18,680 ,
18,670 ^
z
o
a
18,660 2
O
O
18,6501:
Fig. 2 — Effect of magnetic disturbances on radio transmission.
waves cannot as yet be successfully employed and short waves alone are
available. However, experience tends to indicate that on North
and South routes such as between North and South America, the
interruptions associated with magnetic storms are less severe and of
shorter duration.
The cycle of events which accompanied a particularly severe mag-
netic storm ^ in July, 1928, is shown graphically in Fig. 2. The light
■* Data regarding other magnetic disturbances are given in a paper by C. N. Ander-
son, entitled "Notes on the Effect of Solar Disturbances on Transatlantic Radio
Transmission," /. R. E. Proc, Vol. 17, No. 9, September, 1929.
266
BELL SYSTEM TECHNICAL JOURNAL
NEW YORK TO LONDON
10
L
n
ONDON
TO NEW
YORK
n
i
1
miM
^^■tii 1 '
^Hiii>y
^■tii II
r
n
I^^^^^B^Hj [■
A.
H
Hi
1
IH
Jill
UJ
5
H|
1
II
i^lii
^Hil
^^Hflili
^^^^■iHi
^^^^^^Hi
nil
■{
I^^^^^^^IH
'^^^^^I'^l'^^^li^l ^H 1 H
Q
B
■■
1^1'
i^^^^l
tr
S 2
Z
<
Jo 12
NOON
Z
CC
HI
1- 10-
1
<
n^
1
8-
■
III!
1
1 i^^^^^^^l
■
M
II
'
i
■
1
II
1
4
JUNE
JULY
1929
AUGUST SEPT.
COMMERCIAL ^
— KEY —
:=3 UNCOMMERCIAL c
=3 NO DATA
Fig. 3 — Chart showing transmission performance of a short-wave transatlantic
telephone circuit.
TRANSOCEANIC TELEPHONE SERVICE 267
dotted curve shows the variation in the horizontal component of the
earth's field. The heavy solid line follows the daily averages of the
short-wave received signal field. It is apparent that the disturbance
took two days to reach its peak and the recovery to normal took nearly
a week. The heavy dotted line shows received field on long waves (60
kilocycles) and indicates that transmission was improved slightly at the
same time the short waves were suffering high attenuation.
The experience with transatlantic telephone service on short waves
covers a period of nearly three years, there having been available a one-
way channel from the United States to England used as an emergency
facility for the first year and a half, a two-way circuit for the next year,
and two circuits since June, 1929. It is only in this later period, how-
ever, that a circuit has been available operating regularly with the
amounts of transmitter power and antenna directivity which have been
mentioned.
The performance of the two one-way channels forming this circuit is
charted in Fig. 3. The charts are plotted between hours of the day
and days in the year so that each unit block represents one hour of
service time. The solid black areas are time in which commercial
operation could be carried on. The dotted strips are uncommercial
time. The blank areas are for time in which, for one reason or another,
the circuit was not operating and no data were obtained. Perhaps the
most outstanding feature of these charts is the tendency of the lost time
to fall in strips over a period of two or three days. These strips coin-
cide approximately for both directions of transmission. The principal
ones are about July 10 and 15 and August 2 and 17. These are charac-
teristic of the interruptions accompanying magnetic disturbances of the
kind which occur at irregular intervals of a few days to several weeks.
They are, of course, not as severe as the disturbance illustrated in Fig. 2.
It is apparent that for these three summer months this new circuit
gave a good account of itself and furnished commercial transmission
for something like 80 per cent of the time that service was demanded of
it. In these same months the long-wave system suffered its greatest
difficulty from static, and we have concretely illustrated the mutual
support which the two types of facilities give each other.
It should not be inferred from these data that the short-wave trans-
atlantic radio links furnish 80 per cent of the time talking circuits as
stable and noise free as good wire lines. Under good conditions they
do provide facilities which compare favorably with good wire facilities.
On the other hand they may at times be maintained in service and
graded "commercial" under conditions of noise or other transmission
defects for which wire lines would be turned down for correction, since
268
BELL SYSTEM TECHNICAL JOURNAL
^^
^-^^
(v-
NETCONG
RECEIVING
STATION
-^^^
■^^
PATERSON
O
MORRISTOWN
O
NEWARK
O
BOUNDBROOK
O
NEW BRUNSWICK
O
LAWRENCEVILLE
TRANSMITTING
STATION
Fig. 4 — Map showing transmission considerations affecting location of stations.
TRANSOCEANIC TELEPHONE SERVICE 269
the obviously undesirable alternative is to give no service at all until
conditions have improved again. The present development effort is
largely directed toward improvements which will insure not only a
greater degree of reliability against interruptions but which also will
improve the grade of service as a whole.
In the foregoing little has been said about the stations and plant
since a description of these and the operation of them are treated in two
companion papers by Messrs. Cowan and Oswald. It may be well,
however, to view the physical scene broadly as set forth on the accom-
panying map, Fig. 4.
The geographical arrangement of the transmitting and receiving
stations was governed among other things by transmission considera-
tions. The two stations were placed about 50 miles apart because this
is approximately the distance for minimum signal and at a lesser or
greater distance the signals from the American transmitter might be
strong enough to offer some interference to receiving the English or
South American stations on adjacent channels. For the same reason
they were placed at considerable distances from the transmitters and
receivers of other communication agencies. The Netcong receiving
station lies to the north of the Lawrenceville transmitting station so as
not to be in paths of strong signals from the directive antennas which
face northeast toward England and southeast toward South America.
This configuration also places the transmitter outside the sensitive
angles of the directive receiving antennas.
18
Transoceanic Telephone Service — Short-Wave Equipment
A. A. OSWALD!
The application of short-wave radio transmission to transoceanic tele-
phone circuits is developing apparatus and stations designed specifically to
meet the needs of these services. This paper describes from the radio point
of view the important technical features and developments incorporated in
the new transmitting and receiving stations of the American Telephone
and Telegraph Company located respectively at Lawrenceville and Netcong,
New Jersey, and it outlines some of the radio problems encountered in the
station design.
*****
SHORTLY after transatlantic telephone service was opened in
January, 1927 the long-wave radio circuit between New York and
London was supplemented, first by an experimental short-wave radio
link in the west-east direction and later by a short-wave link in the
east-west direction.^ From this beginning, as an auxiliary to the long-
wave circuit, the short-wave system has been improved steadily so that
its average performance throughout the year now more nearly ap-
proaches that of the long-wave system and it has become an important
part of the transoceanic facilities. The relative merits of the two sys-
tems, their combined usefulness, and their transmission features are
the subject of another paper and will not be discussed here. For the
present purpose it will be sufficient to note that there are now in opera-
tion between New York and London, one long-wave and three short-
wave two-way circuits and that within a few weeks a short-wave circuit
will be available between New York and Buenos Aires.
The radio transmitting units for the New York end of these four
circuits are located at the new station which the American Telephone
and Telegraph Company has recently established at Lawrenceville,
New Jersey. The receiving units are concentrated at Netcong, New
Jersey. The factors entering into the selection of these station loca-
tions are outlined in another paper and therefore need not be men-
tioned further. This paper is limited in scope to a necessarily brief
description of the transmitting and receiving systems and apparatus, a
discussion of technical features in the station layouts, and an outline
of the major problems encountered in the station design. Comprehen-
sive treatment of individual units is properly left for other entire papers.
It will be convenient to deal with the transmitting and receiving sta-
^ Presented at the Winter Convention of the A. I. E. E., New York, N. Y., Jan.,
1930.
a O. B. Blackwcll, A. I. E. E. Journal, May 1928. B. S. T. J. April 1928.
270
TRANSOCEANIC TELEPHONE SERVICE
271
tions separately and in each case to consider briefly the system and
apparatus of one channel before describing the general station plan.
Transmitting System
The four channels at Lawrenceville are equipped with independent
transmitters using certain auxiliary apparatus in common. Each
channel involves a radio transmitter with its associated power plant
and wire equipment, and a group of directive antennas designed and
adjusted for the specific wave-length assignments of the channel.
The general method of transmission, with the exception of directional
sending, is the same as that employed for program broadcasting sta-
tions in that the radiated signal contains the carrier and both sidebands.
Systems in which one or more of these components are suppressed at
the transmitter appear to offer further means of improving short-wave
transmission, and the necessary apparatus for the practical application
of such systems when operating at frequencies in the order of 20,000
DIRECTIONAL
ANTENMA
TELEPHONE
LINE
LINE
F
TERMINA
AND
iEPEATEF
DON
Iv
ONITORIN
CONTROL
DESK
JG
1
i
"WO STAC
AUDIO
^MPLIFIEF
'
TRANSMISSION LINE.^
CONTROLLED
OSCILLATOR
J
I
RADIO F
A
rWO STAG
-REQUENC
MPLIFIEF^
E
.Y POWER
PIEZO-ELECTRIC
OSCILLATOR
HARMONIC
GENERATOR*!
HARMONIC
GENERATOR *2
Fig. 1 — Block schematic of transmitting system.
kilocycles is undergoing development. However, throughout the
development of the transmitters as now installed at Lawrenceville the
possibility of future major modifications in the method of transmission
has been kept in mind. For this reason the modulator-amplifier
system was adopted. In this system the signal which is to be radiated,
is prepared by modulation processes at relatively low power levels and
thereafter amplified the requisite amount. The amplifier and its
power plant, representing a large proportion of the investment in
equipment, can be continued in service with no appreciable alterations,
even though the system of transmission and the modulating apparatus
undergo radical changes.
The general scheme of transmission is shown in Fig. 1. After passing
through the line terminal and control apparatus, which includes
272 BELL SYSTEM TECHNICAL JOURNAL
Standard repeaters, the voice currents are further amplified and em-
ployed to modulate the plate voltage of an oscillator consisting of two
250-watt tubes connected in a push-pull circuit and oscillating at the
frequency of the carrier which is to be transmitted. The frequency
of such an oscillator, if not carefully controlled, will wander outside of
the assigned frequency band, thus causing interference with other
services and it will also suffer variations during the modulation cycle
which contribute to fading phenomena encountered at the distant
receiving station. In order to reduce these effects the oscillator is
held in step at the desired carrier frequency by means of a second
oscillator which is electrically removed from the reactions normally
influencing and tending to vary the frequency of the controlled oscil-
lator. Every precaution is taken to maintain accurately the frequency
of the second oscillator and among other things it is governed by a
piezo-electric quartz crystal whose temperature is regulated closely.
Since it is impractical to use crystals cut sufficiently thin to oscillate
directly at frequencies in the range 10,000 to 20,000 kilocycles, thicker
crystals of lower frequency are used in combination with harmonic
generators which multiply the crystal frequency first by two or three
and then by one or two as the case requires. By virtue of the wide
differences between the input and output frequencies of the harmonic
generators these intermediate steps tend to isolate the crystal oscillator
from the other radio circuits and thus aid in stabilizing the frequency.
The modulated radio frequency output of the controlled oscillator is
applied to the grids of a two-stage power amplifier employing water-
cooled tubes designed for operation at these frequencies. The first
stage contains two tubes and the second stage contains six. The
tubes are arranged in push-pull circuits, the entire system being care-
fully balanced to ground. The carrier output power from the last
stage is 15 kw. With 100 per cent modulation this corresponds to 60
kw. at the peaks of the modulation cycle. In other words, a radio tele-
phone amplifier of this type, rated at 15 kw. when provided with a
sufficiently large d-c. power source, could be used as a 10,000-kilocycle
continuous wave generator of 60 kw. capacity.
The radio signal delivered by the amplifier is conveyed to the antenna
by means of a 600-ohm open wire transmission line. The antenna
itself is both a very efficient radiator and a highly directive one.
Transmitting Equipment
At the transmitting station the apparatus for each channel comprises,
(1) wire terminal equipment and repeaters, (2) a voice frequency control
TRANSOCEANIC TELEPHONE SERVICE 273
desk, (3) the radio transmitting set containing the oscillators, modu-
lators, and power amplifier, (4) a power control board, (5) rectifying
apparatus and filters for supplying direct current at 10,000 volts, (6)
motor-generators for providing various circuits with direct current, (7)
water circulating pumps, tanks, and cooling units.
The wire terminal equipment and repeaters at the transmitting
station are standard units mounted on relay racks beside the voice
frequency testing apparatus common for all channels.
The voice frequency control desk provides facilities by which the
attendant can monitor the incoming voice currents and the outgoing
radio signal. Means are provided for observing the volume of these
signals. Oscillators are provided for the purpose of quickly checking
the performance of the system during line-up periods and for sending
Morse signals over the radio link when required. The control desk
Fig. 2- — Front view of short-wave radio transmitter of type used at Lawrenceville.
is also equipped with apparatus for direct telegraph communication
with the technical operator at New York.
The radio transmitter consists of seven independently shielded units
mounted on a common sub-base to form a single assembly, 4 ft. by 20
ft. by 7 ft. high. Some of the units are subdivided into several small
shielded compartments. Very effective electrical screening or shield-
ing between the various parts of a short-wave transmitter is essential.
Otherwise stray fields introduce unwanted feedback couplings which
produce distortion effects and spurious oscillations. Fig. 2 is a front
view of the transmitter. Beginning at the left there are two units for
speech amplification, one for radio frequency generation and modula-
tion, one unit each for the first stage, the interstage circuit and the
last stage of radio-amplification, and a double-sized unit for the output
circuit. It is interesting to note that the over-all length of this as-
274 BELL SYSTEM TECHNICAL JOURNAL
sembly is as much as five eighths of a wave-length at the highest fre-
quency in its operating range, which is 9000 to 21,000 kilocycles. Each
transmitter is required to operate at several assigned frequencies within
this range and to change in a few minutes from one to another. This is
done by changing coils and varying condensers in the oscillator and
amplifier circuits and switching to different quartz crystals. Except in
cases where two assigned frequencies are in harmonic relationship, it is
necessary to provide a crystal for each of the frequencies. The crystals
are mounted in an oven and continuously maintained at 50 deg. ±
0.05 deg. cent, by recording regulators. In order to avoid long inter-
ruptions to service in the event of a crystal failure or other circum-
stance requiring the opening of the oven and the subsequent re-
establishment of temperature equilibrium, the ovens and crystals are
provided in duplicate.
The electrical problems which are encountered by the engineer
designing a power amplifier for these high frequencies arise largely from
the inherent stray or distributed capacities and inductances which are
far less important at lower radio frequencies. For example, between
the anodes of the amplifier circuit there exist capacities, which are
composed of capacities within the tube itself, the direct capacities be-
tween the tube water jackets, the mounting plates and the like. The
total value of this composite capacity in the last stage is approximately
100 m.m.f. This value cannot be appreciably reduced by any change
in design which now seems desirable. The reactance of 100 m.m.f. at
20,000 kilocycles is about 80 ohms. Thus the engineer is confronted
at the outset with a generator (the tubes) which has an internal impe-
dance in the order of 2000 ohms but across whose terminal is shunted
inherently an 80-ohm reactance. Fortunately, this obstacle can be
surmounted by introducing resonance effects but nevertheless it places
very important limitations on the design of the associated circuits.
These problems become more difficult with increase of either power or
frequency. Increase in power requires higher voltages and currents
and thus larger elements, spaced farther apart. The augmented bulk
increases both stray capacities and unwanted inductance of leads.
Higher frequencies increase the magnitude and therefore the relative
importance of these effects.
The power control board has nine panels equipped with the necessary
instruments and apparatus for controlling and distributing all power
to the transmitter. The motor-generators, pumps, fans, oil circuit
breakers, and other apparatus are remotely controlled from this point.
A system of relays and signal lamps provides protection and indicates
the location and general nature of any trouble. With the exception of
TRANSOCEANIC TELEPHONE SERVICE 275
the application of high-voltage direct current, the entire system starts
up and shuts down in the proper sequence in response to the manipula-
tion of a master control switch.
Direct current at 10,000 volts is supplied to the anodes of the power
amplifier tubes by a transformer and rectifier using six standard two-
electrode thermionic tubes. The rectified current is filtered separately
for each stage of the amplifier. This is necessary to prevent distortion
by interstage modulation caused by the common impedance of the
rectifier. Effects of this nature become important as the requirements
placed on unwanted modulation products become more stringent.
Transmitting Antennas
The antennas at Lawrenceville all have comparatively sharp direc-
tional properties. Such antennas are readily realized when dealing
with radio waves of very short wave-lengths. Although the funda-
mental principles involved in producing these directional effects have
been known for many years, economic limitations effectively prevented
their application to transmitting antennas for long wave-lengths.
These limitations are altered immensely in the case of antennas for
short wavelengths and, when the useful propagation properties of short
waves became known, great stimulus was given to the development of
antennas for directional sending and receiving. The same type of
antenna can be used, of course, for both purposes but, since the objec-
tives when sending and receiving are somewhat different, the tendency
has been to develop arrangements adapted to each case.
Directional transmission is a very large subject and will only be
touched upon suflficiently to describe in a very general way the anten-
nas at Lawrenceville. There are many possible arrangements and
combinations and the engineers must choose from these the ones most
suitable for their purpose. In general all of the schemes depend upon
producing interference patterns which increase the signal intensity in
the chosen direction and reduce It to comparatively small values in
other directions.
One of the methods of obtaining a sharply directive characteristic is
to arrange a large number of radiating elements in a vertical plane
array, spacing them at suitable distances and interconnecting them in
such a manner that the currents in all the radiating members are in
phase. A simple way of accomplishing this result and the one which
is now being employed at Lawrenceville depends upon the manner in
which standing waves are formed on conductors. It is generally known
that current nodes and current maxima will recur along a straight
conductor whose length is an exact multiple of one half the wave-length
276
BELL SYSTEM TECHNICAL JOURNAL
of the exciting e.m.f. and that the phase difference between successive
current maxima is 180 deg.^ Such a conductor when folded in a vertical
plane as shown in Fig. 3 and with its length adjusted slightly to com-
pensate for the effects of folding, satisfies the aforementioned require-
ments for producing directional radiation. The arrows in Fig. 3 indi-
cate the relative directions of current flow and the dotted line indicates
Fig. 3- — Conductor bent to form one section of simple directive antenna. The type
used for transmitting at Lawrenceville.
the current amplitudes along the conductor. It will be noted that the
instantaneous currents in all the vertical members are in the same direc-
tion and that in the cross members their directions are opposed. Due
to these current relations and the physical positions of the elements,
the cross members radiate a negligible amount of energy whereas the
vertical members combine their effects for the directions perpendicular
3 This assumes of course that the conductor is in space free frorn objects affecting
its electrical properties and that the ends are free or properly terminated to produce
reflections.
TRANSOCEANIC TELEPHONE SERVICE 277
to the plane of the conductor. In other directions destructive inter-
ference reduces the radiation from the vertical members. The system
is equivalent to four Hertz oscillators driven in phase, and arranged in
two groups one half wave-length apart, the two oscillators of each group
being placed one above the other. Both computation and experiment
have shown that with this system of radiation there is an improvement
of approximately 6 db. In other words the same signal intensity in the
chosen direction is obtained with one fourth of the power required by a
one-element radiator. A second similar conductor system placed
directly behind the first in a parallel plane one quarter wave-length
away, will be excited parasitically from the first conductor and will act
as a reflector, thereby creating a unidirectional system. It has been
found that the reflector further reduces by 3 db the power required to
maintain a given signal intensity in the desired direction, thus bringing
the total gain for the system up to 9 db. This is also in agreement with
the theoretical computations.
It is obvious that the system in Fig. 3 can be extended vertically to
include more radiating elements by increasing the length of the con-
ductor and it can be enlarged horizontally by placing several units
alongside each other, care being taken to obtain the desired phase rela-
tions by transmission lines of the proper length. In this way large
power savings may be effected. At Lawrenceville the maximum gain
is about 17 db (a power ratio of 50) over a vertical halfwave oscillator.
The enlarged system lends itself readily to mechanical support and
forms so-called exciter and reflector "curtains" which are suspended
between steel towers appropriately spaced. Aside from other con-
siderations, which will be mentioned in connection with station layout,
the size of the antenna is influenced by the complex and variable nature
of the wave propagation through space. At present this determines
the degree of directivity which is most useful for the average conditions.^
The closed loops of each unit corresponding to Fig. 3 greatly facilitate
the removal of sleet. In addition to loading the antenna mechanically,
ice, having a dielectric constant of 2.2 at these high frequencies, ad-
versely affects the tuning. At Lawrenceville sleet is removed by heat-
ing the wires with current at 60 cycles. This is accomplished without
interfering with the service by employing one of the less familiar
properties of a transmission line. The same property also is used to
effect impedance matches wherever the transmission lines are branched.
If a line, exactly one quarter wave-length long, of surge impedance Z^ is
terminated with a load Z^, the sending-end impedance Z, is equal to
*J. C. Schelleng, "Some Problems in Short Wave Telephone Transmission."
Presented to the Institute of Radio Engineers at a meeting Nov. 6, 1929.
278
BELL SYSTEM TECHNICAL JOURNAL
Zg-jZji. If Z^ is a pure resistance the sending-end impedance is a
pure resistance. Hence a quarter wave-length line may be used to
connect two circuits of different impedances and these impedances may
be matched by controlling the value of Zg either by varying the diam-
eter of the conductors or their spacing. Likewise, if Z,, is fixed and Z^
is made very small, then Zg will be extremely large.
60 CYCLE
SOURCE
RADIO
TRANSMITTER
Fig. 4 — ^Antenna sleet-melting circuit.
In Fig. 4 two units of the type shown in Fig. 3 are excited through
transmission lines 1 and 2 of equal length in order to give the correct
phase relations in the radiating elements. The lines are joined in
parallel by condensers of low impedance at radio frequencies and they
are connected in series for 60-cycle currents by the quarter wave-length
line A which, being short-circuited at the one end, presents a very high
impedance to radio frequency currents at the other end and therefore
behaves like an anti-resonant circuit. The quarter wave-length line B
serves as a transformer and is adjusted to match the impedance at the
junction of lines 1 and 2 with that of the radio transmitter. The
quarter wave-length line C is effectively short-circuited for radio fre-
quencies by the condenser D and acts the same as A. These quarter
TRANSOCEANIC TELEPHONE SERVICE
279
wave lines consist of short lengths of pipe mounted on frames under the
antenna curtains as shown in Fig. 5.
Transmitting Station
Among the first radio problems encountered in the design of a trans-
mitting station for several channels are those concerning the size,
shape, and number of antennas, their directions of transmission, their
Fig. 5 — Section of antenna system at Lawrenceville, showing lower portion of
curtains and quarter wave transmission lines used as transformers and anti-resonant
circuits.
relative positions from the point of view of mutual interference, and
their grouping around the transmitters.
The number of antennas required for each channel is determined by
the hours of operation and the average grade of service which the sys-
tem is expected to render. For service covering a large portion of each
day several wave-lengths are necessary. Transmitters Nos. 1, 3, and 4
at Lawrenceville each are assigned three frequencies. No. 2 has five
assignments in order to improve the likelihood of at least one channel
being available throughout the entire day at all seasons.
The size and shape of the antennas are, of course, determined by the
directivity wanted, by the type employed, the frequency assignments,
and by considerations of cost. They are governed also by the necessity
of connecting several antennas to the same transmitting set. This
involves both the spacing and arrangement of antennas to avoid
280
BELL SYSTEM TECHNICAL JOURNAL
adverse mutual reactions and it requires that attention be given to the
losses in the connecting transmission lines, which are by no means
negligible. Operating economies suggest concentrating all the trans-
mitters at one point but the cost per kilowatt hour of modulated
high-frequency power must be taken into account when considering the
use of long transmission lines. It should be recognized, of course, that
in the early applications of a comparatively new art, it is impossible to
approach anything like accurate evaluation of all the factors entering
into economic balances and furthermore very considerable weight
needs to be given to the probable future trend of developments.
M4.4M
TRANSMITTING STATION
AMERICAN TELEPHONE & TELEGRAPH CO
LAWRENCEVILLE N.J.
32.7 M GREAT CIRCLE
TO LONDON
30.7 M
20.7 M
5.6 M
Fig. 6 — ^Arrangement of antennas at Lawrenceville transmitting station.
At Lawrenceville all of the antennas for the three channels to Eng-
land are arranged in a straight line about one mile long. The direction
of this line is perpendicular to the great circle path to Baldock, England,
where the signals are received, (Fig. 6). The antennas for the fourth
channel are similarly arranged in a line 1500 ft. long and they are
directed for transmission to Buenos Aires, Argentine.
Placing several antennas in a single line reduces the cost of the sup-
porting structure, and all the antennas have a clear sweep in the direc-
tion of transmission. By locating them in proper sequence with re-
TRANSOCEANIC TELEPHONE SERVICE 281
spect to wave-lengths It is possible without objectionable interference,
to place the antennas end-to-end and thus use supporting towers in
common. Due to the wide difference in wave-length between adjacent
antennas and their right-angle position with respect to the line of
transmission, their proximity has no appreciable effect different from
that of the towers. The proper selection of tower spacing in respect
to wave-lengths makes it possible to erect a uniform supporting struc-
ture. This has the advantage of flexibility and will permit future
alterations of either the location or size of a given antenna. At present,
each antenna occupies the space between three towers.
In order to avoid undue loss In the transmission lines the radio trans-
mitters are grouped In two buildings. The buildings each contain two
transmitters and are identical In layout, in so far as the radio equipment
is concerned. Building No. 1 has additional space for the central wire
terminating and testing equipment. This apparatus is contained In an
electrically screened room which effectively prevents high-frequency
fields from interfering with the proper functioning of the apparatus.
Receiving System
Short-wave reception is characterized by less difficulty with static
than that encountered with long waves. On the other hand it suffers
interference from sources such as the ignition systems of passing air-
planes and automobiles, which ordinarily do not disturb long-wave
systems. Frequently the Incoming radio waves suffer wide and rapid
swings in Intensity and there are variations in the apparent direction
of arrival. On account of the extremely high frequencies the ap-
paratus and antenna structures are very different from those for the
long waves; otherwise the general schemes of reception are similar,
directional effects and double detection methods being employed for
both.
The radio wave is collected by means of a directional antenna array
whose prime function is to improve the ratio between the desired signal
and unwanted noise or other Interference. This It does in two ways:
viz., (1) by increasing the total signal energy delivered to the receiver
and (2) by discriminating against waves whose directions of arrival
differ from the chosen one. Increasing the total energy collected from
the incoming message wave permits the detection of correspondingly
weaker signals because there is an apparently irreducible minimum of
noise inherent to the input circuits of the first vacuum tube in the
receiver '" and this noise establishes a lower limit below which signals
cannot be received satisfactorily. Since, under many conditions, the
^ J. B. Johnson, Physical Rev., July 1928.
282
BELL SYSTEM TECHNICAL JOURNAL
directions of arrival of static and other disturbances including unwanted
radio signals are random, it is obvious that sharp directive discrimina-
tion aids very materially in excluding them from the receiver. On
the other hand, the antennas are not sharply resonant systems and they
do not distinguish between waves from substantially the same direction
and closely adjacent in frequency. This duty is left to the circuits
of the radio receiver.
BEATING
OSCILLATOR
DIRECTIONAL
ANTENNA
TRANSMISSION
^ LINE
LINE
TERMINATION
AND REPEATER
FIRST
DETECTOR
SECOND
DETECTOR
TWO
STAGES
RADIO
AMPLIFIER
INTERMEDIATE
FREQUENCY AMPLIFIER
AND FILTER
TELEPHONE
LINE
AUDIO AMPLIFIER
AND MONITORING
APPARATUS
AUTOMATIC
TWO STAGES
GAIN
INTERMEDIATE
CONTROL
FREQUENCY AMPLIFIER
Fig. 7- — Block schematic of receiving system.
Having collected the signal with a directional antenna the energy is
conveyed to the receiving set by means of concentric pipe transmission
lines of small diameter. The use of concentric conductors simplifies
the prevention of direct signal pick-up by the lines, it reduces losses
and prevents external objects from influencing the transmission proper-
ties, thus allowing the line to be buried in the ground or placed a few
inches above the surface where it will have no appreciable adverse
effect on the antenna performance.
Referring now to Fig. 7, the radio currents arriving over the trans-
mission line are first amplified by two stages of radio amplification
involving tuned circuits which discriminate further in favor of the
wanted signal. The signal delivered by the radio amplifier is at
a suitable level for efficient demodulation and is applied to the grid of
the first detector. By means of a beating oscillator whose frequency is
suitably adjusted, the first detector steps the signal carrier frequency
down to a fixed value of 400 kilocycles from one in the range 9000 to
21,000 kilocycles which depends, of course, on the distant transmitting
station assignment. The intermediate frequency signal at 400 kilo-
TRANSOCEANIC TELEPHONE SERVICE 283
cycles then passes through a combination of amplifiers and filters which
further exclude the unwanted interference. The wanted signal reaches
the second detector where it is demodulated and the voice currents
reproduced. The latter are then amplified and applied to the telephone
lines.
A portion of the output from the intermediate amplifier which would
normally go to the second detector grid, is diverted and further am-
plified. It is then supplied to a device which automatically tends to
maintain the receiver output volume constant by controlling the bias
potential of the first detector grid circuit. The time constants are
adjusted so that this gain control does not respond to the normal varia-
tion in signal power corresponding to speech modulation. Otherwise
of course, there would be serious distortion effects. This device
partially offsets the ill effects of wide fluctuations in signal intensity
but it does not overcome the deterioration in signal quality which
usually accompanies the low field strengths during such fluctuations.
Receiving Equipment
At the receiving station the apparatus for each channel comprises (1)
the radio receiving set, (2) a power plant for the receiver, (3) wire
terminating equipment and repeaters. The latter are located at a
central point in the station along with certain voice frequency testing
apparatus used in common by all channels and supplied with power
from a common source.
A radio receiving set which embodies the above described system and
of the type installed at Netcong is shown in Fig. 8. It consists of a
large number of individually shielded units mounted on panels and
assembled on three self supporting racks of the type commonly em-
ployed in the telephone plant. This permits the use without modifica-
tion of certain standard pieces of equipment, such as jack strips, fuse
panels, meter panels, audio frequency filters, and the like. It also
permits the removal and repair or substitution of units with a minimum
of delay. The set is required to receive signals at three fixed frequen-
cies in the range 9000 to 21,000 kilocycles. This involves connections
with three antennas through three separate transmission lines. The
tuning of the antenna and transmission line terminations are rather
lengthy processes requiring precise adjustments. In order to facilitate
quick changes from one operating frequency to another without intri-
cate tuning operations, the first stage of radio amplification is provided
in triplicate and the switching is done between the first and second
stage. Thus the antennas are permanently connected to the set and
their adjustments remain undisturbed. The circuits of the second
284
BELL SYSTEM TECHNICAL JOURNAL
ca
u
nS
CQ
>
c
o
1-4
>
<1>
CJ
(U
o
13
I
+->
O
in
bO
S
TRANSOCEANIC TELEPHONE SERVICE 285
stage require tuning when the frequency is changed. Hence to tune
the receiver on any one of the assigned frequencies the attendant
merely moves the dials of the second stage to predetermined settings,
switches the grid circuit to a first stage which is already tuned and
connected with the proper antenna and he adjusts the beating oscillator
to obtain an intermediate frequency of 400 kilocycles. Screened grid
tubes are used for the first two stages of amplification. A key shelf is
provided with telephone and telegraph facilities. The power plant
consists of standard 24-volt and 130 batteries, rectifier charging units
and automatic regulators.
Receiving Antennas
In discussing antennas for directional sending it was mentioned
that an identical antenna could be used for receiving purposes, but
since the requirements in the two cases are not the same, quite dif-
ferent structures have been developed, although the methods of ob-
taining directivity are alike. In the sending case the reduction of
random radiation ceases to be profitable when the increment thus added
to the energy, which is radiated in the direction of the distant receiving
station, is a relatively small part of the total. In the receiving case,
although the response to the wanted signal may not be increased
appreciably by further improvement in the directive pattern, the
reduction in noise and interference from random directions justifies
additional improvement. Expressed another way, the objective in
the transmitting case is a high gain compared to a nondirectional
antenna, whereas in the receiving case the objectives are, first, a
high average signal-to-noise ratio and, second, a gain sufficient to
override the noise inherent to the receiving set. Satisfying the first
accomplishes the second.
Improvement of the average directional discrimination means a
nearer approach to ideal conditions. Whereas steel towers, section-
alized cables, guys and the like, when properly located relative to the
conductors of a sending antenna, do not cause any appreciable power
loss, their presence near the receiving antenna may prevent the
realization of the extreme directive properties which are wanted.
Moreover, there is need for much greater rigidity in the positions of the
conductors. For this reason the antennas at Netcong are supported
on wooden frames constructed like large crates.
Due to the variable conditions surrounding the propagation of short
waves in space, the vertical angle of arrival of the signal wave at the
receiving station frequently changes considerably throughout a twenty-
four hour period and is not always the same from day to day. In
19
286
BELL SYSTEM TECHNICAL JOURNAL
order to combat this variable condition, it appears desirable to select
an antenna arrangement which does not have sharp directional proper-
ties in a vertical plane passed through the horizontal direction of arrival.
The type of antenna selected for Netcong meets this requirement by
having only a single horizontal row of quarter-wave vertical elements in
one plane. Another solution, of course, would be to provide several
antennas of different characteristics and to shift about from one
antenna to another as the conditions warranted.
Fig. 9 is a general view of one of the Netcong receiving antennas.
Like the transmitting antennas, the conductors are arranged in two
parallel planes one quarter wave-length apart in order to obtain a
unidirectional system. The conductor in each plane is bent and ter-
Fig. 9 — One of receiving antennas at Netcong. (24.7 meter wave-length.)
minated as indicated in Fig. 10 but is much longer than that shown.
The vertical members are marked A. As in the transmitting case the
directional effect depends upon the manner in which standing waves
occur along the conductor. A signal wave arriving broadside to the
array, induces voltages in the vertical members which are identical in
phase and amplitude.
Because the vertical members are interconnected alternately at the
top and bottom by members of one quarter wave-length and the last
horizontal members are one eighth wave-length, the net effect of the
induced voltages is the establishment of standing current and voltage
waves along the conductor. The receiver is connected at a voltage
anti-node and the current which flows through it is proportional to
the sum of the voltages induced in the vertical members. In the case of
TRANSOCEANIC TELEPHONE SERVICE
287
a signal wave arriving from the horizontal directions parallel to the
plane of the array, the voltages in the vertical members are in succes-
sive quarter-phase relationships, no standing waves are produced, and
no current flows through the receiver. Because current nodes occur
at the center of each horizontal member, the loss by reradiation from
these members is negligible. This is an important feature which
contributed to the selection of this type of antenna for Netcong.
The size of the antenna is determined largely by the manner in which
the signal waves arrive although costs cannot be wholly neglected.
The useful length is limited by the fact that random fading occurs at
distances as short as ten wave-lengths and it is doubtful if an antenna
this long would realize the computed improvement. The cost per
decibel gained is small for the initial steps, but it mounts very rapidly
as the length of antenna increases. The height also is limited by cost
Vs
^
r
\
\
1
a!
^^^^^n^
CURRENT
NODE
%.
VOLTAGE
ANTI-NODE
t
%»
CURRENT
NODE
^8
Fig. 10 — Diagram of simple directive receiving antenna.
and by the necessity of allowing for considerable variation in the verti-
cal angle of arrival as discussed in a previous paragraph.
The antennas at Netcong are six wave-lengths long and the lowest
conductors are about 10 ft. off of the ground. The gains over that of a
half wave vertical antenna are in the order of 16 db (power ratio of 40).
The average improvement in signal-to-noise ratio is of the same order.
There are certain null points toward the sides and rear for which the
ratio of directional discrimination is very large.
The transmission lines are constructed of inner and outer copper
tubes respectively 3/16 in. outside diameter and 5/8 in. inside diameter
The tubes are held concentric by torroidal shaped insulators made of
Isolantite, a ceramic product similar to porcelain and well adapted for
high-frequency voltages. This same material is used for insulating
purposes throughout the transmitting and receiving antennas. Trans-
mission lines are supported a few inches above the ground and are
connected to earth at short intervals. The lines vary in length from
288 BELL SYSTEM TECHNICAL JOURNAL
200 to 1500 ft. One of the interesting problems in connection with
their design is the provision of means for allowing variation in length
with temperature. Ordinary expansion joints introduce difficulties
with electrical contacts and impedance irregularities. To avoid these
the lines are made 10 per cent longer than otherwise necessary and they
follow a sinuous course which permits the necessary bending. Sharp
turns are not permissible because experiments have shown that they
cause reflection disturbances. The measured loss in 1000 ft. of line
at 20,000 kilocycles is 2 db.
Receiving Station
The radio problems encountered in the layout of the receiving sta-
tion, in general, include most of those already mentioned in connection
with the transmitting station, but their solution in some instances is
quite different. In addition there are requirements imposed by sources
of radio noise both within the station itself, and in the surrounding
area which is beyond the control of the station.
The number of antennas is determined, of course, by the frequency
assignments of the distant transmitting station. Where two assign-
ments are within 100 kilocycles it is possible to use the same antenna for
both, but thus far, this has not been done at Netcong.
The size of the antennas is not limited appreciably by the length of
transmission lines because other factors make it necessary to separate
them rather widely. On this account and also because the receiving
apparatus and its power plant are small, comparatively inexpensive
units, it is economical to place the receivers in small buildings centrally
located with respect to the group of antennas for one channel. In this
case the lengths of transmission lines are not controlling factors and the
dimensions of antennas are governed primarily by the considerations
previously outlined when describing the individual antenna. The
small height of the antenna permits them to be placed in the line of
reception of other antennas spaced ten wave-lengths or more away and
of widely different frequencies such as those of one channel. Antennas
adjusted for the same order of frequency are separated more than this.
On the other hand, to avoid adverse reactions no two are placed adja-
cent and end-to-end as at the transmitting station. The end-to-end
separation at Netcong is in the order of four wave-lengths. The areas
surrounding antennas are cleared of trees and kept free of all overhead
wires or conducting structures to avoid reflection effects which disturb
the directional characteristic of the antenna systems.
The locations of antennas are also influenced materially by the neces-
sity of avoiding interference from the ignition systems of internal
TRANSOCEANIC TELEPHONE SERVICE
289
combustion engines. This imposes a requirement that the station site
be isolated from air routes and roads carrying heavy traffic. The
antennas are placed as far as possible from secondary roads which
cross their line of reception.
The layout at Netcong is shown in Fig. 11. There are thirteen
antennas arranged in four groups with a receiver building for each
group. A headquarters building located at the road entrance contains
the wire terminating equipment, line repeaters, and voice frequency
GREAT CIRCLE
TO LONDON
42.9M
CENTRAL TERMINAL
BUILDING
RECE>\;'ER
20.6 m
33.2M
16.0 M
303 M
RECEIVER
*2
/
I4.2M
GREAT CIRCLE TO
BUENOS AIRES
207M
Fig. 11 — Arrangement of receiving antennas at Netcong receiving station.
testing apparatus. The power plant at each receiver and the entire
central terminal apparatus at the headquarters building are placed in
electrically shielded rooms to prevent radio noise disturbances emanat-
ing from them and reaching the receivers directly or via the antennas.
The radio stations described herein are pioneer commercial applica-
tions in the development of short wave telephone transmission.
Although progress has been rapid and far-reaching our knowledge of
the behavior of short waves is by no means complete. It is reasonable,
therefore, to expect that the future holds many improvements and that
the information obtained by further fundamental investigations may
materially alter both our views of the transmission phenomena and
our ideas of what the apparatus and stations should be.
The Words and Sounds of Telephone Conversations
By NORMAN R. FRENCH, CHARLES W. CARTER, JR.,
and WALTER KOENIG, JR.
This paper presents data concerning the vocabulary and the relative fre-
quency of occurrence of the speech sounds of telephone conversation.
Tables are given showing the most frequently used words, the syllabic struc-
ture of the words, the relative occurrences of the sounds, and, for each vowel,
the percentage distribution of the consonants which precede and follow it.
Comparisons are made with the vocabulary and relative occurrence of speech
sounds in written English.
Introduction
CONVERSATION resembles other forms of communication in its
use of symbols, in themselves merely physical phenomena, but
which combined in sequence are by convention endowed with meaning.
The elementary symbols used in conversation are the acoustic dis-
turbances called speech sounds. A language is characterized by the
speech sounds which it uses and by the combinations of speech sounds
which form syllables and words. The physical description of a lan-
guage involves a statement of the characteristics of the individual
sounds and also of the frequency of occurrence of each sound and
combination of sounds. The latter or statistical aspect of conversation
is treated in this paper. ^
Studies of the relative frequency of English speech sounds have been
made previously, but they have been confined, so far as the writers
have ascertained, to the analysis of written matter. Of these an
extended investigation is that made by Godfrey Dewey. ^ For peda-
gogical purposes in connection with difficulties in spelling and in
developing methods of shorthand writing, which seem to have been
the aims in the previous studies, written matter is the natural point
of departure.
There are obvious differences between English when read aloud from
printed matter and English used as a medium of conversation, which
might be expected to produce differences between analyses based on the
two forms. Written matter is permanent and, to some degree, self-
conscious; it receives qualification by dependent clauses and preposi-
1 Some of the results of this study were presented at the May, 1929, meeting of the
Acoustical Society of America. See French and Koenig, JourjialA. S. of A., October,
1929, p. 110.
' " Relative Frequency of English Speech Sounds," Harvard Studies in Education,
IV. Harvard University Press, 1923.
290
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 291
tional phrases and it makes use of synonyms and a vocabulary more or
less ample according to the writer's fancy and ability. In conversation
attention seems to be paid more to the thought than the form of ex-
pression, with the exception, perhaps, that certain modes acceptable in
writing may be considered as too formal for conversation. It is doubt-
ful, however, that conversation should be described as more concise
than written matter. The sentences are, indeed, likely to be shorter.
They are often incomplete, in fact. But often in conversation even a
single statement is completed only after a number of fumbling attempts,
an oral manifestation of crystallizing thought, whereas in written
matter the final expression alone would appear. In repetition of a
thought, synonyms are less likely to be found in conversation than in
written matter. Dependent clauses are less frequent than in written
matter. Qualification and description often take the form of separate
sentences, so that those words characteristic of involved construction
tend to be less prominent in conversation, while the framework words,
such as the auxiliary verbs and pronouns, are more intensively used.
These differences, which tend to restrict the vocabulary, will be found
reflected in the comparisons given later in this paper.
The material for the present study was obtained from telephone
conversations over typical toll circuits terminating in the city of New
York. The process of noting the words of the conversations was
carried out in the following manner: During one week the observer
recorded nothing but the nouns used, during another week she re-
corded only verbs, and during a third week only adjectives and adverbs.
This routine was repeated until observations had been made on 500
conversations for nouns, 500 conversations for verbs, and 500 con-
versations for adjectives and adverbs. Three other classes of words
were recorded: prepositions and conjunctions, pronouns, and articles;
but for these classes approximately 150 conversations in each case
were judged to be sufficient.
Certain classes of words were, for various reasons, omitted entirely.
These are names, titles, exclamations, letters, numbers and the name-
less sound which may be transliterated as "er" or "uh," so frequently
punctuating a haltingly expressed sentence. A more comprehensive
method, but based on a much smaller number of conversations, indi-
cates that the ratio of the total number of occurrences of words
in the omitted classes to the number of occurrences of the words
discussed in this paper is about one to four. Within the omitted
group the division is roughly as follows: proper names and titles, 20
per cent; exclamations and interjections, such as "yes," "no," "well,"
"yeah," "uh-huh," "oh," "all right," "hello," "good-by," laughter
292 BELL SYSTEM TECHNICAL JOURNAL
and profanity, 40 per cent; letters and numbers, 25 per cent; and the
sound "er," 15 per cent.
The words which were obtained by the process of sampling con-
versations for specific parts of speech are not, of course, identical with
those which would have been obtained had the entire conversation
been recorded. The representativeness of the most frequent words,
which largely determine the relative frequency of the speech sounds,
was investigated by a later test in which a different observer recorded
the verbs from 250 conversations. These results will be discussed
later, but it may be pointed out here that the word list obtained by the
two observers corresponded so closely that it is felt that the samples of
parts of speech were recorded with sufficient accuracy and were
sufficiently large to justify taking the words obtained as a good repre-
sentation of the main body of telephone conversation.
The kinds of conversations encountered are shown in Table I. The
great preponderance of business calls is reflected, as will be shown later,
TABLE I
Types of Telephone Calls on which Observations Were Made
a. Material
Business Calls 89.0 per cent
All other Calls 11.0 per cent
b. Speakers
Two Men 86.5 per cent
Two Women 10.4 per cent
Man and Woman 3.1 per cent
in the vocabulary. If a smaller percentage of the calls had been busi-
ness in nature and if a larger percentage had been between women the
vocabulary would probably have been different. Whether any marked
change would have been found is open to some doubt when it is re-
called that business may cover a wide range of topics and that in the
1,900 conversations from which samples were taken there may have
been as many as 3,800 different speakers. Evidence will be given,
however, which indicates that the relative frequency of the speech
sounds would have been changed very little.
Words
The number of conversations on which observations were made was
regulated to some extent by the ratio of the number of total words to
the number of different words recorded in each class. In the early
stages of observing many of the total words recorded were different,
making this ratio low, but as the observations continued fewer and
fewer new words were encountered. In Figure 1 curves are given
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 293
which show, for two classes of words, the way in which the number of
different words in each class varied as the total number of words in
that class increased. To take the nouns, of the first 200 about half
were different, of the first 1,000 about a third were different, of the
total of 11,660 nouns recorded about one tenth were different. An
extrapolation of the curve indicates that the observations would need
2000
(/)
Q
CC
O
5
I-
z
UJ
a
u
u.
o
1000
500
i 100
z
50
"''
v
*^
^'i> -
^^-'
•aC
>^^
fr
.h
^
j>^
X
fl
b
^.^^f--^
,/
r-
^
^
r'
f
^-^
^
/
y
^
^
^
^
v'
^r
.^
• y
^
^
100
1,000 10,000
TOTAL NUMBER OF WORDS IN EACH CLASS
100,000
Fig. 1 — The number of different words occurring in a given total number of words,
for nouns and for verbs.
to be increased tenfold from this point in order to double the number
of different nouns. Approximately the same extension of the ob-
servations would be required to double the number of different verbs.
In neither case, however, could a material change in the relative oc-
currence of the speech sounds be expected if the observations were so
extended. This will be shown below.
Table II shows the total number of words and the number of dif-
ferent words for each part of speech separately. The verbs and aux-
iliary verbs, which were recorded together, have been separated in the
table. The numbers of total words for the other three minor classes
have been found by multiplying the observed figures by the ratio of 500
to the actual number of conversations (about 150) on which observa-
tions were made for these classes. The numbers of different words for
these classes are not similarly increased since virtually all the possible
different words were obtained in the observations. In finding the
number of different words the various forms of the words, such as the
plural form of the nouns, the different tenses of the verbs, and the
comparative and superlative forms of the adjectives, have not been
counted as separate words, although they were recorded and are
294
BELL SYSTEM TECHNICAL JOURNAL
TABLE II
Occurrence of Parts of Speech
Parts of Speech
Number of Words
Ratio Total
Total
Different
to Different
Nouns
Adjectives and Adverbs
Verbs
Auxiliary Verbs ...
11,660
9,880
12,550
9,450
17,900
12,400
5,550
1,029
634
456
37
45
36
3
11.3
15.6
27.5
255
Pronouns *
Prepositions and Conjunctions *
Articles *
398.
344.
1850.
79,390
2,240
35.4
* Derived from data on less than 500 conversations.
treated separately in the analysis for speech sounds. An exception to
this is that each form of the auxiliary verbs "be," "can," "may," etc.,
was counted as a separate word.
It is of interest to find that of approximately 80,000 words so ob-
tained, only 2,240, or less than 3 per cent, are different words. If each
of the modifications of a word is counted as a different word the number
of different words is increased to 2,822 ; but even on this basis less than
4 per cent of the total words are different words. Even among the
nouns the number of different nouns is only a tenth of the total number
of nouns. The five minor parts of speech shown in the last four lines
of Table II form only 5 per cent of the different words and yet make
up 57 per cent of the total words. The nouns, which constitute 46
per cent of the different words, contribute only 15 per cent of the total
words. Such figures indicate clearly that conversation is based on a
framework built up of a relatively small number of different words,
arranged in many patterns, which supports the more variegated words
which convey most of the meaning.
A more detailed idea of this framework is given by Tables III-o and
1 1 1-6, which contain a list of the words which were observed in at least
1 per cent of the conversations. In Table Ill-a the words are arranged
in order according to the total number of times they were recorded.
This is approximately, but not quite, the same as the order of the num-
ber of conversations in which they occurred as may be seen by exam-
ining the numbers following each word. In Table 1 1 1-6 the same words
are arranged alphabetically, for ease in reference. The list comprises
737 words out of the 2,240 different words recorded. The importance
of the list lies in the fact, as will be shown later, that these words almost
completely determine the relative frequency with which the elementary
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 295
Column A:
Column B:
TABLE Ill-a
Word List — Numerical Order *
Words which Occurred in One Per Cent or More
of the 500 Telephone Conversations Analyzed
Total number of times the word (or some form of it) was used.
Number of conversations in which the word occurred.
A
B
A
B
A
B
3,990
I
467
336
what
193
170
anything
100
3,540
you
499
330
morning
191
170
my
97
3,110
the
496
326
an
178
2,060
a
487
321
just
211
168
night
107
2,046
on
458
317
over
208
159
call, n.
111
1,942
to
472
296
be
175
157
your
100
1,792
that
397
156
little
117
1,605
it
417
295
or
178
146
stuff, n.
92
1,506
is
419
295
take
207
146
won't
115
1,363
and
391
276
am
172
140
last, a.
106
274
come
168
140
she
50
1,360
get
393
274
make, v.
169
139
all
100
1,305
will, aux.
402
273
give
172
139
better
103
1,190
of
396
268
very
165
1,170
in
408
264
send
172
139
number, n.
80
1,115
he
297
262
as
125
138
out
90
1,100
we
294
259
right, a.
173
137
try
100
913
they
253
133
ask
101
887
see
328
247
order, n.
119
133
sell
81
883
have
367
243
good
149
131
not
96
823
for
330
241
minute
155
130
those
100
241
price
123
125
only
84
753
know
325
238
here
157
121
business
83
640
don't
301
234
car
88
120
office
83
638
do
302
230
had
151
618
are
293
229
time
165
118
late
94
599
want
297
228
can't
132
118
no, a.
77
597
go
280
226
much
160
117
all right
74
553
tell
264
115
pretty
92
518
with
263
224
there
144
115
shipment
80
496
me
283
222
week
120
113
back, a.
79
486
him
223
215
let
148
112
look, V.
85
214
letter
112
112
mean, v.
82
480
about
266
209
any
140
112
off
67
476
at
238
200
did
144
109
hear
85
474
think
232
199
more
134
473
this
240
195
didn't
142
108
ship, V.
68
458
day
251
193
talk, V.
131
108
way
81
418
thing
235
193
today
124
106
his
70
410
say
211
105
dollar
66
396
can, aux.
221
190
other
128
105
too
77
386
call, V.
200
186
company
111
105
wire, n.
78
379
would
207
186
fine, a.
122
104
haven't
83
184
could
124
104
then
88
370
them
170
183
same
127
103
how
78
358
was
194
179
put
114
103
who
74
339
now
216
178
wait, V.
135
338
from
196
176
has
114
98
buy
60
*In ambiguous cases the part of speech is denoted as follows: noun, n.; verb, v.;
adjective or adverb, a.; auxiliary verb, aux.; preposition, prep.
296
BELL SYSTEM TECHNICAL JOURNAL
TABLE Ill-a (Cont'd)
A
B
A
B
A
B
97
man
67
64
some
43
43
chance
37
97
wouldn't
79
63
been
53
43
coffee
8
96
before
79
63
but
54
43
every
36
96
first
71
63
contract
31
42
stock, n.
33
96
market
51
63
out of
25
42
than
33
93
something
67
63
sample, n.
37
92
month
70
63
these
57
41
feel
30
92
well, a.
71
40
different
30
89
case
47
61
few, a.
50
40
meet
28
61
ton
27
40
reason, n.
32
89
find, V.
72
61
train, n.
33
40
show, V.
30
88
by
75
60
best
52
40
which
33
88
probably
69
60
everything
47
40
yesterday
35
87
afternoon
62
60
may
50
39
pound, n.
21
87
line
60
60
thank
56
38
doing
34
87
name, n.
52
59
check, n.
35
38
keep
28
86
like, V.
71
58
along, prep.
33
86
sure
72
58
job
44
38
old
22
86
yet
67
37
awful
31
85
fellow
70
58
tonight
40
37
bag
24
58
up, prep.
42
37
certainly
28
85
pay, V.
55
57
home
35
37
difference
31
84
talk, n.
67
57
our
47
37
information
29
84
write
61
56
another
46
37
matter, n.
31
83
new
62
56
away
45
37
must
30
83
next
61
56
should
43
37
phone, n.
35
83
were
66
55
expect
49
37
seem
33
82
understand
63
54
around
54
82
when
69
54
copy, n.
36
36
boy
31
79
people
59
36
hand, n.
31
78
year
53
54
idea
38
36
hour
30
53
bad
46
36
house
27
77
us
63
53
couldn't
47
36
mind, n.
30
76
soon
63
52
bill, n.
39
35
early
26
75
place, n.
55
52
nice
38
35
figure, V.
24
73
money
56
52
tomorrow
36
35
oil, n.
18
72
guess, V.
63
52
word
45
35
question, n.
31
71
after
54
51
big
42
35
quite
32
71
hold, V.
60
51
where
42
71
through
46
51
whole
40
34
ahead
27
69
isn't
52
34
point, n.
27
69
leave
54
50
cent
31
34
wonder
28
49
figure, n.
32
33
offer, n.
20
68
coal
27
49
glad
39
33
speak
31
68
might, aux.
59
49
ship, n.
28
33
unless
33
68
work, V.
50
48
report, n.
29
32
bid, n.
24
67
again
62
47
suppose
41
32
deliver
24
67
her
30
46
into
42
32
less
23
67
its
67
45
boat
22
32
possible
24
67
so
53
45
couple
38
67
their
63
45
high
34
31
believe
25
66
long
55
31
check, V.
28
65
because
47
45
ought
37
31
low
22
45
trouble
33
31
situation
25
65
use, V.
50
44
barrel
20
31
touch, n.
29
65
work, n.
49
44
delivery
36
31
why
25
64
listen
55
43
anybody
40
30
basis
21
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION
TABLE Ill-a (Cont'd)
297
A
B
A
B
A
B
30
fix, V.
21
24
hope.^v.
19
18
steel
10
30
move
23
18
trip, n.
17
30
ready
25
24
near
23
18
wasn't
18
24
piece
16
17
above
13
30
receive
22
24
start, v.
22
17
accept
14
30
sorry
25
24
wrong
20
17
against
13
30
town
22
23
busy
17
17
amount
15
29
between
29
23
ever
22
17
appointment
14
29
does
27
23
foot
13
17
cable
10
29
dope, n.
24
23
lot
19
17
cover, v.
14
29
mail, n.
22
22
card
9
29
many
25
22
forget
18
17
definite
13
29
moment
26
17
goods
13
29
need, v.
22
22
friend
14
17
plant, n.
9
22
special
15
17
possibility
15
29
paper
17
22
wire, V.
18
17
size
12
29
telegram
19
21
balance, n.
15
17
somebody
17
29
telephone, n.
27
21
change, n.
15
17
still, a.
16
29
though
29
21
loan
5
17
story
12
28
able
26
21
mail, V.
16
17
ticket
9
28
customer
22
21
welcome, a.
21
17
within
17
28
instruction
20
20
account, n.
16
28
note, n.
24
20
agreement
8
16
handle, v.
14
28
ring, n.
23
16
like, a.
16
28
room
19
20
anyhow
17
16
part, n.
15
20
cut, V.
18
16
quote
14
28
sale
25
20
exactly
17
16
tank
7
27
arrange
23
20
happen
15
16
truck
13
27
bring
24
20
list, n.
13
15
along, a.
13
27
doesn't
23
20
niessage
11
15
also
11
27
done
23
20
most
15
15
answer, n.
15
27
maybe
26
20
record, n.
18
15
board
8
27
never
23
20
stop, V.
18
27
order, v.
24
20
terrible
13
15
cargo
8
27
really
25
15
clean, v.
14
27
share, n.
10
19
address, n.
15
15
clear, a.
14
19
department
16
15
cocoa
9
27
stay
23
19
far
15
15
cost, v.
15
27
wish, v.
22
19
hold, n.
17
15
date
14
26
book
17
19
load, V.
16
15
interest, n.
9
26
inch
7
19
meeting
9
15
item
10
26
machine
14
19
nearly
19
15
station
8
26
proposition
21
19
plan, n.
12
15
spend
11
26
railroad
19
19
position
15
26
run, V.
20
19
rate
11
15
worry, v.
14
26
short
21
14
already
14
25
bank
12
19
straight
15
14
arrangement
11
18
anyway
16
14
bid, V.
11
25
change, v.
22
18
cheap
13
14
club
7
25
city
18
18
even
17
14
extra
11
25
hasn't
25
18
imagine
17
14
fact
14
25
help, y.
19
18
lunch
18
14
finish, V.
10
25
material
14
18
pier
10
14
full
12
24
absolutely
21
18
possibly
14
14
help, n.
10
24
care, v.
24
18
quotation
13
24
down
20
18
small
17
14
hotel
11
24
hard
22
14
open, a.
12
298
BELL SYSTEM TECHNICAL JOURNAL
TABLE Ill-a (Cont'd)
A
B
A
B
A
B
14
operator
10
12
real
11
10
sheet
8
14
particular
13
12
satisfactory'
11
10
street
8
14
perfectly
12
12
several
11
10
territory
5
14
profit
11
12
somewhere
12
10
together
8
14
read
11
12
steamer
10
14
report, v.
12
12
warehouse
8
10
transfer, n.
8
14
second, n.
12
10
warm
7
14
set, n.
8
11
afraid
11
10
whatever
10
11
almost
11
10
woman
5
14
sign, V.
12
11
arrive
10
10
yourself
10
14
stand, V.
14
11
both
10
9
build
7
14
surely
14
11
box, n.
7
9
care, n.
6
14
turn, V.
11
11
cold, n.
6
9
careful
9
13
across
13
11
complete, v.
8
9
certain
8
13
answer, v.
9
11
concern, n.
10
9
charge, v.
8
13
bond
8
11
confirm
7
13
building
11
11
definitely
10
9
color
8
13
charge, n.
8
9
complete, a.
8
13
condition
12
11
detail
10
9
conference
7
11
drawing
8
9
decide
9
13
connection
12
11
funny
11
9
end, n.
7
13
deal, n.
12
11
light, a.
7
9
express, n.
7
13
direct, a.
11
11
mile
8
9
game
8
13
drop, V.
12
11
motor
7
9
hospital
6
13
further
11
11
personally
8
9
immediately
7
13
general, a.
8
11
quality
10
9
large
8
13
himself
13
11
rather
11
13
insurance
11
11
use, n.
10
9
mention
7
13
interested
10
9
necessary
9
13
least
12
10
air
6
9
outside
9
10
awfully
10
9
personal
9
13
luck
12
10
bother, v.
9
9
remember
8
13
notify
6
10
carload
9
9
sit
8
13
offer, V.
12
10
cold, a.
7
9
sometime
9
13
party
12
10
crazy
8
9
statement
9
13
person
13
10
dinner
7
9
suggestion
8
13
quick
13
10
double
7
9
supply, v.
7
13
test, n.
8
10
easily
9
13
without
13
10
either
9
9
true
9
12
agree
11
9
up, a.
8
12
always
10
10
enough
10
9
weren't
7
10
everybody
10
9
willing
7
12
appreciate
11
10
explain
9
9
wise
7
12
bed
10
10
final
6
8
additional
8
12
brother
11
10
freight
8
8
advise
7
12
close, V.
11
10
having
10
8
agent
6
12
consider
9
10
head
9
8
agreeable
7
12
else
12
10
important
10
8
anxious
7
12
expense
10
10
kind, n.
9
12
fair
12
10
limit, n.
8
8
average, n.
7
12
great
11
8
beyond
8
12
loss
10
10
load, n.
8
8
carry
7
10
mark, n.
8
8
certificate
5
12
original
10
10
particularly
9
8
close, a.
8
12
per cent
8
10
positively
10
8
each
8
12
pick, V.
11
10
power
5
8
easy
7
12
policy
6
10
service
10
8
engineer
5
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 299
TABLE Ill-a (Cont'd)
A
B
A
B
.4
B
8
except
8
7
locate
7
6
offhand
6
8
fill
8
7
lovely
6
6
operate
6
7
mind, v.
7
6
opportunity
6
8
firm, a.
5
7
mother
7
6
package
6
8
girl
6
7
once
5
6
practically
6
8
guarantee, n.
7
7
ours
7
6
promise
6
8
heavy
6
6
realize
5
8
look, n.
8
7
phone, V.
6
6
represent
6
8
middle, a.
7
7
proper
7
6
shall
6
8
mistake, n.
7
7
sake
6
6
simple
6
8
news
7
7
satisfied
7
8
ordinary
6
7
side
7
6
straighten
6
8
owe
6
7
state, n.
6
6
such
6
7
store, n.
5
6
thanks
6
8
plan, V.
8
7
supply, n.
7
6
touch, V.
6
8
push, y.
6
7
throat
5
6
unload
5
8
quantity
6
7
wonderful
6
5
advisable
5
8
reasonable
7
5
allow
5
8
regular
8
7
yard
5
5
approval
5
8
reply, n.
7
6
advice
6
5
catch
5
8
sail, V.
7
6
afford
5
5
conversation
5
8
second, a.
7
6
appear
5
8
settle
7
6
argument
6
5
correct
5
8
shape
8
6
begin
6
5
crowd
5
6
broker
5
5
difficulty
5
8
simply
8
6
bunch
5
5
disa pointed
5
8
single
7
6
cancel
5
5
discuss
5
8
suggest
8
6
claim, V.
5
5
doctor
5
8
sweet
7
5
estimate, v.
5
8
weather
5
6
clear, v.
5
5
grade
5
8
weight
5
6
collect
6
5
holiday
5
8
whether
7
6
competition
5
5
increase, v.
5
8
world
8
6
cost, n.
5
7
actual
5
6
dandy, a.
6
5
inform
5
7
ago
5
6
dealer
5
5
insist
5
6
delay, v.
6
5
instead
5
7
apparently
6
6
depend
6
5
intend
5
7
available
5
6
fairly
6
5
interesting
5
7
buyer
5
5
form, n.
5
5
mix
5
7
clean, a.
7
5
operation
5
7
cover, n.
7
6
impossible
5
5
pardon, n.
5
7
desk
7
6
indeed
6
5
payment
5
7
evening
7
6
inquiry
6
5
reach
5
7
event
7
6
issue, n.
5
7
evidently
7
6
lay
6
5
reduction
5
7
exact
7
6
lose
5
5
return
5
6
mark, v.
6
5
show, n.
5
7
favor
7
6
memorandum
6
5
sort, n.
5
7
follow
7
6
notice, n.
6
5
specification
5
7
indicate
6
6
notice, v.
6
5
surprised
5
7
life
7
5
until
5
300
BELL SYSTEM TECHNICAL JOURNAL
Column A:
Column B:
TABLE Ul-b
Word List — ^Alphabetical Order *
Words Which Occurred in One Per Cent or More
of the 500 Telephone Conversations Analyzed
Total number of times the word (or some form of it) was used.
Number of conversations in which the word occurred.
A
B
A
B
A
B
A
18
anyway
16
10
bother, v.
9
7
apparently
6
11
box, n.
7
2,060
a
487
6
appear
5
36
boy
31
28
able
26
17
appointment
14
27
bring
24
480
about
266
12
appreciate
11
6
broker
5
17
above
13
5
approval
5
12
brother
11
24
absolutely
21
618
are
293
9
build
7
17
accept
14
6
argument
6
13
building
11
20
account, n.
16
54
around
54
6
bunch
5
13
across
13
27
arrange
23
121
business
83
7
actual
5
14
arrangement
11
23
busy
17
8
additional
8
11
arrive
10
63
but
54
19
address, n.
15
262
as
125
98
buy
60
6
advice
6
133
ask
101
7
buyer
5
5
advisable
5
476
at
238
88
by
75
8
advise
7
7
available
5
6
afford
5
8
average, n.
7
C
11
afraid
11
56
away
45
71
after
54
37
awful
31
17
cable
10
87
afternoon
62
10
awfully
10
386
call, v.
200
67
again
62
159
call, n.
111
17
against
13
B
396
can, aux.
221
8
agent
6
6
cancel
5
7
ago
5
113
back, a.
79
228
can't
132
12
agree
11
53
bad
46
234
car
88
8
agreeable
7
37
bag
24
22
card
9
20
agreement
8
21
balance, n.
15
24
care, v.
24
34
ahead
27
25
bank
12
9
care, n.
6
10
air
6
44
barrel
20
9
careful
9
139
all
100
30
basis
21
15
cargo
8
5
allow
5
296
be
175
10
carload
9
117
all right
74
65
because
47
8
carry
7
11
almost
11
12
bed
10
89
case
47
58
along, prep.
33
63
been
53
5
catch
5
15
along, a.
13
96
before
79
50
cent
31
14
already
14
6
begin
6
9
certain
8
15
also
11
31
believe
25
37
certainly
28
12
always
10
60
best
52
8
certificate
5
276
am
172
139
better
103
43
chance
37
17
amount
15
29
between
29
25
change, v.
22
326
an
178
8
beyond
8
21
change, n.
15
1,363
and
391
32
bid, n.
24
13
charge, n.
8
56
another
46
14
bid, V.
11
9
charge, v.
8
15
answer, n.
15
51
big
42
18
cheap
13
13
answer, v.
9
52
bill, n.
39
59
check, n.
35
8
anxious
7
15
board
8
31
check, V.
28
209
any
140
45
boat
22
25
city
18
43
anybody
40
13
bond
8
6
claim, V.
5
20
anyhow
17
26
book
17
15
clean, v.
14
170
anything
100
11
both
10
7
clean, a.
7
* In ambiguous cases the part of speech is denoted as follows: noun, n.; verb, v.;
adjective or adverb, a.; auxiliary verb, aux.; preposition, prep.
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 301
TABLE Ul-b (Cont'd)
.1
B
.4
B
A
B
15
clear, a.
14
195
didn't
142
7
favor
7
6
clear, v.
5
37
difference
31
41
feel
30
12
close, V.
11
40
different
30
23
foot
13
8
close, a.
8
5
difficulty
5
85
fellow
70
14
club
7
10
dinner
7
61
few, a.
50
68
coal
27
13
direct, a.
11
49
figure, n.
32
15
cocoa
Q
5
disappointed
5
35
figure, V.
24
43
coffee
8
5
discuss
5
8
fill
8
10
cold, a.
7
638
do
302
10
final
6
11
cold, n.
6
5
doctor
5
89
find, V.
72
6
collect
6
29
does
27
186
fine, a.
122
9
color
8
27
doesn't
23
14
finish, V.
10
274
come
168
38
doing
34
8
firm, a.
5
186
company
111
105
dollar
66
96
first
71
11
complete, v.
8
27
done
23
30
fix, V.
21
9
complete, a.
8
640
don't
301
7
follow
7
6
competition
5
29
dope, n.
24
823
for
330
11
concern, n.
10
10
double
7
22
forget
18
13
condition
12
24
down
20
6
form, n.
5
9
conference
7
11
drawing
8
10
freight
8
11
confirm
7
13
drop, V.
12
22
friend
14
13
connection
12
338
from
196
12
consider
9
E
14
full
12
63
contract
31
11
funny
11
5
con-\-ersation
5
8
each
8
13
further
11
54
copy, n.
36
35
early
26
5
correct
5
10
easily
9
G
15
cost, V.
15
8
easy
7
6
cost, n.
5
10
either
9
9
game
8
184
could
124
12
else
12
13
general, a.
8
53
couldn't
47
9
end, n.
7
1,360
get
393
45
couple
38
8
engineer
5
273
give
172
7
cover, n.
7
10
enough
10
8
girl
6
17
cover, V.
14
5
estimate, v.
5
49
glad
39
10
crazy
8
18
even
17
597
go
280
5
crowd
5
7
evening
7
243
good
149
28
customer
22
7
event
7
17
goods
13
20
cut, V.
18
23
ever
22
5
grade
5
43
e^-ery
36
12
great
11
D
10
e^'erybody
10
8
guarantee, n.
7
60
e^-eryth^ng
47
72
guess, V.
63
6
dandy, a.
6
7
evidently
7
15
date
14
7
exact
7
H
458
day
251
20
exactly
17
13
deal, n.
12
8
except
8
230
had
151
6
dealer
5
55
expect
49
36
hand, n.
31
9
decide
9
12
expense
10
16
handle, v.
14
17
definite
13
10
explain
9
20
happen
15
11
definitely
10
9
express, n.
7
24
hard
22
6
delay, v.
6
14
extra
11
176
has
114
32
deliver
24
25
hasn't
25
44
delivery
36
F
883
have
367
19
department
16
104
haven't
83
6
depend
6
14
fact
14
10
having
10
7
desk
7
12
fair
12
10
head
9
11
detail
10
6
fairly
6
109
hear
85
200
did
144
19
far
15
8
hea\-y
6
20
302
BELL SYSTEM TECHNICAL JOURNAL
TABLE Ul-b (Cont'd)
A
B
A
B
A
B
1,115
he
297
K
112
mean, v.
82
25
help, V.
19
40
meet
28
14
help, n.
10
38
keep
28
19
meeting
9
67
her
30
10
kind
9
6
memorandum
6
238
here
157
753
know
325
9
mention
7
45
high
34
[20
niessage
11
486
him
223
L
8
middle, a.
7
13
himself
13
68
nu'ght, aux.
59
106
his
70
9
large
8
11
mile
8
71
hold, \'.
60
140
last, a.
106
36
mind, n.
30
19
hold, n.
17
118
late
94
7
mind, v.
7
5
holiday
5
6
lay
6
241
minute
155
57
home
35
13
least
12
8
mistake, n.
7
24
hope, V.
19
69
leave
54
5
mix
5
9
hospital
6
32
less
23
29
moment
26
14
hotel
11
215
let
148
73
money
56
36
hour
30
214
letter
112
92
month
70
36
house
27
7
life
7
199
more
134
103
how
78
11
light, a.
7
330
morning
191
86
like, V.
71
20
most
15
/
16
like, a.
16
7
mother
7
10
limit, n.
8
11
motor
7
3,990
54
18
9
10
6
1,170
26
5
' '?
5
37
I
idea
467
38
17
7
10
5
408
7
5
6
6
5
29
87
20
line, n.
list, n.
60
13
30
226
move
much
23
160
64
listen
55
37
must
30
imagine
immediately
important
impossible
in
inch
156
10
19
21
7
66
little
load, n.
load, V.
loan
locate
long
117
8
16
5
7
55
170
87
24
my
N
name, n.
near
97
52
23
increase, v.
indeed
indicate
inform
information
112
look, V.
85
19
nearly
19
8
6
12
23
look, n.
lose
loss
lot
8
5
10
19
9
29
27
83
necessary
need, v.
never
new
9
22
23
62
6
5
5
inquiry
insist
instead
6
5
5
7
31
13
lovely
low
luck
6
22
12
8
83
52
news
next
nice
7
61
38
28
4 -y
instruction
20
18
lunch
18
168
night
107
13
5
insurance
intend
11
5
M
118
131
no, a.
not
77
96
15
13
interest, n.
interested, a.
9
10
26
machine
14
28
6
note, n.
notice, v.
24
6
5
46
1,506
69
interesting, a.
into
is
5
42
419
52
29
21
274
mail, n.
mail, V.
make, v.
22
16
169
6
13
339
notice, n.
notify
now
6
6
216
isn t
97
man
67
139
number, n.
80
6
1,605
15
67
issue, n.
it
item
its
5
417
10
67
29
10
6
96
many
mark, n.
mark, v.
market
25
8
6
51
1,190
of
396
25
material
14
112
off
67
J
37
matter, n.
31
ii
offer, n.
20
60
may
50
13
offer, V.
12
58
job
44
27
maybe
26
6
offhand
6
321
just
211
496
me
283
120
office
83
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 303
TABLE 1 1 1-6 (Cont'd)
A
B
.4
B
A
B
35
oil, n.
18
18
possibly
14
183
same
127
38
old
22
39
pound, n.
21
63
sample, n.
37
2,046
on
458
10
power
|5
12
satisfactory
11
7
once
5
6
practically
'6
7
satisfied
7
125
only
84
115
pretty
92
410
say
211
14
open, a.
12
241
price
123
14
second, n.
12
6
operate
6
88
probably
69
8
second, a.
7
5
operation
5
14
profit
11
887
see
328
14
operator
10
6
promise
6
37
seem
33
6
opportunity
6
7
proper
7
133
sell
81
295
or
178
26
proposition
21
264
send
172
247
order, n.
119
8
push, V.
6
10
service
10
27
order, v.
24
179
put
114
14
set, n.
8
8
ordinary
6
8
settle
7
12
original
10
Q
12
several
11
190
other
128
6
shall
6
45
ought
37
11
quality
10
8
shape
8
57
our
47
8
quantity
6
27
share, n.
10
7
ours
7
35
question, n.
31
140
she
50
138
out
90
13
quick
13
10
sheet
8
63
out of
25
35
quite
32
108
ship, V.
68
9
outside
9
18
quotation
13
49
ship, n.
28
317
over
208
16
quote
14
115
shipment
80
8
owe
P
6
R
26
50
40
short
should
show, V.
21
43
30
26
railroad
19
5
show, n.
5
6
package
6
19
rate
11
7
side
7
29
paper
17
11
rather
11
14
sign, V.
12
5
pardon, n.
5
5
reach
5
6
simple
6
16
part, n.
15
14
read
11
8
simply
8
14
particular
13
30
ready
25
8
single
7
10
particularly
9
12
real
11
9
sit
8
13
party
12
6
realize
5
31
situation
25
85
pay, V.
55
27
really
25
17
size
12
5
payment
5
40
reason, n.
32
18
small
17
79
people
59
8
reasonable
7
67
so
53
12
per cent
8
30
receive
22
64
some
43
14
perfectly
12
20
record, n.
18
17
somebody
17
13
person
13
5
reduction
5
93
something
67
9
personal
9
8
regular
8
9
sometime
9
11
personally
8
9
remember
8
12
somewhere
12
37
phone, n.
35
8
reply, n.
7
76
soon
63
7
phone, V.
6
48
report, n.
29
30
sorry
25
12
pick, V.
11
14
report, v.
12
5
sort, n.
5
24
piece
16
6
represent
6
33
speak
31
18
pier
10
5
return
5
22
special
15
75
place, n.
55
259
right, a.
173
5
specification
5
19
plan, n.
12
28
ring, n.
23
15
spend
11
8
plan, V.
8
28
room
19
14
stand, V.
14
17
plant, n.
9
26
run, v.
20
24
start, V.
22
34
point, n.
27
7
state, n.
6
12
policy
6
5
9
statement
9
19
position
15
15
station
8
10
positively
10
8
sail, v.
7
27
stay
23
17
possibility
15
7
sake
6
12
steamer
10
32
possible
24
28
sale
25
18
steel
10
304
BELL SYSTEM TECHNICAL JOURNAL
TABLE III-& (Cont'd)
A
B
.4
B
A
B
17
still, a.
16
71
through
46
222
week
120
42
stock, n.
M
17
ticket
9
8
weight
5
20
stop, V.
18
229
time
165
21
welcome, a.
21
7
store, n.
5
1,942
to
472
92
well, a.
71
17
story
12
193
today
124
83
were
66
19
straight
15
10
together
8
9
weren't
7
6
straighten
6
52
tomorrow
36
336
what
193
10
street
8
60
ton
27
10
whatever
10
146
stuff, n.
92
58
to-night
40
82
when
69
6
such
6
105
too
77
51
where
42
8
suggest _
8
31
touch, n.
29
9
whether
7
9
suggestion
8
6
touch, V.
6
40
which
33
9
supply, V.
7
30
town
22
103
who
74
7
supply, n.
7
61
train, n.
33
51
whole
40
47
suppose
41
10
transfer, n.
8
31
why
25
86
sure
72
18
trip, n.
17
1,305
will, aux.
402
14
surely
14
45
trouble
33
9
willing
7
5
surprised
5
16
truck
13
105
wire, n.
78
8
sweet
7
9
true
9
22
wire, V.
18
137
try
100
9
wise
7
T
14
turn, V.
11
27
518
wish, V.
with
22
263
295
take
207
U
17
within
17
193
talk, V.
131
13
without
13
84
talk, n.
67
82
understand
63
10
woman
5
16
tank
7
2,i
unless
a
34
wonder
28
29
telegram
19
6
unload
5
7
wonderful
6
29
telephone, n.
27
5
until
5
146
won't
115
553
tell
264
58
up, prep.
42
52
word
45
20
terrible
13
9
up, a.
8
68
work, V.
50
10
territory
15
77
us
63
65
work, n.
49
13
test, n.
8
11
use, n.
10
8
world
8
42
than
ii
65
use, V.
50
15
worry, v.
14
60
thank
56
379
would
207
6
thanks
6
V
97
wouldn't
79
1,792
that
397
84
write
61
3,110
the
496
268
very
165
24
wrong
20
67
their
63
370
them
170
W
Y
104
then
88
224
there
144
178
wait, V.
135
7
yard
5
63
these
57
599
want
297
78
year
53
913
they
253
12
warehouse
8
40
yesterday
35
418
thing
235
10
warm
7
86
yet
67
474
think
232
358
was
194
3,540
you
499
473
this
240
18
wasn't
18
157
your
100
130
those
100
108
way
81
10
yourself
10
29
though
29
1,100
we
294
7
throat
5
8
weather
5
speech sounds occur. They form 96 per cent of the total occurrences
of the words. It is to be noticed that no word was observed to occur
in all the conversations.
Of the 1,503 different words not shown on the list, 819 were ob-
served only once and 320 only twice. It is quite likely that if the
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 305
observations were repeated this part of the list would be dupUcated
very imperfectly, since these words, while in general well-known, tend
to be technical or specific, hence dependent on particular types of
subject matter. All except ten of the omitted words are nouns, verbs,
adjectives or adverbs.
100
g80
o
i^ 60
o
40
u
a.
uj 20
y
y
y'
■0'
/
y
4
A
y
^^
Fig. 2-
1 5 10 50 100 500 1000
NUMBER OF DIFFERENT WORDS IN ORDER OF OCCURENCE
-The cumulative curve obtained when the different words are arranged in order
of occurrence.
The importance of a relatively small number of different words
which are used very frequently is shown graphically in Fig. 2. The
curves shown are cumulative, giving the percentage of the total words
contributed by the different words when arranged in the order of their
occurrence. The curve labeled "Written" is based on the list given
in the study by Dewey, cited above. The economy exercised in con-
versation, or the poverty of conversational expression, according to
the point of view, contrasts sharply with written English. In conver-
sation 30 words account for half the total, in written English 69 words;
in conversation 155 words form 80 per cent of the total, in written
English 640.
The 50 most common words in telephone conversation and in written
English are shown in Table IV, arranged in their order of frequency of
occurrence. These words form 60 per cent of the total in conversation
and 46 per cent in written English. There are 29 words which are
common to the two lists. The personal nature of telephone con-
versation is shown in the two words which head the list. The most
striking difference between the two is the large number of active verbs
which occur in the list for conversation: "get," "see," "know, etc.,
306
BELL SYSTEM TECHNICAL JOURNAL
TABLE IV
Fifty Commonest Words in Telephone Conversation
Compared with Written English
Telephone
Written
Telephone
Written
Conversation
English
Conversation
English
1.
I
the
26.
GO
HIS
2.
you
of
27.
TELL
BUT
3.
the
and
28.
with
they
4.
a
to
29.
me
ALL
5.
on
a
30.
HIM
OR
6.
to
in
31.
ABOUT
WHICH
7.
that
that
32.
at
will
8.
it
it
2>i.
THINK
from
9.
is
is
34.
this
HAD
10.
and
I
35.
DAY
HAS
11.
GET
for
36.
THING
ONE
12.
will
be
37.
SAY
OUR
13.
of
was
38.
CAN
an
14.
in
AS
39.
CALL
BEEN
15.
he
you
40.
would
NO
16.
we
with
41.
THEM
THEIR
17.
they
he
42.
was
THERE
18.
SEE
on
43.
NOW
WERE
19.
have
have
44.
from
SO
20.
for
BY
45.
what
MY
21.
KNOW
NOT
46.
MORNING
IF
22.
DON'T
at
47.
an
me
23.
DO
this
48.
JUST
what
24.
are
are
49.
OVER
would
25.
WANT
we
50.
be
WHO
The 21 words not common to both lists appear in capital letters.
12 in all. None of these appears among the 50 commonest words of
written English. Three nouns, "day," "thing" and "morning,"
appear in the conversational list, none in the other. Only one con-
junction is found in the conversational list, while five appear in the
list for written English.
When the first 100 words in telephone conversation are compared
with the first 100 in written English two somewhat unexpected facts
emerge. In telephone conversation 14 out of the first 100 are words of
more than one syllable; in written English there are ten. Four two-
syllable words appear among the first 50 telephone words; the first 59
of written English are monosyllables. A more striking difference
concerns the origin of the words. Among the first 100 telephone words
there are 11 which are derived through old French from the Latin; in
written English there are only two from the Latin. Six of the 11 words
occur in the first 65 telephone words, while the first word of Latin
origin in written English is the 70th. The telephone words of Latin
origin are, in order of occurrence: "just," "very," "order," "minute,"
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 307
"price," "car," "letter," "fine," "company," "stuff," "number";
in written English these words are "people" and "very." The pre-
dominance of business words in this list for telephone conversation
suggests the influence of trade between England and France in the
Middle Ages.
More detailed comparisons may be drawn from Table V, which
lists the first 25 nouns, the first 25 verbs and the first 25 adjectives and
TABLE V
Twenty-five Commonest Words by Parts of Speech
Compared with Written English
Nouns
Verbs
Adjectives and Adverbs
Telephone
Written
Telephone
Written
Telephone
Written
Conversation
English
Conversation
English
Conversation
English
day-
man
get
say
now
not
thing
time
see
make
just
all
MORNING
WAR
know
come
very
no
ORDER
PEOPLE
want
take
RIGHT
SO
MINUTE
day
go
know
good
WHEN
PRICE
YEAR
tell
go
HERE
any
CAR
thing
think
see
MUCH
more
time
way
say
get
THERE
now
WEEK
WORLD
call
give
any
UP
LETTER
COUNTRY
take
think
more
out
COMPANY
PART
make
LIKE
TODAY
other
NIGHT
business
come
tell
other
only
CALL
LIFE
give
USE
FINE
GREAT
STUFF
FACT
SEND
call
SAME
SOME
NUMBER
LINE
LET
want
little
HOW
business
GUN
TALK
GOVERN
LAST
very
OFFICE
case
PUT
STAND
BETTER
SUCH
SHIPMENT
HOME
WAIT
ask
all
FIRST
way
CENT
TRY
SEEM
out
good
WIRE
POWER
ask
SHOW
not
EVERY
DOLLAR
PRESENT
SELL
look
onlv
THEN
man
HOUSE
look
NEED
LATE
little
MARKET
LOSS
MEAN
SAVE
no
here
month
month
HEAR
WORK
ALL RIGHT
just
case
PEACE
SHIP
BELIEVE
PRETTY
WELL
The words not common to both lists appear in capital letters.
adverbs, for both telephone conversation and written matter. Among
the nouns only eight are common to the two lists. The effects of
business are apparent in the telephone list. On the other hand, the
nouns of the written English list reflect the fact, pointed out by Dewey,
that the list was obtained from a study made soon after the war.
Among the verbs 15 words are common to the two lists and those which
differ are concentrated at the end. Approximately half the adjectives
and adverbs appear in both lists. The nouns from telephone conver-
308 BELL SYSTEM TECHNICAL JOURNAL
sation shown in this table form 2.4 per cent of the different nouns and
40 per cent of the total nouns; the verbs form 5.5 per cent of the
different verbs and 72 per cent of the total verbs; while the adjectives
and adverbs form 3.9 per cent of the different adjectives and adverbs,
but 48 per cent of the total.
An examination of the origin of the words in Table V shows that the
influence of Latin on the frequently used words is largely confined to
nouns. Eleven of the first 25 nouns of telephone conversation, and
eight of the first 25 nouns of written English come from the Latin.
Among the first 25 telephone nouns, aside from the eight nouns men-
tioned above among the first 100 words, there are: "office," "market"
and "case"; among the first 25 nouns of written English the following
are of Latin origin: "people," "country," "part," "fact," "cent,"
"power," "present" and "peace." Only one of the first 25 telephone
verbs comes from Latin: "try," and three of those in written English:
"use," "govern" and "save." Among the adjectives and adverbs
there are found in the telephone list: "just," "very" and "fine," as
above, and in the written English list the word "just" is added to
"very," which was in the first 100 words.
Referring once more to the small number of different words found
it may be pointed out that this shows how difficult it would be to
estimate the size of vocabularies by recording spoken words. The
80,000 words of this study are equivalent to a complete record of seven
hours' conversation, taking a rate of 200 words per minute. As noted
before, the number of different words was only 2,240, even though the
conversations covered a wide range of topics by many different
speakers. To increase this number notably, the curves of Figure 1
indicate that the observations would need to be very extended, since
the rate at which new words appear has already become very low.
For example, if the conversations were to go on continuously for a
week at the above rate a total of 2,000,000 words might be expected.
By extrapolating the curves of Fig. 1, and using a similar curve for
adjectives and adverbs, which lies between the curves shown, it may
be estimated that only about 5,000 of these words would be different
words. Extrapolation is a rough tool, but even with its inaccuracies in
mind, the conclusion seems safe that to measure a vocabulary by
recording spoken words involves the risk of gross underestimation
unless the observations are exceedingly prolonged.
It is suggested that teachers of languages may find the 737 words in
Tables lll-a and IH-^ to be of practical use in their profession. Pre-
sumably the progress of a student in speaking a foreign language would
be materially assisted by a thorough knowledge, early in his course,
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 309
of the words which are met with great frequency. The present
methods of teaching the spoken language no doubt approximate to
this, as a result of experience. It is suggested that the present word
list, which contains the words used so frequently as to form 96 per cent
of the total number observ^ed in this study, provides a guide for the
selection of important words to be taught. Additions are needed to
the list as it stands, in order to care for certain obvious situations not
encountered in telephone conversation concerning, for example, hotels,
restaurants and trains. With these points in mind, the list given has
the advantage of being founded on a study of actual conversation.
Syllables
As a preliminary to analysis of the words into their component sounds
the words were divided into syllables. With regard to the fact that
the study concerned conversation the division was made on phonetic
lines, which, as unabridged dictionaries show, differ from the ortho-
graphical divisions. Likewise a few words such as "every," "prefer-
ence," "average" and the like were divided into two syllables, accord-
ing to the usual colloquial pronunciation.
TABLE VI
The Syllabic Structure of Coxversatioxal Vocabulary
Parts of Speech
Per Cent of Words Having
Number of Syllables Sho\vn
Average
Number of
1
2
3
4
5
6
Syllables
Nouns
Verbs
53.3
81.9
57.8
94.8
82.0
33.8
15.0
30.7
4.7
13.8
9.7
2.8
8.0
0.6
3.2
2.7
0.3
2.8
0.1
0.86
0.47
0.66
0.15
0.03
0.02
0.01
1.63
1.21
Adjectives and Adverbs
Minor
All Words
1.58
1.06
1.23
In Table VI a summary is given of the syllabic structure of words,
based on the total occurrence of the words. It may be noticed that
words longer than two syllables make up only a trifle more than 4
per cent of the words observed. Nouns tend to be more polysyllabic
than other classes, but even so the nouns having more than two sylla-
bles occur so infrequently as to form only 13 per cent of all the nouns
observed.
The types of phonetic syllables which are found range in complexity
from a single vowel through various combinations of consonants with
a vowel. The relative number of the different types is shown in Table
\'II. The letters V and C represent "vowel" and "consonant,"
310 BELL SYSTEM TECHNICAL JOURNAL
TABLE VII
Types of Phonetic Syllables in telephone Conversation
Relative Occurrence per Hundred
Type Occurrence
V 9.7
VC 20.3
CV 21.8
CVC 33.5
VCC 2.8
CCV 0.8
CVCC 7.8
CCVC 2.8
CCVCC 00.5
100.0
respectively, and the letters CC are used to denote a compound con-
sonant form, that is, two or more consecutive consonants. It may be
seen that the typical syllable is of the CVC type, closely followed in
importance by the CV and VC types. The syllables having two or
more consecutive consonants form about one seventh of the total.
Speech Sounds
The analysis of the words into their constituent sounds was at-
tended by certain difficulties which should be borne in mind in consider-
ing the tables which follow. It was not feasible to record the original
words phonetically, just as they were pronounced by the telephone
subscriber. Instead the words were recorded and their phonetic
values assigned later. In so doing the dictionary was not adopted as
an authority for the pronunciation since in the informality of conversa-
tion, even among educated persons, there are elisions and changes of
stress which cause departures from the dictionary standard. Certain
very common words, for example, receive various treatments in con-
versation, depending on their situation In the sentence, the emphasis
desired and the speed of talking. The word "and" may be pronounced
as spelled, but quite often it is reduced to " 'nd" or even " 'n'. "
The prepositions "to" and "of" are similarly varied. Altogether
about 40 common words were found, of this type, each of which seemed
subject to several different pronunciations, even in speech which would
not be regarded as unduly careless. These were all from the minor
classes: auxiliary verbs, pronouns, prepositions and conjunctions.
The modification, in general, is such as to give the vowel its unstressed
value. In the analysis these different forms are included, the weight-
ing for each modification necessarily being a matter of judgment.
The remaining words were each assigned a single pronunciation,
selecting that which we regarded as being the typical pronunciation
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 311
heard in reasonably enunciated conversation among educated persons
in New York. The departures from dictionary standards are largely
confined to the vowels. As a result the analysis is affected to some
degree by the speech habits of the writers.^ It is regrettable that some
arbitrariness should be introduced, but this seems to be a difficulty
common to discussions of vowel sounds. Some of the difficulty is
avoided by making separate classifications for vowels for which the
pronunciation is indefinite, such as the vowels in unstressed positions.
The articles "the," "a" and "an" were omitted entirely from the
analysis on account of the large number of variant pronunciations to
which they are subject.
The results of the analysis into speech sounds are shown in Table
VIII. Three divisions are given: vowels, initial consonants and final
consonants, based on the division into phonetic syllables. The method
followed was: first, to divide the words into phonetic syllables, second,
to assign phonetic symbols to the sounds and third, to weight each
sound by the total number of times the word was recorded. The
sounds are identified in the table, where necessary, by key words.
No difficulties were encountered in analysis of the consonants, but a
few special points which arose in assigning the vowel qualities may be
noted. The key word "pot" is used to denote a vowel sound which is
pronounced differently by many natives of New England and those
whose habits of speech were formed elsewhere.* With these New
Englanders the sound tends toward the quality of the vowel in "pawn,"
although shorter in duration. The same New Englanders make a real
distinction between the vowel of "pot" and the vowel of "palm." By
many speakers elsewhere no such distinction is made and the two are
lumped into a single intermediate sound which is neither the New
Englander's "pot" nor "palm." To avoid confusion the class denoted
by "pot" has been made to include "not" and many other monosylla-
bles of the same ending, as well as "on," "job," "stock," etc., which
grouping is believed to be homogeneous on either basis. The few
words of the class of "palm" which were encountered have been in-
cluded under "par." The class denoted by "par" may be subdivided
into: "par," 1.24; "palm," 0.07. The class denoted by "palm"
would be somewhat larger if the class which we may denote by " path,"
such as "can't," "last," "ask," etc., had not been classified under
^ For the benefit of phoneticians who may be interested it may be stated that the
writers are residents of Greater New York of more than six years' standing, that their
boyhoods were spent in Maine, Illinois and New Jersey, respectively, and their college
years at Maine and Princeton, Harvard and Oxford, New York University and
Harvard, respectively; this seems a background sufficiently varied to bring to light
many of the principal variants of American speech.
* Just what the geographical lines may be, the writers do not pretend to know. A
phonetic map would be of interest.
312
BELL SYSTEM TECHNICAL JOURNAL
TABLE VIII
Relative Occurrence of Speech Sounds in Telephone Conversation
All Words {Except Articles)
Vowels
pin . .
pine . .
pan. .
pen. .
peel . .
pool. .
pot. .
pane,
pole. .
pawn .
pun. .
pull.,
pout .
par. .
pair . .
purr,
pew. .
poise .
75.36
Unaccented Vowels
possible 5.5^
about,
differ. . .
receive .
notion .
wanted .
peop/e . .
5.33
4.56
3.78
2.65
1.83
.97
24.64
100.00
Total Number of
Sounds 92,522
Initial Consofiants
10.27
W
7.58
T
6.89
TH" (then
6.60
Y
6.44
D
6.26
AI
5.21
H
4.78
K
4.74
S
4.15
N
4.14
B
2.96
G(gun;
1.69
L
1.31
F
1.09
R
.80
P
.26
TH' (thin
.19
SH
J
CH
Z
ZH
NG
9.
7.
6.
6.
6.
5.
5.
5.
PR
HW
ST
TR
FR
PL
KW
BL
SP
KL
Others
.38
.86
.72
.48
.21
.89
.75
.55
5.46
4.99
4.64
4.33
4.31
3.96
2.78
2.54
2.02
1.74
1.25
.83
.55
.34
.02
93.60
Compounds
1.06
.91
.87
.69
.62
.36
.28
.23
.19
.18
Final Consonants
t
. . . 14.30
r
. . . 13.05
n
. . . 12.52
1
. . . 8.40
z
. . . 6.01
m
. . . 5.48
d
. . . 4.44
V
. . . 4.23
ng
. . . 3.57
s
... 3.13
k
. . . 2.85
f
... 1.37
th" (with)
1.25
P
1.24
ch
.53
b
.42
g
.38
sh
.32
]
.14
th'(myth)
.04
zh (azure)
.01
h
— ■
w
— ■
y
. . . —
Compounds
nt
nd
St
ts
nk
Id
rz
ks
kt
rd
1.01 Others
6.40
83.68
4.40
2.56
1.18
1.11
.76
.75
.57
.47
.42
.37
3.73
16.32
100.00
64,043
100.00
65,544
"pan," such being the more common American pronunciation.
Actually the occurrence of words in the class of "path" is not high; if
they had been given a special class in Table VIII their relative occur-
rence figure would have been 0.78, reducing the figure for "pan" to
6.11. Special categories are given to the vowel sounds in the classes
denoted by "pair" and "purr" since there is often disagreement con-
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 313
cerning the quality of a vowel which precedes "r." Likewise it was
found expedient to make a number of classifications of vowels in
unaccented positions.
Since the figures of Table VIII are likely to find application as
weighting factors it is convenient to have them add exactly to 100 per
cent, consequently they are given to two places of decimals. An
estimate of the representativeness of these figures may be obtained
from the data presented in Table IX, which were worked out from
TABLE IX
Comparison with Check Test
Relative Occurrence <
of Consonants in Verbs
Sound
First
Observations
Check
Test
Difference
Sound
First
Observations
Check
Test
Difference
B
1.02
1.02
.00
S
10.31
9.67
-.64
D
4.46
4.83
+ .37
T
16.97
17.39
+ .42
F
1.73
2.18
+ .45
V
2.36
2.20
-.16
G
11.15
9.40
-1.75
W
4.87
4.54
-.33
H
1.40
1.66
+ .26
Y
.55
.53
-.02
T
.22
.23
+ .01
Z
1.09
1.44
+ .35
K
8.90
8.74
-.16
CH
.32
.75
+ .43
L
7.70
7.94
+ .24
SH
1.35
1.17
-.18
M
4.45
3.96
-.49
TH'
2.53
2.85
+ .32
N
6.87
7.10
+ .23
TH"
.05
.06
+ .01
P
3.34
3.47
+ .13
ZH
.00
.00
.00
R
3.97
3.88
-.09
NG
4.39
5.00
+ .61
100.00
100.00
observations mentioned before, conducted by a different observer at a
different time, but on the same set of toll circuits. Records were made
only of verbs, and for 250 instead of 500 conversations. The vocabu-
lary collected in the check test resembled that of the first observations
closely. Arranging the words in the order of occurrence, the first 17
words of the first observations are also the first 17 of the check test,
although the order is not repeated exactly. In the first observations
the first few words run: "get," "know," "see," "want," "go," "tell,"
"think" and "say"; in the check test the order is: "get," "see,"
"know," "want," "tell," "think," "go" and "say." Table IX shows
the analysis of the words as to the simple consonants, lumping initial
and final consonants together. Only one of the differences is greater
than 1 per cent and all but three are less than 0.5 per cent. One check
test is not sufficient for a final statement, but judging by these results
the observing method and the samples taken seem to justify considering
the figures of Table VIII as representative as far as the figures in the
314
BELL SYSTEM TECHNICAL JOURNAL
digits position for most of the sounds and as to order of magnitude for
the infrequent sounds.
The effects of restricting the word Ust in various ways are shown in
Figure 3. The first hne shows graphically the relative occurrence per
^
RELATIVE OCCURENCE OF THE INITIAL CONSONANTS
IN TELEPHONE CONVERSATION
P T K F TH' S SH CH M N NG L R W Y H Compound
B D G V TH" Z ZH J
n t-in ,-. _
„nn V^r^ n
■— -L-l
1—
5
n n .— . n rn^n i— i r~l
— |r— 1
0_i__u..,^l^m 'i— i|_j
5 _
LJ
Fig. 3 — The relative occurrence of initial consonants — effects of restricting the
word list.
Line I — Relative occurrence of initial consonants for all words.
Line II — Differences resulting from omission of minor parts of speech (118 words).
Line III — Differences resulting from omission of the 100 commonest words.
Line IV- — Differences resulting from omission of the 1,500 least common words.
hundred for initial consonants as in Table VIII. If the minor parts of
speech are excluded before the analysis, which eliminates only 118
different words, but nearly half the total words, the resulting changes
are shown on the second line. Notable decreases occur for "th," "w"
and "y," which may largely be traced to the omission of "that,"
"they," "this," etc.; "will," "with," "would," etc.; and "you,"
respectively. These elisions enhance the relative contributions from
"get," "see" and "know." When the 100 most common words are
omitted there are also large changes, as shown by the third line. Since
50 of the 100 most common words are of the minor parts of speech the
similarity of this line to the second is not surprising. The omission,
on the other hand, of the 1,500 least common words, namely, those
omitted from the vocabulary of Table III, changes the distribution by
negligible amounts as shown in the fourth line. Since, then, the 737
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 315
commonest words seem to determine the relative frequency of the
sounds of conversation, the writers are encouraged to believe that if
this study were repeated on telephone calls of which a greater propor-
tion were social rather than business in nature the analysis into sounds
would be changed very little. Likewise the conclusion is drawn that
if the study were prolonged tenfold so as to double the number of dif-
ferent words no material change in the relative frequency of the sounds
would be found.
TABLE X
Relative Occurrence of Speech Sounds in Telephone Conversations
Nouns, Verbs, Adjectives and Adverbs
VoiL'els
Initial Consonants
Final Consonants
10.63
7.56
7.19
6.38
5.96
5.78
5.40
5.25
4.59
4.01
2.49
1.83
1.58
1.53
1.39
1.33
.49
.35
73.74
Unaccented Vowels
differ 5.79
receive 5.73
possible 4.82
obout 3.96
wanted 2.54
peop/f- 1.76
notion 1.66
pen. .
pin. . .
pane,
pole . .
pawn .
peel . .
pine,
pun. .
pot . .
pan. .
pull.,
pout .
pair. .
par. .
pool . .
purr. .
pew . .
poise .
Total number.
26.26
100.00
50,161
S
N
T
M
G
K
L
D
W
B
R
P
F
SH
TH'
H
J
Y
V
TH"
CH
Z
ZH
NG
Compounds
PR .
ST .
TR .
PL .
HVV .
KW .
BL .
SP .
KL .
GR .
Others.
8.34
7.94
7.55
7.40
6.87
6.70
6.65
5.25
4.86
4.38
4.11
4.06
3.88
2.42
2.41
2.38
1.33
1.25
1.21
.97
.87
.55
.03
91.86
1.69
1.39
1.11
.58
.49
.44
.37
.30
.29
.27
1.21
8.14
k
m
s
d
z
P
V
f
th"
ch
b
g
sh
J
th'
zh
h
w
Compounds
nt
St
nd .
nk .
Id
rz
ks .
kt .
rd .
ns
Others.
14.64
13.53
9.99
8.62
5.10
4.96
4.50
3.89
3.66
2.42
2.00
1.94
1.28
.82
.81
.73
.66
.57
.24
.07
.02
80.45
3.37
2.07
1.66
1.32
1.31
.98
.82
.73
.64
.54
6.11
19.55
100.00
40.107
100.00
37,493
316 BELL SYSTEM TECHNICAL JOURNAL
For some purposes weighting lists based on the words of speech which
carry the meaning are appropriate. This is approximated to by the
figures given in Table X, in which the sounds are analyzed for nouns,
verbs, adjectives and adverbs only. The outstanding changes in the
weighting of initial consonants have just been commented on in con-
TABLE XI
Relative Occurrenxe of Speech Sounds in Telephone Conversations
Conversational Weighting
Note: The sounds of each word are weighted b}' the number of conversations in
which the word is used, instead of by the total occurrences of the word.
Vowels Initial Consonants Final Consonants
pin 11.22 W 8.26 r 13.87
pen 7.90 T 7.09 t 11.98
pan 6.40 M 6.69 n 10.92
peel 6.21 D 6.52 1 8.13
pine 5.95 K 5.90 m 5.43
pane 5.60 S 5.90 d 5.20
pole 5.18 L 5.44 z 5.13
pun 4.66 B 5.32 ng 4.05
pawn 4.64 H 5.31 v 3.72
pot 4.08 N 5.09 s 3.64
pool 3.40 TH" 5.01 k 3.41
pull 3.24 F 4.10 p 1.55
pout 1.89 G 4.00 f 1.41
par 1.33 R 3.53 th" 1.18
pair 1.31 Y 3.17 ch 65
purr 1.11 P 3.09 b 52
pew 38 SH 2.09 g 49
poise 24 TH' 2.06 sh 45
V 1.43 j 17
74.74 J 94 th' 06
CH 74 zh 02
Unaccented Vowels Z .47 h • —
ZH 03 w —
about 5.39 XG — y —
differ 5.35
receive 4.85 92.18 81.98
possible 3.57
notion 2.69 Compounds Compounds
wanted 2.16
people 1.25 PR 1.27 nt 4.68
ST 1.06 nd 2.09
25.26 HW 1.03 st 1.43
TR 82 ts 1.14
FR 62 Id 89
100.00 PL 49 nk 71
Total Xumber. . . 54,656 KW 40 rz 63
BL 30 ks 61
SP 26 k-t 56
KL 26 rd 51
Others 131 Others 4.77
7.82 18.02
100.00 100.00
39,924 40,993
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 3l7
nection with Fig. 3, line 2. The vowel weighting reflects the enhanced
importance of the "e" in "pen" from the verbs "get," "tell," "send"
and shows a considerable reduction in the vowel in "pool," largely
from the loss of "you" and "to." The unstressed vowels, especially
as in "about" are also diminished. Among the final consonants the
largest change is a reduction in "z," which results from the elimination
of "is," "was," "as," etc. Comparing the distributions of Table VIII
and Table X as a whole, however, both show about the same degree of
non-uniformity; the maximum and minimum weightings do not differ
greatly.
One more type of analysis is given in Table XI. In this case the
sounds of each word are weighted by the number of conversations in
which the word occurred, instead of by the total number of times the
word was used. This seems to be a somewhat radical change in
method, involving, as it does, a considerable reduction in the weighting
of the words at the head of the list. When the effect of eliminating
the first 100 words entirely, shown in Fig. 3, line 3, is recalled, large
changes might be expected. Actually the result is remarkably similar
to the figures of Table VIII. The relatively diminished importance of
"you" and "to" is seen in the vowel list, of "you" again among the
initial consonants, and of "it," "that" and "get" in the list of final
consonants. The range covered by the relative weightings is still
much the same as in Tables VIII and X,
Comparisons with Written English
Some of the differences between the vocabularies of telephone
conversation and written English have been pointed out. The effects
of these differences may be seen in the relative occurrence of the sounds
as shown by Figures 4, 5 and 6 for vowels, initial consonants and final
consonants, respectively, using the analysis based on all the words
(except articles). The upper line in each case is a graphical representa-
tion of the corresponding data of Table VIII, after certain changes have
been made to put them on the same basis as the tables given by Dewey
for written English. In the case of the consonants the only change
needed was omission of the compound consonants. In the case of the
vowels it was necessary to combine some of our classifications, since
Dewey made but 17 distinctions among the vowels. We believe the
combinations made are those followed by Dewey himself, as ascer-
tained from examples given by him in his text. The phonetic symbols
given in Figure 4 are those used by him. The combinations made were
as follows: "pen" and "wanted"; "pane" and "pair"; "pin," "pos-
sible" and "receive"; "pun" and "purr"; "about," "differ," "peop/g"
21
318
BELL SYSTEM TECHNICAL JOURNAL
and "notion." The comparisons are made with Table XVI of Dewey's
book, which does not include "the," and from which we have sub-
tracted the article "a."
The outstanding differences between the vowel frequencies in
telephone conversation and written matter are the excess in conversa-
tion of "about," "pine," "pool" and the deficiencies in "pan," "pin"
%
20
15
10
5
RELATIVE OCCURENCE OF THE VOWEL SOUNDS
IN TELEPHONE CONVERSATION
i
i
M
m
^ ==
^
n
i
I
<^y/A Y/A
I
I
i
5
-5
Fig.
A IX ty ©■ AT
vr ju, JUL <x (j^^oju Ju.
2 z o, ya z uui _J a H z -za. i- cc, z u. ui _i _i
< u y z^ ?r s^ ^ u < o > 3q: 3 u oi -j -j -i o
0. Q.I- <^ °-5|UJ U OL Q. 5 0.3 OU- ;:Idl' o d o
8S'
a Q. Q.
U U I- §
Z (O 3 u
D- O ? O-
TT
n
-rnn n
TJ
PERCENTAGES IN EXCESS OF THOSE FOR WRITTEN MATTER
4 — Comparison with written English — relative occurrence of the vowels.
and "pot." The greater occurrences of "pine" and "pool" are almost
entirely accounted for by the greater use of the words " I " and "you."
The deficiencies mentioned do not, on analysis, seem to depend on one
or two words, but rather on the whole vocabulary, except that of the
increase in the unstressed vowel denoted by "about" nearly 1.7 per
cent comes from the vowels of words which in the study of written
English were classified under "pan."
Among the initial consonants (Fig. 5) the greatest change is in the
occurrence of "y," which is much more frequent in conversation.
This again is largely caused by the pronoun "you." Much of the
increase in "g" may be traced to the greater use of "get" and "go."
The sounds "w" and "t" are the most frequent sounds in written
English, as well as in conversation.
Figure 6 shows that in the case of the final consonants the sounds
"t" and "1" are notably more frequent in conversation than in written
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 319
matter. The increase in "t" arises almost entirely from "that,"
"it" and "get" which combined have a contribution about 4.9 per cent
larger in conversation than in written matter. About half the increase
RELATIVE OCCURRENCE OF THE INITIAL CONSONANTS
IN TELEPHONE CONVERSATION
(compound CONSONANTS NOT INCLUDEd)
fS
MNNGLRWYH
nr-.n. , ^nn , ,^^
TIT
"ODD
izr
PERCENTAGES IN EXCESS OF THOSE FOR WRITTEN MATTER
Fig. 5 — Comparison with written English — relative occurrence of the initial
consonants.
RELATIVE OCCURRENCE OF THE FINAL CONSONANTS
IN TELEPHONE CONVERSATION
(compound CONSONANTS NOT INCLUDED)
''b "^ D *" G ^ v"'''tH^ Z^^Zh'^'^jMNNGL RWHY
IT
'UU^
n _
U
PERCENTAGES IN EXCESS OF THOSE FOR WRITTEN MATTER
Fig. 6 — Comparison with written English — relative occurrence of the final consonants.
in "1" is attributable to the words "will" and "tell." Some of the
deficiency in " v" may be traced to the word "of" which has a contribu-
tion 1.8 per cent greater in written matter. On the other hand the
320 BELL SYSTEM TECHNICAL JOURNAL
words "have" and "give" together contribute 1.1 per cent more to
conversation, so that the net difference in "v" is to be traced to small
accretions from the whole vocabulary rather than a few specific words.
Relative Occurrence of Combinations of Sounds
A more elaborate analysis of the phonetic syllable is given in Table
XII, which shows, for each vowel, the frequency of occurrence of the
consonants preceding the vowel and also of the consonants which follow
the vowel. The complete word list (except articles) was used as a
basis. The cases in which no consonant occurs in front of the vowel
are included, as well as the cases in which there is no following con-
sonant. These figures are shown as a double column under the key
word denoting the vowel sound. In each double column the figures
on the left apply to initial consonants and on the right to final con-
sonants. The figures are given in per cent, so that each column adds
to 100. The consonants are grouped by phonetic classes. The table
is to be read as follows: of the syllables in which the vowel sound is that
in "pan," 28 per cent begin with "th" (as in "that"), 26 per cent have
no initial consonant, 16 per cent begin with "h," 7 per cent with "k,"
6 per cent with compound consonants, 5 per cent with "b," etc.; while
29 per cent end with compound consonants, 27 per cent with "t,"
etc. Where no figure is entered the occurrences were less than 0.5
per cent ; where a dash is shown no combinations of the kind indicated
were observed. If the figures are taken by rows instead of columns
no meaning can be attached to them before they are multiplied by the
relative occurrence of the different vowels.
In studying this table it is to be remembered that because the dif-
ferent vowels have very different frequencies of occurrence the sub-
divided data shown in different columns cannot be considered as equally
representative. Syllables having the vowel as in "pin," for example,
were present, as shown in Table VIII, to the number of 0.1027 X
92,522, or 9,500. The syllables in this class which begin with "t" are
shown in Table XII to be 1 per cent, representing 95 occurrences. In
the class having the vowel of "poise," however, there were only 176
examples, so that the 37 per cent of these syllables beginning with "p"
result from only 65 occurrences.
It is to be seen that only one vowel, "pool," is preceded by a par-
ticular sound more than 50 per cent of the time, this sound naturally
being "y." Six vowels are preceded by particular sounds more than
25 per cent of the time. The sounds of "pair," "purr," "par" and
"differ" must be followed by "r," a blank, or a compound consonant
beginning with "r," as a result of the way in which the analysis was
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 321
Q. -P -i£
Q. 1- X
-Q "D CP
to Q U
<;- r in X r
V <0
I II
U. 1- l/l w> (J
> X N X -1
+) N
I I
> 1- N N -)
£ C U) ^ L 5 J
2 z z _) tr s >-
H
BLANK
BLANK
COMPOUND
COMPOUND
oil
1 1 1
.n
1 1 1
1 — I
^ C: >n
1 1 1
-co -
1 1 1
1
1 1 1
rvi - 1
OJ -
r 1 1
't 2
1 1 1 1 1
1 - ?^ ru
1 1 f 1 1
- 1 (M 1 -
1 1 c\j 1 ■<r
1 o 1
1 1 1 1 1
1 1 1 1 1
- 1 CM 1
1 1 1 1 1
1 1 1 1 (0
"£31111
1 1 1
1 1 1 1 1
<0 - 1
1 § 1 1 I 1 1
1 1 1 1 1 1 1
1 1 1 gi 1 1 '
CM t 1 1 (M 1
in ^ 1 - 1 1 1
- - 1 - -J
1 1 1 1 Oi 1 1
- - 1 - - It
1 1 ^
' ' in
1 "1
1 1 a>
'J -
1 01
gai
1—
S
<:i
1 ^ '
1 lO
(V -
1 1 1
r). p
1 ^ 1
r
I 1
11-11
- 1 t») 1 -
1 1 r~ 1
m CM - 1
1 1 1 1 1
- 1 CM
in 5 1
- ^ 1 -
1 — 1 1
11-11
- CM 1 1 1 II
5 CO 1 .n 1
^ (3> « en , , 1
1^ 1 oj 5 1
1 1 1 1 1 1 1
CO 1 ^n CM 1 1
1 * -
1 ^ -
, <*>
1 in
COUJ
OCD.
cx:lu
o
UJ
z
Q.
(—
2
= C\J
r) oj OJ
1 1 1
O 1
C\J
1 - 1
fl) fM 1
I - r
Si 1 .
1 ^ 1 1
in ( —
1 1 ro 1 1
— 1 II
1 1 >n 1 1
1 1 1 1 1
- 1 — 1 1
1 1 1
1 1 1 1 1
1 - 1 1 1
1 1 i~- 1 1
(0 1 1 1 (0
<0 \f 1 Oi 1 1 1
^ ^ 1 (o r^ CM I
1 -^ 1 1 1 1 1
1 1 cy 1 1 1
1 CO 1 — 1 1 1
'J "
- S <°
i ^ 2
- CO _
, rvj ro
„CM ^
1 o
' CM -
2
LLI
Q.
1 - 1
- ro (0
r- Kl --
1 t 1
1 — 1
(\l - lO
- r- 01
1 — 1
OJ 1
11-11
III 1
^ 1 <o
1 1
1 CM
1 1 in 1 1
— 1 1 1 —
1 1 eo 1 1
1 1 1 1
- 1 1
1 1 1
- 1 1 - CM 1 1
CM ^ 1 1 I 1 1
1 P ^ fM 1 1
1 t 1 ^ 00
— ro 1 c^ 1 1 1
- CM 1 ^
1 *
' CO
in 1
' ?! ^
CM ro
_J
_l
_J
UJ
-J
- CM
- ^ rl
1 -* r-
OJ (D 5
^ c5 c\J
(M "J
1 (M 1
_ in rvj
1 1
— 1
CNj c\j C^
1 1 1
CM 1 ^ — 1
rO t — 1 1
^ CM - - 1
PI 1 CM
1 1
<0 1 1
1 ^} > 1
1 1 CM 1 1
1 1 1
1 1 1 1 -
— CO 1 ro 1 1 1
CM CM 1 <n ro — 1
1 ^^ 5 ^ 1 1
S 1 in - - 1
- § 1 o 1 1 1
m \ - Q
, CM lO
' in CM
10 in (0
1 O 01
z <^
1 CM '£
i
1
2
1 1 1
fO — OJ
1 1 1
CO fo -
1 1 1
fy -
OJ o
1 1 1
1 1 1 1 1
CM 1 1 CM
T} CM r-
- 1 (0 1
1 1 1 1 1
CM = 1
till
1 1 1
t t-- t 1 1
II 1 c;<
1 1 1 1 1
- 1
1 1 1 1 ?! 1 1
<£> \ \ — 1 —
^ O -III
= It 1 - - 'J
. 1 1 1 f;^ - 1
1 > ^ >
^ S -^
1 <o
- s ?^
1 1 en
(0 in
2
CC
- 55 ^
- t^
- s <0
- <J>
1 1 1
- ■<J ^
.n -
>r> -
1 in -
^ _ ffl
1 1 1
t
in —
- - CM -
<o -
CM 1 <5 CM
1 1 1 1 1
CM 1 - fO 1
<0 'J 1 1
CO _ ,
r) - -
(D 1 -
1 1 1 1 1
(D 10 - - - 1 1
CM 1 CM II
(^ in 1 CM 1 1 1
CM CM 1 01 ro ro CM
i . . 1 S; • '
--II CM 1
1 - <"
' OJ
« S CO
o
' CM
CM in in
1 1 n
— CM 1^
UJ
Q.
2
= 2
O 1^ ^
n 2! -
— ro \t
(\j - -
1 n 1
_ « _
>n -
■^ <o n
1 (M 1
m - [
1 Tj 1 CM
- 1 2
- 1 .n CM -
CM r CM ro
t 1
— iQ ro
CM 1 1 1
till
^ m =
ro - 1
(0 1 in 1 1
— 1 1
r- CM 1 ro 1 1 1
2 (M 1 Tj ro CJ> 1
If) (g r~ CO II
(0 1 ij - 2 1
- CO 1 ^ 2. 1 1
<n - 1 CM CM Kj -
CO 00
1 1 01
^ ?) <^
1 © M
^ - "^
_l
UJ
Q. V -n
Q. 1- x:
-Q T) 01
(0 Q o
c^ r in X j:
f in
I II
LL H c/) cn o
> £ N ^ -
I I
> 1- N N -)
E C CP -- L S 31
5 z z -I cr § >•
•Ci:x:Qc3
<<5d
-l_lOO
I (CcDaCL
Q
o
►J
w
>
a
w
o
?
o
J
J
o
l-H Q
1=1 Z
Xj <
«< u
f-H W
c/)
o
z
o
o
b
O
o
H
H
CT!
322 BELL SYSTEM TECHNICAL JOURNAL
made, similarly for the "1" of "peop/e" and the "n" of "notion."
Aside from these no single consonant occurs as often as 50 per cent of
the time after a particular vowel. With five vowels a particular
consonant ends the syllable more than 25 per cent of the time. In
nearly every case the most frequent combinations can be traced directly
to the first 50 words of the vocabulary. Five vowels are preceded by
blanks more than 50 per cent of the time and eight are followed by
blanks in more than 50 per cent of the cases. The combinations of
different vowels with compound consonants vary considerably in
importance, ranging in the final position from practically none with the
vowel of "pew" up to the vowel of "pun," which is terminated by a
compound consonant 63 per cent of the time.
Conclusion
In concluding, a brief review is presented of the main points of
interest. The paper has for its basis a Hst of 80,000 words obtained
from telephone conversations. This list has been studied with respect
to the number of different words contained in it, the relative occurrence
of the different speech sounds and the combinations of sounds which
form syllables. In so far as the authors know this is the first study
of this type based on conversations as contrasted with written matter.
Perhaps the most striking aspect of the word list is the small number
of different words contained in it, only 2,240 out of the total of 80,000.
Of these 2,240 words 819 occur only once. The balance, or 1,421 words,
constitute practically 99 per cent of the total words recorded ; of these
the 121 different words which constitute the minor parts of speech
form 45,000 of the total occurrences. The pronouns "I" and "you"
together occur over 7,500 times.
This intensiveness with which a small number of words is used in
conversation is considerably greater than in the written English ana-
lyzed by Dewey. In conversation the 155 most frequently used words
make up 80 per cent of the total occurrences; to reach the same per-
centage in the written English analyzed by Dewey 640 words must be
included. The frequently used words of conversation are character-
ized, as compared with written English, by the greater prominence of
certain active verbs, such as "get," "see," "know," etc., 12 of which
occur in the first 50 words of conversation, while there are none in the
first 50 words of written English. The most frequent words of con-
versation differ from written English also in the greater number of
words of Latin origin which appear frequently in conversation: 11 from
the first 100 of the list for conversation, as compared with two from
the first 100 of written English.
THE WORDS AND SOUNDS OF TELEPHONE CONVERSATION 323
The word list is characterized by a large percentage of monosyllables.
Over four fifths of the 80,000 occurrences are of this type, a result
largely brought about by the frequent repetition of the minor parts of
speech, among which 95 per cent are monosyllables. When the words
are analyzed into phonetic syllables about one fifth are found to be of
the type vowel-consonant, about one fifth consonant- vowel, and a
third of the type consonant-vowel-consonant.
The relative occurrences of the different speech sounds were obtained
by assigning phonetic values to the sounds of the phonetic syllables
and weighting each by the total number of times it was used. Twenty-
five categories are used for the vowels. Seven of these are for vowels
in unaccented positions, which make up, altogether, about 25 per cent
of the vowels. The relative occurrences of the individual sounds differ
greatly for different vowels. The range extends from about 10 per
cent for the vowel of "pin," and about 8 per cent for the vowel of
"pine," down to 0.3 per cent for "pew" and 0.2 per cent for "poise."
Among the initial consonants 94 per cent are single sounds, and the
remaining are compounds of two or more successive consonants. The
range extends from about 9 per cent for "w," and about 8 per cent for
"t" down to about 0.3 per cent for "z" and the slightest trace, .02
per cent for "zh." The most frequent compound initial consonant is
"pr," with an occurrence of 1 per cent. Among the final consonants
the compounds are somewhat more prominent, forming 16 per cent.
The most frequent final consonant is "t," 14 per cent, the next is "r,"
13 per cent, the range extending down to 0.1 per cent for "zh." The
most frequent compound final consonant is "nt," 4.4 per cent, and the
next is "nd," 2.6 per cent.
Considering the marked differences between the word lists for
conversa4;ion and for written English, a comparison of the relative
frequency of the speech sounds in the two cases is perhaps more remark-
able for the likenesses than the difTerences. About the same range of
percentages is covered in both cases. Certain sounds do show marked
difTerences. Among the vowels the unaccented vowel denoted by
"about" is more frequent in conversation and the vowel of "pan" less
frequent. The initial "y" and the final "t" are also more frequent in
conversation. Many of the differences can be traced directly to one or
two words which in their frequent use are typical of conversation.
In considering the occurrence of speech sounds in telephone conversa-
tions from the point of view of their contribution to the ease or diflfi-
culty of carrying on conversations it seemed of interest to determine
how the occurrence of the speech sounds was affected by changing the
list in certain ways. Omission of the minor parts of speech changes the
324 BELL SYSTEM TECHNICAL JOURNAL
relative occurrence of a number of the sounds materially, although the
general range of percentages covered is changed very little. Omission
of the 1,500 least common words has a negligible effect. When the
words are weighted by the number of conversations in which they
occurred, out of 500, instead of by their total occurrence, the effect
on the distribution of sounds is surprisingly small, considering the
radical change in method.
While the analysis into speech sounds for purposes connected with
the design of telephone circuits was the real goal of this study, it is
hoped that the information concerning both words and sounds will be
of service also to those working in the fields of phonetics and philology.
The Reciprocal Energy Theorem
By JOHN R. CARSON
This paper gives a simple theorem determining relative transmission
efficiencies in a two-way transducer, and showing that the conditions for
equal efficiencies of transmission in the two directions are simply those for
maximum output and maximum reception of energy. The theorem is then
applied to radio communication and a second theorem stated and proved by
which the ratio of the transmitting efficiences of any two antenna systems is
expressed in terms of their receiving efficiences. The paper closes with a
mathematical note on a generalization of Rayleigh's Reciprocal Theorem.
THE Reciprocal Theorem, originally enunciated by Rayleigh, which
has proved so useful to communication engineers, may be stated,
with sufficient generality for engineering purposes, as follows:
Let an e.m.f. E] , inserted in any branch, designated as No. 1, of a
transducer,^ produce a current I2 in any other branch No. 2; correspond-
ingly let an e.m.f. Ei" inserted in branch No. 2 produce a current I\'
in branch No. 1; then
I\ E\ = 1-2 El
■'1
and when £/ = Eo" the currents in the two branches are equal.
The engineer, however, is primarily interested in energy rather than
current relations, whereas the theorem says nothing explicitly regarding
energy relations and relative efficiencies in two-way transmission.
It is, however, a simple matter to deduce from it quite general and
useful formulas relating to relative transmission efficiencies. In the
present paper there will be formulated and proved a reciprocal energy
theorem for the general transducer, after which it will be applied to the
question of antenna transmission efficiency in radio communication.
Consider a transducer having two sets of accessible terminals 1,1
and 2,2. With terminals 2,2 closed by an impedance S2 = ^2 + ixt,
let the driving point impedance, as measured from terminals 1,1 be
denoted by Zn = Rn + iXn; similarly with terminals 1,1 closed by
an impedance Zi = r^ -\- ixi, let the driving point impedance, as
measured from terminals 2,2 be denoted by Z22 = -R22 + iX22. Now
with the terminals closed by the impedances z-i and 22, let an e.m.f.
£1 be inserted in series with the terminal impedance Zi; then the
current In, delivered to the transducer at the sending terminals 1,1 is
* A transducer is defined as a complete transmission system which may or may
not include a radio link, which has two accessible branches, either of which may act
as the transmitting branch while the other acts as the receiving branch. These
branches may be designated as operating branches.
325
326 BELL SYSTEM TECHNICAL JOURNAL
^"=F^ (1)
and the current /12, received by the terminal or load impedance, 22,
is given by
/12 = ^ ' (2)
Here Z12 is the transfer impedance of the transducer for the specified
terminations.
The power Pu" developed by the generator of e.m.f. Ex is
The power Pji delivered to the transducer is
P„ = K„|/„l= = ^;^£,= (4)
and the power P12 delivered to the load impedance z-i is
P,2 = r,\Ii2? =y^,E:~. (5)
Now reverse the direction of transmission; that is insert an e.m.f.
E2 in series with the terminal impedance z^; corresponding to equa-
tions (3)-(5) we have then
^'' -iz22 + 22r^' ^^
I Z22 + Z2 r
P21=-^£2^ (8)
As a consequence of the Reciprocal Theorem the transfer impedances
are equal ; that is
Z'ii = Z12. (")
From the preceding we get at once the following expressions for
the ratios of the powers delivered to the load impedances;
Pi2^r2/£iY
P2X rAEi)
THE RECIPROCAL ENERGY THEOREM
327
rj
r
2 \ / JV22
i/Un
Ri2 + r^X Zu + zi
+ n
^22 I 22
Pu"
22
i?
22
Ru
Zn + 2i
Z22 + Z2
P22
(10)
(11)
From (10) it follows that for equal total generated powers, the relative
transmission efficiency in the two directions is given by
12
P21
R22 + ^2
Ru + r,
Zn +
Z22 + S2
(12)
while on the basis of equal powers delivered to the transducer, the
relative transmission efficiency is, by (11)
V =
12
P21
R22
Ru
Zu + Zi
Z22 + Z2
(13)
Now in correctly designed communication transmission systems,
the terminal impedances are so proportioned with reference to the
characteristics of the transducer itself as to secure maximum output
and maximum transfer of power from generator to load; the required
condition is that the terminal impedances 2) and 22 be the 'conjugate
image impedances' of the transducer; analytically stated
Zi = Ru — iX
u
and
Z2 = R22 — iX
22-
Introducing these relations into (12) and (13), we have
^^ = r/ = 1 (14)
and the relative transmission efficiencies are the same in the two
directions. We thus have the following propositions: —
If a transducer is terminated in its conjugate image impedances — the
condition for maximum output and maximum transfer of power — the
efficiency of transmission is the same in the two directions.
We shall now apply the preceding to the derivation of a simple
formula which enables us to determine the relative transmission
efficiencies of any two long wave radio antennas.-
Consider any antenna, designated as No. 1, and let it be acting as
* As pointed out in the paper on "Reciprocal Tlieorems in Radio Transmission "
Proc. I. R. E., the Reciprocal Theorem does not hold rigorously in radio transmission
if the earth's magnetic field plays an appreciable part in the transmission phenomena.
Consequently the formula and proposition which follow apply rigorously only to'
long wave transmission; they are probably, however, approximately correct for short
wave transmission except in the neighborhood of the critical wave-length 214 meters.
See a paper by Nichols and Shelling, "Propagation of Electric Waves over the Earth ''
B. S. T. J., April 1925.
328 BELL SYSTEM TECHNICAL JOURNAL
a transmitter to a reference antenna, designated as No. 3, which is
located at any desired point 3. Let £13 denote the intensity of the
(vertical) electric field produced at point 3 by antenna No. 1. Then
the current induced in the receiving branch of No. 3 will be azEu,
the parameter as being the receiving sensitivity of antenna No. 3.
The power Pn transferred from 1 to 3 is then
Pi3 = rzaz^Eu^,
where r^ is the equivalent resistance of the receiving branch of antenna
No. 3.
Now reverse the direction of transmission; we have
Pzi = riarEzi^.
We now suppose that the terminal impedances are adjusted for maxi-
mum output and maximum transfer of power and that the power Pu
developed by No. 1 when transmitting is equal to the power P33
developed by No. 3 when transmitting. Then it follows at once from
the reciprocal energy theorem, that Pu = Pzu and
ExzY ^i«i! .
2
£31 / ''30:3'
Now replace antenna No. 1 by any other antenna, designated as No. 2 ;
we then have from the foregoing
E,zY _r,a.^ _
Ez2 / rzdz^
By virtue of the terminal impedances specified, r^ = Ri and
^2 = R2 where Ri and R2 are the resistances of the two antennas as
measured from their operating terminals. Consequently, since
£32 = Ezi, we have
EuV Riar i?l/^l-
where hi and h-z are the equivalent heights of the two antennas.
The ratio rjn will be termed the 'relative transmission figure of
merit' of the two antennas No. 1 and No. 2 with respect to trans-
mission between any two specified points. For directional antennas,
the parameters cci and a^ will depend on the direction of transmission;
that is, the location of the receiving with respect to the transmitting
point.
THE RECIPROCAL ENERGY THEOREM 329
The foregoing may be summed up in the following proposition.
The relative transmission figure of merit oj any two antennas with
respect to transmission from a given transmitting point to a given re-
ceiving point is equal to the ratio of their resistances as measured from
their operating branches, multiplied by the square of the ratio of their
receiving sensitivities with respect to transmission from the receiving
point to the transmitting point.
This theorem has a considerable field of practical utility. For
example it enables us to deduce the relative transmitting properties
and efficiency of any antenna system from its receiving efficiency.
It has already been so applied in one actual case of large importance.
Note on the Reciprocal Theorem
The proof of the Reciprocal Theorem, as given originally by Lord
Rayleigh, was applicable only to 'quasi-stationary' transducers, that
is transducers which obey the simple laws of electric circuit theory.
In the July 1924 issue of the Bell System Technical Journal the writer
stated and proved a generalized theorem subject, however, to the
restriction that the permeability ix of the medium shall be everywhere
unity. The theorem referred to is
Let a distribution of impressed periodic electric intensity F' = F'{x, y, z)
produce a corresponding distribution of current intensity u' = u'{x, y, z),
and let a second distribution of equi-periodic impressed electric intensity
F" = F"{x, y, z) produce a second distribution of current intensity
u" = u"{x, y, z), then
f{F'-u")dv = f(F"-u')dv,
the volume integration being extended over all conducting and dielectric
media. F and u are vectors and the expression (F-u) denotes the scalar
product of the two vectors.
Later Pleijel ^ stated the theorem for unrestricted values of n. In
discussing reciprocal theorems in the June 1929 issue of the Proc.
I. R. E. the writer expressed some doubt as to the validity of Pleijel's
proof (which is entirely different from my own). Subsequent study,
however, has convinced me that except for minor and easily remedied
errors, the proof is entirely sound. Later the writer discovered that
the restriction n = 1 can easily be removed from his own original
proof as will now be shown.^
'"Two Reciprocal Theorems in Electricity," Ingeniors V'etenskaps Akademien
Nr. 68, 1927.
* Another and somewhat different extension of the proof has been derived by my
associate Dr. W. H. Wise.
330 BELL SYSTEM TECHNICAL JOURNAL
If ju 5^ 1 everywhere and if we write
u; = u + curl M = \E + curl M (V)
equation (8) of my paper becomes ^
1 , too r W I iwr\ , ^ , 1 , njr
_^ + _J_exp^--jJ. = G+^curlM
(2')
X
and correspondingly equation (9) becomes
f{w'-G") - {w"-G')]dv
+ I ^ {u;'-curl M") - (w;"-curl M')]dv = 0. (3')
If now in (3') we replace u; by w + curl M and note that u/X = E,
(3') reduces to
f{iu'-G") - {u"-G')]dv
- f{{G'- curl M") - (G" • curl M') } dv (4')
+ f{E' -cml M") - (E"-cur\ M')]dv = 0.
Finally since E - G = -—A, (4') reduces to
f{{u'-G") - {u"-G')}dv
^l^fUA'- curl M") - {A" ■ curl M') ]dv ^ {). (5')
c
But
/(A' -curl M")dv = /(M"-curl A')dv
= T- f- ^(5"-curl^0^i;
47r J II
= -T- f ^^-^ (curl A" ■ curl A')dv,
47r J n
SO that the second integral of (5') vanishes and
f{{u'-G") - {u"-G')}dv = 0, (60
which is equation (9) of the original paper. The rest of the proof of
the theorem is now simply that of the original paper.
It will be observed the theorem is stated for the current u = \E;
that is the conduction (plus polarization) current. Ballantine ^ in
^ The paper itself must be consulted for the significance of the symbols and the
method of attack and proof.
6 June 1929 issue of Proc. I. R. E.
THE RECIPROCAL ENERGY THEOREM 331
discussing this subject states that the theorem holds for the current
w = \E -\- curl M. This cannot be true in general, however, because
from the foregoing in order that the theorem should hold for the
current w, it is clearly necessary that
f{{F'-cm\ M") - (f"-curl M')]dv = 0.
This is only true in the exceptional cases where the impressed force is
derivable from a potential; that is, curl Z' = 0, or else f = where
M ?^ 0.
The Approximate Networks of Acoustic Filters
By W. P. MASON
The approximate equivalent electrical networks of acoustic filters are
developed in this paper, from the lumped-constant approximation networks
for electric lines. In terms of this network, design formulae have been
developed for all single band pass filters. It is possible, from these formulae,
to determine the physical dimensions of an acoustic filter necessary to have
a given attenuation and impedance characteristic.
THE original theory of acoustic filters given by Stewart ^ is based
upon the representation of such filters by means of lumped
constants in the form of a 7" network. More recently, the writer ^
has presented a theory of acoustic filters, showing that they are
equivalent to a combination of electric lines. Lines, as an approxi-
mation, can be represented by networks with lumped constants, and
hence an acoustic filter has a lumped-constant approximation network,
which should represent the filter well at low frequencies. It is here
shown that the network proposed by Stewart is a first approximation
to the network of electric lines given in the former paper.^-^ This
first approximation represents the low pass filter well at low fre-
quencies, but does not very adequately represent the band-pass filters.
Accordingly, a second approximation is developed. All of the single
band-pass filters have been analyzed and design formulae are given
for them in terms of the second approximation network.
The Approximate Lumped-Constant
Networks of Acoustic Filters
An acoustic filter, as developed so far, consists of a main conducting
tube, and a side branch. In a symmetrical filter, the side branch is
connected to the main conducting tube half-way between the two ends,
as shown on Fig. 1 . The type of filter obtained depends primarily on
SIDE BRANCH
/
^MAIN CONDUCTING TUBE
Fig. 1
1 Stewart, Phys. Rev., 20, pp. 528-551, 1922. Phys. Rev., 25, pp. 90-98, 1925.
* Mason, Bell System Technical Journal, 6, pp. 258-294, 1927.
^ This fact has also been pointed out by Stewart, Journal of the Optical Society,
July 1929, and by Lindsay, Phys. Rev., 25, pp. 652-655, 1929.
332
APPROXIMATE NETWORKS OF ACOUSTIC FILTERS
?,?>?>
what type of side branch is used, the resonances of the latter deter-
mining the frequencies of maximum suppression.
The equivalent electrical circuit for an acoustic filter, was shown in
a previous paper ^ to be two lines shunted by the impedance of the side
branch. This representation is shown on Fig. 2. To obtain a lumped-
Fig. 2
constant representation for this network, it is necessary first to con-
sider the lumped-constant representation of a line, which is discussed
below.
A. Lumped- Constant Representation of a Line
In a previous paper - it was shown that the propagation constant of
a tube is given by the equation
p-. _
-^[(•-tVS)-^V£^]. ^'^
while the characteristic impedance is given by the expression
Z =
pc'P
(2)
In these equations co is 27r times the frequency, c the velocity of sound,
Po the perimeter of the tube, 5 its area, p the density of the medium
and 7'-, a constant 1 elated to the viscosity, which for air has the value
4.25 X 10-* in c.g.s. units.
A tube is the analogue of an electric line with distributed resistance,
inductance, and capacity. No quantity corresponding to leakance is
present. To determine the values of these quantities, use is made of
the well known equations for a line
Z =
R -f jcoL
G + jcoC '
P = V(i^-f jcoL)(G-f jcQ,
(3)
where R, L, G and C are respectively the distributed resistance, induc-
tance, leakance, and capacity of the line per unit length. Comparing
- Loc. cit.
22
334
BELL SYSTEM TECHNICAL JOURNAL
(3) with (1) and (2), it is found that
R
L
C
G
P
(4)
pc
0,
2 »
neglecting small correction terms. These are the equivalent distrib-
uted constants per unit length of the pipe expressed in acoustic
impedance units.
The representation of lines with distributed constants by means of
networks containing lumped constants has received considerable atten-
tion.^ With three impedances, either the T or it network representa-
tion shown on Fig. 3, can be used.
/OL
2S
Ppi-Ml'^P^
X
SL
PC'
2 S
X
X
PoL\/r^fpuj _
25^ V 2
Fig. 3
The impedances of short or open circuited lines can be represented
approximately by fewer elements than three. The first approximation
for a short circuited line is an inductance and resistance equal to the
sum of the distributed inductances and resistances of a line, while the
first approximation for an open circuited line will be a capacity equal
to the distributed capacities of the line. These approximations hold
for very low frequencies. The second approximation for open and
short circuited lines can be obtained with three impedances, as shown
4 A. E. Kennelly "Artificial Electric Lines, 1917."
K. S. Johnson "Transmission Circuits for Telephone Communication, 1925,"
page 151.
APPROXIMATE NETWORKS OF ACOUSTIC FILTERS
335
on Fig. 4. These representations follow directly from the T or ir
2S2V 2
— wwv —
PL
2S
SL
-OC2
Fig. 4
yOL
s2 V 2
rVWW — O^MXIS
SL
20C2
network representation shown on Fig. 3, by open or short circuiting
the T and tt networks, respectively.
B. Lumped-Constant Representation oj an Acoustic Filter
In his theory of acoustic filters, Stewart has represented an acoustic
filter by the network shown on Fig. 5, where Z^ is the impedance of the
/PL
s,
PL
Fig. 5
side branch. Stewart has represented the side branch impedance, by
either one or two elements, depending on the side branch, and the main
branch by a single inductance, equal to the sum of the distributed
inductances of the tube. This corresponds to the first approximation
of the representation of a line by lumped constants. This repre-
sentation gives good results for the low pass filter, but does not repre-
sent, very adequately, the band-pass filters.
The best second approximation for an acoustic filter, employing two
elements to represent the main conducting tube, is shown on Fig. 6.
PL
S|
■ S|L
PC2
;z2
■ S|L
"PC2
Fig. 6
336
BELL SYSTEM TECHNICAL JOURNAL
The main conducting tube is represented by an L network containing
the total distributed capacity of the tube in the shunt arm, and the
total distributed inductance of the tube in the series arm. The side
branch impedance shunts the two L networks at their center.
The propagation constant and characteristic impedance of this
structure are given by the expressions
cosh
P = 1
2ix)'-L'- jicpL I (jo'-L'~
/^2^ 1
1 +
joopL
2Zo.S',
J PC-' 1
1 -
•) f •' \ -1
wLSi{2Z2)
L-
1 -
CO
'-IJ
{^)
where S\ is the area of the main branch.
If these equations are compared with those given in the former
paper,- it is seen that they are approximately those obtained by taking
the first two terms of the expansions of the trigonometrical functions.
The characteristics of the filter are not very readily seen from equation
(5), but can be readily found by transforming the network shown on
Fig. 6, into the much more general lattice network shown in Fig. 7.
APPROXIMATE NETWORKS OF ACOUSTIC FILTERS
337
That the network shown on Fig. 7 is the equivalent in characteristic
impedance and propagation constant of that shown on Fig. 6, can
readily be verified by substituting the impedances of the lattice net-
work into the formulae for a lattice network
Z = ^ZaZb; cosh P
Zn + Za
Zb ~ Za
(^')
where Za is the impedance of one of the series arms, and Zb that of one
of (he lattice arms. A lattice network has a pass band when (he reac-
SERIES y/
ARM ~~y^ /
FREQUENCY
/ LATTICE
/*— ARM
r'
Fig. 8
tance of the series arm is of opposite sign to that of the lattice arm.
When the reactances of the two arms have the same sign, an attenua-
tion band results, while when the reactances of the two arms are equal,
an infinite attenuation constant results, since here the lattice will be a
balanced Wheatstone bridge.
For example, suppose that a side branch impedance, equivalent to
an inductance and capacity in series, is used. The impedance of the
lattice arm has two zero impedance points — one of which is at an infi-
nite frequency — and two infinite impedance points- — one of which is at
zero frequency — as shown on Fig. 8. The impedance of the series arm
3.^8 BELL SYSTEM TECHNICAL JOURNAL
Is that of an anti-resonant circuit, as shown on Fig. 8. There are
two possible impedance characteristics for the series arm, in relation
to the lattice arm, which will give a single band filter. One of these is
obtained by letting the series arm have an infinite impedance when the
lattice arm has a zero impedance, which results in a low pass filter.
The second relation — which is that shown on Fig. 8 — is obtained by
letting the series arm have an infinite impedance when the lattice arm
has an infinite impedance. The pass band is between zero frequency,
and the frequency at which the lattice arm resonates.
In a similar manner, the other types of acoustic filters can be ana-
lyzed.
C. Side Branch Impedances
The possible types of side branches can be divided into two classes,
those which are entirely enclosed, and those which are open to the air.
The first kind are characterized by a series capacity, while the second
kind always have a shunt inductance.
One of the simplest side branch impedances is a short tube open on
the end. The first approximation to this side branch is an inductance,
as shown on Table I, No. 1, equal to the total distributed inductance of
the tube. This approximation holds well if the product of the tube
length by the frequency, is not too large. A longer tube, open on the
end, can be represented by an inductance and capacity in parallel as
discussed in Section A and shown on Table I, No. 2. A tube closed
on the end can be represented by an inductance and capacity in series
as shown on Table I, No. 4.
When these tubes are used as side branches, an additional factor
comes in — an end correction. That is, the side branch must be con-
sidered as extending into the main branch for a distance proportional
to the radius, because a motion of air in the direction of the side
branch, occurs in the main branch. The value of this effect has been
investigated by Rayleigh, who found that this effect can be calculated
by increasing the length of the tube by a length equal to .785 times the
radius. Another correction applies to an open ended tube, which has
been determined experimentally as .57 times the radius. Hence the
length of an open ended tube must be considered as
/' = / + (.785 + .57)r.
A straight tube can give all the combinations of side branch imped-
ances, but one of its dimensions is necessarily limited, namely the
area. For the area cannot become larger than the area of the main
tube, since otherwise it could not be connected to the main tube. By
APPROXIMATE NETWORKS OF ACOUSTIC FILTERS 339
ELEMENT
-^IKJCOM^-
STRUCTURE NO. I
L =
P\'
VALUES OF CONSTANTS
V = 1 -+- l.355r
S = TTr2
ELEMENT
L
c
—
L =
PX Xi
S ^ = 2^
>
C2
STRUCTURE N0.2
/J
VALUES OF CONSTANTS
V = 1 + 1.355 r
S = TTr2
ELEMENT
p\- I'S
STRUCTURE
NO.
4
I
1'
I
'1
r
VALUES OF CONSTANTS
V= X + 0.785 r
S = TTr2
ELEMENT
L=^ C=^
2S /0C2
STRUCTURE NO. 5
-^t t--
2/-h— '^
VALUES OF CONSTANTS
Log
m).
0.46
t
r2t
(r2-r2)T.lX
(^i-f)
T + -
vft
S = n
Log
m)^°-^^
ELEMENT
STRUCTURE NO. 3
-\2^
<Ty ^<Z0^=
V, = l|+o.785ri S|=Trr
■I- M
I -^1-'" I
/OV Vs
L=^ Cr p , -,
S 2pc-^ \p = lp+0.57r? Sp=nr|
ELEMENT
'-2S %C2
STRUCTURE NO. 6
Y\-Y^Y
12-12""
I 2 02— It f 2
V| = l|+ 0.785r, S|r:Tirf
VALUES OF CONSTANTS
l'2-\2
VALUES OF CONSTANTS
S2:TTr2
1='
i2^+3i'fr2
^si^^^
|1'|)+1'3S2]
(i',3+3i'i r|)s I S2+3sfvf 1-2+1' i s|
3(S21|+1'2S|J
3S2[S|1',+S2l2]
S = '
/2sfs2[S|S2(v|+3Vfl2)+3sfl',l'|+s|lf]
3(S2V, + V2S,)^
s =
382(5,1', +8212) -
(V,3^3l', I'l)
[vr+3V, I'l ) s,S2+382v2v2 + r|s||
Table I
340 BELL SYSTEM TECHNICAL JOURNAL
using other types of side branches, this difficulty can at least be
partially eliminated. For example, a concentric tube closed on the
end is, to a first approximation, equivalent to an inductance and
capacity in series, and it can be made to have a larger area relative to
the main branch tube, than can the straight tube.
The choice of the forms of the structures to give the simplest
impedance elements, is large. For example, Stewart represents a
shunt inductance and capacity in parallel, by a concentric tube closed
on the end, and a straight tube open on the end, joined together to the
main conducting tube at a common point.^ Other methods for
representing two elements are shown on Table I. In these structures,
the equivalent length and equivalent areas have been calculated cor-
responding to these values for a straight tube. These elements have
been calculated by calculating the impedances looking into the
structures and taking the second approximations.
D. Design Formulx for Acoustic Filters
Using the side branch impedances shown in Table I, in the lattice
network shown by Fig. 7, the resulting characteristics can readily be
obtained. A large number of multiband characteristics can be secured
by using various combinations of side branches, but only five single
band filters (to the degree of approximation considered here) have been
found. The attenuation characteristics of these filters and the design
formulae for them are shown on Table II. In designing a filter, it is
usual to obtain the dimensions in terms of the singular frequencies
which determine the action of the filter. One other parameter appears,
Zo, which represents the characteristic impedance of the filter at the
mean frequency of the band i.e. fm = V/1/2. It is usual to match,
approximately, the impedance terminations of the filter to the value Zq.
All of these filters have been calculated for side branch tubes, of con-
stant cross section but any of the other side branches shown on Table
I can be used by employing the equivalent values of /' and 5 shown
there.
The frequency /a appearing in the filter No. 1 has no significance for
the attenuation constant. It determines the frequency at which the
characteristic impedance equals infinity. Considering the loss caused
by inserting the filter between two impedances equal approximately
to Zo, an additional loss occurs at the frequency /„, due to a mismatch
of the impedance of the filter and the terminating impedances. Filter
No. 4 of Table II is similar to No. 3 except that it has twice the attenu-
ation constant. It is then equivalent to two sections of the No. 3
filter.
* See for example Journal of the Optical Society, July 1929, page 18.
m
-</)-*
<
E
u
ffi
5
Z
<r
>-
o
(T
8 NOIiVONjliV
O
8 N0llVnN3iiV
8 N0IXVnN3iIV
8 NOIlVnN31iV o
tyw
C4-
u
t=
(\J
^1?f
^8
I2i
(M
wp
■^c^
<.8
1?
c~
tC
CJ
1
<V8
(J
o
N
l*)|(M
f«C\J
+
CM-
U
I
o
N
Q.
CVJ
o
+
^
>
1
ur
Q.
^
o
N
u^
1
(v_8
wp
0J_
1
Q.
T
't
o
fsi
8 N0IIVnN3ilV
o
^
t
f '
Q. c/i
8 N0llVnN3IlV
'^ri^-^
Ci u)
+
lOlftI
T-
8 N0llvnN3iiV
Qk/l
^
tf»
1=
«-l^
8 NOiivnNiii* °
al'
"iLSi^T,
~|cv|8
nMvi~8
~^|(vi8
i-iLS
~rt°
Ol -
•3
8 N0livnN311VO
^
is;
t- <
< X
<£5
Contemporary Advances in Physics, XX
Ionization of Gases by Light
By KARL K. DARROW
The subject of this article, the ionization of gases by ultraviolet light,
is a narrow but singularly inviting department of modern physics. The
obstacles to experiment are so great that they are only now being overcome
by the latest improvements in laboratory technique; nearly all the valu-
able data are of recent acquisition, and the period of discoveries is not yet
past. Some of the results afford excellent confirmation of present atomic
theory; others are still obscure and challenging.
THE subject of this article is more narrowly restricted than those
of many others of the series. It is narrower even than the title
might imply; for by "ionization" I mean for the present only the
detachment of the most loosely bound electron of a molecule or an
atom, and by "light" only the waves of the visible spectrum and that
adjoining range of wavelengths to which the name of ultraviolet is
customarily confined. Either of these limitations is implicit in the
other; for though most molecules are fashioned with electrons bound
with varying degrees of tightness, and the removal of any one thereof
is an act of ionization, it is beyond the power of such light-waves to
abstract any except the loosest. Perhaps it will be found instructive
if for so definite and circumscribed a problem I relate the methods
of experiment, the data of the experiments, the simple theory, and
the artifices which have been conceived to reconcile the theory and the
data, sometimes with success and at other times in vain.
Most of the really valuable data are of recent acquisition, for there
are difficulties hampering the attack upon the problem, which the
progress of laboratory technique is only gradually clearing away.
Consider, for instance, the question of providing the light. It is de-
sirable to be able to illuminate the gas with monochromatic light of
any wavelength, photons of any energy. When ionization by electrons
is being studied, one varies the energy and the wavelength at will by
varying the voltage impressed on the electrons. With light, this is not
within our power; one has to take the quanta as they are supplied by a
luminous source.
If the spectrum of the source consists of bright lines widely separated,
the ionization which any of them alone produces may be measured,
and the energy of the quanta is very narrowly defined. On the
curve of ionization vs. frequency, then, every spectrum line supplies an
341
23
Contemporary Advances in Physics, XX
Ionization of Gases by Light
By KARL K. DARROW
The subject of this article, the ionization of gases by ultraviolet light,
is a narrow but singularly inviting department of modern physics. The
obstacles to experiment are so great that they are only now being overcome
by the latest improvements in laboratory technique; nearly all the valu-
able data are of recent acquisition, and the period of discoveries is not yet
past. Some of the results afford excellent confirmation of present atomic
theory; others are still obscure and challenging.
THE subject of this article is more narrowly restricted than those
of many others of the series. It is narrower even than the title
might imply; for by "ionization" I mean for the present only the
detachment of the most loosely bound electron of a molecule or an
atom, and by "light" only the waves of the visible spectrum and that
adjoining range of wavelengths to which the name of ultraviolet is
customarily confined. Either of these limitations is implicit in the
other; for though most molecules are fashioned with electrons bound
with varying degrees of tightness, and the removal of any one thereof
is an act of ionization, it is beyond the power of such light- waves to
abstract any except the loosest. Perhaps it will be found instructive
if for so definite and circumscribed a problem I relate the methods
of experiment, the data of the experiments, the simple theory, and
the artifices which have been conceived to reconcile the theory and the
data, sometimes with success and at other times in vain.
Most of the really valuable data are of recent acquisition, for there
are difficulties hampering the attack upon the problem, which the
progress of laboratory technique is only gradually clearing away.
Consider, for instance, the question of providing the light. It is de-
sirable to be able to illuminate the gas with monochromatic light of
any wavelength, photons of any energy. When ionization by electrons
is being studied, one varies the energy and the wavelength at will by
varying the voltage impressed on the electrons. With light, this is not
within our power ; one has to take the quanta as they are supplied by a
luminous source.
If the spectrum of the source consists of bright lines widely separated,
the ionization which any of them alone produces may be measured,
and the energy of the quanta is very narrowly defined. On the
curve of ionization vs. frequency, then, every spectrum line supplies an
341
23
342 BELL SYSTEM TECHNICAL JOURNAL
experimental point of which at least the abscissa Is certain. But at the
frequencies between the lines, there is no way of getting information;
and many a published curve is traced by guesswork right across the
regions of major importance where data are essential, simply because
Nature left those regions vacant of lines in the spectrum of the mercury
arc!
If on the other hand the source has a continuous spectrum or one
crowded with bright lines, the device for resolving or filtering the light
will transmit to the ionizable gas photons not of a single wavelength,
but of a range of many Angstroms — dozens or scores of Angstroms,
perhaps even a hundred. A single measurement may be, and usually
is, plotted as if it belonged to the central wavelength of the transmitted
band; but actually the ionization results from waves of all the wide
range, and not even from a uniform distribution of energy across the
range, but from a distribution affected by the qualities both of the
source and of the resolving apparatus. With chemical filters, i.e.
with coloured absorbing liquids, bands of transmitted light may be
formed in various parts of the spectrum. They are likely to be broad
and hazily bounded, and limited in number; but the liquid filters are
inexpensive and easy to handle, and in tracing backward the sequence
of observations on any one substance, one often finds that the very
earliest were made with filters. Monochromators— which is to say,
spectroscopes — form bands of which the central wavelength and the
width may be varied at will. This sounds ideal; but in practice, of
course, the narrower the transmitted range of wavelengths, the
scantier the transmitted energy; and one must compromise as best
one may between a band too narrow to produce a measurable degree
of ionization, and one so broad that it is hard to apportion the credit
for the effect which it produces among the frequencies which make it
up. It will be evident that the ideal curve, drawn through experi-
mental points scattered thickly all through the spectrum and each
corresponding to a single wavelength, is difficult of attainment and
even of approach !
As the atmosphere of the earth prevents us from observing the
spectra of the stars at shorter wavelengths than some 280w/i, so the
opacity of all terrestrial solids prevents us from projecting quanta of
smaller wavelength than \2Smix, into an enclosure. Indeed, it is only
from fluorite and only from occasional samples of fiuorite that one can
make windows which are transparent so far out; the next best and much
the commoner substance, quartz, ceases to transmit at about 145m/z.
We are thus almost entirely debarred from observations on the noble
gases and on the common diatomic gases, which is deplorable.
CONTEMPORARY ADVANCES IN PHYSICS 343
The desired photons being successfully fired into the gas, the next
problem is that of distinguishing the ionization they may cause in it
from whatever other liberation of charge they may effect in striking
walls, electrodes, or any of the other furniture within the tube. Light
of sufficient frequency to ionize a gas will usually be able to produce an
outflow of electrons from almost any metal. One takes of course the
elementary precaution of designing one's tube in such a way, that the
beam of light traverses it from entrance-window to exit-window
without touching any electrode; but the disparity of the effects is
nevertheless so great, that a modicum of scattered light may evoke
more electrons from the metal than the primary beam in all its strength
detaches from atoms of the gas. Some experimenters use alternatively
two electrodes of very unequal size, expecting that a current due to
ionization of the gas will be the same in magnitude whichever they use
as cathode, while a current due to light falling on the electrodes will
be greater when the larger is the cathode. Some vary the density of
the gas, assuming that if the current is proportional to density it must
be due to the effect which they seek; but it would also vary as the
density, if instead it consisted of electrons expelled from the electrodes
by light proceeding from excited atoms of the gas. Some finally have
so designed their apparatus that they perceive positive ions only; this
seems to be the safest way.
Like Hughes,* I will divide the data according to the character of the
gases to which they refer: the common or "permanent" (chiefly dia-
tomic) gases first, then mercury, finally the alkali metals.
The permanent gases can be disposed of in short order, for our
knowledge in this field is scanty, though surprising. Measurements of
ionization by electron-impacts, and what little has yet been deduced
from spectra, agree in indicating that for all of them (with the slight
exception of nitric oxide NO) the ionizing-potential is greater than 10
volts. Translating this figure into wavelengths of light, we infer that
only photons of smaller wavelength than 125w7/x can ionize such a
molecule in a single impact. Therefore light which is able to traverse
any window of solid substance should be unable to ionize any perma-
nent gas (except NO) ; in other words, any such gas enclosed in a tube
should be immune to ionization by any radiation entering from outside.
Yet there is unimpeachable evidence that air, and oxygen and nitrogen
separately, and possibly hydrogen and iodine, are ionized by light
which has penetrated fluorite. The threshold for these gases must
* A. L. Hughes, Ionization of Gases and Vapors by Light (Washington University
Studies, 1929). 1 have benefited much by this article, and also by that of F. L.
iVIohler, Recombination and Photoio7iization (Reviews of Modern Physics 1, 216-227
(1929)).
344 BELL SYSTEM TECHNICAL JOURNAL
therefore lie on the long- wave side of 125m/x. For air it is presumed to
lie between this and 145m/x, since a plate of quartz holds back the ioniz-
ing rays. The discrepancy between these and the expected thresholds
may not seem large, but it is important. Naturally one has recourse
to the idea of a two-stage ionization, occurring when two quanta in
succession are absorbed by the same molecule— an idea which, we
shall see, is frequently invoked in other cases. If this is valid, the
ionization should increase as the square of the intensity of the light.
There seem to be no data bearing on this point. To quote from
Hughes, "further investigations in this field are badly needed."^
With mercury the situation is much more definite, but for those who
like to have simple theories verified completely it is no more satisfac-
tory.
The spectrum of the mercury atom is well mapped and well inter-
preted, and the ionizing-potential for electron-impacts has been deter-
mined over and over again. From both of these it follows that ioniza-
tion by single photons should be possible only at wavelengths smaller
than 1188A. However it is certain that the light of the famous res-
onance-line of mercury, 253 7A, is able to ionize the vapor of the ele-
ment whence it proceeds.-
This seems the natural equivalent of the well-known fact that when
mercury atoms are bombarded by a sufficiently dense electron-stream,
ionization begins at the resonance-potential. The quanta of the
wavelength 2537A have 4.9 equivalent volts of energy. Such a
photon strikes an atom, and excites it transferring if from the normal
into a certain excited state, denoted by the symbol 2^Pi; a second
comes along and likewise is absorbed, bringing the energy of excitation
of the atom up to twice 4.9 equivalent volts; this amount falls short of
the ionizing potential by only 0.6, and a third photon more than sup-
plies what is required. So runs the simple interpretation; but we shall
see that only the first of these steps is confirmed by further experiment,
and that the rest of the process must happen in some other way,
though the way is far from clear.
1 For references to the literature I refer to Hughes {I.e.) Among the latest papers
are those of A. L. Hughes {Proc. Camb. Phil. Soc, 15, pp. 483-491 (1910)); F. Palmer
(Phys. Rev. 32, pp. 1-22 (1911)); E. B. Ludlam {Phil. Mag. (6) 23, pp. 757-772
(1912)); W. West, E. B. Ludlam {Proc. Roy. Soc. Edinb., 45, pp. 34-41 (1925)).
Some of the early work on air was confused by what appears to have been a photoelec-
tric effect of particles of colloid size ("nuclei" or Kerne) produced by the action of
ultraviolet light on impurities in the air — one of the once-popular and now forgotten
problems of physics.
2 Literature: G. F. Rouse and G. W. Giddings, Proc. Nat. Acad. Sci., 11, pp.
514-517 (1925); 12, pp. 447-448 (1926). F. G. Houtermans, ZS. f. Phys., 32, pp.
619-635 (1925). Twenty years ago W. Steubing {Phys. ZS., 10, pp. 787-793 (1909))
observed that light coming from a mercury arc and passing through quartz was able
to produce a current in a tube containing mercury vapor; but his result has been
impugned.
CONTEMPORARY ADVANCES IN PHYSICS 345
It was the beautiful experiment of Rouse and Giddings which con-
firmed that the first of the steps is the entry of the atom into the
2'P] state; for they showed that ionization of the gas occurs only when
the impinging quanta have just the energy required for that transition,
not when they have a little less or even a little more. This they were
able to show because of the phenomenon of "self-reversal." When a
luminous gas becomes dense and hot, the lines of its emission-spectrum
broaden ; for the atoms perturb one another, the energy- values of their
stationary states are changed by various amounts, and the frequencies
of many of the quanta which emerge are appreciably shifted upwards
or downwards from the original or "standard" values appropriate to
isolated atoms. If in addition the region of density and heat is sur-
rounded by another where the gas is cooler and more rarefied, the
atoms in this outer zone, being relatively unperturbed, will be able
to absorb the quanta having the standard frequencies, but not those
others of which the frequencies are shifted. In technical language, the
"core" of the line is absorbed; only the "wings" pass through; the Hne
exhibits "self -reversal." In the spectrum of the ordinary mercury-
vapor lamp, the line 2537 is notably self -reversed. Cooling the lamp
with flowing water or an air-blast, however, abolishes the effect; the
line shrinks to its normal narrowness, the wings disappear, but the
photons of the core are able to escape from the tube. Any action there-
fore which is performed by the light of a cooled mercury-vapor lamp,
but ceases when the cooling is suspended, must be due to quanta having
energies adjusted exactly to the values which are able to excite isolated
atoms of mercury. Rouse and Giddings found that ionization of
gaseous mercury is precisely such an action.^
We cannot so readily conclude that the second step in the ionization-
process is the absorption of a second 4.9-volt quantum. It would be
rather of an odd coincidence, if there were an excited state of the
mercury atom differing in energy from the 2^P] state by just so much as
this latter differs from the normal state — not, however, an impossible
coincidence. Another and a stronger argument is furnished by the
fact that when quanta of various wavelengths shorter than 2537 — in-
cluding some which could transfer the atom from the 2^Pi into other
known excited states — are projected into the gas along with 2537,
the rate of ionization is not augmented. If none of these can help the
electron to escape, it is not so likely that a second quantum of precisely
the wavelength 2537 can achieve it. Moreover the duration of the
' There was still a residual current in the irradiated tube when the cooling of the
lamp was discountinued; but it depended on the size of the cathode in such a way as
to suggest that it was due to light falling on that electrode (cf. page 343). Houter-
mans later verified this result.
346 BELL SYSTEM TECHNICAL JOURNAL
2^Pi state is known to be so short (of the order of 10"'^ second) that
under the actual conditions of some of the experiments an atom would
not often meet two quanta in such quick succession that at the advent
of the second it would still be in the 2^P] state into which the first had
put it. In other words, the number of 2^P] atoms in the gas at any
moment is too small.
This last would be a serious obstacle to any theory, but for the fact
that the mercury atom possesses another stationary state slightly
below the 2^Pi, the which is metastable. This is the 2^Pq state, of 4.7
equivalent volts; its mean duration may amount to something like
a hundredth of a second. Now collisions of mercury atoms in the 2^Pi
state with atoms of certain other kinds, argon notably, may cause the
former to pass over into 2^Po- This is an instance of "collisions of the
second kind." When mercury-vapor is mixed with a much larger
quantity of argon and is illuminated with 2537 light, the number of
2^Po atoms is at any moment much greater than the number of 2^Pi
atoms would be, if the argon were absent ; further, it is proportional to
the amount of argon.
Now F. G. Houtermans found that the rate of ionization, in mercury
mixed with argon and irradiated by 2537, is proportional to the amount
of argon. Therefore, in one stage of its progress from a normal atom
to an ion, the mercury atom must be in the 2^Po state. It enters this
state from the 2^Pi because of a collision with an argon atom. How
does it leave? by absorption of a second 4.9-volt quantum? Two of
the considerations of the last paragraph but one speak against this
idea, and Houtermans thinks that the 2^Po atom collides with another
which is in the 2^Pi state, and there is an interaction — this would be
another sort of "collision of the second kind" — in which one of them
adds to its store of energy all or most of what the other possesses. So
it arrives within an equivalent volt or so of the state of ionization; if
one were to take over all the energy of excitation of the other, it would
have 4.7 + 4.9 = 9.6 equivalent volts, out of the 10.4 required. Still
a third step seems to be essential.
The reader may have wondered that I have as yet said nothing
about the dependence of ionization on intensity of light, for evidently
the former should increase as the cube of the latter if the process is a
three-stage one as I have sketched. The matter has been tested by
experiment; the answer was unexpected, for the ionization varies as the
square of the light — as though the process were of two stages.^ We
* This simple result was obtained only over certain ranges of temperature and
pressure of the vapor, but these were precisely the ranges where both are low, and we
should expect the result to be most reliable and least subject to confusion by secondary
effects. As the pressure rises so does the exponent n in the relation ionization =
{intensity)'^.
CONTEMPORARY ADVANCES IN PHYSICS 347
cannot suppose that the atom after its second gulp of energy picks up
the remaining 0.8 volt in a collision with a fast-moving ordinary atom,
for at normal temperatures such fast-moving atoms would be exces-
sively rare. Houtermans suggests that when a 2^i'o and a 2^Pi atom
collide with one another, they unite to form an ionized molecule Hg2+.
This is far from being the only case in which a molecule is invoked
as the deus ex machina to help out with an otherwise untenable theory.
We turn now to the alkali metals, or rather to the three heavier
among them, caesium, rubidium, and potassium. With these it is
more nearly possible to get a full view of the situation. The phenom-
ena are not confined to spectrum ranges in or beyond the remotest
attainable fringes of the ultra violet. Indeed, in these four cases,
even the wavelength where ionization by single impact should begin
is well within reach, being in the nearer ultra violet; 241 2A for Na,
2856 for K, 2968 for Rb and 3183 for Cs. Ionization currents are
provoked by light at even greater wavelengths ; this resembles the case
of mercury irradiated by 2537, and is equally perplexing, indeed more so.
They are however much greater, near or beyond the limiting wave-
length for one-stage ionization; and there, we seem to be witnessing
the simplest process of all. With caesium, rubidium and sodium, the
data in this range conform to simple theories in a gratifying way. I
will consider these first, and then the most mysterious case of all, that
of potassium.
The vapor pressures of the alkali metals increase with atomic num-
ber, and for rubidium and caesium are great enough to permit the
methods employed with the gases mentioned above: which is to say,
that stationary vapor of known density may be illuminated by light
of known intensity, and the amount of ionization be measured absolutely
by drawing off all the ions. This I denote as the "absolute" method.
There is another, the "method of space-charge annulment." The
tube containing the vapor contains also a hot-filament cathode and
some form of anode, and the filament is kept so hot, the P.D. between
it and the anode kept so low, that the electron-borne current between
cathode and anode is limited by its own space-charge. W^hen
positive ions are formed in the vapor, as in these experiments they are
by light, a fraction of the negative space-charge is annulled, and the
current increases. The change in the current is a measure of the num-
ber of positive ions formed. Nothing of the sort results if light falls
on solid objects in the tube and ejects electrons, an insensitiveness
which is a great advantage of the method. For positive ions it is a
very sensitive method; one finds such statements as "each positive ion
formed causes a million extra electrons to flow from cathode to anode,"
348
BELL SYSTEM TECHNICAL JOURNAL
and Foote and Mohler, who were the first to apply this method to
ionization by Hght, perceived the effect at pressures of mercury vapor
as low as .002 mm. Hg. It does not permit of absolute measurements;
but one may use it to make accurate measurements of the relative
ionizing-power of light of any number of wavelengths, and then stand-
ardize them en bloc by a determination at a single wavelength with the
absolute method.
3200
3000
2800 2600
WAVE LENGTH
2400
2200
Fig. 1 — Ionization by light plotted as function of wavelength for caesium (Critical
wavelength: 3184A). (F. L. Mohler, C. Boeckner).
I now reproduce two of the most recently published curves of ioniza-
tion vs. wavelength: Fig. 1 for caesium, from F. L. Mohler and C.
Boeckner;^ Fig. 2 for rubidium, from E. O. Lawrence and N. E.
Edlefsen.^ It is the downward trend of these curves from the limiting
wavelength towards shorter waves which interests us now. Ionization
by light of a given intensity is most abundant when the quanta have
just the energy required to detach the electron, and no more. The
more the energy of the photon exceeds the strictly necessary value,
the less it is likely to be captured and have its energy spent for ioni-
zation.
The various theories, except for one, predict a steady downward
trend; one in particular, that of R. Becker, supplies the broken curve
5 Bur. Stand. Journ. Res., 3, pp. 303-314 (1929).
^Phys. Rev., (2) 34, pp. 233-242 (1929).
CONTEMPORARY ADVANCES IN PHYSICS
349
of Lawrence and Edlefsen's figure (relative ordinates have no signifi-
cance, it is only the trends of the curves which should be compared).
Mohler and Boeckner also measured the actual number of ions
produced by light of known intensity in a known quantity of gas,
using of course the absolute method, and expressing their results in the
following way. Suppose a thin stratum of gas, of thickness dx and
area A. Denote by N the number of atoms per unit volume of the
2400 2600 2800 3000 3200
ANGSTROMS
3400
3600
3800
Fig. 2 — Ionization by light plotted as function of wavelength for rubidium (Criti-
cal wavelength at 2968A). Circles and crosses correspond to different densities.
(E. O. Lawrence, N. Edlefsen.)
gas; then NAdx will stand for the number in the stratum. Denote
by Q the total number of photons striking the stratum in unit time;
suppose that they fall upon it perpendicularly, and are evenly dis-
tributed over its area. The number of ions formed in the stratum
in unit time, / will be proportional to NAdx and to Q/A. Write:
/ = kNQdx
the coefficient k is the quantity of which the experiments are designed
to reveal the value. (We should not be entitled to expect this to be
constant, if more than a small fraction of the quanta were spent in
ionization; but in the practical cases we may.) The values which
they give are (2.3 ± 0.2) • 10~^^ for caesium and 1.1 • 10~^^ for rubidium,
350 BELL SYSTEM TECHNICAL JOURNAL
at the limiting frequency in each case. Earlier E. M. Little ' had got
a value two orders of magnitude lower for caesium; this difference has
not been reconciled. These values will later be compared with those
to which the theories lead.
The upturn in the curve of Fig. 1 on the shortwave side of 2600A
may serve ^ as an introduction to the case of potassium. Adjourning
therefore the discussion of the righthand part of the curve of Fig. 2,
I take up next this strange and singular case.
The first who plotted an ionization-vs-wavelength curve for potas-
sium was E. O. Lawrence.^ The vapor-pressure of potassium being
low, he so designed his tube that the beam of light passed across the
vertically-rising jet of gas distilling from a pool of highly-heated metal.
This expedient was used by all the other physicists who worked upon
potassium, and was at one time held responsible for the curious results,
until finally Mohler and Boeckner confirmed the previous data by
measurements on stagnant vapor. The ionization-current was col-
lected by electrodes placed on either side of the jet and away from
the light; so the method is fit to give the relative ionizing-powers of
light of various wavelengths, though not an absolute measurement,
the density in the jet being unknown. Lawrence's monochromator
provided beams of light extending over some 80A of the spectrum.
Few data can have been more unexpected, indeed more positively
unwelcome, than those which he obtained; for what they intimated
was, that ionization begins, or at least the sharp increase of ionization
occurs, at a wavelength definitely too small. It seems as though a photon
could not ionize a potassium atom without having definitely more than
the necessary energy; a conclusion which would be in disaccord with
fundamental theory, and with the (subsequent) experiments upon
rubidium and casium.
New experiments upon potassium gave comfort to the theory, but
also demonstrated the anomaly which Lawrence had discovered.'" The
7 Phys. Rev., (2) 30, pp. 109-118, pp. 963-964 (1927).
* However it does not appear in the corresponding curve obtained by Lawrence
and Ediefsen.
^Phil. Mag., 50, pp. 345-359 (1925). There had been four precursors: S. H.
Anderson, L. A. Gilbieath, R. C. Williamson, R. Samuel (for the references, see
Hughes, I.e.). The earliest two reported ionization at wavelengths where it now
seems unlikely that true ionization of the vapor would have been perceptible; the
others used chemical filters and so were unable to plot a curve, but seem to have
observed the weak ionization produced between 2800 and 3100A.
" Such a proof would relieve us from one of the greater difficulties of the " molecule "
hypothesis — the necessity of assuming that ionization of a K2 molecule by light is an
event thousands of times as probable as that of a K atom, for in potassium vapor
under the actual conditions free atoms are believed to be a thousand times more abun-
dant than molecules, and yet the ions which we are ascribing to the latter are much
more plentiful. R. W. Ditchburn and F. L. Arnot {Proc. Roy. Soc. 123, pp. 516-536
(1929)) found nothing but K+ ions in the ionized vapor, thus disposing of the notion
that the process might consist in the detachment of an electron from a thenceforward
stable K2 particle.
CONTEMPORARY ADVANCES IN PHYSICS
351
curve which I display as Fig. 3 is taken from the latest paper,' ^ but
the marked points comprise those of Lawrence's first article (large
circles) and those obtained in the interim by R. C. Williamson. '- The
monochromators used in these late researches gave narrower wave-
POTASSIUM
2200
2400
2600
ANGSTROMS
2800
3000
Fig. 3 — Ionization by light plotted as function of wavelength for potassium
(Critical wavelength: 2856A). Circlets, crosses and large circles correspond to
different sets of observations by Lawrence & Edlefsen, Williamson & Lawrence.
(E. O. Lawrence, N. Edlefsen.)
length-bands than those used formerly, and so revealed the small peak
at the proper limiting-frequency which had eluded Lawrence at the
outset. The much more prominent peak at shorter waves remains
outstanding. The data, be it mentioned, are here reduced to equal
intensities of light for the various wavelengths.
11 Lawrence & Edlefsen, Phys. Rev., (2) 34, pp. 1056-1060 (1929).
i2Proc. Nat. Acad. Set., 14, pp. 793-799 (1928).
352 BELL SYSTEM TECHNICAL JOURNAL
The molecule was invoked at once as the deus ex machina; the ioniza-
tion beginning beyond the proper wavelengths was supposed to be
ionization of molecules, with or without dissociation. So long as the
threshold was thought to be near 2600 or 2550, this idea was fortified by
the following calculation. Suppose that a photon of wavelength
2555A has just the energy required to split a K2 molecule into a
K atom, a K+ ion, and a free electron; and that a photon of 2856A
has just the energy required to split a K atom into a K+ ion and a free
electron. One easily sees that then the difference between the energies
of these two photons would be just the energy required to split a
K molecule into two neutral K atoms. The difference amounts to
0.5 equivalent volt. This figure agrees ^^ with independent estimates
of the value of the latter quantity, which is the heat of dissociation of
K2. The force of this agreement has just been weakened by the
curve of Fig. 3, showing as it does that the ionization in question
begins near 2700A — weakened, but not destroyed, for the ions pro-
duced by waves shorter than 2555 might be explained in a way which
the reader will easily imagine after the next two paragraphs. The
other alternative is, to hope that quantum mechanics will presently
prove that the ionization-vs-frequency curve for the potassium atom
ought to display both the maxima which are found.
Return now to the curve of Fig. 2 for rubidium. On the long-wave
side of the limiting-frequency there is a series of peaks ; they lie at the
frequencies of the various members of the principal series of lines in
the Rb spectrum. Even more striking peaks of this sort were earlier
obtained with caesium by Foote, Mohler, and R. L. Chenault;^^ the
relevant part of one of their curves is shown as Fig. 4.
Palpably these are phenomena of the same sort as one meets when
mercury is irradiated by 2537; and they signify an ionizing-process of
two or more stages, the first of which is excitation by the absorption of
a photon. There is probably no need to suppose more than two stages;
the energy received by the atom from the photon is always much more
than half of what is required to ionize. It is supposed by those who
have obtained the data that the process is completed by an impact of
fast-moving atom, one of those which by virtue of Maxwell's distribu-
tion have the necessary excess of energy over the relatively modest
mean value corresponding to the actual temperature. The relative
heights of the peaks would then be determined partly by the relative
abundance of atoms having the necessary energies, and partly by the
relative probabilities of the corresponding types of excitation, which
" R. W. Ditchburn, Proc. Camb. Phil. Soc, 24, pp. 320-327 (1928).
^^Phys. Rev., (2) 26, pp. 195-207 (1925); 27, pp. 37-50 (1926).
CONTEMPORARY ADVANCES IN PHYSICS
353
are quantities with which the theories deal. According to Foote,
Mohler and Chenault, these relative heights are in fair accordance
with the theories. The actual heights, however, depend on the mean
duration of the excited states. I do not know whether it has been
proved that these last long enough to permit the explanation.
36
32
28-
IS-4p
Ui
>-
t
> 24
20
z
u
<rt
o
I-
o
I
a
u 16
>
i 12
_L
162°C -
.28
- 24
- .20
.16
.12
08
.04
.00
3900 3800 3700
3600 3500
\ IN A.U.
3400
3300 3200
Fig. 4 — Ionization of caesium vapor by light, at wavelengths greater than the
critical (3184A). (Aiohler, Foote & Chenault.)
Since the quanta spent in ionization vanish from the light, the
transmitted beam when spread into a spectrum reveals absorption
at their frequencies. These absorption-spectra supply all that is
known as yet about the process of ionization by light in sodium and
in atomic hydrogen and valuable additions to the data for the three
heavier alkali metals.
It will be remembered that the lines of a line-series in an absorption-
spectrum occur because the photons of the corresponding frequencies
can be absorbed by atoms in a particular initial state (normal or
excited) which thereupon pass over into higher states of excitation; that
as the lines converge upon the limit of the series, the corresponding
terminal states approach that of ionization; that the limiting or
354 BELL SYSTEM TECHNICAL JOURNAL
convergence-frequency itself, multiplied by //, give the energy required
to ionize an atom from that initial state which is common to the entire
series.
Thus photons having the convergence-frequency of any series are
just able to detach an electron from an atom in the corresponding
state. Consequently photons having any greater frequency have
energy sufficient to detach an electron, and give it some kinetic energy
in addition. Now we are not aware of any "quantum" limitations on
the amount of energy which a freed electron may receive. We thus
infer that light of any frequency superior to a convergence-frequency
will be able to ionize atoms and to be absorbed in doing so, and that
there will be a continuous region of absorption in the spectrum extend-
ing upwards from the limit of each series. For such a region I will
use the terms continuous band and continuum.
Bohr drew this inference in the first of his epoch-making papers on
the interpretation of spectra. He was able then to point to only one
example; a continuum beyond the limit of the principal series of
sodium, observed by R. W. Wood.^^ Afterwards J. Hartmann ^^
searched the spectrograms of the stars, and in those of the so-called
"hydrogen stars" he found a continuous band beyond the limit of
the Balmer series. This, be it noted, is the sign of ionization of
hydrogen atoms initially not in the normal, but in a certain excited
state. The continua beyond the principal series of the alkali metals,
however, are due to ionization of normal atoms. Those of sodium and
potassium were studied by Holtsmark; ^"^ those of caesium and rubid-
ium have been discerned (Harrison, I.e. infra); and the former two
were measured, that is to say the variation of absorption-coefficient
with frequency was measured, for sodium by G. R. Harrison ^^ and
B. Trumpy,^^ and for potassium by R. W. Ditchburn.^^
Obviously if the fundamental theory is correct, absorption is pro-
portional to ionization, and the curves representing the two as functions
of wavelength should coincide everywhere if scaled to coincide at any
one point; and measurements of either should make the other nugatory.
Unfortunately it is difficult to measure the absorption properly, perhaps
impossible to do it with anything like the precision feasible with the
other measurement.'-^ Harrison managed to get smooth absorption-
'^Phil. Mag., (6) 18, pp. 530-534 (1909).
^^Phys. ZS., 18, pp. 429-432 (1917).
^"^ Phys. Rev., 20, pp. 88-92 (1919).
^^Phys. Rev., (2) 24, pp. 466-477 (1924).
13 Z5./. Phys., 47, pp. 804-813 (1928).
2° Proc. Roy. Soc, 117, pp. 486-508 (1928).
^1 Mohler and his colleagues state that with an amount of ionization tenfold greater
than that which is observed with caesium at the series-limit, a stratum of the gas at
230° would have to be nitie metres deep to give a 50 per cent absorption.
CONTEMPORARY ADVANCES IN PHYSICS 355
curves (obtained of course by applying the densitometer to the spec-
trogram) with sodium. On the other hand, the experiences of Ditch-
burn with potassium are not encouraging. Not only did he have to
shoot a jet of rapidly-distilling vapor across the beam of light, but he
was obliged to swamp it in a vast excess of nitrogen — partly to keep the
metal from boiling away in a rush, partly it seems to prevent the vapor
from attacking the quartz windows. The curves are very crinkly,
and it is difficult to tell what share of the absorption should be credited
to molecules and what to atoms.
Nevertheless Ditchburn was able to deduce a value of the coefficient
k having the same order of magnitude — 10~^^ — as those which Mohler
and Boeckner had obtained with caesium and with rubidium when
they were measuring, not the disappearance of photons from the beam,
but the advent of ions in the gas. Mohler and Boeckner themselves
observed the absorption of light in caesium, and they found for k the
value 4-10~^^, — a good agreement, but they qualify it with the words
"subject to great uncertainty because of the low value of the total
absorption." Let it be pointed out in closing, that agreements such
as these are proof that in this region of the spectrum, photons ionize
when they are absorbed, and absorption is due to ionization. To
physicists familiar with the new atomic theories, this seems self-
evident, and scarcely worth the proving; but it is not self-evident,
and there was a time, not many years ago, when such a proof would
have been a sensational event.
Motion of Telephone Wires in Wind
By D. A. QUARLES
This paper deals with the position of equilibrium of a loop of wire in a
steady transverse wind and with the swinging of such a loop in one or more
gusts of wind. In the first part, the loop is assumed to be inelastic and to
swing as a rigid body. Under these conditions, nomograms are given from
which may be read the deflection of loops of wire .104" or .165" in diameter
as a function of steady wind velocity. The maximum additional swing of
such a loop with a single gust and with a succession of gusts of given peak
velocities may also be read from the nomograms. A chart is also included
giving the effect of wind velocity on the sag of .104" and .165" hard drawn
copper wires at tensions and span lengths common in the telephone plant.
UNTIL recent years, most of the important open wire toll circuits
of the Bell System had the two wires of a pair spaced 12 inches
apart. This wide spacing, with the consequent high mutual induct-
ance between the several pairs on a pole line, limited the use of the
lines for multiplex transmission with high frequency or "carrier"
currents. A reduction in the separation of the wires of a pair with
the retention of the present center-to-center spacing of the pairs was
one of the measures which offered the opportunity of increasing the
message carrying capacity of a pole line. The controlling factor in
limiting such a reduction in spacing was the hazard of the wires of a
pair swinging together in the wind thus interrupting or impairing
the transmitted messages.
About two years ago the 12-inch spacing was reduced to 8 inches in
some cases. This was considered to be as great a change as could be
safely taken from a mechanical point of view, based on the available
data. These data consisted in part of experiments made on an
experimental line and in part of an analysis of the performance of
certain working wires in the telephone plant which, for various reasons,
had been installed on a close spaced basis.
It was realized that if the wires of pairs could be placed even closer
together, materially lower crosstalk between the circuits would result,
thus increasing the circuit capacity of open wire lines, and therefore
effecting economies. Accordingly, a comprehensive investigation of
the wire spacing problem was begun. As some of the factors involved
in a theoretical determination of the chance of two parallel wires
swinging together in the wind were rather obscure and difficult of
evaluation, it was decided to attack the problem experimentally. A
field site was selected some distance from New York, where the terrain
and weather conditions were suitable for such an investigation, and
an experimental station was constructed and appropriately equipped.
356
MOTION OF TELEPHONE WIRES IN WIND
2,hl
Some time will be required, however, before definite conclusions can
be drawn from the experimental work of this new laboratory.
As an aid in the interpretation of the experimental results, certain
theoretical work has been done on the dynamics of a wire loop swinging
in the wind. It is this phase of the problem that is dealt with in this
article.
In the first part of this discussion, the wire loop is treated as an
inelastic, rigid body.^ As it was later found that under the conditions
applying in our problem there was a considerable increase in the sag of
the wire due to the wind, an investigation was made of the magnitude
of the correction required when the elasticity of the wire is taken into
account, the results of which are given in the latter part of this article.
Fig. 1.
Consider an element of the wire, shown in Fig. 1 in cross-section,
swinging about axis 0, at a radius y. The wind is assumed horizontal
and transverse to the axis. The sag a is also assumed small compared
with the span length so that to a sufficient approximation the length
of the wire is equal to the length of the span and the surface of the wire
opposing the wind is independent of the angle of deflection {a) of the
wire in the wind.
The velocity of the element of wire relative to axes fixed with respect
to the earth is ya. The wind velocity relative to the same coordinate
^ An article entitled "The Behavior of Overhead Transmission Lines in High Winds "
by Professor E. H. Lamb, which appeared in the October 1928 Journal of the Institu-
tion of Electrical Engineers, gives an analysis of the inelastic, rigid loop problem which
has been followed in general outline in the present treatment. There is disagreement,
however, with one of the fundamental assumptions upon which Professor Lamb's
analysis is based and our formulae are therefore generally at variance with those
derived in his article.
Mr. R. L. Peek, Jr. of Bell Telephone Laboratories, working independently, ar-
rived at results in agreement with those given in the present article.
24
358 BELL SYSTEM TECHNICAL JOURNAL
system is V and the wind velocity relative to the wire at any instant is
therefore the vector difference V — ya which has the magnitude
VF" + {yaY — IVya COS a.
It is assumed that the wind pressure against this element is propor-
tional to the square of this vector and acts along its direction. The
moment about the axis of the wind pressure on the element ds is
therefore given by:
^[F- + (3'a)" — ly'aV COS a\y cos ^ds,
where U is the ratio of wind pressure per unit length to square of veloc-
ity. Evaluating cos ^ and noting that ya is small compared with F,
this reduces to:
kdsy V^ cos a 1 — 4t ( cos a -\ i •
L F \ cos a / J
Putting
y =
and
ds = dx
and integrating, the total moment of wind pressure is
4 16
-kV^aC cos a — -r-EkVa^Ca (cos^ a -{- 1).
If the line through the supports is inclined to the horizontal by angle
7 this expression becomes:
4 16
-kV^aC cos a cos y — ■r-rkVa'^Ca (cos- a -\- 1) cos^ y
The dynamic equation for the motion of the loop then becomes:
a H (1 + COS" a)a -\- -r- s\n a = -; ,
m 4 a 4 ma cos y
where m is the mass of unit length of wire.
Static equilibrium is then given by:
tan a =
mg cos 7
Proceeding with the analysis, an equation is found for small motions
MOTION OF TELEPHONE WIRES IN WIND 359
about any position of equilibrium (deflection a) of the form
'(p -\- 2eip + n-ip = 0,
where
_ (t + cos^c^) kV
^ ~ 2m
and
7 £, «
W
4o cos a
For cases of practical interest in this investigation n^ > e^ and the
motion about equilibrium is periodic and of period
rp _ ^TT _ kfl cos a
where a is the sag in feet and g the acceleration of gravity in feet per
second per second. The ratio of the period of small oscillations about
equilibrium to the period when a is zero is given by T/Tq = Vcos a.
The damping as measured by the ratio of successive half swings, X,
is given by
log, X = , = - €
■\n^ — e^ n
If a wire, held at a deflection a by a steady wind F, is subjected to a
gust of wind having maximum velocity V\, the additional throw of the
wire will depend on the duration of the gust and may in general be
either greater than or less than the increase in steady deflection which
Vi, if sustained, would produce. The maximum throw will be given
by a gust of most favorable duration and /x,„g has been defined as the
ratio of this maximum throw to the increase in deflection that would
result if the peak velocity were sustained. Similarly, for a periodic
succession of gusts, there is a most favorable timing which in general
will produce displacements greater than would a wind which sustained
the velocity of the gust peaks. The ratio of the throws produced by a
most favorably timed succession of gusts to the increase in deflection
which would result if the peak velocity of the gusts were sustained, has
been defined as ^imp-
The formulae derived above have been applied to the practical
conditions of the telephone line problem,- where our interest is centered
in hard drawn copper wire, commonly of .104" or .165" diameter, with
spans ordinarily from 90 to 200 feet and sags commonly from 7" to 20"
2 This work was carried out in the Bell Telephone Laboratories by Mr. V.
Nekrassoff.
360 BELL SYSTEM TECHNICAL JOURNAL
though occasionally considerably greater. The method, which will
be described in more detail elsewhere, was to reduce the expressions for
wind pressure per unit length of wire, F, angular displacement a,
periods of small oscillations, T and Tq, damping constant X, and the
effects of single and periodic gusts, /i,„s and ^mp, to explicit functions of the
wind velocity in miles per hour, the diameter of the wire in inches, sag of
the wire in inches and trigonometric functions of the deflection of the
loop a and inclination of the loop 7. The factor k does not appear
directly in the equations, having been replaced by fractional powers of
wind velocity and wire diameter derived from the experimental results
of Relf.3
The following nomograms have been constructed by this method.
Nomogram No. 1 (Fig. 2) gives the steady deflection « of a span of wire
inclined to the horizontal at an angle 7 and the force in pounds per
linear foot of wire for a normal wind of velocity V. It also gives the
ratio of the period of small oscillations about the equilibrium position
to the natural period about the vertical position, this ratio depending
only on a. The actual value of the period in seconds may be read on
nomogram No. 2 (Fig. 3).
By the use of nomogram No. 3 (Fig. 4), the damping constant X, and
the gust ratios n,ns and ju,„p may be computed from the sag a, the wind
velocity V and the diameter D.
These nomograms in short give the numerical solution for our prob-
lem for wires of the two diameters assumed, namely .104" and .165".
Two major assumptions should be noted, first, that the wire loop
swings in a plane and second, that the wire is inelastic. The first
assumption has a certain justification in that each element of wire if
independent of adjacent elements would be in equilibrium in the same
deflected angle a as is found for the loop as a whole. Expressing this
in another way — if it be assumed that the wind is uniform along the
span there would be no forces, considering only first order effects, to
distort the loop out of a plane.
The second assumption is not so readily justified, in fact the sag of
the wire may be greatly affected by the wind pressure. The equili-
brium deflection a is, however, independent of the sag of the wire and is
found to be the same when the elasticity of the wire is taken into ac-
count as that derived for an inelastic wire.
Considering only the case where the line through the supports is
horizontal (7 = 0), we define 2r as the unstressed length of wire in the
loop and note that this may be either greater or less than the span
length 2c depending upon the tension at which the wire is suspended.
^ British Advisory Committee for Aeronautics — Report Xo. 102.
o
6
hK
.n(ni-o)c«)<t-<DO>o»o otftjui 1^
(\K^n<Ttf)«>r^r^a)(r rt i m
■nuitnininin>nininin«) io<t>io e
cibcidododood 666 6
«0 (fl r-
» « ffi
(o «e (0
6 6 6
n 00
- (\j
d d
d
n
(^
6
o
CO
d
o
o
CO
d
o
a>
6
'J
to
6
n
<o
CD
6
6
o
o
o>
6
CD
6
6
(0
•
6
Ok
6
o»
o>
o
o»
6
as.
ooiOinoom^r- no on CD in
ti>— (T>^nco>rt(M<D >nr\j (Mn — oo
— 0000^*^^*0 coaj h- p^ h- 0)
— rrroooooo ooo Q
ooodobo ocidd odd o
>n o <j>
in n CO
ggg
coo
•n oj
tn m
o O
d d
g
o
d
Q
o
t
o
d
in
3
6
n
<o
s
o
s
3
6
s
6
n
(0
rvi
Q
O
o
(0
Ot
O
6
«
6
o
ft)
P
o
<o
8
6
o
p
o
8
6
°«
-f -■j-j'^ J- ."^ ^ J? J- j^ -f
ijnnrvjw — o <r> o> a> t^ vi in ^
tn <\i —
10 (£1 to
O) tO
tn (A
1. 1
■A
,1
L
1
,1
t
1,
1
J
s
1 ,
n
1
n
J
II
M
1
J?
1
tn
d
L.
«
:?^?SSp£$S'8£|s"SSS 5 I S "^ I
1 1
1
1
1
1
■to
1
r
n
1
r
1
1
1
1
^1
or-^oo-ooo-no inooinotfi'O "O .n to n
©^••oooo^o-noi to « o> w iM n ij «o © -
••r-*Jt^TO(M-o (J> (0 r^ « to lO ■^ m i\i - —
S g
eO
(0
CO
O
d
(0
o
d
I-
o
d
s
o
d
n
o
d
Si
o
6
o
d
in
n
o
d
tn
o
d
lA
tn
o
6
o
8
o
6
doooodododd d d d d d d 6 6 d 6 d
d d
l-lH"
^-r-<»«*-««o>oo> oJiOh-onf^ n d) ■^ n
n,^rnirft«>*t^«» n^tntocDo. - ty v to
.n>i<n.ngiiAiAintno<p<eO(OiO(ptD (D r- <^ r~- r-
oooobbddddoddoodd 6 o d o 6
o o
m o
r-. (0
d d
o
<o
d
<o
d
lO
(O
d
6
o
o
<J>
d
<o
Ot
6
<o
d
d
ai
d
5
6
o
d
So o o O O o
^ lO to r^ <D (j>
nj (\1 fvi rvi r\j (\j (\1
(*) (O p) o n
o o O w
a> 01 o — Ai n ^
n fi -vi -4 •* -ir r
tW •* ■« V lo
ri' I ' I' I ' I ' I '
OO O O O O Q
>no o o o o o
r-i^ to t/X ^ m r\^
AJnjni N rg n] r\j
01 ____;:!— _^_ w cjf^^
I I ' l 'l I I | m l ) ll M i nu lllll|ll ^M lll|lll>ll^ M /ll > ll^l W | H ll 'n lll V^l^ll l l l|lll'l i^ li 'N' l l^ il'' ' ll /l 'l' ' l '''V l V | ' l ^^V l'<ll'l' / l n^ ll ^ l | 'll ^ l' ll 'll V l|l M l' l ll | lll l ^
O O OOOO OO
O O O (O O lO O lo
)0 "O ^rtf^NOJ —
+j a
— ^ r^
ti
Y
c
J ^^
I/)
s
WO ^
£ § vo
:- a; mLi
>\"
o ,>i o f<
.2 M>><
„ <3 o >
II
II II
— — :— ~ ^ — fi td
OfVKOOiOO r^ <M CO >o
-<
<nrt>00> ^ O t~ IB
omorf) — t^ w «)
©"louj — t» (\J ■
I'l'i'i ' t'J^t vti
O<M<0p to O
gg0>O<Vl («) >0
--<\I(\J IM CJ
rvi <D
innj9to«i*) (O V <D o*
i^-oio^ ^ >o o> to «
— — - — — — — <\i oj
-<
>ONoio>in<o < CO at
- — :— — — — «0 oj
ONiOO to O )^ <M <0
•ooionjnin <o a at
— — OJNOlNN N OJ
omou) — >^ w
■j Jj . T.T.T.Jv ,
(OfOioo> <t o r- S
0"'io>'i — r- N «>
-<
o(Mtf>p <n o *~
oQa>o Ai K) o <o
■A rviot a« i*> <0 ^ 40
^0(0^ ^ U) 01 <o
n^^to <D h- to o
■n
(VI
o
■<
■n
V
o
-<
7d«-« W O o> V
--fUNNN ni N
--0 — mietooi-— ntnr^fft— ^^p^ «j
^ 4 W 4 VW ^ to mS "i o mS |C o <b t^ 1^ i-^
(Vt « o
n n rt o
■•otoAi^ioono — <>J - _
s asa?s?s8sssg
^ ^ ^ O^ ^
r- a A o -
o o o —
t * * ^
i^r^^gs 4 O r> O
fu ni — o O'O'OiB*)* f«. h- ^. * "O « >«
— — — — oppopoooooo oo
p o o o o
a — V r^ <t\
N - — -
o o o o o o
4, '^ ' ^ ' ^ '' ^ y ^ s^^^ H£ ■ ■^Y ''■''' ' ^'^'^'/■''^ 'i ^ ^ ' 'i^ ' ' ii ''^^^^^V ^'^^l^f^^■'l h ^^l <^N^^^^^S^ ^ h' ^'^ i \ \ \ ^^ ^'^\^^ ^H \ \\ '\ 'h¥ ^ ^kM '**' i ^ ^ * ^^^^
(^niO * * o
OiOO>0 — f~
i00<0^ ^ «
t r> o fO 'H o>«Ai«)i/
(^ (>J CV — O 99i90a
. - - OOOOC
. . . lO (M O
OOOOOOOO
O O O O O O
N 0>
(\J04WN- - - —
SO(MU>0>Qr^^ON^tO«»0— N4«0 — (*) "> »* <y — '^
*oiniO<D<o«ior~r.r^r~f^K«icee29,etO<» «<» O p - -ji.3.__^_
rss
iO(\I^» rt lO
•f o>o^ ^ to
n*^«o • *^
lOON — t*) 0«ltO'')NOo>OOtOOOO
M«^«-« r--«>«No-o«to<n«'«
to t^ o — <n tt«i> — rtiot-9— ^h-o«
rt <S -4 4 ^ •e^^uiuitri uSui « to lb K K
o o o o o lo to o to o p o p 5 «i ^ 5 5 p
-q9\
MOTION OF TELEPHONE WIRES IN WIND
361
50
0.104" DIAMETER WIRE
0.165" DIAMETER WIRE
45
+ POINT WHERE FORCE
OF WIND EQUALS
WEIGHT OF WIRE
40
35
1
1
1
30
1
1
1
1
if)
1
1
1
1 1
I
u
1
1/
z
1
1
/
r
z
'^/l
$25
1
1
o°
//
1
1
^/
'
1/
1
/ 1
/
t
1
/
1
>
f
1
/
1
1
7
20
/
/
1
f/'
/
/
/
1
4
^
/
f
//
/
y
/
1
1
/',
/
_ *
-^
X
1
/
1
/
15
1
/,
'oV
/
J
1
"^h
1
/
/
1
1
7
/
/
/
;^
/
1
10
/
1
1
/
^^_
^
y
1
1
— f
1
J
x'
1
f
/
5
^
^
X
10
1 1 1 1 1 1 1 1 1
-0.104" DIAMETER WIRE —
-- ^ I
65
1
1
+ POINT WHERE FO
OF WIND EQUA
RTF
1
LS
1
WEIGHT OF WIRE
60
/
1
r
1
1
1
',/
/
f
S5
1
1
1
f
1
1
1
'
1
//
1
1
//
1
50
1
It
1
-f —
/
1
1
' oV
1
1 1
1
oV
f
//
1
*v
/
45
1
/
/
-i
1
/
/
/
/
1
f
/
/
/
1
/
/
/
1/
4
' +>
f
'
/
40
/
/
/
1
y
' /
/
^y
y
f-
1
o^l'
/
/ ^/l
1 ."/'
1
35
i
r,
1
1
/
/
1
/
1 oV
1
1
/-
^
f
//
'■7
30
/
1
' /
/
/
/
/
/
/
/
/
1
/
/
/
/
r
t f
/
/
/
/
/
/
^/
/
/
/
/
/
/
i^
.
1
PO
/
/
/
t
f
'2
+
t
15
7
1 /
J.
y
10
20 40 60 80 100
WIND VELOCITY IN MILES PER HOUR
20 40 60 80 100
WIND VELOCITY IN MILES PER HOUR
SPAN OF 130 FEET
Fig. 5-.4.
SPAN OF 200 FEET
Fig. 5-5.
MOTION OF TELEPHONE WIRES IN WIND
361
50
0.104" DIAMETER WIRE
0.165" DIAMETER WIRE
45
+ POINT WHERE FORCE
OF WIND EQUALS
WEIGHT OF WIRE
40
35
1
1
1
30
1
1
/
/
10
1
1
I
o
t
1
/
z
1
1
i
f
z
'^/l
o
< 2S
1
1
o°
I
1
t
in'-^
1
f
A?/
1
1
f
/
/
/
/
1
J
f
/
/
1
J
20
/
/
1
9'
f
/
1
1
'^A
1
f
,+
/
1
1
/
'/
/
I
n
/
^ ^>
y^
y
/
/
1
/
15
/
/<
%'/
/
/
' /
'^/
1
/
/
/
/
7
/ ,/
f ,
/
f
/
1
/
10
/
1
1
/
^ ^
^y'
y
/
1
f
-i
1
/
/
A
/
s
^^^
1 1 1 1 1 1 1 1
-\JA\J'\ UIMMtP LK WIHL j
65
I
1
+ POINT WHERE FORCE
OF WIND EQUALS
WEIGHT OF WIRE
1
1
1
60
1
—h
1
1
1
'
/
/
/
55
1
/
1
f —
1
1
//
/
1
//
/
50
1
1 1
f-
1
t
1
1
1
1
"7
■f —
/
1
1
o7
f
1
1
/y
i
/
45
1
/
1
1
/
1
/
1
1
f
/
/
1
1
/
/
1
1/
4
f
r
/
40
/
/
/
/
J
' /
r
/
f
/
,y
y
/
o'^y
/
/ - /'
/
/ ."/'
/
35
i
r,
J
1
y
1
1
/
' "V
1
1
/-
r
f
//
H
30
1
i
' 1
^ I
1
1
/
1
1
/
1
1
1
/
1
1
i
25
t /+
1
/
/
/
1
/
^/
1
/
/
/
1
/
__^
i^
,
1
?n
/
/
/
/
/
t
'J
+'
t
1 s
7
1 .^
^'
y
10
20 40 60 80 100
WIND VELOCITY IN MILES PER HOUR
20 40 60 80 100
WIND VELOCITY IN MILES PER HOUR
SPAN OF 130 FEET
Fig. 5-^.
SPAN OF 200 FEET
Fig. 5-B.
362
BELL SYSTEM TECHNICAL JOURNAL
If E is the modulus of elasticity in pounds per square inch cross-section,
D the diameter in inches, a the sag in feet and m the weight of wire per
linear foot, the approximate relationship^ is:
a' + — (c - r)a =
ttD^E
As only horizontal winds normal to the line of supports are being
considered, the wind pressure when the loop is in equilibrium is
horizontal. The weight of the wire being vertical the two forces add
at right angles, their resultant being the square root of the sum of their
squares. This resultant lies of course in the plane of equilibrium of
Fig. 6 — Test House and Line.
the loop. The wind pressure component is about equal to the gravity
component for a velocity of 38 m.p.h. in the case of .104" wire and about
47 m.p.h. in the case of .165" wire. The effective weight of the wire
under these conditions would be greater by a factor of V2 than the true
weight. In general, ni in the above formula is the effective weight of
the wire per unit length.
A wire having a sag of 5" in a 130' span with a temperature of — 10°
F. would have a sag of about 9" at 50° F. and about 16" at 100° F. due
to thermal expansion. The sag of such a wire would be increased by
wind pressure as shown in Fig. 5-A , the wind being given in true normal
velocity. The figure shows the increase to be most marked for low
temperatures and small diameters as would be expected. Similar
^ Due to Mr. J. A. Carr of Bell Telephone Laboratories.
MOTION OF TELEPHONE WIRES IN WIND 363
results are shown in Fig. S-B for a span of 200'. Both indicate the
marked increase of sag under not uncommon wind conditions.
While the above formula and charts give a fairly definite picture of
the effect of elasticity on the solution of the problem of static equili-
brium, the much more complex problem of the motion of an elastic
loop in a varying wind has not been attacked.^ The necessity for such
additional refinements can probably not be determined until the field
experiments above referred to have progressed to the point where
fairly comprehensive data are available for analysis and for a check
of the theoretical conclusions arrived at in this paper.
* An article by Karl Wolf in Zeitschrift f iir Angewandte Mathematik und Mechanik
of April 1927 treats certain aspects of the dynamics of an elastic loop, with particular
reference, however, to power lines. As yet, no attempt has been made to apply the
results of this work to our particular problems.
Economic Quality Control of Manufactured Product^
By W. A. SHEWHART
That we cannot make all pieces of a given kind of product identically alike
is accepted as a general truth. It follows that the qualities of jjieces of the
same kind of product differ among themselves, or, in other words, the quality
of product must be expected to vary. The causes of this variability are, in
general, unknown.
The present paper presents a scientific basis for determining when we
have gone as far as it is economically feasible to go in eliminating these un-
known or chance causes of variability in the quality of a product. When
this state has been reached, the product is said to be controlled because it is
then possible to set up limits within which the quality may be expected to
remain in the future. By securing control, we attain the five economic ad-
vantages discussed in Part III.
I Introduction
1. What is the Problem of Control?
WHAT is the problem involved in the control of quality of manu-
factured product? To answer this question, let us put our-
selves in the position of a manufacturer turning out millions of the same
kind of thing every year. Whether it be lead pencils, chewing gum,
bars of soap, telephones or automobiles, the problem is much the same.
He sets up a standard for the quality of his product and then tries to
make all pieces of product conform with this standard. Here his
troubles begin. For him standard quality is a bull's-eye, but like a
marksman shooting at such a target, he often misses. As is the case in
everything we do, unknown or chance causes exert their influence.
The problem then is: how much may the quality of a product vary and
yet be controlled? In other words, how much variation should we
leave to chance?
To make a thing the way we want to make it is one popular concep-
tion of control. We have been trying to do this for a good many years
and we see the fruition of this effort in the marvelous industrial develop-
ment around us. We have accepted the idea of applying scientific
principles but now a change is coming about in the principles them-
selves which necessitates a new concept of control.
A few years ago we were inclined to look forward to the time when a
manufacturer would be able to do just what he wanted to do. We
shared the enthusiasm of Pope when he said "All chance is but direction
thou canst not see," and we looked forward to the time when we would
see that direction. In other words, emphasis was laid on the exactness
^ Paper presented before A. A. A. S. on December 28, 1929, at Des Moines, Iowa.
364
ECONOMIC QUALITY CONTROL OF PRODUCT 365
of physical laws. Today, however, the emphasis is placed elsewhere as
is indicated by the following quotation from a recent issue, July, 1927,
of the journal Engineering:
"Today the mathematical physicist seems more and more inclined to the opinion
that each of the so-called laws of nature is essentially statistical, and that all our
equations and theories can do, is to provide us with a series of orbits of varying
probabilities."
The breakdown of the old orthodox scientific theory which formed
the basis of applied science in the past necessitates the introduction of
certain new concepts into industrial development. Along with this
change must come a revision in our ideas of such things as a controlled
product, an economic standard of quality and the method of detecting
lack of control or those variations which should not be left to chance.
Realizing, then, the statistical nature of modern science, it is but
logical for the manufacturer to turn his attention to the consideration
of available ways and means of handling statistical problems. The
necessity for doing this is pointed out in the recent book on the "Ap-
plication of Statistics in Mass Production," by Becker, Plant and
Runge. They say:
"It is therefore important to every technician who is dealing with problems of
manufacturing control to know the laws of statistics and to be able to apply them
correctly to his problems."
Another German writer, K. H. Daeves, writing on somewhat the same
subject says:
"Statistical research is a logical method for the control of operations, for the re-
search engineer, the plant superintendent, and the production executive."
This statement is of particular interest because its author has for
several years been associated with the application of statistical methods
in the steel industry.
The problem of control viewed from this angle is a comparatively new
one. In fact, very little has been written on the subject. Progress in
modifying our concept of control has been and will be comparatively
slow. In the first place, it requires the application of certain modern
physical concepts and in the second place, it requires the application of
statistical methods which up to the present time have been for the most
part left undisturbed in the journals in which they appeared. This
situation is admirably summed up by the magazine Nature of
January, 1926, as follows:
"A large amount of work has been done in developing statistical methods on the
scientific side, and it is natural for any one interested in science to hope that all this
work may be utilized in commerce and industry. There are signs that such a move-
ment has started, and it would be unfortunate indeed if those responsible in practical
affairs fail to take advantage of the improved statistical machinery now available."
366 BELL SYSTEM TECHNICAL JOURNAL
2. Object
The object of this paper is the presentation of a scientific basis for
interpreting the significance of chance variations in quality of product
and for eliminating causes of variability which need not be left to
chance, making possible more uniform quality and thereby effecting
certain economies.
3. Nature of Control
Let us consider a very simple example of our inability to do exactly
what we want to do and thereby illustrate two characteristics of a
controlled product.
Write the letter a on a piece of paper. Now make another a just like
the first one; then another and another until you have a series of a's,
a, a, a, a, . . . . You try to make all the a's alike but you don't; you
can't. You are willing to accept this as an empirically established
fact. But what of it? Let us see just what this means in respect to
control. Why can we not do a simple thing like making all the a's just
alike? Your answer leads to a generalization which all of us are per-
haps willing to accept. It is that there are many causes of variability
among the a's: the paper was not smooth, the lead in the pencil was not
uniform and the unavoidable variability in your external surroundings
reacted upon you to introduce variations in the a's. But are these the
only causes of variability in the a's? Probably not.
We accept our human limitations and say that likely there are many
other factors. If we could but name all the reasons why we cannot
make the a's alike, we would most assuredly have a better understand-
ing of a certain part of nature than we now have. Of course this
conception of what it means to be able to do what we want to do is not
new; it does not belong exclusively to any one field of human thought;
it is a commonly accepted conception.
The point to be made in this simple illustration is that we are limited
in doing what we want to do ; that to do what we set out to do, even in so
simple a thing as making a's that are alike requires almost infinite
knowledge compared with that which we now possess. It follows,
therefore, since we are thus willing to accept as axiomatic that we
cannot do what we want to do and that we cannot hope to understand
why we cannot, that we must also accept as axiomatic that a controlled
quality will not be a constant quality. Instead a controlled quality
must be a variable quality. This is the first characteristic.
But let us go back to the results of the experiment on the a's and we
shall find out something more about control. Your a's are different
from my a's; there is something about your a's which makes them yours
ECONOMIC QUALITY CONTROL OF PRODUCT 367
and something about my a's that makes them mine. True, not all of
your a's are alike. Neither are all of my a's alike. Each group of a's
varies within a certain range and yet each group is distinguishable from
the others. This distinguishable and, as it were, constant variability
within limits is the second characteristic of control.
4. Definition of Control
For our present purpose a phenomenon will be said to be controlled
when, through the use of past experience, we can predict, at least
within limits, how the phenomenon will be expected to vary in the
future. Here it is understood that prediction within limits means that
we can state, at least approximately, the probability that the observed
phenomenon will fall within the given limits.
In this sense the time of the eclipse of the sun is a predictable
phenomenon. So also is the distance covered in successive intervals
of time by a freely falling body. In fact, the prediction in such cases is
extremely precise. It is an entirely different matter, however, to
predict the expected length of life of an individual at a given age; the
velocity of a molecule at a given instant of time ; the breaking strength
of a steel wire of known cross section; or numerous other phenomena
of like character. In fact, a prediction of the type illustrated by fore-
casting the time of an eclipse of the sun is almost the exception rather
than the rule in scientific and industrial work.
In all forms of prediction an element of chance enters. The specific
problem which concerns us at the present moment is the formulation of
a scientific basis for prediction, taking into account the element of
chance, where, for the purpose of our discussion, any unknown cause of
a phenomenon will be termed a chance cause.
II. Scientific Basis for Control
1. Three Important Postulates
What can we say about the future behavior of a phenomenon act-
ing under the influence of unknown or chance causes? I doubt that,
in general, we can say anything. For example, let me ask: "What will
be the price of your favorite stock thirty years from today?" Are you
willing to gamble much on your powers of prediction in such a case?
Probably not. However, if I ask: "Suppose you were to toss a penny
one hundred times, thirty years from today, what proportion of heads
would you expect to find?," your willingness to gamble on your powers
of prediction would be of an entirely difterent order than in the previous
case.
368
BELL SYSTEM TECHNICAL JOURNAL
The recognized difference between these two situations leads us to
make the following simple postulate:
Postulate 1. All chance systems of causes are not alike in the
sense that they enable us to predict the future in terms of the past.
Hence, if we are to be able to predict the quality of product at least
within limits, we must find some criterion to apply to observed vari-
ability in quality to determine whether or not the cause system pro-
ducing it is such as to make possible future predictions.
Perhaps the natural course to follow is to glean what we can about
the workings of unknown chance causes which are generally acknowl-
edged to be controlled in the sense that they permit of prediction within
limits. Perhaps no better examples could be considered than those
which influence length of human life and molecular motion, for it often
appears that nothing is more uncertain than life itself, unless perhaps
it be molecular motion. Yet there is something certain about these
uncertainties. In the assumed laws of mortality and distribution of
molecular displacement, we find some of the essential characteristics
of control within limits.
A . Law of Mortality
The date of death always has seemed to be fixed by chance even
though great human effort has been expended in trying to rob chance
UJ
O
<
UJ
>
«>-
t^
< CD
^m
±o
z
o
I-
u
<
a.
45 50
AGE IN YEARS
Fig. 1— Law of mortality — ^law of fluctuations controlled within limits.
of this prerogative. We come into this world and from that very in-
stant on are surrounded by causes of death seeking our life. Who
knows whether or not death will overtake us within the next year?
ECONOMIC QUALITY CONTROL OF PRODUCT 369
If SO, what will be the cause? These questions we cannot answer.
Some of us are to fall at one time from one cause, others at another
time from another cause. In this fight for life we see then the element
of uncertainty and the interplay of numerous unknown or chance
causes.
However, when we study the effect of these chance causes in produc-
ing deaths in large groups of individuals, we find some indication of a
controlled condition. We find that this hidden host of causes produce
deaths at an average rate which does not differ much over long periods
of time. From such observations we are led to believe that, as we
approach the condition of homogeneity of population and surroundings,
we approach what is customarily termed a "Law of mortality" such as
indicated schematically in Fig. 1. In other words, we believe that in
the limiting case of homogeneity the causes of death function so as to
make the probability, let us call it dy, of dying within given age limits,
such as forty-five to fifty, constant: That is, we believe these causes are
controlled. In other words, we assume the existence of a kind of
statistical equilibrium among the effects of such an unknown system
of chance causes expressable in the assumption that the probability of
dying within a given age limit, under the assumed conditions, is
an objective and constant reality.
B. Molecular Motion
Just about a century ago, in 1827 to be exact, an English botanist.
Brown, saw something through his microscope that caught his interest.
It was motion going on among the suspended particles almost as though
they were alive. In a way it resembled the dance of dust particles in
sunlight, so familiar to us, but this dance differed from that of the dust
particles in important respects — for example, adjacent particles seen
under the microscope did not necessarily move in even approximately
the same direction, as do adjacent dust particles suspended in the air.
Watch such motion for several minutes. So long as the temperature
remains constant, there is no change. Watch it for hours, the motion
remains characteristically the same. Watch it for days, we see no
difference. Even particles suspended in liquids enclosed in quartz
crystals for thousands of years show exactly the same kind of motion.
Therefore, to the best of our knowledge there is remarkable permanence
to this motion. Its characteristics remain constant. Here we cer-
tainly find a remarkable degree of constancy exhibited by a chance
system of causes.
Suppose we follow the motion of one particle to get a better picture
of this constancy. This has been done for us by several investigators,
370
BELL SYSTEM TECHNICAL JOURNAL
notably Perrin. In such an experiment he noted the position of a
particle at the end of equal intervals of time, Fig. 2. He found that
Fig. 2 — ^A close-up of molecular motion appearing absolutely irregular, yet controlled
within limits.
the direction of this motion observed in one interval differed, in general,
from that in the next succeeding interval. He found that the direction
of the motion presents what we instinctively call absolute irregularity.
Let us ask ourselves certain questions about this motion.
Suppose we fix our attention on the particle at the point A. What
made it move to B in the next interval of time? Of course we answer
ECONOMIC QUALITY CONTROL OF PRODUCT 371
by saying that a particle moves at a given instant in a given direction,
say AB, because the resultant force of the molecules hitting it in a plane
perpendicular to this direction from the side away from B is greater
than that on the side toward B ; but at any given instant of time there
is no way of telling what molecules are engaged in giving it such mo-
tion. We do not even know how many molecules are taking part.
Do what we will, so long as the temperature is kept constant, we can-
not change this motion in a given system. It cannot be said, for ex-
ample, when the particle is at the point B that during the next interval
of time it will move to C. We can do nothing to control the motion
in the matter of displacement or in the matter of the direction of this
displacement.
Let us consider either the x or y components of the segments of the
paths. Within recent years we find abundant evidence indicating that
these displacements appear to be distributed about zero in accord with
what is called the normal law. That is to say, if x represents the
deviation from the mean displacement, zero in this case, the probability
dy of X lying within the range x to x -\- dx is given by
dy = ^^e-^^'l^'^'Hx, a)
where a is the root mean square deviation.
Such evidence as that provided by the law of mortality and the law
of distribution of molecular displacements leads us to assume that there
exist in nature phenomena controlled by systems of chance causes such
that the probability dy of the magnitude X of a characteristic of some
such phenomenon falling within the interval X to X + dX is express-
able as a function / of the quantity X and certain parameters repre-
sented symbolically in the equation
dy=f(X,K\,, ■■■,\J,dX, (2)
where the X's denote the parameters. Such a system of causes we
shall term constant because the probability dy is independent of time.
W^e shall take as our second postulate:
Postulate Z — Constant systems of chance causes do exist in
nature.
To say that such systems of causes exist in nature, however, is one
thing; to say that such systems of causes exist in a production process
is quite another thing. Less than ten years ago it seemed reasonable
to assume that such systems of causes existed in the production of
telephone equipment. Today we have abundant evidence of their
372
BELL SYSTEM TECHNICAL JOURNAL
existence. The practical situation, however, is that in the majority
of cases there are unknown causes of variability in the quality of a
product which do not belong to a constant system. This fact was dis-
covered very early in the development of control methods, and these
causes were called assignable. The question naturally arose as to
whether it was possible, in general, to find and eliminate causes of
variability which did not form a part of a constant system. Less than
ten years ago it seemed reasonable to assume that this could be done.
Today we have abundant evidence to justify this assumption. We
shall, therefore, adopt as our third postulate:
Postulate 3 — Assignable causes of variation may be found and
eliminated.
Hence, to secure control, the manufacturer must seek to find and
eliminate assignable causes. In practice, however, he has the difiiculty
of judging from an observed set of data, whether or not assignable
causes are present. A simple illustration will make this point clear.
2. When Do Fluctuations Indicate Trouble?
In many instances the quality of the product is measured by the
fraction non-conforming to engineering specifications or as we say the
fraction defective. Table 1 gives for a period of 12 months the ob-
TABLE 1
Apparatus Type A
Apparatus Type B
Month
n
No.
Insp.
Hi
No.
Def.
p = mln
Fraction
Def.
Month
n
No.
Insp.
No.
Def.
p = n\ln
Fraction
Def.
Jan
Feb
Mar
Apr
May
June
July
Aug
Sept
Oct
Nov
Dec
527
610
428
400
498
500
395
393
625
465
446
510
4
5
5
2
15
3
3
2
3
13
5
3
.0076
.0082
.0017
.0050
.0301
.0060
.0076
.0051
.0058
.0280
.0112
.0059
Jan
Feb
Mar
Apr
May
June
July
Aug
Sept
Oct
Nov
Dec
169
99
208
196
132
89
167
200
171
122
107
132
1
3
1
1
1
1
1
1
2
1
3
1
.0059
.0303
.0048
.0051
.0076
.0112
.0060
.0050
.0117
.0082
.0280
.0076
Average ....
483.08
5.25
.0109
149.33
1.42
.0095
served fluctuations in this fraction for two kinds of product designated
here as Type A and Type B. For each month we have the sample
size n, the number defective Wi and the fraction p = ni/n. We can
ECONOMIC QUALITY CONTROL OF PRODUCT
373
better visualize the extent of these fluctuations in fraction defective by
plotting the data as in Fig. 3-a and Fig. 3-b.
0.04
0.03 -
UJ
>
o
UJ
^. 0.02
z
o
H
o
<
DC
0.01
I t
• • •
. • • •
0.041-
a 0.03
>
H-
O
UJ
u.
UJ 0.02
Q
Z
o
1-
o
<
0.01
M
M
J J A
MONTHS
O
J L
M
M J J A S
MONTHS
N
J L
Fig. 3 — Should these variations be left to chance?
a. Apparatus Type A.
b. Apparatus Type B.
What we need is some yardstick to detect in such variations any
evidence of the presence of assignable causes. Can we find such a
25
374
BELL SYSTEM TECHNICAL JOURNAL
yardstick? Experience of the kind soon to be considered indicates an
affirmative answer. It leads us to conclude that it is feasible to es-
tablish criteria useful in detecting the presence of assignable causes of
0-04r
£;0.03
UJ
>
u
UJ
LI.
^ 0.02
z
o
\-
o
0.04 r
a 0.03
UJ
>
I-
o
liJ
lii 0.02
o
z
o
I-
o
<
a.
u.
0.01
J L
« t
J I I I I L
J FMAMJ JASOND
MONTHS
M
M
J J A
MONTHS
Fig. 4 — Should these variations be left to chance?
a. Xo.
b. Yes.
variation or, in other words, criteria which when applied to a set of
observed values will indicate whether or not it is reasonable to believe
ECONOMIC QUALITY CONTROL OF PRODUCT 375
that the causes of variabiHty should be left to chance. Such criteria
are basic to any method of securing control within limits. Let us,
therefore, consider them critically. It is too much to expect that the
criteria will be infallible. We are amply rewarded if they appear to
to work in the majority of cases.
Generally speaking, the criteria are' of the nature of limits derived
from past experience showing within what range the fluctuations in
quality should remain, provided they are to be left to chance. For
example, when such limits are placed on the fluctuations in the qualities
shown in Fig. 3, we find (see Fig. 4) that in one case two points fall out-
side the limits and in the other case no point falls outside the limits.
Upon the basis of the use of such limits, we look for trouble in the form
V
H
_J
<
a
O
z
o
I-
o
z
D
u.
Ul
1
O
in
• • •
ONLY SUCH VARIATIONS •SHOULD BE « LEFT* TO , CHANCE
AS TIME GOES ON
Fig. 5 — Art plus modern statistical machinery makes possible the establishment of
such limits.
of assignable causes in one case but not in the other. However, to be
of really practical interest, we should be able to answer the following
question: Can we expect to be able to find and eliminate causes of
variability only when deviations fall outside the limits? First, let us
see what statistical theory has to say in answer to this question.
Upon the basis of postulate 3, it follows that we can find and remove
causes of variability until the remaining system of causes is constant
or until we reach that state where the probability that the deviations in
quality remain within any two fixed limits (Fig. 5) is constant. How-
ever, this assumption alone does not tell us that there are certain limits
within which all observed values of quality should remain provided the
causes cannot be found and eliminated. In fact so long as the limits
are set so that the probability of falling within the limits is less than
376
BELL SYSTEM TECHNICAL JOURNAL
unity, we may alweiys expect a certain percentage of observations to
fall outside the limits even though the system of causes be constant.
In other words, the acceptance of this assumption gives us a right to
believe that there is an objective state of control within limits but
in itself it does not furnish the practical criterion for determining when
variations in quality, such as given in Fig. 3, should be left to chance.
Furthermore, we may say that mathematical statistics as such does
not give us the desired criterion. What does this situation mean in
plain every day engineering English? Simply this: such criteria, if
they exist, cannot be shown to exist by any theorizing alone, no matter
how well equipped the theorist is in respect to probability or statistical
theory. We see in this situation the long recognized dividing line
4.0
3.5
UJ
3.0
>
1-
O
UJ
2.5
a.
UJ
Q
2.0
1-
Z
o
1.5
a.
Q.
1.0
0.5
AVERAGE
AVERAGE
AVERAGE
1923 - 1924
1925
YEAR
1926 (9 months)
Fig. 6 — Evidence of improvement in quality with approach to control.
between theory and practice. The available statistical machinery
referred to by the magazine Nature is, as we might expect, not an
end in itself but merely a means to an end. In other words, the fact
that the criterion which we happen to use has a fine ancestry of high-
brow statistical theorems does not justify its use. Such justification
must come from empirical evidence that it works. As the practical
engineer might say, the proof of the pudding is in the eating. Let us
therefore look for the proof.
3. Evidence that Criteria Exist for Detecting Assignable Causes
A . Fig. 6 shows the results of one of the first large scale experiments to
determine whether or not indications given by such a criterion applied to
quality measured in terms of fraction defective were justified by experi-
ECONOMIC QUALITY CONTROL OF PRODUCT
377
ence. About thirty typical items used in the telephone plant and pro-
duced in lots running into the millions per year were made the basis for
this study. As shown in this figure during 1923-24, these items showed
68 per cent control about a relatively low average of 1.4 per cent defec-
tive.^ However, as the assignable causes indicated by deviations in
the observed monthly fraction defective falling outside of control
limits were found and eliminated, the quality of product approached
the state of control as indicated by an increase of from 68 per cent to
84 per cent control by the latter part of 1926. At the same time the
quality improved; in 1923-24 the average per cent defective was 1.4
per cent whereas by 1926 this had been reduced to .8 per cent. Here
we get some typical evidence that, in general, as the assignable causes
are removed, the variations tend to fall more nearly within the limits as
indicated by an increase from 68 per cent to 84 per cent. Such evi-
dence is, of course, one sided. It shows that when points fall outside
the limits, experience indicates that we can find assignable causes, but
it does not indicate that when points fall within such limits, we cannot
find causes of variability. However, this kind of evidence is provided
by the following two typical illustrations.
TABLE 2
Electrical Resistance of Insulations in Megohms,
Should Such Variations be Left to Chance?
5045
4635
4700
4650
4640
3940
4570
4560
4450
4500
5075
4500
4350
5100
4600
4170
4335
3700
4570
3075
4450
4770
4925
4850
4350
5450
4110
4255
5000
3650
4855
2965
4850
5150
5075
4930
3975
4635
4410
4170
4615
4445
4160
4080
4450
4850
4925
4700
4290
4720
4180
4375
4215
4000
4325
4080
3635
4700
5250
4890
4430
4810
4790
4175
4275
4845
4125
4425
3635
5000
4915
4625
4485
4565
4790
4550
4275
5000
4100
4300
3635
5000
5600
4425
4285
4410
4340
4450
5000
4560
4340
4430
3900
5000
5075
4135
3980
4065
4895
2855
4615
4700
4575
4840
4340
4700
4450
4190
3925
4565
5750
2920
4735
4310
3875
4840
4340
4500
4215
4080
3645
4190
4740
4375
4215
4310
4050
4310
3665
4840
4325
3690
3760
4725
5000
4375
4700
5000
4050
4185
3775
5075
4665
5050
3300
4640
4895
4355
4700
4575
4685
4570
5000
5000
4615
4625
3685
4640
4255
4090
4700
4700
4685
4700
4850
4770
4615
5150
3463
4895
4170
5000
4700
4430
4430
4440
4775
4570
4500
5250
5200
4790
3850
4335
4095
4850
4300
4850
4500
4925
4765
5000
5100
4845
4445
5000
4095
4850
4690
4125
4770
4775
4500
5000
B. In the production of a certain kind of equipment, considerable
cost was involved in securing the necessary electrical insulation by
means of materials previously used for that purpose. A research pro-
gram was started to secure a cheaper material. After a long series of
preliminary experiments, a tentative substitute was chosen and an
1 Jones, R. L., "Quality of Telephone Materials," Bell Telephone Quarterly, June,
1927.
378
BELL SYSTEM TECHNICAL JOURNAL
extensive series of tests of insulation resistance were made on this
material, care being taken to eliminate all known causes of variability.
Table 2 gives the results of 204 observations of resistance in megohms
taken on as many samples of the proposed substitute material.
Reading from top to bottom beginning at the left column and con-
tinuing throughout the table gives the order in which the observations
were made. The question is: "Should such variations be left to
chance?"
No a priori reason existed for believing that the measurements form-
ing one portion of this series should be different from those in any other
portion. In other words, there was no rational basis for dividing the
SHOULD THESE VARIATIONS BE LEFT TO CHANCE?
52
10
50
5
I
O
O 48
UJ
2 46
UJ
O 44
Z
<n 42
10
"^ 40
o
<
[£ 38
UJ
>
34
•• •
•-•
-•-s?
• •
52
■
(0 50
?
I
g48
-
UJ
5
46
.
Z
••
UJ ^ >
• •« •
z
^42
in
Ul
a. 40
1-
UJ
o
< 3fl
—
a.
UJ
>
< 36
~
34
1 1 1 1
10 20 30
SAMPLE
ANSWER: "NO"
40
50
10
SAMPLE
ANSWER; "YES"
20
Fig. 7.
total set of data into groups of a given number of observations except
that it was reasonable to believe that the system of causes might have
changed from day to day as a result of changes in such things as atmos-
pheric conditions, observers, and materials. In general, if such a
change is to take place, we may readily detect its effect provided we
divide the total number of observations into comparatively small sub-
groups. In this particular instance, the size of the sub-group was taken
as four and the black dots in Fig. 1-a show the successive averages of
four observations in the order in which they were taken. The
dotted lines are the limits within which experience has shown that
these observations should fall, taking into account the size of the sam-
ECONOMIC QUALITY CONTROL OF PRODUCT
379
pie, provided the variability should be left to chance. Several of the
observed values lie outside these limits. This was taken as an indica-
tion of the existence of causes of variability which could be found and
eliminated.
Further research was instituted at this point to find these causes of
variability. Several were found and after these had been eliminated,
another series of observed values gave the results indicated in Fig. 1-b.
Here we see that all of the points lie within the limits. We assumed,
therefore, upon the basis of this test, that it was not feasible for
research to go much further in eliminating causes of variability.
Because of the importance of this particular experiment, however,
48.0
o
X
a>
a.
o
47.8
O
O
<
cr
tu
>
<
47.6
L
4 8 12 16
SAMPLE
Fig. 8 — \'ariations that should be left to chance. Does the criterion work? "Yes."
considerably more work was done, but it failed to reveal causes of
variability. Here then is a typical case where the criterion indicates
when variability should be left to chance.
C. Suppose now that we take another illustration where it is reason-
able to believe that almost everything humanly possible has been done
to remove the assignable causes of variation in a set of data. Perhaps
the outstanding series of observations of this type is that given by
Millikan in his famous measurement of the charge on an electron.
Treating his data in a manner similar to that indicated above, we get
the results shown in Fig. 8. All of the points are within the dotted
limits. Hence the indication of the test is consistent with the ac-
cepted conclusion that those factors which need not be left to chance
had been eliminated before this particular set of data were taken.
380 BELL SYSTEM TECHNICAL JOURNAL
4. Role Played by Statistical Theory
It may appear thus far that mathematical statistics plays a relatively
minor role in laying a basis for economic control of quality. Such,
however, is not the case. In fact, a central concept in engineering work
of today is that almost every physical property is a statistical distribu-
tion. In other words, an observed set of data constitutes a sample of
the effects of unknown chance causes. 1 1 is at once apparent, therefore,
that sampling theory should prove a valuable tool in testing engineering
60 r
50 60 70 80 90 100
MODULUS OF RUPTURE IN 100 POUNDS PER SQUARE INCH
Fig. 9 — Variability in modulus of rupture of clear specimens of green sitka spruce
typical of the statistical nature of physical properties
hypotheses. Here it is that much of the most recent mathematical
theory becomes of value particularly in analysis involving the use of
comparatively small numbers of observations.
Let us consider, for example, some property such as the tensile
strength of a material. Provided our previous assumptions are justi-
fied, it follows that after we have done everything we can to eliminate
assignable causes of variation, there will still remain a certain amount
of variability exhibiting the state of control. Let us consider an ex-
tensive series of data recently published by a member of the Forest
Products Laboratories'- (Fig. 9). Here we have the results of tests for
tensile strength on L304 small test specimens of sitka spruce, the kind
- Xewlin, J. A., Proceedings of the American Societv of Civil Engineers, September,
1926, pp. 1436-1443.
ECONOMIC QUALITY CONTROL OF PRODUCT 381
of material used in aeroplane propellers during the war. The wide
variability is certainly striking. The smooth solid curve is an approx-
imation to the distribution function for this particular property repre-
senting at least approximately a state of control. The importance of
going from the sample to the smooth distribution is at once apparent
and in this case a comparatively small amount of refinement in statisti-
cal machinery is required.
Suppose, however, that instead of more than a thousand measure-
ments we had only a very small number, such as is so often the case in
engineering work. Our estimation of the variability of the distribution
function, representing the state of control, upon the basis of the inform-
ation given by the sample would necessarily be quite different from that
ordinarily used by engineers (see Fig. 10). This is true even though
i.Sr
Z 1.6
o
1-
<
>
LU
Q
gl.4
<
Z
<
I-
(/)
ti.
O 1.2
UJ
I-
<
\-
LJ
LOh
INEW
OLD
i- 1 ' 1 1 1 ^ L_^ 1 1 \ I I I 1 I I I I
3 5 10 20 50 100
n
Fig. lO^Correction factors made possible by modern statistical theorj' are often
large. — ^Typical Illustration.
we make the same kind of assumption to begin with as engineers have
been accustomed to do in the past. This we may take as a typical
example of the fact that the production engineer finds it to his advan-
tage to keep abreast of the developments in statistical theory. Here
we use new in the sense that much of modern statistical machinery is
new to most engineers.
382
BELL SYSTEM TECHNICAL JOURNAL
5 . Conclusion
Based upon evidence such as already presented, it appears to be
practicable to set up criteria by which to determine when assignable
causes of variations in quality have been eliminated so that the product
may then be considered to be controlled within limits. This state of
control appears to be, in general, a kind of limit to which we may expect
to go economically in finding and removing causes of variability without
changing a major portion of the manufacturing process as, for example,
would be involved in the substitution of new materials or designs.
III. Advantages Secured through Control
1. Reduction in the Cost of Inspection
If we can be assured that something we use is produced under con-
trolled conditions, we do not feel the need for inspecting it as much as
fi,6
A S O N
1927
M J J /
1928
MONTHS
S O N D J F
MAM
1929
0.6
<-0,2
_1_
_!_
JASON
1927
_1_
_1_
J_
_1_
_1_
_1_
MAMJ JASON
1926
MONTHS
F M A M
1929
Fig. 11 — Approach to stable equilibrium or control as assignable causes are weeded
out, thus reducing the need for inspection.
we would if we did not have this assurance. For example, we do not
waste our money on doctors' bills so long as we are willing to attribute
the variability in our health to the effects of what in our present termin-
ology corresponds to a constant system of chance causes.
In the early stages of production there are usually causes of varia-
bility which must be weeded out through the process of inspection. As
ECONOMIC QUALITY CONTROL OF PRODUCT 383
we proceed to eliminate assignable causes, the quality of product usu-
ally approaches a state of stable equilibrium somewhat after the man-
ner of the two specific illustrations presented in Fig. 11. In both
instances, the record goes back for more than two years and the process
of elimination in each case covers a period of more than a year.
It is evident that as the quality approaches what appears to be a
comparatively stable state, the need for inspection is reduced.
2. Reduction in the Cost of Rejections
That we may better visualize the economic significance of control,
we shall now view the production process as a whole. We take as a
specific illustration the manufacture of telephone equipment. Picture,
if you will, the twenty or more raw materials such as gold, platinum,
silver, copper, tin, lead, wool, rubber, silk, and so forth, literally col-
lected from the four corners of the earth and poured into the manu-
facturing process. The telephone instrument as it emerges at the end
of the production process is not so simple as it looks. In it there are
201 parts, and in the line and equipment making possible the connec-
tion of one telephone to another, there are approximately 110,000 more
parts. The annual production of most of these parts runs into the
millions so that the total annual production of parts runs into the
billions.
How shall the production process for such a complicated mechanism
be engineered so as to secure the economies of quantity production and
at the same time a finished product with quality characteristics lying
within specified tolerances? One such scheme is illustrated in Fig. 12.
Here the manufacturing process is indicated schematically as a funnel,
at the small end of which we have the 100 per cent inspection screen to
protect the consumer by assuring that the quality of the finished
product is satisfactory. Obviously, however, it is often more econom-
ical to throw out defective material at some of the initial stages in
production rather than to let it pass on to the final stage where it would
likely cause the rejection of a finished unit of product. For example,
we see to the right of the funnel, piles of defectives, which must be
junked or reclaimed at considerable cost.
It may be shown theoretically that, by eliminating assignable causes
of variability, we arrive at a limit to which it is feasible to go in reducing
the fraction defective. It must sufiice here to call attention to the kind
of evidence indicating that this limiting situation is actually approached
in practice as we remove the assignable causes of variability.
Let us refer to the information given in Fig. 6 which is particularly
significant because it represents the results of a large scale experiment
384
BELL SYSTEM TECHNICAL JOURNAL
carried on under commercial conditions. As the black sectors in the
pie charts decrease in size, indicating progress in the removal of as-
signable causes, we find simultaneously a decrease in the average frac-
RAW MATERIAL
INSPECTION
TO REDUCE COST
OF PRODUCTION
PARTS
PARTS
too <Vb INSPECTION
TO
PROTECT CONSUMER
PARTIAL ASSEMBLIES
FINISHED PRODUCT
GOOD
UNITS
Ro] TELEPHONE
AND
LINE
Fig. 12 — An economic production scheme.
tion defective from .014 to .008. Here we see how control works to
reduce the amount of defective material. However, this is such an
important point that it is perhaps interesting to consider an illustration
from outside the telephone field.
Recent work of the Food Research Institute of Stanford University
shows that the loss from stale bread constitutes an important item of
cost for a great number of wholesale as well as some retail bakeries.
They estimate that this factor alone costs people of the United States
millions of dollars per year. The sales manager of every baking cor-
ECONOMIC QUALITY CONTROL OF PRODUCT
385
poration is interested, therefore, in detecting and finding assignable
causes of variation in the returns of stale bread provided that by so
doing he may reduce to a minimum the loss arising in this way.
6.63
6.16
3.81
uj 4.52
_l
<
a.
o
I 99
UJ
>
t-
O
UJ
u.
LU
Q 6.20
I-
Z
UJ
o
a.
UJ
°- 11.70
• •
,••
• . • • • •
^ • • * *»• •
• ••
••••
• •
' • •
■ •
-— — w-r*-
— « >« ' — •^-tt"* «-• ^. — "
vx=
• • •
BAKERY I
BAKERY 2
BAKERY 3
BAKERY 4
BAKERY 5
BAKERY 6
• •
••-. ♦.
" • 9 • •» »-^» = BAKERY 7
• •
3.78
4.90
4.82
•• • • •••
^•^«
• . •
• • •^
=F5^
• •
• •• • •
• - _- . • • •
•• • •
BAKERY 8
BAKERY 9
••
■ ,*> ■
• •
• •
= BAKERY 10
I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I
5 10 15 20 25 30 35 40
WEEKS
Fig. 13 — Results showing how control effects a reduction in the cost of rejections.
Some time ago it became possible to secure the weekly record of
return of stale bread for ten different bakeries operating in a certain
metropolitan district. These observed results are shown graphically
in Fig. 13. At once we see that there is a definite lack of control on the
386 BELL SYSTEM TECHNICAL JOURNAL
part of each bakery. The important thing for us to note, however, is
that the bakery having the h^west percentage return, 1.99 per cent, also
shows better control than the other bakeries as judged by the number
of points falling outside the control limits in the period of 36 weeks.
3. Attainment of Maximum Benefits from Quantity Production
The quality of the finished product depends upon the qualities of raw
materials, piece parts and the assembling process. It follows from
simple theory that so long as such quality characteristics are controlled,
the quality of the finished unit will be controlled, and will therefore
exhibit minimum variability. Other advantages also result. For
example, by gaining control, it is as we have already seen, possible to
establish standard statistical distributions for the many quality char-
acteristics involved in design. Very briefly, let us see just how these
statistical distributions, representing states of control, become useful
in securing an economic design and production scheme.
Suppose we consider a simple problem in which we assume that the
quality characteristic Y in the finished product is a function / of m
different quality characteristics, Xi, X^, • • • , X^, representable
symbolically by Equation (3).
Y =f{X„ X,, ■■■, XJ. (3)
For example, one of the X's might be a modulus of rupture, another a
diameter of cross section, and Y a breaking load. Engineering re-
quirements generally place certain tolerances on the variability in the
resultant quality characteristic F, which variability is in turn a func-
tion of the variabilities in each of the m different quality characteristics.
As already stated, the quality characteristic Y will be controlled
provided the m independent characteristics are controlled. Knowing
the distribution functions for each of the m different independent
variables, it is possible to approximate very closely the per cent of the
finished product which may be expected to have a quality characteristic
Y within the specified tolerances. If it is desirable to minimize the
variability in the resultant quality Y by proper choice of materials, for
example, and, if standard distribution functions for the given quality
characteristics are available for each of several materials, it is possible
to choose that particular material which will minimize the variability
of the resultant quality at a minimum of cost.
4. Attainment of Uniform Quality Even TJiough Inspection Test Is
Destructive
So often the quality of a material of the greatest importance to the
individual is one which cannot be measured directly without destroying
ECONOMIC QUALITY CONTROL OF PRODUCT 387
the material itself. So it is with the fuse that protects your home; with
the steering rod on your car; with the rails that hold the locomotive in
its course; with the propeller of an aeroplane, and so on indefinitely.
How are we to know that a product which cannot be tested in respect
to a given quality is satisfactory in respect to this same quality? How
are we to know that the fuse will blow at a given current; that the steer-
ing rod of your car will not break under maximum load placed upon it?
To answer such questions, we must rely upon previous experience. In
such a case, causes of variation in quality are unknown and yet we are
concerned in assuring ourselves that the quality is satisfactory.
Enough has been said to show that here is one of the very important
applications of the theory of control. By weeding out assignable causes
of variability, the manufacturer goes to the feasible limit in assuring
uniform quality.
5. Reduction in Tolerance Limits
By securing control and by making use of modern statistical tools,
the manufacturer not only is able to assure quality, even though it
cannot be measured directly, but is also often able to reduce the
tolerance limits in that quality as one very simple illustration will serve
to indicate.
Let us again consider tensile strength of material. Here the measure
of either hardness or density is often used to indicate tensile strength.
In such cases, it is customary practice to use calibration curves based
upon the concept of functional relationship between such characteris-
tics. If instead of basing our use of these tests upon the concept of
functional relationship, we base it upon the concept of statistical rela-
tionship, we can make use of planes and surfaces of regression as a
means of calibration, thus in general making possible a reduction in the
error of measurement of the tensile strength and hence the establish-
ment of closer tolerances. It follows that this is true because, when
quality can be measured directly and accurately, w^e can separate those
samples of a material for which the quality lies within given tolerance
limits from all others. Now, when the method of measurement is
indirect and also subject to error, this separation can only be carried
on in the probability sense assuming the errors of measurement are
controlled by a constant system of chance causes. It is obvious that,
corresponding to a given probability, the tolerance limits may be re-
duced as we reduce the error of measurement.
Fig. 14 gives a simple illustration. Here the comparative magni-
tudes of the standard deviations of strength about the two lines of
regression and the plane ^ of regression are shown schematically by the
^ For definition of these terms see any elementary text book on statistics.
388
BELL SYSTEM TECHNICAL JOURNAL
a
i/i 40,000
I
I-
o
g 30,000
q:
I-
<0
to
Z 20,000 L
t; 40
60 60 100 2.4 2.6 2.8 3.0
ROCKWELL HARDNESS (E) DENSITY IN GRAMS PER CENTIMETER CUBE
A B
Z = TENSILE STRENGTH IN LBS./SQ. IN.
Y=ROCKWELL HARDNESS (E)
X = DENSITY IN GRAMS PER
CENTIMETER CUBE
Fig. 14 — How control makes possible improved quality through reduction in tolerance
limits.
ECONOMIC QUALITY CONTROL OF PRODUCT 389
lines in Fig. 14-J. The lengths of these are proportional to the allow-
able tolerance limits corresponding to a given probability. Customary
practice is to use the line of regression between tensile strength and
hardness. Note the improvement effected by using the plane of regres-
sion. By using the hardness and density together as a measure of
tensile strength in this case, the tolerance limits on tensile strength
corresponding to a given probability can be reduced to approximately
one-half what they would be if either of these measures were used alone.
IV. Conclusion
It seems reasonable to believe that there is an objective state of control,
making possible the prediction of quality within limits even though the
causes of variability are unknown. Evidence has been given to indi-
cate that through the use of statistical machinery in the hands of an
engineer artful in making the right kind of hypothesis, it appears
possible to establish criteria which indicate when the state of control
has been reached. It has been shown that by securing this state of
control, we can secure the following advantages:
1. Reduction in the cost of inspection.
2. Reduction in the cost of rejections.
3. Attainment of maximum benefits from quantity production.
4. Attainment of uniform quality even though inspection test is
destructive.
5. Reduction in tolerance limits where quality measurement is
indirect.
26
optimum Reverberation Time for Auditoriums
By WALTER A. MAC N AIR i
The suggestion is made that the sound damping material in an auditorium
should be such that the loudness of tones will decay at the same rate for all
frequencies. To attain this the reverberation time at 80 cycles must be
twice what it is at 1000 cycles.
The change of optimum reverberation time with volume is shown to be
derivable from a single hypothesis.
I. Reverberation Time vs. Frequency
THERE is very little published data in regard to the change in
reverberation time with frequency in auditoriums which are
considered near ideal. It is often mentioned by engineers and physi-
cists that to secure the best acoustical results, the reverberation time
should be the same for all frequencies in any one room. This specifies
that the sensation level shall decay at the same rate for all frequencies
of interest.
It seems more reasonable, however, to specify that the loudness of all
pure tones shall decay at the same rate for all frequencies since it is the
loudness of a tone which takes into consideration not only the energy
level but also its ultimate effect upon one's brain. In Fig. 1 - are
plotted data which show the relation between the loudness as judged
by a considerable number of observers and the sensation level. It
will be seen that for frequencies between 700 and 4000 cycles per
second these two quantities are equal to each other so that the two
points of view mentioned above demand identical conditions through-
out this frequency band. Outside of this band, however, any change
in the sensation level gives a greater change in the loudness, as may
be seen.
The maximum loudness in which we are interested at present is about
73.^ In the figure the curves may be replaced by straight lines which
represent fair approximations to the observed data up to this loudness.
This family of straight lines may be represented by the expression
Lt = AfSu (1)
where Aj \s the slope of the line adopted to fit the data for the fre-
1 Presented before Acoustical Soc. of Amer., Dec, 1929. Jour. Acou. Soc. Amer.,
Jan., 1930.
-This is Fig. 108 from "Speech and Hearing" by H. Fletcher.
^ This is the loudness that the source chosen in Part II of this paper will produce
in a room of 1000 cubic feet having a reverberation time of 0.8 seconds.
390
OPTIMUM REVERBERATION TIME FOR AUDITORIUMS 391
quency /. The values of A; chosen from this figure are given by the
next, Fig. 2. This approximation simpHfies our calculations very
much and introduces errors which are not intolerable.
Referring back to Fig. 1, if we wish to adjust the absorption of the
room so that the loudness of all pure tones will decay at the same rate,
say for the moment 60 units per second, it is seen that the sensation
level must drop 60 db per second for frequencies between 700 and
«,
V
/
y
y
<^
/>
y
«
^
/
/
/
/,
<>,
y
/
^y
/'
/
'a
'/
AD
/
/
;>
1/
'A
/
A
^
/
A
<s
'h
y.
d
D
/
/
A
r
/
A
4
,>°
/
/
/
f
/
/
/
A?
fA An
,f
'+
A
f
A
f
k
/
/
UJ
Z
r
J
/
A
/
/
a
f
/
/
/
y
%'
/
-J
f
/
/
/
/
V
f
^k
y
/
/
/
/
V.
/
**n
l\
^
V
L
"a
y
1 —
pn
/^
y
^
J
in
i
A
»
f
/
10 20 30 40 50 60 70
SENSATION LEVEL
80
90
100
Fig. 1. — Loudness of pure tones.
4000 cycles and for other frequencies the sensation level must drop
60/.(4/ db per second; or in other words, the reverberation time for
frequencies between 700 and 4000 cycles should be one second and
outside of this band it should be Aj seconds. Fig. 2, then, which is a
plot oi A} vs. frequency now becomes also an illustration of the shape
of the reverberation time vs. frequency curve which a room should
have in order that the loudness of pure tones of all frequencies shall
decay at the same rate.
392
BELL SYSTEM TECHNICAL JOURNAL
According to Sabine's well known formula the reverberation time is
inversely proportional to the number of absorption units in the room so
that, if we assume this, we may immediately infer the shape of the
curve which represents the number of absorption units necessary at
any frequency, referred to the amount required at 1000 cycles, to
obtain our required condition. These values are plotted in Fig. 3. If
it should happen that the greater part of the sound absorption in a
room is caused by one particular kind of surface, then the curve in
Fig. 3 is the shape of the absorption curve that this material should
have.
A pertinent observation on which every one seems to agree is that if
2.4
2 34568 2 34568
10 100 1,000 10,000
FREQUENCY IN CYCLES PER SECOND
Fig. 2. — \ 'allies of Af vs. frequency.
an auditorium has an unusually long reverberation time and conse-
quently is of little use, when empty, it attains excellent acoustic
conditions when filled with a large audience. In these cases a very
large part of the absorption is caused by the audience. The absorp-
tion of an average audience has been measured by W. C. Sabine ^ and
his results are also plotted in Fig. 3. The close agreement between this
curve and the one we have obtained from our hypothesis gives con-
siderable confidence in our general viewpoint.
II. Reverberation Time vs. Volume
It is generally accepted that the best acoustical conditions in a room
are obtained when the reverberation time is adjusted to a definite value
known as the optimum reverberation time. Observations reported in
literature agree that the value of the optimum reverberation time in-
^ "Collected Papers on Acoustics," page 86.
OPTIMUM REVERBERATION TIME FOR AUDITORIUMS 393
creases with the size of the room in the way shown in Fig. 4 where the
curves are the choices reported by Watson,^ Lifschitz,^ and Sabine.*'
These experimental results have served as the basis of successful
adjustment and design of many auditoriums. One naturally seeks
the factor which determines a choice of reverberation time of two
seconds for a million cubic feet theatre and on the other hand a choice
of near one second for a 10,000 cubic foot music room. It is our pur-
pose now to point out the factor which apparently does this.
We will set down a condition which we believe to be this factor and
then will show that the requirements demanded by it agree quite
100
90
/•
?
>
^
^
\
s
y
y
Y
/"
s
70
/
60
50
y
1
y
40
y
i
1
)EAL SPECIFIED BY THE PRESENT WORK
30
20
in
PRODUCED BY AN AUDIENCE
DATA OF W.C.SABINE
VALUE 100 ASSIGNED FOR 1000 CYCLES
10
4568 2 34568
100 1,000
FREQUENCY IN CYCLES PER SECOND
4 5 6 8
10,000
Fig. 3. — Relative number of absorption units vs. frequency.
closely with the empirical results illustrated in Fig. 4 and mentioned
above. The condition is
Ltdt = - K,
(2)
in which ^o is the time a sustained source of sound E is cut off, /i the
time the sound becomes inaudible, Lt the loudness of the sound at
any instant t, and K a constant. As shown in Fig. 1, the loudness of a
one thousand cycle note is equal to the sensation level, that is,
Lt = Si for 1000 cycles.
■•Watson, Architecture, May, 1927.
* Lifschitz, Phys. Rev., 27, 618, 1926.
« Sabine, Trans, of S.M.P.E., XII, 35, 1928.
394 BELL SYSTEM TECHNICAL JOURNAL
Since, during the time of decay, Si decreases uniformly with time, and
therefore Lt also, then for a thousand cycle note, evaluating our
integral we have
Lt,T, = 2K (3)
or
St^Ti = 2K,
where
T, = t,- to.
This last expression is practically in the form in which this condition
was first stated by Lifschitz.'' In (3) there are three unknowns and a
fourth is implied, namely, the power of the source, E.
We now turn our attention to finding the relation between the
volume of a room and the reverberation time dictated by the stated
condition. Following P. E. Sabine let us take the rate of emission of
the source, E to be 10^° cubic meters (35.3 X 10^" cubic feet) of sound
of threshold density per second. Now ^
4V, 4 X 35.3 X 10'"
Ti = — loge ,
ca c • a
where F is the volume of the room in cubic feet.
c is the velocity of sound, 1120 feet per second.
a is the number of absorption units in sq. feet and '^
T c ini 4 X 35.3 X 10'"
Li, = St, = 10 logio ~
If we should substitute these values in (3) we would obtain a relation
between V, a, and K which must be satisfied when condition (2) is
satisfied. In other words, this relation would specify the amount of
absorption, for a one thousand cycle note, a room should have if it
complies with (2).
If we assume Sabine's well known formula, namely,
^ _ .057
where T is the reverberation time in seconds we may express this
relation in terms of V, T, and K with the result
(2KYI-
10.40 + log To, - log V = ^ ^g3 j.^ ,., , (4)
' See Crandall "Theory of Vibrating Systems and Sounds," page 211.
* See Crandall "Theory of Vibrating Systems and Sounds," page 210, and the
definition of sensation level.
OPTIMUM REVERBERATION TIME FOR AUDITORIUMS 395
where Top is the value of T imposed by our condition (2) for a thousand
cycle tone.
Referring to Fig. 4 it will be seen that all three observers agree
rather closely that the reverberation time for an auditorium of
1,000,000 cubic feet should be 2.0 seconds. This value refers to a tone
of 512 cycles, the customary frequency used for experimental obser-
vation. It has been shown above that the reverberation time for
8
10,000 100,000
VOLUME IN CUBIC FEET
4 5 6 8
l,000,000'
Fig. 4. — Optimum reverberation time \s. volume in cubic feet for 512 cycles.
1000 cycles should be 92.5 per cent of the reverberation time for a
512 cycle tone, so that the 2.0 seconds above corresponds to 1.85
seconds for 1000 cycles. We can evaluate K in (4) by adapting this
latter value of Top for a volume of 1,000,000 cubic feet. This
gives K = 32.6. Substituting this value in (4) we obtain
6.35
log V = 10.40 + log Top -
T 1/2
■* op
(5)
1.8
1.6
§1.2
o
O 1.0
tiJ
-
^
^
ir^
-H
-
-
-"'^
.
—
-
—
0.8
z
" 0.6
0.4
0.2
1,000
4568 2 34568
10,000 100,000
VOLUME IN CUBIC FEET
4 5 6 8
1,000,000
Fig. 5. — Optimum reverberation time vs. volume in cubic feet for 1000 cycles.
From (5) we may obtain Top for 1000 cycles for any volume. See
Fig. 5. As mentioned above these values of Top are 92.5% of Top
396 BELL SYSTEM TECHNICAL JOURNAL
for 512 cycles so that these latter may be easily deduced for com-
parative purposes. These values are plotted to give curve number 4
in Fig. 4. It is seen that this curve agrees very well with those
showing the choices of competent judges.
III. The Mork General Hypothesis
Equation (2) may be written as follows, since we have assigned a
value to K:
Ltdt = - 32.6 (6)
L
and it will be remembered that we have considered Li^ to be the loud-
ness set up by a certain standard source. Allowing V to vary with /
constant (1000 cycles) we have obtained a relation between the opti-
mum reverberation time and volume of rooms for 1000 cycles. We
wish to point out now that exactly this same condition (6) with V
constant and / variable, will give the same results that we have ob-
tained in Part I of this paper with the only further requirement that
for other frequencies than 1000 cycles the strength of the source E
shall be such that the loudness L/^ set up in the room at the frequency
considered shall be exactly the same as the loudness which our standard
source would set up at 1000 cycles.
In Part I of this paper our stated condition was that the loudness of
all pure tones shall decay at the same rate for all frequencies. Since
we have specified that the loudness at the time /o shall be the same for
all test frequencies and also that the loudness at the time ti shall be
zero for all frequencies, it is quite evident that the above integral can
have the same value at all these frequencies only when the loudness
decays at the same rate for all frequencies concerned. In other words,
this condition stated as an integral specifies exactly the same require-
ment on the decay of loudness that we expressed in our statement early
in Part I of this paper.
IV. Conclusions
To recapitulate, we have set down an equation, together with a
specification of the strength of the virtual source in each case, from
which we obtain the value the reverberation time for any frequency
tone should have in any sized room according to the condition which
apparently controls the choice of observers.
One naturally turns to see what meaning may be attached to this
significant expression, namely, the integral of the loudness taken
OPTIMUM REVERBERATION TIME FOR AUDITORIUMS 397
throughout the time of decay to inaudibility. Since this integral has
the same value for all auditoriums which are considered ideal, it implies
that one's brain is a ballistic instrument which is concerned with not
only the maximum value of loudness but also with the effect of loudness
integrated throughout a considerable interval of time.
Abstracts of Technical Articles from Bell System Sources
Phenomena in Oxide Coated Filaments.^ Joseph A. Becker. A
theory of the changes in activity in oxide coated filaments is proposed.
From a comparison of the behavior of these filaments and filaments
with composite surfaces such as thorium on tungsten, caesium on
tungsten, and casium on oxygen on tungsten it appears probable
that oxide coated filaments owe their high activity to adsorbed metallic
barium. The changes in emission from a coated filament produced
by changes in plate potential and by currents sent into or drawn from
it, are ascribed to electrolysis of the oxide. When electrons are sent
into a coated filament barium is deposited on the surface and the
activity increases until an optimum is reached beyond which the
activity decreases. When current is drawn from the oxide, oxygen is
deposited on the surface. If the oxygen is beneath the adsorbed
barium, it increases the activity; if it is above the barium, it decreases
the activity. Both barium and oxygen diffuse readily from the sur-
face into the oxide and vice versa. This theory is tested, confirmed,
and extended by numerous experiments.
An experimental technique is employed by which relative rates of
evaporation of small amounts of electropositive and electronegative
materials can be determined with considerable precision. The same
technique might be useful in a number of similar investigations.
Metallic barium or oxygen which evaporate from a coated filament
are allowed to deposit on one side of a flat tungsten ribbon whose
thermionic activity is followed. When the plausible assumption is
made that an optimum activity is obtained when the tungsten is
covered with a single layer of electropositive material, the relative
rates of evaporation can be converted to absolute rates. This tech-
nique is also employed to determine the factors which control the
evaporation of oxygen from a coated filament.
Estimation of the Volatile Wood Acids Corrosive to Lead Cable Sheath.^
R. M. Burns and B. L. Clarke. The detection of volatile acids in
the air drawn from creosoted wood conduit corrosive to lead cable
sheath made desirable the development of a suitable method for the
extraction and estimation of volatile wood acids. Such a method
consists in the condensation of the volatile constituents of wood
sawdust removed under reduced pressure and titration of the conden-
1 ThePhys. Rev. Nov. 1929.
' Jndust. and Eng. Chem., Jan. 1930.
398
ABSTRACTS OF TECHNICAL ARTICLES 399
sate using a modified differential potentiometric electrode. Acidity
data have been obtained for Douglas fir, western hemlock, southern
yellow pine, western pine, spruce, redwood, cedar, and oak, and a
correlation is attempted between these acidities and the observed
corrosive character of the woods.
Electron Waves.^ C. J. Davisson. This paper is a brief review of
the experiments made on the diffraction of electrons by crystal during
the first two years following the discovery of this phenomenon, and
an indication of the paths along which future experimentation may
be expected to proceed.
Television in Colors by a Beam Scanning Method.'^ Herbert E. Ives
and A. L. Johxsrud. It has been recognized ever since the practical
achievement of television, and indeed before, that television might
be achieved in colors by utilizing the principles used in three-color
photography. The requirements in the two cases are very closely
parallel. Three-color photography had to wait for its practical
achievement, on photographic materials sensitive to all colors of the
visible spectrum. The parallel requirement in the case of television is
for photoelectric cells similarly color sensitive. The requirements of
television as to primary colors to be used for the synthesis of the colored
image are relatively more difficult of fulfillment than in the case of
color photography because in television we need not merely colored
light sources, but light sources which shall be capable of following the
variations of the television signal current with high speed. If, how-
ever, these two requirements, namely color sensitive photoelectric
cells and high speed-colored lights, are met, television in color could
conceivably be realized by utilizing any one of a number of devices
for analyzing and recombining images which have been successfully
applied in three-color photography.
Air Transport Communication.'" R. L. Jones and F. M. Ryan.
The successful operation of an air transportation system depends in
no small degree on the communication facilities at its command.
Rapid and dependable communication between transport planes in
flight and the ground is essential. Two-way radio telephony provides
this necessary plane-to-ground contact.
The design of a radio telephone system for this service requires
quantitative knowledge of the transmission conditions encountered in
^ Jour. The Franklin Inst., Nov. 1929.
* Jour. Opt. Soc. of Amer., Jan. 1930.
'Jour. A. /.£.£., Jan. 1930.
400 BELL SYSTEM TECHNICAL JOURNAL
plane-to-ground communication. An experimental investigation of
these conditions over the available frequency range has been carried out
and the results are described.
A complete aircraft radio telephone system designed for the use of
air transport lines and an airplane radio receiver designed for reception
of government radio aids to air navigation are also described.
A Study of Noise in Vacuum Tubes and Attached Circuits. '^ F. B.
Llewellyn. The noises originating in vacuum tubes and the attached
circuits are investigated theoretically and experimentally under three
headings: (1) shot effect with space charge, (2) thermal agitation of
electricity in conductors, (3) noise from ions and secondary electrons
produced within the tube.
A theoretical explanation of the shot effect in the presence of space
charge is given which agrees with experiment insofar as a direct deter-
mination is possible. It is shown that the tubes used should be capable
of operating at full temperature saturation of the filament in order
to reduce the shot effect.
In the computation of the thermal noise originating on the plate
side of a vacuum tube, the internal plate resistance of the tube is to
be regarded as having the same temperature as the filament.
Noise produced by ions within the tube increases as the grid is
made more negative.
With tubes properly designed to operate at temperature saturation
it is possible to reduce the noise on the plate side to such an extent
that the high impedance circuits employed on the grid side of the
first tube of a high gain receiving system contribute practically all of
the noise by virtue of the thermal agitation phenomenon.
On the Nature of ''Active'' Carbon? H. H. Lowry. Practically
all investigators have used for their measure of "activity" the adsorp-
tive capacity of the carbon (charcoal) under certain arbitrary condi-
tions. In several previous papers data have been given which indicate
that the adsorptive capacity of carbon is increased by any process
which increases either the total surface per unit weight or the degree
of unsaturation of the surface atoms, or both. No exceptions to this
generalization have been encountered. Since the adsorptive capacity
is dependent on two factors which may be independently varied, it
seems hardly logical to continue its use as a measure of the activity of
carbon. Since it is generally recognized that the forces effective in
adsorption processes are a result of the unsaturation of the surface
* Proc. The Inst. Radio Engiiieers, Feb. 1930.
''Jour, of Fhys. Cliem., Jan. 1930.
ABSTRACTS OF TECHNICAL ARTICLES 401
atoms, the ratio of the adsorptive capacity to the total adsorbing sur-
face would appear to be much more satisfactory for a measure of the
activity.
The data shown graphically in this paper show that starting with a
given raw material, i.e., an anthracite coal, an increase in the tempera-
ture to which the material is heated above 1000° decreases the adsorp-
tive capacity per unit pore volume. It is pointed out that the pore
volume may be considered a measure of the extent of adsorbing sur-
face and that the activity of an adsorbent carbon (charcoal) should be
measured by the amount of gas adsorbed per unit area of its surface.
The data, therefore, indicate that the activity of a charcoal is indepen-
dent of the atmosphere in which it is prepared and dependent only on
the maximum temperature to which it is heated. At any temperature
between 900 and 1300° an increase in the adsorptive capacity is most
probably accompanied by a proportional increase in the extent of the
adsorbing surface. For example, although the adsorptive capacity
of the samples prepared at 1100° ranged from 1.8 to 23.1 c.c. carbon
dioxide per gram at 0° and atmospheric pressure, the actually meas-
ured values of activity ranged from 0.201 to 0.295, while the weighted
average for all the samples prepared at the same temperature was 0.27 :
the variations observed are believed to be due to the limitations, which
have been discussed, of the measure of the surface area rather than to
a real difference in the activity.
The Operation of Modulators from a Physical Viewpoint.^ E.
Peterson and F. B. Llewellyn. The mathematical expressions
which occur in the treatment of non-linear devices as circuit elements
are interpreted in terms of a graphical physical picture of the processes
involved. This picture suggests, in turn, several useful ways of apply-
ing the equations in cases where the driving forces are so large that the
ordinary power series treatment becomes prohibitively cumbersome.
In particular, the application has been made in detail to the calculation
of the intermediate-frequency output to be expected from a heterodyne
detector having an incoming radio signal and locally generated beating
oscillator voltage applied on its grid and a circuit of finite impedance
to the intermediate frequency attached to its plate.
A Study of the Output Power Obtained from Vacuum Tubes of Different
Types} H. A. Pidgeon and J. O. McNally. Economical operation
of the large number of tubes involved in the Bell System makes nec-
essary the adoption of common supply voltages. This requires that
* Proc. The Inst. Radio Engineers, Jan. 1930.
^ Proc. The Inst. Radio Engineers, Feb. 1930.
402 BELL SYSTEM TECHNICAL JOURNAL
repeater tubes of various types be designed to operate at a fixed plate
voltage. For this reason the design of ampHfier tubes to give as
large a power output as possible at the operating plate voltage is of
considerable importance.
In the case of three-electrode tubes it is possible from theoretical
considerations to compute, approximately, the electrical parameters
a tube must have in order to give the maximum output power of a
given quality obtainable under fixed operating conditions.
The electrical characteristics and output of fundamental, second,
and third harmonics of two of the more common telephone repeater
tubes are given.
It is of considerable interest to determine whether greater power
output of comparable quality can be obtained from tubes containing
more than one grid. Since no sufficiently exact theoretical analysis
of multi-grid tubes is yet available to permit the determination of
the parameters of optimum tubes, a comparative experimental inves-
tigation of a number of such structures has been undertaken. The
electrical characteristics and output of fundamental, second, and third
harmonics of several such experimental tubes are given. The power
output of multi-grid tubes and of three-element tubes is compared.
The reasons for the comparatively large power output of certain types
of multi-grid tubes are discussed.
Effect of Small Quantities of Third Elements on the Aging of Lead-
Antimony Alloys}^ Earle E. Schumacher, G. M. Bouton, and
Lawrence Ferguson. The data presented in this paper definitely
show that small quantities of certain elements when added to lead — 1
per cent antimony alloys have a very marked effect on the rate at which
antimony is precipitated from supersaturated solid solution. Some
suggestions of the mechanism of this change can be had from a consider-
ation of the experimental findings in conjunction with the pertinent
equilibrium diagrams.
Although the literature shows that the third elements studied are
insoluble in lead in the solid phase, no results have been reported on
alloys containing these elements in very low concentrations. Even
though they should be insoluble in lead, antimony may so change the
lead lattice that they become soluble in lead-antimony. Furthermore,
since these elements form either compounds or solid solutions with anti-
mony, there are forces of attraction between them which may be
strong enough to carry small quantities of the third elements, along
with the antimony, into solid solution in the lead. The resulting ter-
^" Indust. and Eng. Cliem., Nov. 1929.
ABSTRACTS OF TECHNICAL ARTICLES 403
nary solutions, by their different energy relations, may cause the ob-
served effects on the rate of precipitation of antimony.
The Tube Method of Measuring Sound Absorption Coefficients.^^
E. C. Wente. The general principles underlying the tube method
of measuring sound absorption can be derived conveniently from the
analogous equations for the electrical transmission line. These equa-
tions show that the actual method of measurement is capable of many
modifications, some of which have already been adopted by various
experimenters. However, if reliable results are to be obtained, it is
important that the apparatus be so designed that the propagation
along the tube be rectilinear and the attenuation small, and that the
tone be kept free from harmonics.
In the tube method the absorption is measured at perpendicular
incidence, whereas in the reverberation method it is measured at ran-
dom incidence. A theoretical study of the absorption of sound by por-
ous materials as a function of the angle of incidence shows that in some
cases there may be a considerable discrepancy between the values
obtained by the two methods. The tube method may also give im-
practicable results for materials which are to be used in the form of
large panels and absorb sound largely by virtue of inelastic bending
rather than because of their porosity.
" Jour, of the Acoust. Sac. of Amer., Oct. 1929.
Contributors to this Issue
Ralph Bown, M.E., 1913, M.M.E., 1915, Ph.D., 1917, Cornell
University, Captain Signal Corps, U. S. Army, 1917-19; Department of
Development and Research, American Telephone and Telegraph
Company, 1919-. Mr. Bown has been in charge of work relating to
radio transmission development problems. He is a Past President of
the Institute of Radio Engineers.
John R. Carson, B.S., Princeton, 1907; E.E., 1909; M.S., 1912;
American Telephone and Telegraph Company, 1914-. Mr. Carson
is well known through his theoretical transmission studies and has
published extensively on electric circuit theory and electric wave
propagation.
Karl K. Darrow, B.S., University of Chicago, 1911; University
of Paris, 1911-12; University of Berlin, 1912; Ph.D., University of
Chicago, 1917; Western Electric Company, 1917-25; Bell Telephone
Laboratories, 1925-. Dr. Darrow has been engaged largely in writing
on various fields of physics and the allied sciences. Some of his earlier
articles on Contemporary Physics form the nucleus of a recently pub-
lished book entitled "Introduction to Contemporary Physics" (D.
Van Nostrand Company). A recent article has been translated and
published in Germany under the title "Einleitung in die Wellen-
mechanik."
William Fondiller, B.S., College of the City of New York, 1903;
E.E., Columbia University, 1909; M.A., Columbia University, 1913;
Engineering Department, Western Electric Co., Inc., 1909-25; Bell
Telephone Laboratories, Inc., 1925-. Mr. Fondiller's work has re-
lated to the development of transmission apparatus, such as loading
coils, filters, transformers, etc. and is now Assistant Director of Ap-
paratus Development of Bell Telephone Laboratories, Inc. In this
capacity he is responsible for the design of telephone apparatus and
investigations of materials.
Norman R. French, A.B., University of Maine, 191-i; A.M., 1916;
Instructor, Physics Department, University of Maine, 1914-16; In-
structor, Princeton University, 1916-17; General Staff, A.E.F., 1917-
18; Commanding Officer, Flash and Sound Ranging Sections, Army
Engineers' School, A.E.F., 1918; American Telephone and Telegraph
Company, Department of Development and Research, 1919-. Mr.
French's work has related chiefly to loading, submarine cables and
transmission quality.
404
CONTRIBUTORS TO THIS ISSUE 405
Charles W. Carter, Jr., A.B., Harvard, 1920;B.Sc., Oxford, 1923;
American Telephone and Telegraph Company, Department of Devel-
opment and Research, 1923- Mr. Carter's work has had to do with
the theory of electrical networks and with problems of telephone
quality.
Walter Koenig, Jr., A.B., Harvard, 1923; Instructor and Re-
search Assistant, Harvard, 1923-24; American Telephone and Tele-
graph Company, Department of Development and Research, 1924-.
Mr. Koenig has been engaged chiefly in studies relating to trans-
mission quality.
W. A. MacNair, B.Sc, Colgate Univ., 1920; Ph.D., Johns Hopkins
Univ. 1925; National Research Fellow in Physics, 1925-27; Bell
Telephone Laboratories, 1929-.
W. P. Mason, B.S., University of Kansas, 1921; M.A., Columbia,
1924; Ph.D., Columbia, 1928. Engineering Department, Western
Electric Company, 1921-25; Bell Telephone Laboratories, 1925-.
Mr. Mason's work has been largely in transmission studies.
A. A. Oswald, B.S., Armour Institute of Technology, 1916; E.E.,
1927. Western Electric Company, Engineering Department, 1916-24;
Bell Telephone Laboratories, Inc., 1925-. Mr. Oswald's work has
been concerned with the development of long and short wave trans-
atlantic and ship-to-shore radio-telephone systems; and, during the
War, of systems for airplane radio-communication and radio-control.
D. A. Quarles, A.B., Yale University, 1916; U. S. Army, 1917-
19; Engineering Department, Western Electric Company, 1919-
25; Bell Telephone Laboratories. 1925-. Mr. Quarles was earlier
engaged in transmission studies of circuits and networks. More
recently he was in charge of inspection engineering on apparatus
products. As Assistant Director of Apparatus Development, he is
now engaged in development work on Outside Plant products.
Walter A. Shewhart, A.B., University of Illinois, 1913; A.M.,
1914; Ph.D., University of California, 1917; Engineering Department,
Western Electric Company and Bell Telephone Laboratories, 191 8-.
Mr. Shewhart is making a special study of the application of probability
theories to inspection engineering.
The Bell System Technical Journal
July, 1930
Radio Telephone Service to Ships at Sea *
By WILLIAM WILSON and LLOYD ESPENSCHIED
The ])aper discusses the American end of the ship-to-shore radio telephone
system and the connecting equipment on board the Leviathan. The most
suitable wavelengths for this service are in the short-wave range, but the use
of these wavelengths complicates the problem, since different wavelengths
are required according to the distance of the ship from shore, the time of day,
season of year, etc. The problem on shipboard is further complicated by the
fact that the transmitting and receiving systems are necessarily near together
and special precautions are necessary to take care of interference from the
radio telephone transmitter and the radio telegraphic services. In addition
to interference from these sources, there is a background of interference in the
ships' electrical equipment, all of which necessitates a much more powerful
land station than is necessary on shipboard.
In the present system, the shore transmitter has a power rating of 15 kw.
and the ship transmitter of 500 watts. The shore transmitting station is lo-
cated at Ocean Gate, N. J., and the receiving station at Forked River, N. J.
At both of these locations, directive antennas are employed which cover the
ships' lanes. The stations are connected by wire to the Long Lines toll office
in New York, and the o\'er-all control of the circuit is carried out from this
])oint. Both the ship and shore transmitters are crystal controlled. The
ship's receiver is highly selective and is of the double-detection type. Com-
munication between the ship and the shore is carried out by use of a pair of
frequencies, one for transmission in each of the two directions, separated
from each other by about three per cent. Ships of a number of nations
are being equipped with wireless telephone apparatus and as the service
expands, it will undoubtedly be necessary to formulate a plan in which inter-
national agreement is reached on the allocations of frequencies for ship-to-
shore telephony and telegraphy, in order that undue interference within the
ser\'ices themselves or between the two services shall not ensue.
IN view of the developments which have recently taken place in the
field of ship-to-shore radio telephony, it would appear appropriate
to review the state of the science and to discuss the problems which
have arisen, the facilities which have been installed, and the general
results obtained.
The ship-to-shore radio telephone system, which is here described,
was opened for public service between the Leviathan and the United
States on December 8, 1929. This was the first extension of the public
telephone service to a ship at sea and enabled calls to be made between
the vessel and any Bell System subscriber. The system as set up is
intended primarily for giving telephone service to the larger passenger-
carrying vessels as an extension to the wire network, and should be
distinguished from the more simple uses which have been made of radio
* Presented at the North Eastern District Meeting of the A. I. E. E., Springfield,
Mass., May 1930.
407
27
408 BELL SYSTEM TECHNICAL JOURNAL
telephony in the marine field, such as that of enabling a coastal station
operator to talk with coast guard vessels, fishing trawlers, etc.
This paper is concerned with the developments which have been
carried out in the United States, including the establishment of trans-
mitting and receiving stations on the New Jersey coast, the equipping
of the Leviathan and the establishment of service to that ship.
It is significant of the wide-spread interest in this type of service that
developments have also gone forward rapidly on the European side
where the British, Germans, and French are preparing coastal stations
and equipping some of the larger ships for public telephone service.
The British have already initiated service to two of the White Star
Liners, the Olympic and the Majestic, and before the summer is over it
is likely that half a dozen of the larger transatlantic vessels will be
undertaking this service, connecting with both the American and the
European networks.^
Early Developments
Attempts to apply telephony in the maritime field date back to the
pioneer work on radio telephony itself, but it was not until the applica-
tion of the vacuum tube were developed that radio telephony for any
service became finally practicable.
Following the long distance, point-to-point radio telephone experi-
ments of 1915, there was carried out in the following year what is
believed to have been the first trial of two-way radio telephony from
the wire telephone system to a vessel at sea. This trial was conducted
by Bell System engineers in cooperation with the Navy Department.
On that occasion the Secretary of the Navy, in his office in Washington,
carried on two-way conversations with the captain of the U. S. S. New
Hampshire oflf Hampton Roads.
Following the further development of radio telephony during the
War, there was undertaken, in the years 1920-1922, an extensive devel-
opment of ship-to-shore radio telephony, looking toward the linking of
ships at sea with the land line telephone network.- At that time there
was built a coastal radio telephone station at Deal Beach, N. J., and
several ships were equipped on a trial basis. Extensive engineering
tests were made and a number of demonstrations carried out which
proved the physical feasibility of establishing such connections.
W^hile the trials were successful from the technical standpoint, the
development was not carried into commercial use because the adverse
economic conditions existing in the post-W^ar period did not appear to
^ Ship-to-shore telephone service is now given (July, 1930) from both U. S.
and British shores to'the Leviathan, Olympic, Majestic and Homeric.
-"Radio Extension"" of the Telephone System to Ships at Sea," by H. W.
Nichols and Lloyd Espenschied, /. R. E. Proceedings, Vol. 11, 1923.
RADIO TELEPHONE SERVICE TO SHIPS AT SEA
409
justify the initiation of the new service at that time. Furthermore,
the waves in the range of 300-500 meters, which had been used in these
early trials, were soon thereafter assigned for broadcasting.
In the last few years the whole outlook has changed considerably.
The development of short-wave radio systems has greatly increased the
message carrying capacity of the radio spectrum and has made it feasi-
ble to maintain communication over a greater range of distances than
was previously practicable for ships. Transoceanic radio telephone
services have been inaugurated, and with the large increase in steam-
ship travel there has arisen a renewed interest in the extension of
telephone service to ships at sea.
When it became evident that short-wave transmission might be
CC
Ul
^-
UJ
a.
ui
a.
If)
b
o
>
o
a.
o
o
z
u
a:
y-
in
a
ui
o
<
10,000
1,000
100
10
^
=
^
^
^
^
^
^
s
s
s
B
S
Nf
=
=
^
—
/
1
TRANSMISSION FROM 2XG TO FT. ST. GEORGE
ENROUTE TO BERMUDA ISLES
4
OCT. 21-24, 1925
_| 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 -
11
=:
^
:^
=
—
. —
—
—
—
—
—
—
—
—
—
zc
*—
J
i
/
\
"N
\y
/
y
\
\
t'
A
^-
—
—
—
—
3
—
■—
—
=^
—
—
-*-
E
—
:^
=::
/
1
/
\
/
/
^
' '
I
^
=
_
1
^
—
—
—
'
3
EE
E
1-
HA
Kill
1 T
HM
^
—
—
—
(n ~
—
—
—
—
— '
—
— 1
—
—
^
HARBOR
f-
—
—
—
^
1
;
^^
^^
^^
/I
^^
1
1 1
20
Ml
250 Ml.
^
8f
),r
^1.
680 Ml.
7«
>o|
Ml
76
5N
!!l
765MI, 1 1
7'p
5N
.11.
12
M
12
MID
M
12
M
12
MID
N
12
M
12
MID
M
Fig. 1 — Received Fields, New York-Bermuda Run 1925.
desirable for ship-to-shore telephone service, there was undertaken a
program involving the measurement of the strength of the electric
fields received aboard ship from a shore transmitter. This work was
part of a general program intended to obtain fundamental data upon
short-wave transmission, for purposes of point-to-point, as well as for
ship-to-shore telephone services. The tests were first made in 1925 on
vessels running between New York and Bermuda. Further measure-
ments were made on other ships in subsequent years.
Fig. 1 is an example of the result of these earlier tests. Transmission
was from Deal Beach on 4.5 megacycles (66 meters). The curve shows
the relatively weak field which was received as the vessel left dock, due
to the considerable stretch of land which intervened in the transmission
410 BELL SYSTEM TECHNICAL JOURNAL
path, the rise of the field to high values as the ship passed out of the
harbor, and the gradual diminution as the vessel continued on her
course. It will he oljserved that transmission on this frequency was
effective at night all the way to Bermuda, but that during the daytime
the transmission failed for distances greater than a few hundred miles.
Corresponding measurements showed that daylight transmission could
be secured by means of a higher frequency, such as 9 megacycles (v33
meters). Measurements of this kind, supplemented by data obtained
for a wide range of distances over land, and for transatlantic distances,
have built up a fairly complete set of quantitative data on short-wave
transmission over different distances and for various times of the day
and year.
Along with this study of transmission conditions, there was carried
on the development of short-wave apparatus technicjue for telephony.
The first application was in the field of point-to-point transatlantic
operation and the considerable art built up there, including the design
of transmitters, receivers, directive antennas, and the working out of
two-way operating methods, served as a very useful basis from which to
develop the coastal and ship stations for the maritime system.
With this background of development, preparations were made to
set up a two-way, short-wave radio telephone system for commercial
service. This service was centered upon New York because of the
large concentration of ocean-going trafftc at that port.
The Technical Problem
One of the most important problems to be solved in the design of a
short-wave system is that of determining the frequencies necessary for
giving the service involved. The frequencies which are best suited to
the different distances, time of day, and season of the year for trans-
mission over the North Atlantic are indicated in the curves of Fig. 2.
The curves for the greater distances refer to the transmission which
appears to take place in the upper regions of the earth's atmosphere
and is usually referred to as sky-wave transmission. Each of the sky-
wave curves traces the optimum frequency-distance relation for the
time of day and season of the year indicated. The curves merely give
a general picture of the frequency relation and do not take account of
other effects, such as fading, magnetic storms, etc.
The figure brings out very clearly the necessity for using a variety of
wavelengths if the ship lanes are to be adequately covered. Fortu-
nately, there is a considerable band on each side of the curves shown, in
which good transmission can be obtained, and this enables one to choose
a small number of frequencies in the short-wave range which are ade-
RADIO TELEPHONE SERVICE TO SHIPS AT SEA
411
quate to cover the conditions. Actually, it is found that a set of about
four frequencies will suffice to cover the North Atlantic. For distances
greater than a few hundred miles this characteristic obtains irrespective
of whether the transmission is over water or over land, by reason of the
fact that the transmission appears to take place in the upper regions of
the earth's atmosphere.
Closer in to the transmitting station, however, there is the so-called
surface component, the attenuation of which is much less over sea water
than over land. It will be seen that the surface wave may be relied
upon for distances of the order of 200-300 miles, for frequencies of
about 4 megacycles. The transmission of this component is much
more stable and reliable than is the transmission of the sky wave. It
O
<
U' Z
3
I
>-l
O
z
UJ
D 8
O
OA
5^
I ,
""^
'
^
■
'r
.^■
r'
2'
/
y
SllM
/E.p;_
,
■
/
/
N(l^
,/
/
^
(
\ X
,
N
GHT
-WIN
TER
'^^
<
'-i
C
£000
DISTANCE. - STATUTE UILE.S
Fig. 2. — Distance-frequency characteristic.
seemed important, therefore, to utilize the surface wave to the maxi-
mum extent possible.
W'ith this in mind, a series of transmission measurements was made
over a stretch of water between New Jersey, Long Island, and Nan-
tucket for the purpose of more accurately evaluating the effectiveness
of the surface wave component, particularly in so far as it bears upon
the question of how close to the water front the coastal station need be
placed. Transportable transmitting and receiving stations were used
in these tests. It was found that as the transmitting or the receiving
station was moved away from the water front, the attenuation in-
creased materially. For example, moving either terminal a mile back
from the coast line increases the attenuation some 8 decibels at 4.5
megacycles. On the other hand, a narrow stretch of land, such as a
sand bar, out a few miles from the coast, introduces relatively little loss.
412
BELL SYSTEM TECHNICAL JOURNAL
These results indicate that if full advantage is to be obtained from the
more reliable surface-wave component, the coastal station should be
immediately upon the seacoast or a salt-water bay.
An important factor in connection with radio reception on ship-
board is that of electrical interference. The modern steamship re-
quires for its operation and its service a large amount of electrical
machinery. In addition to this, it is equipped with various radio
telegraphic services. The operation of all of this electrical equipment
produces interference in a receiver which is much in excess of that
normally encountered in a shore receiving station which can be so
located as to be reasonably free from electrical disturbances. Further-
more, there is on the ship another source of disturbance which is due to
Fig. 3 — U. S. coastal station, circuit between New York and ship.
charging and discharging of various parts of the rigging in the strong
electromagnetic fields of the various radio transmitters. These various
sources of disturbance were found in the earlier shipboard experiments
and the high noise levels are, in general, the predominant factor in
limiting the communication range. These factors made it desirable
to employ at the shore end as powerful a transmitter as was available
and to use whatever benefit could be obtained from antennas designed
to be roughly directive along the transatlantic ship lanes. A trans-
mitting set of the type used in transatlantic communication, but
adapted for the ship-to-shore wavelengths, was therefore employed.
Since the shore receiver can be located in a comparativ^ely quiet
situation and since use can also be made of roughly directive receiving
antennas, there is no advantage in transmitting as large an amount of
RADIO TELEPHONE SERVICE TO SHIPS AT SEA
413
1
o
m
vD
■4^
t
rO
rO
>^
^^--^^ \
\
^"^^'^^'^
\
SC 2
-1 xQ
\
>v
^
\
in
O
In
8^
-\ ^_,^---^
\
^^^-\^'^^
mi /
U.' >- A
o< JX--
\
z^j^lY
J.^^^V^^
(D / ' \ " -\
\
O
V^
Ae
^^^T;
BAY"
BISC
^
\ _l \j \
:^^=^^
^^>=^s^^^
— V \
\
^^li---"
\ ■
in
'^.c^"
\\ ^
^^X^r"
\
in
^ — ^ —
\ " "'-.
*\
— \
o
— - —
*""' v\
\
^_— -Pt-
'0' \
\__- —
-" '^
1
\
m
—
(M
\
_____ \j\:
O
-4
\ \
w
\
1 ^
U^
in
C\J
Si
'
1ST
5T 8.
ST
1ST
\\^\
<n
UJ
cr 4r
O
ro
\\ Wv
M
rO
::- ,0
\\ \\\
<
1 .-i -
rO
O
~]
D JUI
SEPT
A,PRI
JAN
— 4U^
^
e
in
CO
J
^2??
UN
\\
t
1
RIL 1 S
LY 1ST
B. 1ST
PT. 1 ST
- \\ J
'^
/___^
®^k
■^
~~~~~-
< -)U. 01
\\
■<J
1 1 1
k \ \j
/
O
If)
-~^_
< 0) (J
li- z / \
.-o.< / \
\V
____/
o
"~~- — ■ — /— —::
•" i^-/J
A /
~T~-
in
iT)
^^
/^"^-v-
_Zt
W
__/
O
^
r^
K/^
■ — ~(£2 /
m
/~
in
/
/y 1-^*-'
■'' / / /
11 1
/
tO
/
/ <^^^c=«iL
1 '^ 1 1
11 1 1
~~~~--~,.._/
Q
to
z"^
Wl
uL
/
in
in
I I
^j'f^,^ "^x yv ?
"tC-v
/
?
"\
^■~-^ \\
^
^^"^hJU//
<D
-^/o-^
/
^^v„V\
\ /
^~~~^-~,^
/
/
0"^----^ / ^
' UTM 1
^r-^$^
in
/
/"^■"^
1
r^
y^
/ "^
1/1^^^
^
^~^\
in
10
O
00
\^
'^lil
s •
7
-^
h
r~
•n
3i
rfi 1
■^
•<t
rO
n
tvj 1
c
a
-M
CO
_o
c
<
■1-1
u
o
414
BELL SYSTEM TECHNICAL JOURNAL
power from the ship as from the shore. The actual power radiated by
the Leviathan s transmitter is of the order of 500 watts. The shore
receiver is of the type used on the transatlantic radio telephone circuits,
working with a directive antenna. The arrangement provides a fairly
well proportioned system, the channels being substantially equally
effective in the two directions.
The Shore System
The general setup of the system is illustrated in Fig. Z. The coastal
stations, sending and receiving, are located about 60 miles south of
New York on the New Jersey shore, at Ocean Gate and Forked River.
The course followed by the transatlantic ships is indicated on the map
of Fig. 4. The directional bearing of this course and the directivity
3200
1.0 .6 .2
75
VOLTAGE
AMPLITUDE
70
65
60 55 50 45 40 35 30 25 20 15 10
DEGREES — LONGITUDE WEST OF GREENWICH
Fig. 5 — Directional bearings.
characteristic of the New Jersey shore station antennas are illustrated
in Fig. 5. It will be observed that the breadth of the transmitted
beam is adequate to take care of the variation of the directional bearing
of the course. For steamship routes other than the transatlantic, as
for example the coastal route to the South, other antenna arrangements
will be required.
In general, the whole coastal station, including the transmitting and
receiving units, taken together with the wire line connections and con-
trol position in New York, is similar to one end of a transatlantic point-
to-point circuit. The transatlantic facilities have been described in
previous papers '^ and reference should l)e made to them for more detail
than is given below. The transmitting set has been adapted to cover
^ Papers on Transatlantic Telephone Service by Messrs. Miller, Bown, Oswald,
and Cowan, presented at Winter Convention of the A. I. E. E., New York, N. Y.,
January 1930. Papers by Messrs. Bown and Oswald, B. S. T. J., Apr. 1930.
RADIO TELEPHONE SERVICE TO SHIPS AT SEA
415
the frequency range used in the service. It has a power of 15 kw. out-
put of unmodulated carrier and is capable of delivering 60 kw. peak
power. A photograph of a similar transmitter at Deal Beach, which is
being used for this service pending the completion of the transmitting
station at Ocean Gate, is shown in Fig. 6. The antennas are simpler
Fig. 6 — Deal Beach transmitting set.
and less directional than those employed in the transatlantic circuit,
and give a transmission gain of 8 to 10 db as compared with a single
half-wave antenna.
The receiving station at Forked River has been in operation since the
416
BELL SYSTEM TECHNICAL JOURNAL
opening of commerical service last December. A photograph of the
receiving set is shown in Fig. 7. The receiver is of the double-detec-
tion type, of high gain and selectivity, and employs screen-grid tubes.
It is provided with automatic gain control. The apparatus shown
includes not only the receiving set proper but also the equipment which
is required for monitoring the circuit and for connecting with the wire
Fig. 7 — Forked River receiving set.
line into New York. The receiving antennas are of the same general
type as those used in the transatlantic system, which consist of a row
of quarter-wavelength verticals connected alternately top and bottom
by quarter-wavelength conductors. In the case of the longer wave-
lengths used in the ship-to-shore service, the vertical conductors are
RADIO TELEPHONE SERVICE TO SHIPS AT SEA
417
reduced in height and the horizontal links correspondingly elongated.
A photograph of the station at Forked River and two of the antennas is
shown in Fig. 8.
The control and operating terminal equipment in New York is
identical with that in use on the transoceanic radio telephone circuits.
The control positions, as they exist in the New York long-distance
telephone building for both transatlantic and ship-to-shore circuits,
are pictured in Fig. 9. These control positions have associated with
them such things as voice-frequency repeaters, indicators of the volume
being transmitted and received over the circuit, gain controls, monitor-
ing and testing facilities, and voice-current operated switching de-
vices. The latter prevent the speech received from the ship from
being reradiated from the shore transmitting station and permit inde-
pendent adjustment of amplification in the circuits leading to the
transmitting and receiving stations. Thus, the volume sent to the
Fig. 8 — Forked River station with antenna.
transmitting station may be kept substantially constant, despite
variations in speech volume received from different land line subscri-
bers and full modulation of the transmitter may be obtained for over-
riding noise on the ship. The function of the technical control operator
is that of maintaining the circuit in the correct technical condition for
talking. In general, it is the intention that the shore transmitting
and receiving stations should function, as far as possible, merely as
repeater stations, with the control of the over-all circuit from New York
to the ship resting in the New York technical operator.
The circuit terminates as an operating facility before a traffic opera-
tor at one of the long-distance telephone boards. In Fig. 10 is shown
an illustration of the traffic positions for the transatlantic radio tele-
phone circuits, including, at the right, two positions devoted to the ship-
to-shore service. The duty of one of these two girls is confined to the
radio circuit itself in that she talks to the ship operator, passes and re-
ceives information as to calls, and is generally responsible for complet-
ing the connection between the ship circuit and the land line subscriber.
418
BELL SYSTEM TECHNICAL JOURNAL
The adjacent operator is concerned more particularly with the land
line subscribers, answering inquiries and recording calls outbound to
ships and, in turn, getting in touch with and holding land line sub-
scribers for inbound calls.
Fig. 9 — New York technical control positions.
The Ship Station
The Leinathans radio transmitter was designed to supply about
500 watts, 80 per cent modulated radio frequency power to an antenna
at frequencies from 3 to 17 megacycles. To insure satisfactory operat-
RADIO TELEPHONE SERVICE TO SHIPS AT SEA
419
ing conditions, the carrier frequency stability was made as good as that
required for point-to-point service and the transmitter has been de-
signed with the object of holding the frequency within 0.01 per cent.
This facilitates the establishment of contact between the ship and shore
and obviates the necessity for frequent retuning of the shore receiver.
The background noise on the unmodulater carrier, due to commutator
ripple, etc., is inappreciable and the audio-frequency characteristics
from 200 to 2750 cycles is flat to within ± 2 db.
In addition to these electrical requirements, the mechanical design
must be such as to withstand ship's vibration, permit easy access to the
Fig. 10 — New York traffic positions.
interior so as to facilitate wave change, and at the same time protect
the operators from electrical shock.
The transmitter consists of a crystal oscillator and associated am-
plifiers. The crystal provides the necessary carrier frequency accuracy
and stability and the amplifiers step up the power of the carrier to the
desired level. Audio-frequency filters are placed in all voltage supply
circuits to eliminate background noise. The modulation system with
its associated transformers is designed to produce the requisite audio-
frequency quality. A diagram of the circuit is shown in Fig. 11.
Very thorough electrical shielding is necessary between amplifier
stages to prevent singing. This shielding makes the changing of coils,
420
BELL SYSTEM TECHNICAL JOURNAL
which is necessary for the changing of wavelengths, very unhandy, and
hence the crystal control and amplifier system, except the last stage
is provided in duplicate, one side being used for the longer and the other
for the two shorter waves. Wave changing, except for the output
circuit of the power stage, is then accomplished by connecting the
proper amplifier to the power stage.
The quartz plates used in the crystal control system are circular,
being approximately one inch in diameter, and are clamped rigidly in
the holder. This clamping serves to prevent any change of frequency
with mechanical vibration. The holder with its crystal is mounted in
a small oven, the temperature of which is held constant at 50 deg. cent,
to better than ±0.1 deg. cent. The thermal system of this oven is so
CRYSTAL
AND OVEN
5 WATT
OSCILLATOR
7j WATT
AMPLIFIER OR
HARMONIC GENERATOR
LJ
50 WATT
AMPLIFIER
500 WATT
POWER AMPLIFIER
Fig. 1 1 — Ship transmitter's schematic diagram.
designed that the change of internal oven temperature with tempera-
ture changes of the surrounding air is negligible.
As shown in the figure, the crystal is connected between the grid and
filament of a 5-watt vacuum tube which, together with the parallel
resonant circuit connected to the output of this tube, forms the crystal
oscillator. The radio-frequency voltage developed by the crystal oscil-
lator is applied directly to the grid of a 7>^-watt screen-grid tube which
can be used either as an amplifier or a frequency doubler. The output
of this tube, except in the case of the higher frequencies, is applied
directly to the grid of a 50-watt screen-grid amplifier. For the higher
frequencies a second frequency doubler can be switched into the circuit.
The output of the 50-watt tube is coupled through a balanced trans-
former to the final amplifier stage. The amplifier or frequency doubler
RADIO TELEPHONE SERVICE TO SHIPS AT SEA
421
stages are separately shielded and radio-frequency filters are provided
in all power supply leads.
The power amplifier consists of an air-cooled, three-element, one-
kilowatt tube. Neutralization is accomplished by the familiar balanc-
ing arrangement shown in the figure. The output circuit of this stage
consists of a parallel resonant circuit with provision for tapping in the
connection to the antenna.
Fig. 12— Leviathan transmitter.
Modulation takes place in the plate circuit of the final amplifier stage,
the plate current supply being fed through a special transformer, the
secondary of which is connected to two 250-watt tubes connected push-
pull and fed by a 50-watt amplifier.
The power supply is obtained from motor-generator sets operated
from the 110-volt, d-c. ship supply. Protection of the operators and
422 BELL SYSTEM TECHNICAL JOURNAL
apparatus is provided by means of relays and contactors in the high-
voltage supply circuits which prevent the high voltages from being
applied if the filament or grid circuits are not closed or if the doors of
the transmitter are open.
An illustration of the ship's transmitter is shown in Fig. 12. The
picture is somewhat out of perspective owing to difficulty in taking the
photograph in the limited space available on shipboard.
The receiving problem on shipboard is complicated by a number of
factors. The transmitting and receiving frequencies must be within a
few per cent of each other, if the best transmission conditions for the
time and place are to be utilized and if the frequencies are to remain in
the bands assigned internationally to the mobile services. This re-
quirement, as well as the noivSe conditions on shipboard, calls for a
receiver of high selectivity, which is obtained, in the present instance,
by the use of a double-detection set. The over-all selectivity is accom-
plished both by having a number of highly selective circuits ahead of
the first detector and by using tuned circuits in the intermediate fre-
quency stages, the high-frequency selectivity being used primarily to
prevent overloading of the first tube and the intermediate frequency
circuits being used to obtain the final selectivity required.
A reduction of the disturbances due to stay noises and better dis-
crimination against the transmitted carrier is obtained if the trans-
mitting and receiving antennas are widely spaced. On the other hand,
for operating reasons, it is desirable to have the transmitter and receiver
located in the same room. In the case of the Leviathan installation,
the transmitting antenna is located directly above the radio room,
between the second and third stacks, and the receiving antenna is
placed as far as possible behind the third stack. The receiving antenna
is connected through a suitable step-down circuit to a shielded trans-
mission line, the other end of which is connected to the receiver, the
receiver itself being very thoroughly shielded to avoid direct interfer-
ence from the transmitter. On account of limited space, only two
antennas are provided to handle the four frequencies, each antenna
representing a compromise between the most efficient antennas which
could be put up to handle the separate wavelengths.
As stated above, the receiver itself is of the double-detection type,
using heater type tubes throughout. Screen-grid tubes are used for
the first detector and intermediate frequency amplifiers and three-
element tubes in the remaining positions. A photograph of the re-
ceiver and associated voice-frecjuency equipment, as it is installed on
the Leviathan, is shown in Fig. 13, and a diagrammatic representation
of the receiver is shown in Fig. 14. The high-frequency selective sys-
RADIO TELEPHONE SERVICE TO SHIPS AT SEA
423
tern consists of four separately shielded tuned circuits, coupled by small
capacities. The use of a screen-grid tube in the detector circuit
gives a two fold advantage over the use of a three-element tube in that a
higher input impedance is maintained at the higher frequencies and
the necessity for neutralizing against the reaction of the beating oscilla-
tor on the input circuit is eliminated. The beating voltage is made of
the order of 75 to 100 volts for the purpose of reducing the effective
tube noise in the detector plate circuit. With this arrangement no
d-c. plate voltage is ordinarily required. The screen voltage is 22
Fig. 13 — Leviathan receiver.
volts. The output circuit is tuned to the intermediate frequency of
300,000 cycles and connection with the first intermediate amplifier is
effected by means of a low impedance transmission line. The inter-
mediate frequency amplifier stages are coupled by means of doubled
tuned circuits. The use of properly designed circuits of this type
makes it possible to obtain a high degree of selectivity against unde-
sired frequencies while obtaining sufficient band width to maintain
ease of tuning and to pass the desired frequencies. The second de-
tector is of the conventional grid bias type. Automatic gain control
28
424
BELL SYSTEM TECHNICAL JOURNAL
is provided in which a certain amount of the carrier is taken at the
end of the intermediate frequency stages, ampHfied and rectified. The
resulting d-c. current produces a voltage drop across a resistance, which
is applied to thegridof the first detector in such a manner that an increase
in the intermediate frequency output brings about a reduction in the
total set gain and vice versa. Manual gain control for following wide
changes in the received fields is accomplished by variation of the
voltages applied to the grid and the screen of the first detector.
The voice-frequency equipment, in addition to the desk telephone
H F FILTCR
i^'' OET. ANO BEATING OSC.
Fig. 14 — Ship receiver schematic diagram.
set located in the subscriber's booth, comprises a technical operator's
position located adjacent to the ship's receiver, and an attendant's desk
located on a lower deck in a room adjacent to the subscriber's booth.
The control equipment consists of repeaters, volume control devices,
and volume indicators, by means of which the levels of the incoming
and outgoing signals can be properly adjusted. Keys are provided
which enable the technical operator to talk either over the radio circuit
or to the ship subscriber. The booth attendant has facilities by which
he can talk either to the subscriber or to the control operator and has a
connection with the ship's telephone system for the purpose of locating
persons on the ship and calling them to the radio telephone booth.
RADIO TELEPHONE SERVICE TO SHIPS AT SEA 425
The subscriber's booth is provided with a desk telephone set having a
high-grade transmitter. The outgoing and incoming circuits are
shielded from each other and brought separately to the transmitter and
receiver of the subscriber's set. An illustration of the subscriber's
booth on the Leviathan is shown in Fig. 15.
Fig. 15 — Subscriber's booth on Leviathan.
The Wavelength Situation and Simultaneous Telephone
AND Telegraph Operation
Communication between ship and shore is carried out by the use of a
pair of frequencies, one for transmission in each of the two directions,
separated from each other by about 3 per cent. The specific frequen-
cies which were first assigned by the Federal Radio Commission to the
shore station and the Leviathan were necessarily chosen on more or less
426 BELL SYSTEM TECHNICAL JOURNAL
of a makeshift basis, in the absence of any comprehensive wavelength
plan for this new service. The Commission has recently had under
study the setting up of more adequate provisions for ship-to-shore
telephone channels, whereby it is hoped a series of frequencies may be
designated for telephone service exclusively and whereby there may
be established the relation between the telephone and the telegraph
frequencies necessary for the avoidance of interference between the
two services. Especially is coordination of the two sets of frequencies
necessary on the larger vessels, in order that simultaneous telegraph
and telephone service may be given without mutual interference. On
the larger liners simultaneous use of the radio telephone and radio
telegraph service must be provided for. This means that the trans-
mitters of both services must keep accurately on their frequencies and
be free of spurious components, and that the receivers must be highly
selective. It further entails that the transmitting and receiving
frequencies in each of the two cases be so coordinated that the trans-
mission frequency of one service does not lie too near the receiving
frequency of the other, and bespeaks a considerable amount of mutual
cooperation between the operating agencies involved. Difficulties of
fitting in the two services were encountered in the early work on the
Leviathan and, although the problem has not been worked out to final
solution, sufficient progress has been made, in cooperation with the
engineers of the Radio Corporation of America, to enable the telegraph
and telephone services to be conducted simultaneously without undue
interference.
In view of the fact that ships of a number of nations are already pre-
paring to give radio telephone service on the transatlantic routes and
with the probability of this service also extending to other parts of the
world, it would appear to be a matter of importance that the whole
question of marine frequency allocations be worked out in the near
future not merely on a national but also on an international basis.
Transmission Results
The transmission results which have been obtained with the Levia-
than on her first trip of commercial service are summarized in Fig. 16.
It will be noted that practically continuous 24-hour communication
was maintained for distances within 1000 miles of the shore, correspond-
ing to two days out. The service at greater distances was more inter-
mittent. This was largely due to the fact that during this first trip the
effort was concentrated on covering reliably the more important nearer-
in distances, and the ship was not prepared to transmit on frequen-
cies above 8 megacycles. The service proved to be much in demand
RADIO TELEPHONE SERVICE TO SHIPS AT SEA
427
EASTBOUND
DEC. 1929
TIME OF DAY — EST
AM NOON PM
6 8 10 12 2 4 6
8 10
NO,
OF
tALLS
-r
"T"
n r
NOON
DIST.
AMBROSE
SUN. N.Y.-SHIP
8 SHIP- NY
LEFT N.Y
12 PM
^^r^
10
MON. NY -SHIP
9 SHIP- NY
13
TUE. N.Y-SHIP
10 SHIP-NY.
I 8 _ iHIGH N OISE
13
HEAVY CRACKLES'
WED. NY-SHIP
II SHIP-N.Y
FIGURES bENOTE APPROXIMATE
FREQUENCY IN MEGACYCLES
NOISE
NOISE
THUR. N.Y-SHIP
12 SHIP-NY
LOW FIELDS, NOISE
?I?i?L
FRI. 13
LOW FIELDS, NOISE --^-r"^—
8 ' 4
ARRIVED CHERBOURG
170
7 20
1270
1820
2390
WESTBOUND
TUE. 17
WED. N.Y-SHIP
18 SHIP-N.Y
THUR, N.Y-SHIP
19 SHIP-N.Y
FRI, N.Y-SHIP
20 SHIP- NY
SAT N.Y-SHIP
21 SHIP-N.Y
SUN. N.Y-SHIP
22 SHIP-NY
MON. N.Y-SHIP
2 3 SHIP-NY
TUE. N.Y-SHIP
24 SHIP-NY.
TOTAL
ENTIRE N>: -SHIP
TRIP SHIP-NY
LEFT CHERBOURG ®
LOW FIELDS, ship's MOTOR NOISE --5--I-?. —
LOW FIELDS — T--r— -
4 ' 8
CRACKLES^"^®^--
HIGH NOISE--
QRM ship's SET i^
LOW FIELDS
8 ' 4
l4iai 4 NOISE
8 W 8
-„„^...„. ^^, QRM0N4MCn
QRM SHIP S SET 13 I 4 iS iM i8i4
.CRACKLES LOW-FllilDSa CRACKLES ^ '« ' ^
13
xl*_A_l.
QRM SHIPS SET^
± i L 8 | 4\ 8_
ARRIVED
AMBROSE LIGHT
_L
J_
a_
_L
_L
_l_
246 8 10 12 2 46 8
AM NOON PM
TIME OF DAY —EST
10
18
28
32
46
151
2640
2100
1530
1110
680
195
DOCK
Fig. 16 — Transmission results between 5. 5. Leviathan and Xew York.
428 BELL SYSTEM TECHNICAL JOURNAL
by the passengers, as is indicated by the number of calls completed
each day, particularly on the return trip. A similar number of test
and demonstration calls was made during the voyage. The calls
were completed without undue delay, there being only one ship in-
volved, and a fairly high grade of communication was obtained.
In conclusion it will be realized that the solution of the techincal
problem of ship-to-shore telephony is now well in hand and has been
carried to the point of having proved the practicability of giving this
service. Further problems are naturally arising in carrying the devel-
opment into more general effect, particularly operating problems and
those concerned with the international coordination of the service.
The indications are that the larger transoceanic ships will be rather
generally equipped for telephony and that the service will become one
of permanent value in the maritime field.
A General Switching Plan for Telephone Toll Service
By H. S. OSBORNE *
This paper outlines a comprehensive plan for improved switching of long
haul toll telephone traffic in the United States and Eastern Canada. A brief
discussion is given of the methods of designing the toll plant to give ade-
quate transmission efficiency for all connections established in accordance
with this plan. This includes a new method of providing amplification at
intermediate switching points replacing the cord circuit repeatei method.
ON January 25, 1915, telephone service was, with due ceremony,
inaugurated between the Atlantic and Pacific Coasts of this
country. This occasion marked a great step forward both technically
and commercially. Before that time, the limit of practicable telephone
transmission had been about 1,500 miles. The transcontinental
service was made possible by the completion of numerous important
developments and particularly by the perfection of telephone repeaters
and of means for applying them to long wire circuits.
Until then the Pacific States and their neighboring states had been
isolated telephonically from the eastern and midwestern parts of the
country. The demonstration of commercially practicable telephone
circuits across the continent gave a great impetus to the idea of
universal service, that is the provision of a telephone plant such that
telephone service could be given at commercially attractive rates
between any two telephones in the country.
In the fifteen years since the opening of the first transcontinental
circuits, the ideal of universal service has to a large extent been
realized. Practically all the telephones of the United States and a
large part of Canada now have provision for connection with the
countrywide toll telephone network, more than 99 per cent being
included. To achieve universal service, however, involves a great
deal more. Circuits must be provided in such numbers and so
arranged that connections between any two telephones can be estab-
lished quickly and without too many intermediate switching points.
Also the telephone plant must be designed for such standards of
transmission that these connections, when established, permit satis-
factory conversation. In general, the technical advances which have
been made during the last fifteen years to achieve the present standards
of toll service have been described from time to time before the
American Institute of Electrical Engineers, and it is not within the
scope of this paper to review them.
* Presented at Convention A. I. E. E., Toronto, June 1930.
429
430
BELL SYSTEM TECHNICAL JOURNAL
Associated Avith this development of the telephone plant has been
a very rapid increase in trat^c. Fig. 1 indicates this increase in
the United States and Canada since 1915. A striking characteristic
of this growth is that the increase has been much more rapid for the
longest lengths of haul than for the shorter lengths of haul. For
example, during the last five years in which the messages on lengths
of haul up to 250 miles approximately doubled, the messages on
hauls from 250 to 1,000 miles increased five times and those over
1,000 miles increased more than ten times. This characteristic is also
1
1000
800
600
400
200
1915
1920
1925
1930
( EST.)
Fig. 1 — Total toll messages in millions per year — Bell system.
illustrated in Figs. 2, 3, and 4 which show respectively the growth
in the number of circuits between Toronto and Detroit 240 miles in
length, Buffalo and Chicago 550 miles in length, and direct circuits
from New York and Chicago to the Pacific Coast, averaging about
2,500 miles in length. This particularly rapid growth in very long
haul trafftc has made it practicable to establish a considerable number
of long haul circuit groups and has greatly assisted in the problem of
handling satisfactorily calls between widely separated points. It has
led to the condition today in which 74 per cent of the long distance
(Long Lines) messages are handled over direct circuits and 20 per cent
with one intermediate switch.
SWITCHING PLAN FOR TELEPHONE TOLL SERVICE
431
The part of the business on which it is most difficult to give a high
grade of service is naturally the scattering business between widely
separated points. In these cases each item of traffic, that is the
business between two specific points, is relatively small but the number
of items of traffic is great. The number of messages involved in
each item of traffic does not justify direct circuits and in very large
LI
15
10
1915
1920
1925
1930
(EST.)
Fig. 2 — Growtli in number of toll circuits — Toronto to Detroit.
numbers of cases it is necessary, in order to provide a connection,
to make more than one intermediate switch. This applies at present
to six per cent of the long distance telephone business of the country.
All measures of the quality of service — speed, accuracy and trans-
mission — show that the difficulty of satisfactorily handling the service
increases rapidly with the number of intermediate switches involved.
The development of the toll business has led to a great increase in
the amount of business between large numbers of widely separated
points. There has also been an extensive trend toward concentration
of the plant used in handling the business in important toll offices
and along important routes. The long haul toll business is now
handled at about 2,500 "toll centers" out of approximately 6,400
432
BELL SYSTEM TECHNICAL JOURNAL
central offices in the United States and eastern Canada. Furthermore,
the technical developments in toll circuits have led to great increases
in the numbers of circuits along a given route. The extension of the
use of carrier telephone has increased the capacity of a 40-wire pole
line from 30 circuits to 70 circuits. On the heaviest toll routes,
moreover, circuits are now provided by means of toll cable construction,
a single cable carrying 200 or 300 circuits. During the past year
14
12
10
1915
1920
1925
1930
(EST.)
Fig. 3 — Growth in number of toll circuits — -Buffalo to Chicago.
or two the growth has been so rapid as to stimulate a very large
amount of construction of underground toll conduit routes, providing
in many cases for several thousands of telephone circuits on a single
route.
General Toll Switching Plan
The conditions outlined above form the background which has made
it both possible and important to adopt a new fundamental arrange-
ment for the layout of toll plant and the routing of toll messages.
This is called the "General Toll Switching Plan." The purpose of
this plan is to provide systematically a basic plant layout designed for
the highest practicable standards of service consistent with economy.
SWITCHING PLAN FOR TELEPHONE TOLL SERVICE
433
including speed, accuracy and directness of routing between any two
points in the country and suitable transmission standards. This
involves the layout of the plant in such a manner as to limit as much
as practicable the number of switches required for providing a con-
nection between any two telephones and the establishment of standards
of design and construction providing satisfactory transmission over
any route thus established. The plan is, therefore, of particular
50
40
30
20
10
1915
1920
1925
1930
(EST.)
Fig. 4 — Growth in number of toll circuits — New York and Chicago
to San Francisco, Los Angeles and Seattle.
value in improving the service conditions of switched toll traffic,
that is, traffic requiring the connection of two or more toll circuits.
The general features of the plan will be understood by reference to
Figs. 5 and 6. Figure 5 shows the application of the plan to a limited
operating area such, for example, as a State. Within the area are
selected a small number of important switching points designated as
"primary outlets." Each toll center is connected directly to at least
one of these outlets and all primary outlets within the area are directly
interconnected. This makes possible the interconnection of any two
toll centers within the area with a maximum of two switches and
within the part of the area served by one primary outlet, with a
maximum of one intermediate switch.
434
BELL SYSTEM TECHNICAL JOURNAL
The primary outlets were selected after a careful study of the
present switching and operating conditions and the probable develop-
ment of toll traffic within the various areas with a view to obtaining
the minimum number of primary outlets capable of handling the
traffic economically. The routings provided by the plan are supple-
mented by direct circuits, or by other routings where the amount of
business justifies such additional circuits as indicated by the dashed
lines in Fig. 5. In general the requirement is made that these supple-
mentarv routes shall be at least as satisfactory, both as regards
SOLID LINES - FUNDAMENTAL ROUTES OF
GENERAL PLAN
DASHED LINES - SUPPLEMENTARY DIRECT
CIRCUIT GROUPS
O PRIMARY OUTLET
• TOLL CENTER
Fig. 5 — General toll switching plan — application in local company area.
number of switches and transmission, as the routes provided by the
fundamental switching plan. However, when the supplementary
routes are used only as alternates to a primary routing they may be
somewhat less satisfactory in these respects.
The tentative selection of primary outlets is shown in Fig. 7. It is
interesting to note that it is found practicable to take care of switching
for the 2,500 toll centers of the United States and eastern Canada
by the establishment of approximately 150 of these as primary outlets.
For handling the business throughout the country eight of the
primary outlets are designated as regional centers, which are indicated
SWITCHING PLAN FOR TELEPHONE TOLL SERVICE
435
in Fig. 7. The method of routing calls is indicated by Fig. 6. Each
primary outlet is connected with at least one regional center and with
as many more as practicable. Each regional center is directly con-
nected to every other regional center in the country. By this means,
any one of the primary outlets, which are the 150 most important
switching centers in the country, can be connected to any other
primary outlet in the country with a maximum of two switches and
within the area served by a regional center with a maximum of one
intermediate switch. As an illustration of the concentration of
switching which results, New York serves as regional center for the
entire northeastern section of the United States and eastern Canada.
SOLID LINES - FUNDAMENTAL ROUTES OF # REGIONAL CENTER
GENERAL PLAN Q PRIMARY OUTLET
DASHED LINES- SUPPLEMENTARY DIRECT • TOLL CENTER
CIRCUIT GROUPS
Fig. 6 — General toll switching plan — illustration of interconnection of important
switching offices throughout Bell system.
The extent to which intermediate switching is limited by the
application of this plan is indicated by Fig. 8, which shows the maxi-
mum number of switches recjuired under the plan between different
types of toll centers. It is estimated that the percentage of long
haul messages requiring more than one intermediate switch will,
by means of this plan, be reduced by more than 50 per cent.
As an example of the benefit resulting from the adoption of this
plan between two remote points, consider a connection which was
requested between Pembroke, Ontario and St. Anthony, Idaho.
Under the old routing instructions such a call required intermediate
436
BELL SYSTEM TECHNICAL JOURNAL
'c
3
O
V--'-
1»
!.--#
1^
■|.°
!•
«
^
o
I.
i2 ^o
'I*
I ^,_
to
a:
~3
C
H
D
O
c
<
z
o
<
OJ
o
C
t')
o
III
ft
PC
a.
0)
•
•
u
tn
■a
^""""^
C
i
rt tn
,'
C -1-1
£ 2»
,/'■
/
.c/5
o
c
.9
03
O
O
nJ
.S
O
■^
i3
c
o
be
SWITCHING PLAN FOR TELEPHONE TOLL SERVICE
437
switches at Ottawa, Toronto, Chicago, Denver, Salt Lake City,
Pocatello and Idaho Falls, a total of seven. The chance of establishing
such a connection within satisfactory limits of time was, of course,
relatively small and the resulting circuit, when established, did not
permit the conversation to be held. Under the general toll switching
plan, this call will be routed with switches at Ottawa, New York,
Denver and Pocatello, a reduction of three switches. Furthermore,
the circuits involved in this connection will be designed with such
transmission standards as to give satisfactory conversation.
To—
From
Same Regional Area
Another Regional Area
Re-
gional
Center
Pri-
mary
Outlet
Toll
Center
Di-
rectly
Con-
nected
to
Re-
gional
Center
Toll
Center
Di-
rectly
Con-
nected
to
Pri-
mary
Outlet
Re-
gional
Center
Pri-
mary
Outlet
Toll
Center
Di-
rectly
Con-
nected
to
Re-
gional
Center
Toll
Center
Di-
rectly
Con-
nected
to
Pri-
mary
Outlet
Regional Center
Primary Outlet
Toll Center (directly
connected to Regional
Center)
1
1
1
2
1
1
2
1
2
2
3
1
1
2
1
2
2
3
1
2
2
3
2
3
3
4
Toll Center (directly
connected to Primary
Outlet)
Fig. 8 — Maximum number of switches under general toll switching plan.
The routes provided by the plan for countrywide service are also
supplemented by more direct routes of equivalent or better service
characteristics in cases where the amount of business is sufficient to
make this economical. Furthermore, the routes to regional centers
are, in some cases, supplemented by alternate routes through what
are called "secondary outlets." These are distinguished from the
primary outlets in that they do not necessarily have direct circuit
connections to all toll centers in their areas but serve a useful purpose
as an alternate route for the toll centers connected to them.
The essential features of the general toll switching plan from the
standpoint of the interconnection of the switching offices may be
summarized as follows:
Regional Centers
Regional centers are switching offices strategically located to cover
the various parts of the country and completely interconnected
with direct circuits, thus forming the basis of a countrywide
toll network.
438 BELL SYSTliM TKCIINICAL JOURNAL
Primary OtUlets
Primary outlets are switching offices having direct circuits to one or
more regional centers and each having direct circuits to all toll
centers in the area for which it is the primary outlet. Also,
each primary outlet is connected to every other outlet within as
large an area as practicable, usually within a State.
Supplementary Offices
Secondary Outlets
Secondary outlets are switching offices having direct circuits to one
or more regional centers and are intended primarily to furnish
alternate routes for toll centers for reaching the regional centers,
thus providing a greater degree of flexibility in the plant.
Secondary Switching Points
Secondary switching points are additional switching offices intended
to provide routes which are more direct thus reducing back haul
for intra-area business.
Transmission Considerations of General Toll Switching Plan
An important part of the development of the plan was the determi-
nation of proper transmission requirements such that any toll con-
nection established in accordance with the plan would have satisfactory
transmission efficiency.
Before the perfection of telephone repeaters, the provision of satis-
factory transmission efficiency depended largely upon limiting the
total attenuation loss of the complete circuit. At the present time
the perfection of repeaters has practically removed that limitation.
For example, the attenuation in a New York-Chicago circuit in cable
is such that without the use of repeaters the ratio of input power to
output power for speech currents transmitted over the circuit would
be 10^^, while by the use of repeaters at the terminals and at 17 inter-
mediate points the ratio actually is 10.
The removal of the limitation formerly set by circuit attenuation
makes possible the increase of the efficiency of circuits to the limit
determined by some other characteristic of the circuit. There are
various things which under different conditions may determine this
limit. One is the effect on transmission of echoes, namely, portions
of the speech currents reflected back from the distant end of the
circuit or from intermediate points. Another is the distortion due to
the building up of greater transmission gain at certain frequencies
than at others, which effect may result if repeaters introduce too
SWITCHING PLAN FOR TELEPHONE TOLL SERVICE 439
great an amplification into the circuit. As an extreme case, this
might result in a sustained oscillation or singing on the circuit. Other
effects which may be important are those of crosstalk between tele-
phone circuits, or of noise induced in the telephone circuits from
outside sources, both of which are increased by increasing repeater
gains. On the longer connections, echoes are almost always the
controlling factor, whereas on the shorter connections, such effects as
crosstalk, singing and noise generally are limiting. A reduction in
any of these effects generally involves more expensive types of con-
struction.
The difference between the attenuation loss of the circuit and the
total transmission gain introduced into the circuit by repeaters is
spoken of as the net equivalent. For long telephone circuits it is
generally economical to provide sufficient repeater gain so that the
circuit can be operated at the minimum net equivalent permissible,
this minimum equivalent being determined by the transmission
factors just mentioned. Therefore, in establishing satisfactory trans-
mission efficiencies for the overall toll connections in accordance with
the toll switching plan, each link must be designed on the basis of
the minimum working net equivalent which it will contribute to an
overall connection made up of several circuits switched together.
The establishment of satisfactory and economical transmission
requirements for the toll circuits laid out in accordance with the plan
involves the following steps:
a. The establishment of satisfactory overall net transmission equiva-
lents.
h. The coordinated design of all classes of toll circuits, and of the
subscribers' circuits, toll switching trunks and tributary trunks
connected to them, in such a way that the desired overall
transmission standards will be given at a minimum total cost
when suitable transmission gains are provided by repeaters
in the toll circuits and at toll switching points.
c. The economical and satisfactory distribution of transmission gain,
permitting all toll circuits to be operated at their minimum
net equivalents when this is desirable.
The overall transmission equivalents to be given under the plan
are based on standards which have heretofore been used for a large
part of the toll business but which it has been impracticable to meet
in many cases between widely separated points. With the means
now available for operating circuits at their minimum working net
equivalents, it was found that satisfactory overall transmission
29
440
BELL SYSTEM TECHNICAL JOURNAL
equivalents could be provided under the plan even for the maximum
number of switches using standards for the construction of toll circuits
very comparable with those already applied to new circuits. Ex-
pressed in terms of the transmission reference standard, the plan set
up gives a maximum of 25 db overall equivalent within one inter-
connected area (two intermediate switches) and a maximum of 31 db
between any two telephones of the United States and eastern Canada.
In order to determine the most economical distribution of these
overall equivalents, a study was made based upon the estimated
total number of toll circuits of each class in 1932 and their distribution
by length. It is also necessary to include the corresponding estimates
for the plant between the toll office and the subscriber, the loss in
this part of the plant being on the average about half of the overall
net equivalent of the connection.
Based upon these estimates, it was possible to determine, by an
economic study, the distribution of the overall minimum net equivalent
between these various parts of the circuit which would give minimum
total expenses. The toll terminal losses and the minimum net equiva-
lents for toll circuits established in this way are shown in Fig. 9.
Classification of Toll Circuit Involved
Minimum
Working Net
Loss of Toll
Circuit — db
Maximum
Via Operating
Equivalent
— db
Transmission
Margin — db
Toll Center to Primary Outlet
Toll Center to Regional Center
Primary Outlet to Regional Center
Regional Center to Regional Center
Primary Outlet to Primary Outlet
Toll Center to Toll Center
3.0
3.5
3.5
4.0
4.0
6.0
9.0
7.0
4.0
4.0
3.0
3.0
3.0
6.0
+ 1.0
+ 0.5
- 0.5
- 1.0* ■
- 1.0
Direct Toll Circuit (for terminal use
only)
Toll Terminal Loss
* Circuits equipped with echo suppressors may be designed with greater negative
margins.
Fig. 9 — Transmission design data of general toll switching plan.
In addition to the circuits involved in multi-switch business, the
studies connected with this plan necessarily include circuits used for
terminal business only, and others for which switching is limited to
a single intermediate switch at points where transmission gain is not
required. These circuits are associated with the plan because the
portions of the circuit between the toll center and the subscriber are
common for these circuits and for circuits directly involved in the
general plan. Design standards for these classes of circuits are also
shown in I'ig. 9.
SWITCHING PLAN FOR TELEPHONE TOLL SERVICE 441
Provision of Transmission Gain at Intermediate
Switching Points
The third step mentioned previously is the determination of the
best distribution of repeater gains to permit the individual circuit to
be operated by itself or in conjunction with other toll circuits at
approximately the minimum net equivalent as determined by the
several effects mentioned previously. In so far as the gain of repeaters
permanently inserted at intermediate points in a toll circuit is con-
cerned, this is a matter of economical design of the circuit and has
been adequately covered in other papers. We are interested here,
however, in considering the provision of gain at the intermediate
switching points when two toll circuits are connected together.
As indicated previously, echo effects are usually controlling on the
longer connections, whereas crosstalk, singing and noise will usually
control on the shorter connections. This is due to the fact that for
the great majority of toll circuits the echo effects on individual circuits
increase more rapidly with length than do crosstalk and noise. Singing
tendencies also increase at a rapid rate with increase in length on
two-wire circuits but tend to be independent of length on four-wire
cable and carrier telephone circuits which are used to a large extent
to provide the circuits between the primary outlets and regional
centers and between the regional centers. Furthermore, when two or
more toll circuits are connected together, the echo effects of the indi-
vidual circuits add together almost directly, whereas the effects of
crosstalk, singing and noise increase at a much less rapid rate. The
result of these general considerations is that when a toll circuit is
switched to another toll circuit, the overall combination can, in
general, be operated at a lower net equivalent as determined by
echo effects than the sum of the two circuits when operated individually
in which case the minimum equivalent is determined by the crosstalk,
singing and noise effects. Therefore, it is necessary in the case of
connections built up by connecting together a number of toll circuits
to introduce repeater gain at the intermediate switching points.
If gain were not introduced at intermediate points, it would be neces-
sary in order to obtain the same overall results on connections involving
more than one toll circuit to design and build a considerably more
expensive type of toll circuit plant in which the crosstalk, singing
and noise effects would be greatly reduced.
In the past, gain was inserted at intermediate switching points by
the use of cord circuit repeaters. These familiar devices consisted
of telephone repeaters inserted in the cord circuits and associated by
means of double plugs with the toll circuits and with individual
442 BELL SYSTEM TECHNICAL JOURNAL
balancing networks designed for each toll circuit. By this means
intermediate gains of from 4 to 10 db were inserted at the switching
points when connection was made between two toll circuits.
The use of cord circuit repeaters has been an outstanding element
in the provision of improved transmission on switched connections.
It has, however, some disadvantages which have increased in impor-
tance with the increase in transmission efficiency of circuits and with
the rapid development of toll business. The routine for inserting the
cord circuit repeaters when needed is necessarily somewhat cumber-
some, involving considerable expense for operators' labor and for
increased use of the toll circuits by operators. Furthermore, under
practical conditions it was found to be not possible to insure that the
cord circuit repeaters would always be used when required by the
routing instructions.
Recent developments in the types of toll circuit have greatly
increased the numbers of toll circuits provided with repeaters at their
terminals as a part of the most economical design of a circuit. When
such repeaters are available, the desired switching gain can be obtained
by making use of the gain available in these repeaters. The great
increase in the number of terminal repeaters required for other reasons,
important reductions in the cost of repeaters and the savings of
operators' labor and circuit time have made it practicable to adopt a
plan of providing, at certain points, terminal repeaters for every
circuit, thus doing away entirely with cord circuit repeaters at these
points. With the terminal repeater arrangement, the insertion of
transmission gain on switched connections is done automatically by
taking out of each circuit on such connections a section of artificial
line. This is, of course, the equivalent of increasing the gain of the
terminal repeater.
Satisfactory transmission results for all connections under the
general toll switching plan involve the insertion of repeater gain on
all connections switched at important switching points. This will be
carried out by the terminal repeater plan just described. The artificial
lines or pads which are cut out of the circuit on switched connections
have losses of from 1 to 4 db, depending upon the circumstances of
each case. This means that when two toll circuits are switched
together, from 2 to 8 db is automatically subtracted from the con-
nection at each switching point. The arrangement is indicated
schematically in Fig. 10. The design of each circuit must, of course,
be such that when either end of the circuit is connected to a subscribers'
station, the repeater gain at that end will not be greater than that
permissible under the terminating condition, but that when two or
SWITCHING PLAN FOR TELEPHONE TOLL SERVICE
443
more of such circuits are connected together for a long built-up toll
connection, the complete circuit will operate at as nearly as practicable
its minimum working net equivalent. While under these conditions
the permissible values of the pads associated with the terminal re-
peaters naturally vary in individual cases, it has been found possible
to work out for general use a series of values which should give satis-
factory results. These are indicated in Fig. 11. It will be noted
that these values are such that a circuit switched at both ends to
other toll circuits is operated at either .5 db or 1 db less than its
minimum working net equivalent, this deficit being made up by a
corresponding margin at the ends of the circuit. For example, by
reference to Fig. 11, it will be noted that whereas the design values of
the three intermediate links of a five-link connection equate to 11 db,
these links will contribute a total loss of only 9 db. On the other
F1G.1
3
o-vwv-ixi-
SWITCHING
OFFICE
3 3
-{XHvw\r-o o-vwHXH
SWITCHING
OFFICE
3 3
-N-^/w\^-o o^vwHXh
3
MXHww-o
FIG.2 o-^wWKl-
3 3
KXhVWV-O CMAA/V
rM-
3 3
-IXhVWV-O O-AAftA,
rM-
3
-[Xpvwv-o
FIG. 3 o-
3 3
-tXl-]V\AAr-0 o-^wv
rN-
Fig. 10 — Illustration of typical transmission data of terminal repeater-;-switching
pad method of operation. Fig. 1^ — -Circuits between switching pad offices in terminal
condition. Fig. 2- — Circuits of Fig. 1 interconnected at switching pad offices.
Fig. 3 — Connection between non-pad offices switched at pad office.
hand, the end links will contribute a total of 8 db, whereas their
design values equate to only 6 db. The 2 db marginal deficiency in
the intermediate links is compensated for by the 2 db marginal surplus
in the end links. When intermediate links are used as end links in
built-up connections, the switching pads at the terminating ends
restore the necessary positive margins.
The design of the very long intermediate circuits, such as some of
those connecting two regional centers, requires special consideration
and treatment to meet the transmission requirements specified. By
making use of a fundamental feature of four-wire circuits equipped
with echo suppressors and by employing circuits with the highest
velocities of propagation for this purpose, these circuits may be
designed in practically all cases to contribute not more than the desired
444
BELL SYSTEM TECHNICAL JOURNAL
operating equivalent for an intermediate link. Four-wire circuits
equipped with echo suppressors are unique in that at the longer
circuit lengths the increase in minimum net equivalent with further
increase in length becomes very slight.
Two general arrangements for removing the switching pads from
and restoring them to the toll line circuits are available depending
upon the type of switchboard facilities involved. Either arrangement
PO
3 3* is>
iVVV^Wp
QNGC
■ TC
«v
SAME AREA
DISTANT AREA
DISTANT AREA
Wr| DISTANT AREA
PO
INTRA AREA
TC
(3)[-0
SAME AREA
■"PO
TC=TOLL CENTER
P0= PRIMARY OUTLET
RC= REGIONAL CENTER
NGC = NON-GAIN SWITCHING
CENTER.
( ) MINIMUM WORKING NET LOSS MAXIMUM TOLL CIRCUIT EQUIVALENT I 7 db
O OPERATING VIA EQUIVALENT MAXIMUM OVERALL CONNECTIONS 3ldb
[] TRANSMISSION MARGIN ASSUMED LIMITING TOLL TERMINAL LOS5--7db
*VALUE or PAD IN TERMINAL LINKS DEPENDENT ON NOISE AND CROSS-TALK CONDITIONS
Fig. 11 — Diagrammatic representation of transmission data for handling switclied
toll traffic under general toil switcliing plan.
requires the modification of both the toll line circuit and the switch-
board circuits. One method controls the switching pad by a marginal
relay in the sleeve of the toll line circuit. In the other arrangement,
the pad is under the control of relays operated by battery supplied
from a simple.x bridge in the connecting circuits.
With the general toll switching plan the number of places in which
switching gain is required is greatly limited, being, as pointed out
above, a total of about 150 out of 2,500 toll centers. This number
will be somewhat increased by secondary switching points in which
it is found economical to insert switching gain in order to save the
back-haul involved in following the routing provided by the plan.
However, the net result is that under the toll switching plan the number
of points at which switching gain is provided will be materially
limited, with corresponding economies.
SWITCHING PLAN FOR TELEPHONE TOLL SERVICE 445
Programming the Establishment of the General Toll
Switching Plan
The full application of the general toll switching plan involves a
large number of individual rearrangements of plant layout, the
establishment of certain new circuit groups and the rerouting of a
considerable amount of switched business, the conversion of the
switching offices to the terminal repeater arrangement, and the
modification of the transmission requirements of certain of the circuits.
The date at which these rearrangements will be completed is naturally
different for different sections of the country and is determined by
the regular program of plant additions and rearrangements to take
care of increasing business and of needed service improvement. The
existence of a comprehensive plan of this sort insures that the program
of rearrangements as carried out will be along the lines of greatest
economy and maximum improvement in service. The present plans
of the telephone companies in the United States and Canada indicate
that the plan as now established will be very closely approximated
by the actual plant in the course of about five years.
Future View
Such a plan as has just been discussed is naturally not a static
thing but is subject to continual modification to bring it into corre-
spondence with changed conditions. In connection with such changes
it is of interest to consider briefly the probable long time trend of the
development of the plan.
One possible ultimate development would be the increasing con-
nection of primary outlets to a single regional center so that ultimately
only one regional center would be necessary. If this were to take
place, the regional center would undoubtedly be Chicago. Fig. 12
is interesting as showing the extent to which the primary outlets
already are connected directly with Chicago, over one half of them
having such direct connection.
If Chicago ultimately became the only regional center, it would
reduce the maximum number of switches to three. It seems evident,
however, that such a plan would have many disadvantages. It seems
clear that with such an arrangement, numerous secondary regional
centers would be necessary to avoid uneconomical back-haul of large
amounts of traffic, and the economies of such an arrangement do not
look promising. Furthermore, it would lead to a tremendous con-
gestion of through switching at one point, this congestion going far
beyond the limits of economical concentration and leading to serious
operating difficulties.
446
BELL SYSTEM TECHNICAL JOURNAL
>
J3
■^s
71
^1
c
0)
ujq:
o
cca.
>— I
u
a
3
O
u
be
3
O
o
bo
c
d
O
'So
c
3
O
3
bo
In
(J
'%
CO
03
i-i
0)
c
O
be
■ «-«
SWITCHING PLAN FOR TELEPHONE TOLL SERVICE 447
A second, and it is believed more promising general trend would
result from the gradual increase in the number of regional centers as
the continued development of business makes this economical. With
this growth would come also a continued increase in the number of
toll centers connected directly to a regional center. By this process
there would be a continued growth in the number of toll centers
which can be interconnected with a maximum of two intermediate
switches, and it is possible that ultimately the primary outlets can
drop out of the picture completely, giving a maximum of two inter-
mediate switches for the entire country. While any such outcome is
evidently many years away, it seems probable that it is along these
lines the growth in development of the plan should be directed.
Although this direction of development avoids the congestion which
would be brought about by the single regional center plan, even under
this plan the rapidly growing amount of toll switching to be done in
large metropolitan centers offers a very important problem for the
future. Toll switching at these points is rapidly outgrowing the
capacity of a single manual switchboard, as the switching of local
calls did long ago. Equipment changes are being made which increase
this capacity, but they can be but a temporary relief. Looking to
the future, an increasing amount of the outgoing traffic will be handled
by operators in the local central offices, reaching the toll line over
toll tandem trunks. It is evident, however, that the ultimate solution
of the problem will involve the use of machine methods for the selection
of the toll line by the operators, as is now done in certain segregated
toll tandem systems.
The entire trend of recent years is thus to decrease the differences
between the handling of exchange messages and of toll messages.
At the present time more than 95 per cent of the toll messages are
completed while the subscriber remains at the telephone, with speeds
of completion only slightly slower than those of exchange messages.
Transmission standards, while naturally somewhat better for the
shorter distances involved in exchange messages, are, nevertheless,
rapidly becoming very comparable. The present view of trends for
the future is for continuation of this process, perhaps even to the use
of similar types of machine equipment at central offices for switching
the various classes of messages.
The author gratefully acknowledges the assistance of many of his
associates in the preparation of this paper, and particularly of Mr.
J. V. Dunn.
Image Transmission System for Two-Way Television*
By HERBERT E. IVES, FRANK GRAY and M. W. BALDWIN
A two-way television system, in combination with a telephone circuit,
has been developed and demonstrated. With this system two people can
both see and talk to each other. It consists in principle of two television
systems of the sort described before the June, 1927, Convention of the
American Institute of Electrical Engineers. Scanning is by the beam
method, using discs containing 72 holes, in place of 50 as heretofore.
Blue light, to which the photoelectric cells are quite sensitive, is used for
scanning, with a resultant minimizing of glare to the eyes. Water-cooled
neon lamps are employed to give an image bright enough to be seen without
interference from the scanning beam. A frequency band of 40,000 cycles
width is required for each of the two television circuits. Synchronization is
effected by transmission of a 1275 cycle alternating current controlling
special synchronous motors rotating 18 times per second. Speech trans-
mission is by microphone and loud speaker concealed in the television
booth so that no telephone instrument interferes with the view of the face.
DITRING the past few years, since the physical possibihty of
television has been established, the chief problems which have
received attention have been those of one-way transmission. In
particular, the experimental work in radio television has had for its
principal goal the broadcasting of television images, which is inher-
ently transmission in one direction. At the time of the initial de-
monstration of television at Bell Telephone Laboratories in 1927,^
one part of the demonstration consisted of the transmission to New
York of the image of a speaker in Washington simultaneously with
the carrying on of a two-way telephone conversation. At that
time it was stated that two-way television as a complete adjunct to
a two-way telephone conversation was a later possibility. It is the
purpose of this paper to describe a two-way television system now
set up and in operation between the main offices of the American
Telephone and Telegraph Company at 195 Broadway and the Bell
Telephone Laboratories at 463 West Street, New York. It con-
sists in principle of two complete television transmitting and re-
ceiving sets of the sort used in the 1927 one-way television demonstra-
tion. In realizing this duplication of apparatus, however, a number
of characteristic special problems arise, and the paper deals chiefly
with matters peculiar to two-way as contrasted with one-way tele-
vision.
* Presented at June, 1930, meeting of A.I.E.E., Toronto, Canada.
' Bell System Technical Joitnial, October, 1927, ]>]>. 551-652.
448
IMAGE TRANSMISSION SYSTEM 449
Physical Arrangement and Operation
The detailed description of the optical and electrical elements of
the two-way television system will be more readily grasped if it is
preceded by an account of the general arrangement of the parts and
of the method of operation of the system from the standpoint of the
user.
The physical arrangement of the two-way television system is shown
by the pictorial sketch Fig. 1, and in the photographs Fig. 2 and
Fig. 3. The terminal apparatus is largely concentrated into a booth,
— the television booth — similar in many respects to the familiar
telephone booth, and a pair of cabinets, which contain the scanning
discs and light sources. As in the 1927 demonstration, scanning
is performed by the beam method, the scanning beam being derived
from an arc lamp whose light passes through a disc furnished with a
spiral of holes and thence through a lens on the level of the eyes of the
person being scanned. The light reflected from the person's face is
picked up by a group of photoelectric cells for subsequent ampli-
fication and transmission to the distant point. The signals received
from the distant point are translated into an image by means of a
neon glow lamp directly behind a second disc driven by a second motor
placed below the first and inclined at a slight angle to it. The two
discs, which are shown in the center cabinet of Fig. 2, are of slightly
different sizes; the upper one 21" in diameter and the lower one 30".
They difl'er from the discs used in the earlier demonstration in that in
place of the 50 spirally arranged holes formerly used, they carry 72
holes whereby the amount of image detail is doubled. While with the
earlier "50 line" picture recognizable images of a face were obtainable,
the aim in this new development was to reproduce the face so clearly
that there would never be any doubt of recognizability, and so that in-
dividual traits and expressions would be unmistakably transmitted.
This doubled number of image elements necessarily requires, for the
same image repetition frequency (18 per second) twice the transmission
band, or approximately 40,000 cycles as against 20,000 for the 1927
image.
The only part of the television apparatus visible to the user is the
array of photoelectric cells which are in the television booth behind
plates of diffusing glass. In addition to the photoelectric cells and
their immediately associated amplifiers, the booth contains a con-
cealed microphone and loud speaker. By means of these, the voice
is transmitted to the distant station and received therefrom without
the interposition of any visible telephone instrument which could
obscure the face.
450
BELL SYSTEM TECHNICAL JOURNAL
in
•■'ri:!'^ .»
,• '- -. ■ -;.'--.^ ' -^-;n ' --'-l'~. ' :^.' .'.?-W^
a.
^
i
a.
J~>
O
>
CD
*
<
0)
_J
z
Id
1
Z
O
.•t;
I
(/)
Q.
■«
LJJ
in
-J
UJ
^
H
(0
_i
-+
_i
UJ
CD
c
.2
'33
'E
■(-'
%
■i->
o
•g
1
in
!2
'C
o
r,
d
CU
u
1
X
r-(
Q.
^
>s
u
-I-"
LxJ
_1
O
LU
^
h-
1-
o
^
z
^
<
v
LU
z
z
O
^
X
%
a.
UJ
_j
LU
p
h-
CD
z
m
<
o>
u
c:
LU
2
<
IMAGE TRANSMISSION SYSTEM
451
From the standpoint of the user, the operation of the combined
television and telephone system is reduced to great simplicity. He
enters the booth, closes the door, seats himself in a revolving chair,
swings around to face a frame through which the scanning beam reaches
his face, and upon seeeing the distant person, he talks in a natural
tone of voice, and hears the image speak. Conversation is carried on
as though across a table.
Fig. 2 — The three major cabinets of the television-telephone apparatus.
Optical Problems
Some of the more special of the problems encountered in two-way
television are primarily optical in character. The principal one is
that of regulating the intensity of the scanning light and of the image
which is viewed so that the eye is not annoyed by the scanning beam
or the neon lamp image rendered difficult of observ^ation. It has
been necessary for the solution of this problem to reduce the visible in-
tensity of the scanning beam considerably below the value formerly
used and to considerably increase the brightness of the neon lamp.
The means adopted consists first, in the use of a scanning light
of a color to which the eye is relativ^ely insensitive but to which
photoelectric cells can be made highly sensitive. For this purpose
blue light has been used, obtained by interposing a blue filter in the
452
BELL SYSTEM TECHNICAL JOURNAL
path of the arc light beam, and potassium photoelectric cells specially
sensitized to blue light and more sensitive than those previously
used have been developed. The number of these cells and their area
has also been increased over those used in the earlier television
apparatus so that the necessary intensity of the scanning beam is
decreased.
The second half of the problem, namely that of securing a max-
Fig. 3 — Interior of the television booth.
imum intensity of the neon lamp, has been attained by the develop-
ment of water-cooled lamps capable of carrying a high current.
The net result of the use of the blue light for scanning, of more
sensitive photoelectric cells, and of the high efficiency neon lamps is that
the user of the apparatus is subjected only to a relatively mild blue
IMAGE TRANSMISSION SYSTEM 453
light sweeping across his face, which he perceives merely as a blue spot
of light lying above the incoming image. Figure 3 shows the in-
terior of the television booth with the frame through which the
observer sees the image of the distant person.
A second optical problem is the arrangement of the photoelectric
cells required in order to obtain proper virtual illumination of the
observer's face. As we have previously pointed out in discussing
the beam scanning method,^ the photoelectric cells act as virtual
light sources and may be manipulated both as to their size and
position like the lights used by a portrait photographer in illum-
inating the face. In the present case, it is desired to have the whole
face illuminated and accordingly photoelectric cells are provided to
either side and above. One practical difficulty which is encountered
is that eyeglasses, which often cause annoying reflections in photog-
raphy are similarly operative here. For this reason, it is important
that the photoelectric cells be placed as far to either side or above
as possible. The banks of photoelectric cells shown in Fig. 3 are
accordingly much farther removed from the axis of the booth than
were the three cells used in the first demonstration. In the position
which has been chosen for the cells, reflections from eyeglasses are
not annoying unless the user turns his face considerably to one side
or the other.
The number of cells has been so chosen as to secure a good bal-
ance of effective illumination from the three sides and it has been
found desirable to partly cover the cells on one side in order to aid
in the modelling of the face by the production of slight shadows in
one direction.
Another optical problem is the illumination of the interior of the
booth. There must, of course, be sufficient illumination for the
user to locate himself, and it is also desirable that the incoming
image and the scanning spot be not seen against an absolutely black
background. The illumination of the booth is by orange light, to which
the cells are practically insensitive, and so arranged that the walls and
floor are well illuminated. In addition to the wall and floor illumination,
a small light is provided on the shelf bar in front of the observer so
as to cast orange light on the front wall surrounding the viewing
frame. This light contributes materially to reducing the glaring
effect of the scanning beam, and to the easy visibility of the incoming
image.
In addition to the optical features which are visible to the person
sitting in the booth, there are very necessary optical elements which
^Jul. optical Soc. of America, Mardi, 1928, p. 177.
454
BELL SYSTEM TECHNICAL JOURNAL
have to do with the positioning of the outgoing and incoming images.
A practical problem which is encountered when customers of various
heights use the apparatus is that the scanning beam, if fixed in its posi-
tion, would strike too high or too low upon many faces. In order
to direct the beam up or down as is required, a variable angle prism,
consisting of two prisms arranged to rotate in opposite directions,
Fig. 4 — Optical means for controlling heights of scanning and viewing beams.
is interposed in the path of the scanning beam. This prism, which
lies directly in front of the projection lens used with the upper disc,
is shown in Fig. 4 at P. Its rotation is controlled by a knob with a
numbered dial. The exact position is determined by the operator
by reference to a monitoring image which will be described below.
Another optical element which serves two purposes, is a large con-
vex lens lying between the receiving disc and the observing frame,
IMAGE TRANSMISSION SYSTEM
455
shown at L, Fig. 4. This lens is used to magnify the incoming image
to such a size that the image structure is just on the verge of visibiUty,
under which condition the face of the distant person appears as though
he were approximately eight feet away. In addition to acting as
a magnifier, this lens serves to position the incoming image to fit the
height of the user. For, by raising or lowering it by means of a
knob, the operator, using the information as to the observer's height
obtained from the position of the scanning beam, locates the lens so
that the virtual image appears in the proper position.
Photoelectric Cells and Associated Circuits
The photoelectric cells used in this apparatus are similar in shape
to those used in the first demonstration, but somewhat larger. Each
100
80
I-
z
u
Cj 60
40
20
^
/c
'\
' 1
\
\
VIOLET
VGREEN
RED
0.3
0.4
0.5 0.6
WAVE LENGTH IN p.
0.7
o.a
Fig. 5 — .4. Relative optical transmission of the blue filter through which the
scanning beam passes. B. Relative sensitivity of the photoelectric cells to various
parts of the spectrum. C. Relative sensitivity of the eye to A-arious parts of the
spectrum.
cell is twenty inches long and four inches in diameter, giving it an
area of approximately eighty square inches for collecting light. The
anode is made in the form of a hollow glass rod wound with wire.
This construction prevents the electrical oscillations that would
otherwise result from mechanical vibrations of the anode. The sen-
sitive cathode consists of a coating, covering the rear wall of the tube,
of potassium sensitized with sulphur.'^ This kind of cell is consider-
ably more sensitive than the older type of potassium-hydride cell
3 A. R. Olpin, Phys. Rev., 33, 1081 (1929).
30
456
BELL SYSTEM TECHNICAL JOURNAL
while still having most of the sensitiveness in the blue region of the
spectrum. Figure 5 shows the response of the photoelectric cells
used to the various parts of the spectrum together with the trans-
mission of the blue filter and the brightness of the various parts of the
spectrum as evaluated by the human eye. The very great efficiency
of the photoelectric cells and the inefficiency of the eye to the light
used are apparent.
To amplify the photoelectric current, the cells are filled with argon
at a low pressure. Photoelectrons passing from the sensitive film to
the anode ionize the gas atoms along their paths and thus cause a
greater flow of current. The ionization of the gas does not, however,
instantaneously follow sudden variations of the true photoelectric
emission from the sensitive film, that is, there is a time lag in the
If)
_i
ffl
u
UJ
Q
+ 4
+ 2
10
(0
O
lU
>
i<
C- EQUALIZED
,,- --^OUTPUT
^~^^^^^^^r"^
\
"^"\
^^^^ A- SINGLE
^ ^^ CELL
''^ "v.
\B- GROUP OF
\ CELLS
\
100
10,000
FREQUENCY IN CYCLES PER SECOND
1,000
100,000
Fig. 6 — Loss in response of photoelectric cells at high frequencies.
ionization of the gas and in the disappearance of ionization. This
lag results in a relative loss and phase shift of the high frequency
components of a television signal with respect to the low frequency
components which become serious in the wider frequency range
utilized in the 72 line image. The relative loss in output from a
single large photoelectric cell at high frequencies is shown in decibles
by curve A of Fig. 6.
In the television booth, the twelve large cells mounted in the walls
of the booth present an area of approximately seven square feet to
collect light reflected from a subject. To secure the desired effective
illumination, the cells are mounted in three groups, comprising a
group of five cells in each of the two side walls of the booth and a
group of two cells in the sloping front wall above the subject. The
twelve cells are enclosed in a large sheet copper box, provided with
IMAGE TRANSMISSION SYSTEM
457
doors to each group. The cells of each group are connected in par-
allel through the input resistance of a two stage resistance-capacity
coupled amplifier similar to those previously used. This raises the
level of the signal to such a point that the output of the three ampli-
fiers may be carried through shielded leads and connected in parallel
to a common amplifier.
The metal anodes and lead wires of the cells in parallel in any
one group give an appreciable capacity to ground, which results in a
further loss in amplitude and phase shift of the high frequency com-
ponents of the signal. The combined loss introduced by ionization of
the gas in the cells and by capacity to ground is shown by curve B of
Fig.
7 — Schematic of interstage amplifier coupling to equalize for the high frequency
losses in the photoelectric cells.
Fig. 6. This combined loss is equalized by an interstage amplifier
coupling, Fig. 7. The equalized output from the photoelectric cells
is shown by curve C, Fig. 6.
Two-Way Image Signal Amplifiers
The vacuum tube system used to amplify the photoelectric cell
currents in two-way television is patterned closely after that used
previously in one-way television, and the description here will be con-
fined chiefly to novel features. These new features are necessary to
take care of the doubled frequency band which results when the
scanning is done with a 72-hole disc rather than with a 50-hole one,
and to provide sufficient power to operate the high intensity neon
lamp which is essential to two-way television. Certain other new
features have been introduced in order to simplify the apparatus and
to reduce the maintenance required to keep it in good working con-
dition.
The vacuum tubes which operate at low energy levels are the
so-called "peanut" type, chosen because of their freedom from
microphonic action and their low interelectrode capacities. Protection
against mechanical and acoustical interference is secured by mounting
these tubes in balsa wood cylinders which are loaded with lead rings
458
BELL SYSTEM TECHNICAL JOURNAL
and cushioned in sponge rubber. The tubes are electrically connected
in cascade by means of resistance-capacity coupling, so that the
whole amplifier system is stable over long periods of time and is also
uniformly efficient over the required frequency band. Grid bias for
the small tubes is supplied by the potential drop across a resistance
in the filament circuit; the power requirements for the low level
stages of the amplifier are filled by 6-volt filament batteries and 135-
volt plate batteries, all located externally where they can be checked
and replaced conveniently.
The amplifier system is divided into units of convenient size as
shown in Fig. 8. Associated with each of the three banks of photo-
o
a.
O
lUiO
U _J
Oo
O
X
0.
CC UJ
o_i
CLU
llJ
2-1
(£0.
"is
O
I- 111
t£
WtL
<l
Zuj
<N
UJ-J
_l<
mo
<o
r<'
-i
r^
r" t"
r
(r<
^sz[
0.
<
Z
o
z
i;;^:^
m
+40
+20
-20
-40
y
y
■ 1
y
/
-60
/
-80
y^
-lOO
10^
10 1
1 CO
O
-2 >
10 -^
10-4
10"^
Fig. 8 — Schematic diagram of the complete television channel and the relative
voltage levels of the signal along the channel.
electric cells is a two-stage unit known as the photoelectric cell amp-
lifier; the combined output of these three units is carried to a four-
stage unit known as the intermediate amplifier whose output is of
sufficiently high level to be carried outside the copper cell cabinet to
the three-stage transmitting power amplifier on the relay rack. A
four-stage unit known as the receiving power amplifier is also on the
rack, and serves to amplify the signal from the other station to a
level which will yield an image of satisfactory contrast when it is
translated into a light variation by means of the neon lamp. The
final stage of this amplifier consists of two special 250-watt tubes in
IMAGE TRANSMISSION SYSTEM 459
parallel. These large tubes are used because their plate impedance
is of the same order of magnitude as the impedance of the neon lamp,
and because they will supply the necessary direct current to the neon
lamp without overheating.
Figure 8 also shows what may be termed a voltage level diagram for
the whole system. Ordinates on this diagram represent voltage
amplitudes at the junctions between units of the system, and by
themselves tell nothing at all about the power conditions in the system,
since the impedances are not specified. It is interesting to observe
that the signal voltage produced by the three banks of photoelectric
cells has an effective value of about 50 microvolts across the 50,000
ohm input resistance; the transmitting amplifier delivers about 1 volt
to the 125-ohm cable circuit, and the receiving amplifier delivers
about 100 volts to the 1,000 ohm neon lamp circuit. The signal cur-
rent through the neon lamp has an effective value about a thous-
and million times greater than that of the current variation in one
of the photoelectric cells.
The most outstanding contribution to the development of tele-
vision amplifiers is the combination of output and input transformers
whose transmission characteristics are shown in Fig. 9, A, and whose
impedance characteristics are shown in Fig. 9, B, and C. The ex-
ceptionally wide frequency range, corresponding to a ratio of. limiting
frequencies of 5,000 to 1, transmitted by these transformers is
due largely to the use of chrome permalloy, a recently developed
core material having very high permeability. The improved char-
acteristics are also the result of refinements in design which involve
the use of adjusted capacities and resistances to control the character-
istics at the higher frequencies. Due to the fact that each terminated
transformer looks like a resistance of 125 ohms over practically the
entire frequency range of the image signal, it makes no difference in
the form of the overall voltage amplification characteristic of the cir-
cuit whether the transformers are connected together directly or by
means of the equalized cable circuit whose characteristic in shown in
Fig. 10. Advantage of this circumstance is taken in providing switch-
ing means whereby each transmitting amplifier may be connected
through a resistance pad to its local receiving amplifier, enabling a
person to see his own image in the television booth, which is a conven-
ience in making apparatus adjustments.
Transformers of this type must be carefully protected against
magnetizing forces which might cause polarization of the core material.
In order to keep the plate current of the final tube of the transmitting
power amplifier from flowing through the winding of the output
460
BELL SYSTEM TECHNICAL JOURNAL
+12
+11
+10
o +9
_i
ill
(D
UJ
Q -18
-19
-20
-21
10
INPUT
_,
^,_ — - —
' ■ ■'
-^
OUTPUT
~~ — ' — ^^
^
■*v.
100
FREQUENCY
1,000 10,000
IN CYCLES PER SECOND
100.000
175
150
125
100
^75
^50
25
-25-
-50
RESISTANCE
^^
■
REACTANCE
^
B
100
FREQUENCY
1,000 10,000
IN CYCLES PER SECOND
100,000
175
150
125
100
-75
I
O 50
25
-25
-50
RESISTANCE
^ —
^~^
.^^
"
REACTANCE
■ ■
10
100
FREQUENCY
1,000 10,000
IN CYCLES PER SECOND
100,000
Fig. 9 — A. Voltage ratio characteristics of VV-TSTQ input transformer and \V-7880
output transformers, each connected between its rated impedance. B. Impedance
characteristic of W-7880 output transformer with 1765 ohm resistance load. C.
Impedance characteristic of \V-7879 input transformer with 20 mmf. capacity load.
IMAGE TRANSMISSION SYSTEM
461
transformer, the transformer winding is shunted by a battery and a
resistance in series. The resistance is made high, so that the trans-
mission loss due to bridging it across the circuit is small; the voltage
of the battery is made equal to the potential drop across the resistance
due to the plate current of the tube, so that the average voltage
across both the battery and the resistance, and hence across the trans-
former winding, is zero.
A vacuum thermocouple is connected in series with the line winding
of the output transformer, serving as level indicator for the trans-
mitting amplifier. The level indicator for the receiving amplifier is
12
II
10
-
U
u
WITH EQUALIZER^
\
—
/
/
/
WITHOUT EQUALIZER^
^
y
i
^ —
10
100 1,000 10,000
FREQUENCY IN CYCLES PER SECOND
100,000
Fig. 10 — Insertion loss characteristic of cable cricuits which transmit the image
signal, measured between 125 ohm resistances.
a vacuum thermocouple in series with the grid resistance of the two
250-watt tubes.
The electrical control panels associated with one terminal of the
television apparatus are shown in Fig. 11.
Transmission Circuits
Two special requirements for the two-way television transmission
circuits are to be emphasized. The first, which has already been
referred to, is the wide frequency transmission band, from 18 cycles
to 40,000 cycles, which must have a high degree of uniformity of trans-
mission efficiency and freedom from phase distortion. The second is
the necessity for two circuits for the television images. This arises
from the fact that the two parties to the conversation must both see
and be seen at all times. There can be no interruption of one face by
the other, comparable with the alternation of the role of speaker and
listener in telephony which permits the use of a single circuit for
ordinary speech communication.
The terminal stations of the two-way television system are con-
462
BELL SYSTEM TECHNICAL JOURNAL
nected by eight underground circuits, each consisting of 13,032 feet
of No. 19 gauge and 390 feet of No. 22 gauge non-loaded cable. Two
circuits are used for transmitting the image signals, two for the accom-
panying speech, one for the synchronizing current, two are used as
Fig. 11 — Control apparatus panels associated with one terminal of the television
apparatus.
order wires, and one is kept as a spare. All of the circuits have identical
transmission characteristics, but equalization is necessary only on the
two which carry the image signals. Figure 10 shows the insertion loss
characteristic of each circuit, and also shows the insertion loss charac-
IMAGE TRANSMISSION SYSTEM
463
teristic of the image circuits when the image line equalizers are in-
cluded.
Although the distance between the stations is small the require-
ments of the television system from the standpoint of freedom from
noise and other interference require that considerable care be given
to the selection of the cable circuits used. All terminal connections
are made through balanced repeating coils or transformers so that all
of the circuits are balanced to ground. Also, in order to insure that
the crosstalk between the various channels be unnoticeable the
terminal equipment is so adjusted that approximately the same amount
of power is transmitted by each circuit.
Neon Lamps and Associated Circuits
After amplification, the received television signal is impressed on
the grids of two power tubes in parallel to furnish current for a neon
receiving lamp. The terminal lamp circuit is shown in Fig. 12.
» _
X
Fig. 12 — Schematic of neon lamp circuit.
The grid bias of the two power tubes is varied by the operator to
control the DC plate current, which replaces the original DC signal
component suppressed at the sending end. The quality of the re-
produced image is determined by the operator's control over the
relative levels of the incoming AC signal and the restored DC current.
The television current from the power tubes is translated back into
light by a water-cooled neon lamp designed to operate on a large
current. The structural details of the lamp are shown in Fig. 13.
Heavy metal bands attach the rectangular cathode to a hollow glass
stem occupying the central portion of the tube. Water from a small
circulating pump flows continuously through the glass stem and cools
the cathode by thermal conduction through the metal bands. To
reduce sputtering of the cathode and consequent blackening of the
464
BELL SYSTEM TECHNICAL JOURNAL
glass walls, the front surface of the cathode is coated with beryllium.
This metal resists the disintegrating action of the glow discharge very
satisfactorily and gives the lamp a prolonged life. Other metal sur-
faces in the tube are shielded from the discharge by mica plates; and
*
Fig. 13 — Water-cooled neon lamp.
the discharge passes from the frame-like anode to the front surface of
the cathode, covering it with a brilliant layer of uniform cathode glow.
Pure neon in a plate type of lamp gives a very inferior reproduction
of an image. The impedance of the lamp is relatively high and com-
prises both a resistance and a reactance which vary with frequency.
The variation in the impedance causes a relative loss in the frequency
components of the signal and also introduces spurious phase shifts.
IMAGE TRANSMISSION SYSTEAI 465
In addition, pure neon has an after-glow; the gas continues to glow
for an appreciable time after current ceases to flow. This after-glow
casts spurious bands of illumination out to one side of the brighter
image details.
A small amount of hydrogen in the neon prevents such an afterglow;
and at the same time improves the circuit characteristics of the lamp.
The total impedance of the lamp is lower, making it a less influential
part of the lamp circuit; and the resistance and reactance vary in
such a manner that the phase shift is more nearly proportional to
frequency (a phase shift proportional to frequency causes no distortion
in the reproduction of an image). Other active gasses may be used
with the neon to improve the operation of a television lamp, but one
or two per cent of hydrogen is most satisfactory.
Since hydrogen is absorbed by the electrodes in a glow discharge, it
slowly disappears from the neon during operation of the lamp. For
this reason the lamp is provided with a small side reservoir of hydrogen.
The lamp and the reservoir carry porous plugs immersed in a pool of
mercury; and a flexible rubber connection permits the two plugs to be
brought into contact at will. Minute quantities of hydrogen may be
introduced into the lamp by simply bringing the two plugs into con-
tact for a short time.
Even with this improvement the circuit characteristics of a lamp are
not ideal. W'ith power tubes it is usually desirable to include a
fixed resistance in series with the lamp to prevent semi-arcing con-
ditions. Such a resistance also makes the lamp a less influential
fraction of the total circuit impedance.
Optical Monitoring System
In order to insure that the incoming and outgoing images are prop-
erly positioned, no matter what the stature of the person sitting in the
booth, and that the images shall be of proper quality, it is essential to
have some means for the operator to observe and adjust these images.
The optical monitoring system provided consists of an outgoing mon-
itor and means for adjusting the scanning beam, and an incoming
monitor and means for adjusting the position of the viewing lens to
suit the height of the sitter.
The outgoing monitoring system is the same as that used in the one-
way television apparatus which has already been described. A small
neon lamp (Fig. 14, at bottom of top disc) is placed behind the sending
disc but displaced several frames from the aperture through which the
arc lamp beam passes. By continuing the spiral of holes part way
around it is possible to see the complete image from the auxiliary neon
466
BELL SYSTEM TECHNICAL JOURNAL
lamp, to which the outgoing signals are also supplied. In order to see
this monitoring lamp from the operator's position, a right-angle prism
and a magnifying lens are placed in front of the disc and the image is
observed through an opening in the side of the motor cabinet. The
Fig. 14 — Sending and receiving discs, with neon lamps and optical arrangements
for image monitoring.
task of the operator is to direct the scanning beam up or down by
means of the variable angle prism until the face of the person in the
booth is centrally located. This adjustment is facilitated by a wire
IMAGE TRANSMISSION SYSTEM 467
which passes across the image and is placed at the height at which the
user's eyes should appear.
The height of the observer's eyes is an indication of the position
which should be taken by the large magnifying lens L, and the operator,
after having properly placed the scanning beam, reads the scale on the
variable angle prism dial, and then sets the magnifying lens by turning
its controlling knob to the same number. When both adjustments
are complete, the person in the booth will not only be properly scanned
but will be in the best position to see the image.
In order to monitor the incoming image, an optical arrangement is
adopted by means of which light from the water-cooled neon lamp is
taken off at the side and reflected through the disc and thence reflected
again, as shown in Fig. 14 (top of bottom disc), through a second,
lower, observing hole on the side of the motor cabinet. Because of
the small area of the side view of the neon lamp, a lens system is
inserted which focusses the image of the lamp at the place to be
occupied by the pupil of the operator's eye. When the eye is properly
placed, the whole of the lens area is seen filled with light and exhibits
the incoming image.
In addition to the monitoring means just described, an additional
view of the incoming image is provided by means of a 45° mirror
which is carried on the back of a movable shutter which is shown
at S in Fig, 4. This shutter carries an illuminated sign on the side
turned to the user with the inscription "Watch this space for
television image." The shutter with its sign covers the image until
the adjustments just described are made, when it is dropped out of
sight. While it is in place, the operator is provided with an additional
monitoring image reflected from the 45" mirror. This view is, of
course, in every respect identical with that which the user sees.
The function of the incoming monitoring system is primarily to
enable the operator to set the electrical controls to give the proper
quality of image. He also has another task which is that of properly
framing the image. This he can do by turning the framing handle,
which is described elsewhere, while watching the image from the 45°
mirror. This framing operation is preferably performed not on a
person sitting in the booth but upon some suitable object such as a
mirror located upon the rear door of the booth. In order to make
this framing adjustment, the operators at the two terminals set their
scanning beam dials to predetermined positions such that the scan-
ning beams place the framing mirrors at the lower edge of the scanning
rectangles, the phases of the incoming discs are then shifted until the
images of the mirrors are seen properly located in the incoming mon-
itors.
468 BELL SYSTEM TECHNICAL JOURNAL
Signalling System
In order to coordinate operations at the two terminal stations, an
order wire system is provided. There are four telephone sets at each
station; one on the attendant's desk in the ante-room, one concealed
inside the television booth, one in the control room, and one at the
control panels for the technical operator, who operates the small
switchboard which is part of the system. Two of the underground
cable circuits connect the two switchboards, so that there may be not
more than two separate conversations between stations at one time.
Ringing is accomplished by means of standard 20-cycle ringing current
furnished by the Telephone Company.
During a demonstration, the attendants' telephones are connected
permanently over one of the cable circuits. To relieve the operators
of the duty of ringing each time the attendants wish to communicate,
a push button and buzzer are provided at each attendant's desk,
operated by the standard ringing currents simplexed on the synchro-
nizing circuit. This arrangement leaves the operators free to manip-
ulate the television apparatus.
The two order wire circuits are each simplexed to provide two
additional circuits which operate signal lamps indicating to both
operators when either chair in the television booths is occupied and
turned in position.
Discussion
The primary objects in developing and installing the two-way
television system have been two. The first was to obtain information
on the value of the addition of sight to sound in person to per-
son communication over the telephone. The second was to learn
the nature of the apparatus and operating problems which are
involved in a complete television-telephone service. While the in-
stallation is entirely experimental, it is being maintained in practically
continuous operation for demonstration to employees and guests of the
Telephone Company, and interesting data are being gathered on all
aspects of the problem.
It may be said without fear of contradiction that the pleasure and
satisfaction of a telephone conversation are enhanced by the ability
ot the participants to see each other. This is, of course, more evident
where there is a strong emotional factor, as in the case of close friends
or members of the same family, particularly if these hav^e not been seen
for some time.
Were the television apparatus and required line facilities of extreme
simplicity and cheapness it would be safe to predict a demand for its
IMAGE TRANSMISSION SYSTEM 469
early use. At the present time, however, the terminal apparatus is
complex and bulky, and requires the services of trained engineers to
maintain and operate it. In addition to the cost of the terminal
apparatus there is the unescapable item of a many-fold greater trans-
mission channel cost. Because of the wide transmission bands re-
quired for the television images, the inherent necessity for a television
channel in each direction, and the extra channels for synchronizing
and signalling, the total transmission facilities used in this demonstra-
tion are those which could, according to current practice, carry about
fifteen ordinary telephone conversations. It is to be expected, of
course, that development work will result in some increase in the
efficiency of the transmitting channels and in simplifications of the
terminal apparatus. It is conceivable, therefore, that our present
conception of the cost of the whole system may ultimately be materially
changed.
Synchronization System for Two-Way Television *
By H. M. STOLLER
In a previous paper presented before the June, 1927 Convention of the
American Institute of Electrical Engineers, the method of securing syn-
chronization of television signals was described as employed in the Bell
System Television demonstration of April, 1927. The present paper de-
scribes the development of a new control circuit which is in use in the new
two-way television system between the Bell Telephone Laboratories at 463
West Street and the American Telephone and Telegraph Company build-
ing at 195 Broadway, New York.
TELEVISION transmission requires not only synchronization of the
transmitting and receiving equipment but such synchronization
must be held to a narrow phase angle so that the scanning discs at the
transmitting and receiving end will never depart more than a small frac-
tion of a picture frame width from the desired position.^ In the 1927
demonstration, 2125 cycle synchronous motors were employed with
supplementary D.C. motors to facilitate starting. This plan required
the use of vacuum tube amplifiers of large size in order to supply suf-
ficient power to the synchronous motors.
Such high frequency synchronous motors, however, are inefficient
and expensive, so that when designing the new system, it was desired
to solve the problem of synchronization with simpler and cheaper
equipment and in a manner which would require less attention in
starting. It was particularly desired to employ a motor which could
be operated directly from the 110 volt lighting circuit without any
auxiliary "A", "B" or "C" batteries for the control equipment.
Description of Motor
Figure 1 shows a photograph of the new television motor and its
associated control equipment.
The motor is a four pole compound wound D.C. motor with the
following special features added.
1. An auxiliary regulating field, the current through which is controlled
by the vacuum tube regulator.
2. A damping winding on the face of the field poles to prevent the field
fiux from shifting (Fig. 3).
* Presented at June, 1930, meeting of A.I.E.E., Toronto, Canada.
1 These requirements are more fully discussed in a previous paper. {Journal of
the A. I. E. E., Vol. 46, page 940, 1927.)
470
5 YNCHRONIZA TION S YSTEM
471
Three slip rings are provided at points 120 electrical degrees apart
for furnishing three phase power to supply plate and filament
voltage for the regulating circuit.
A pilot generator of the inductor type is built into the motor frame
and delivers a frequency proportional to the motor speed for
actuating the control circuit.
Fig. 1 — New television motor and vacuum tube control circuit.
5. A hydraulically damped coupling is provided between the motor
shaft and the scanning disc. (Fig. 4.)
The motor frame was made from a standard 36 tooth stator punching
by cutting out three teeth per pole, thus forming four polar areas of
six teeth each. The shunt, series and regulating fields enclose the
31
472
BELL SYSTEM TECHNICAL JOURNAL
entire polar areas. The damping winding consists of insulated closed
turns of heavy copper wire distributed over the pole faces in the slots
as shown in Fig. 3. It will be noted that this damping winding has
no effect upon the flux through the poles as long as the flux density
over the polar surface does not shift. In other words, the damping
winding permits the total flux of the motor to increase or decrease as
required by the regulating circuit but will oppose any tendency of the
flux to shift back and forth across the pole face. As will be explained
SCANNING DISK
Fig. 2 — Schematic diagram of control circuit.
later on, this feature is essential in order to prevent hunting or insta-
bility of the image.
The hydraulically damped coupling between the motor shaft and
the scanning disc is also essential in order to avoid hunting. It employs
flexible metal bellows filled with oil and connected by a small pipe
equipped with a needle valve for adjusting the amount of damping.
Figure 4 shows its construction. The scanning disc itself is centered
on a ball bearing which allows the disc to rotate with respect to the
shaft within approximately ± 5 degrees mechanical movement.
5 YNCHRONIZA TION S YSTEM
473
Control Circuit
Figure 2 shows a schematic diagram of the control circuit. When
the motor is operating at full speed the pilot generator delivers approx-
imately 1 watt of power at 300 volts, 1275 cycles to the plates of a pair
of push-pull detector tubes. The grids of these tubes are supplied
with an e.m.f. of the same frequency from an oscillator or other
source of power having a sufficiently constant frequency. The amount
of power required for this grid circuit is only a few thousandths of a
watt. The detector tubes rectify the plate voltage producing a po-
tential drop across the coupling resistance R]. If the plate and grid
voltages are in phase, so that the grids of the tubes are positive at the
SLOT INSULATION
Fig. 3 — Damping winding preventing shifting of field flux.
same instant that the plates are positive, the plate current will be a
maximum. If the grid voltage is negative when the plate voltage is
positive the plate current is practically zero, so that the magnitude of
this current is a function of the phase relationship between the grid
and plate voltages as shown in Fig. 5.
The voltage drop across the coupling resistance i?i is applied to the
grid circuits of three regulator tubes. These tubes derive their plate
voltage supply from a three phase transformer fed with power from the
three slip rings provided on the motor. These tubes act as a rectifier
whose output is controlled by the potential impressed upon the grids
from the coupling resistance R\. The current of the regulator tubes
is passed through the regulating field provided on the motor. This
field is in a direction to aid the shunt field and series fields of the motor.
474
BELL SYSTEM TECHNICAL JOURNAL
The operation of the circuit is as follows: In starting switch ^i is
closed which applies direct current to the shunt field and armature
circuits of the motor. The motor accelerates as an ordinary compound
wound motor. Switch S2 is then closed applying three phase power
from the slip rings of the motor to the transformer. As the speed of
the motor approaches the operating point, the beat frequency between
the pilot generator and the oscillator will cause beats in the current
through the regulating field which are visible on the meter Mi. Let us
assume that the field rheostat has been previously adjusted so that with
SCANNING DISK-22"DIA.
BELLOWS
m
ss
ss
Fig. 4— Hydraulically dampedcoupling'to prevent hunting of motor.
the shunt field alone the motor will tend to run slightly over the desired
operating speed. When the exact operating speed is obtained, the
beat frequency in the regulating field will be zero and as the motor
tends to accelerate, the phase relationship between the pilot generator
and the oscillator will reach a point tending to give maximum strength
to the regulating field. When this point is reached, the acceleration of
the motor will be checked by the increased field and the speed will tend
to fall until the phase of the pilot generator with respect to the
oscillator has shifted sufficiently so that the regulating field current is
5 YNCHRONIZA TION S YSTEM
475
reduced to an equilibrium value, after which the motor continues to
run at constant speed in accordance with the frequency of the os-
cillator.
Operating tests on the circuit show that the motor will hold in step
over line voltage ranges from 100 to 125 volts and will be self-synchron-
izing over somewhat narrower voltage limits. Thus, under normal
conditions all that is necessary from an operating standpoint is to
close the switch and wait for the motor to pull into step.
Control Oscillator
The control oscillator is a standard type of vacuum tube oscillator
having a frequency precision of the order of 1 part in 1000, when
PLATE
CURRENT
/effective
I VALUE
90°
PHASE ANGLE BETWEEN
PLATE AND GRID VOLTAGES
Fig. 5 — Phase detector tube characteristic.
180'
delivering the negligible output of .005 watts to the grid circuit of the
detector tubes. This frequency is delivered directly to the motor
circuits at one end of the line and is transmitted over a separate cable
pair to the control circuits at the other end of the line. It was found
that the detector tubes would operate successfully over a considerable
variation in power level, provided the minimum oscillator output was
sufficient.
An interesting alternative method was developed in which the
synchronizing channel between stations may be omitted entirely,
but this method was not used in the present system as the additional
cost was not justified. The method, however, is described as it may
prove of value if television transmission over long distances is con-
sidered.
476 BELL SYSTEM TECHNICAL JOURNAL
Mr. W. A. Marrison in his paper "A High Precision Standard of
Frequency," Proceedings I. R. E. July, 1929, described a crystal
controlled oscillator which would maintain a precision as to frequency
of 1 part in 10,000,000. This oscillator employs a quartz crystal as
its primary means of control and by means of secondary circuits the
natural period of the crystal, which is approximately 100,000 cycles,
may be stepped down to lower frequencies which are more convenient
for such purposes as motor control.
By this means, a frequency of the desired value may be obtained
with a precision so great that the speed of the scanning discs under
control of the above described circuit will be so nearly perfect that no
synchronization channel at all is required. For example, if the period
of observation of the television image is 5 minutes, the scanning disc
will make 5300 revolutions when operating at a speed of 1060 r. p.m.
Assuming a precision of control of 1 part in 10,000,000, the maximum
error during the 5 minute interval will be 5300 divided by 10,000,000
or about 1/2000 of 1 revolution. Expressed in degrees on the periphery
of the disc, this is equivalent to approximately 1/6 of 1 degree or since
the width of the television image with 72 holes in the scanning disc is
5 degrees, the image will drift 1/30 of a frame width during the 5
minute interval. If the speed of the scanning disc at the other end
drifts an equal amount in the opposite direction, the displacement of
the television image will be only 1/15 of a frame width, which is a
tolerable amount of drift.
From a practical viewpoint, however, it is apparent that the addi-
tional cost of very precise independent oscillators would be greater than
the cost of providing the synchronization channel, except possibly for
transmission over long distances.
Framing
Referring to Fig. 2, it will be noted that a phase shifter is provided
between the oscillator and the input terminals to the control circuit.
This phase shifter is designed with a split phase primary member
producing a rotating magnetic field. The secondary member is sin-
gle phase and is mounted on a shaft provided with a handle. By
rotating the handle of the phase shifter in the desired direction, the
frequency delivered from the phase shifter will be the algebraic sum
of the frequencies of the oscillator plus the frequency of rotation of the
armature of the phase shifter. It is, therefore, a simple matter for
the operator at the receiving end to momentarily increase or decrease
the control frequency and thus bring the picture into frame.
SYNCHRONIZATION SYSTEM 477
Discussion
During the development of the control system, one of the first
difficulties encountered was hunting of the controlled motor. The
problem of hunting, of course, becomes more difficult of solution the
greater the precision of speed regulation desired and the greater the
moment of inertia of the load connected to the motor, the latter state-
ment applying only to controlled systems of the synchronous type.
Since the moment of inertia of the scanning disc is large relative to
that of the motor armature, it is seen that the conditions for securing
stable rotation would be unfavorable in both the above mentioned
respects if the scanning disc were mounted directly on the motor shaft.
The hydraulically damped type of coupling above described was,
therefore, inserted between the motor shaft and the scanning disc.
It was found, however, that hunting still occurred. A further analysis
of the problem showed that the axis of the field flux of the motor was
shifting back and forth across the pole faces. The damping winding
shown in Fig. 4 was then added with a marked improvement. It was
also observed that a strong series field on the motor assisted in secur-
ing stability and it was, in fact, necessary to employ all three ex-
pedients to secure satisfactory performance. In the system as finally
developed the television image, if disturbed by a momentary load
such as the pressure of the hand against the disc, would come back to
rest within approximately one second, there being two oscillations
during this interval. In actual operation, it was found that the
normal fluctuations in line voltage occurring on the commercial power
supply produced no transients of sufficient magnitude to cause any
objectional instability in the received image.
In conclusion, it should be pointed out that this type of control
system could be equally well employed with larger motors for other
applications requiring precise speed regulation. While the circuit
described is applicable only to a direct current motor, a similar system
may be applied to an alternating current motor substituting a saturat-
ing reactor in place of the regulating field winding in the manner
described by the author in his paper ^ presented before the Society of
Motion Picture Engineers, September, 1928.
2 S. M. P. E. Transactions, Vol. 12, No. 35, page 696.
Sound Transmission System for Two- Way Television*
By D. G. BLATTNER and L. G. BOSTWICK
In this paper is described the speech transmission part of the two-way
tele\'ision system described in companion papers. The system is designed
to produce the best possible illusion of face-to-face communication between
speakers located at a distance. Some of the novel features of the system
described include the use of distant pick-up transmitters and loud speakers
concealed in the wall of the booth, also the use of heavy glass windows
through which the scanning beam and the reproduced image are projected
as a means of preventing the admission of noise into the booth.
IN the design of a sound transmission system to be correlated
with a visual system, the requirements as to perfection of results
desired are no more stringent than for other high grade sound rep-
producing systems ^ that have been described in the literature from
time to time. Rather in this case the peculiarities of the system are
largely those incidental to the adaptation of old technique to meet
new conditions.
The principal limitation of the sound system imposed by the visual
system is that the user be relieved of all necessity of holding a telephone
in close proximity to the head. Such a limitation is highly desirable
in order to secure the most natural pose of the features and the most
satisfactory scanning. Obviously, the best way of meeting this lim-
itation is by the use of telephone instruments of the type adapted for
picking up and reproducing sounds at a distance. The use of such
instruments has the further advantage that they can be located near
the vision screen and so reproduce any peculiarities in tone quality
that would result if the speaker were actually located at the position
of the image. Of, course, the sharpness of this perspective effect
obtained is influenced by the loudness of both the original and the
reproduced sounds but the matter of location of instruments is also
very important.
It would thus seem that the use of distant pick-up and distant pro-
jecting instruments offers certain rather fundamental advantages but
it is also true that it presents certain other disadvantages. One of the
disadvantages is that the distant pick-up microphone gives less output
than a close-up device because of the reduced sound pressure on the
diaphragm; also a sound producing device to give suitable reception
* Presented at June 1930 meeting of A. I. E. E., Toronto, Canada.
1 "Public Address Systems" by J. P- Maxfield and I. W. Green in A. I. E. E.,
Feb. 14, 1923. Also "High Quality Recording and Reproducing of Music and
Speech" by J. P. Maxfield and H. C. Harrison in A. I. E. E., Feb. 1926.
478
SOUND TRANSMISSION SYSTEM 479
at a distance must be supplied with a higher transmission level than
would a close-up instrument. It thus becomes necessary to provide
for greater gain in transmission and greater electrical power capacity
than would be required were the instruments held close to the head.
The use of the more elaborate transmission facilities is in itself dis-
advantageous but it also tends to increase the feed-back from the loud
speaker to the microphone; also the effect of any noise at the micro-
phone position or at the listening position tends to interfere more seri-
ously with the successful conduct of conversation. In the design of the
two-way television system recently installed between the Bell Telephone
Laboratories at 463 West Street and the American Telephone &
Telegraph Co. at 195 Broadway in New York City, it was felt that
it would be possible to overcome these technical objections to the
distant type instruments and that the advantages mentioned would
justify any measures necessary to do so.
The question of instruments was solved by the use of the Western
Electric 394 condenser type transmitter - and a dynamic direct radiator
loud speaker. The transmitter is one of the type generally used for
phonograph and sound picture recording and for other purposes where
good quality, high stability and quietness of operation are essential.
The direct radiator type of loud speaker was used instead of the usual
horn type because of the limited mounting space available. It con-
sists of a dynamic structure with a rigid duralumin diaphragm about
3" in diameter flexibly supported at the edge and radiating directly
into free air. \\'hile such a structure is not highly efficient and permits
of only a small sound power output these considerations are of second-
ary importance in this case. The instruments were located in the
front wall of the booth about 2' from the position of the user and ad-
jacent to the viewing screen in order to enhance the perspective as
described above, the microphone being above and the loud speaker below
as shown in Fig. 1. These instruments were (in this particular case)
connected into a four-wire circuit although in certain cases it might be
desirable to use a 2-wire circuit. Such a change would of course be en-
tirely feasible. The remainder of the apparatus used consisted of ampli-
fiers located at the transmitting end of each channel and an attenuator at
the receiving end, the two ends being connected by means of a loop of
approximately 3 miles of non-loaded non-equalized cable. The
amplifiers and the attenuators were each readily adjustable so that the
sounds of different speakers could be reproduced at the optimum loud-
ness. Observation of the performance of the system was made
possible in each of the control rooms by means of a monitoring head
^ E. C. Wentc in Physical Review of May 1922.
480
BELL SYSTEM TECHNICAL JOURNAL
type receiver bridged across the mid-point of an attenuator tying the
two channels together. The attenuation used in the monitoring cir-
cuit was such as to give no audible feed-back in either booth. The
results obtained with this set-up were considered satisfactory from the
standpoint of both volume and quality. Ready recognition of familiar
Fig- 1 — Microphone and loud speaker in position above and below television scanning
and viewing aperture.
voices and the association of the source of the reproduced sounds with
the image were the usual occurrence. Figure 2 shows in block form
the complete circuit set-up and Figure 3 shows the combined response
frequency characteristic of the microphone, amplifier and loud speaker.
SOUND TRANSMISSION SYSTEM
481
The ordinates of this curve represent variations in sound pressure from
the loud speaker for constant pressure on the transmitter diaphragm.
These data were obtained with the loud speaker located in a heavily
damped room. The measurements were made on the sound axis at a
distance of 2', representing the relative position of the observer under
conditions of actual use.
In setting up such a system the chief consideration is in regard to the
acoustic feed-back from the loud speaker to the microphone and in
this connection the design of the booth is an important factor. The
booth must necessarily be so shaped that the user, looking at the view-
ing screen, can be satisfactorily scanned and the light reflected from
463 WEST ST.
I CONTROL ROOM
CONDENSER
MICRO- I-
PMONE
CONDENSER
MICRO-
PHONE
AMPLIFIER
LOUD
SPEAKER
VOLUME
INDICATOR
MAIN
AMPLIFIER
MONITORING
RECEIVER
MONITORING
ATTENUATOR
195 BROADWAY
CONTROL ROOM
RECEIVING
ATTENUATOR
MONITORING
ATTENUATOR
MONITORING
RECEIVER
LOUD
SPEAKER
MAIN
AMPLIFIER
VOLUME
INDICATOR
CONDENSER
MICRO-
PHONE
AMPLIFIER
CONDENSER
MICRO-
PHONE
Fig. 2 — Circuit diagram for sound transmission system for two-way television.
the scanned areas will strike the banks of photoelectric cells required
for the reproduction of the visual likeness. This requires that the
person scanned be located in close proximity to the scanning disc and
to the photoelectric cells as well as to the microphone and loud speaker.
Such an arrangement is objectionable from an acoustic standpoint in
that in the present state of development the cells are necessarily large
and poor absorbers of sound. They thus tend to cause part of the
sound output from the loud speaker to reflect back into the microphone.
If the sound so reflected or fed back is equal or greater in magnitude
than the original sound picked up and is of the proper phase relation,
the system will "sing" and the sound system become of no practical
use. A further eff"ect of the design of the booth is that as a closed
cavity, it tends to cause sounds of a certain pitch range to be accen-
482
BELL SYSTEM TECHNICAL JOURNAL
tuated. To reduce these effects as far as possible, the television booths
were made as large as other considerations would permit and all surfaces
were covered where possible with acoustic absorbing material. They
have a floor area of about 35 sq. ft. and are about 8 ft. high. Because
of the increased transmission required for the proper interpretation
of sounds in the presence of noise, the booths were made of heavy
masonry material to insulate the user and the microphone from
the noise incidental to the rotating parts of the television apparatus.
It was thus necessary to project the scanning beam and to view the
illuminated image through a window located in the front wall. The
microphone and the loud speaker were fitted into this wall, which was
then covered over with a thin screen to improve the appearance as
(/I
■I- 10
u +5
UJ
O
z
o
1/1
lA
2
<n
z
<
tx.
-5
I- -10
z
0-I5
a.
20
j^r
.^ - \
\
^
I A
/I
j\/
\
^v
V
\7
^1/
U
\
r
^
100
3456789 2 3
1,000
FREQUENCY IN CYCLES PER SECONDS
5 6 7 8 9
10,000
Fig. 3 — Response frequency characteristic of microphone, amplifier and loud speaker.
shown in Fig. 3 of companion paper, "Image Transmission System for
Two-Way Television." These means effectively reduced the noise in
the booth to an unnoticeable amount and reduced reflection effects to
the extent that the average speaker talking in a conversational manner
could be reproduced at a loudness best suited to the general effect.
The optimum loudness seemed to be about the same as would be ob-
tained from the speaker direct at a distance of about 10 feet, the ap-
parent distance between the image and the observer. At this loudness
the gain of the amplifiers was 12 db less than that required to cause
singing.
While the system demonstrated was operated over a distance of only
a few miles, it will be appreciated that the same terminal facilities might
have been used over much greater distances. Thus for the first time
in the history of electrical communication it can be said that complete
freedom of exchange of both visual and aural expression between dis-
tant users of the telephone has been made possible.
Transmitted Frequency Range for Telephone Message
Circuits
By W. H. MARTIN
IN the Bell System the general objective which has been set up for
the transmitted frequency range for new designs of telephone mes-
sage circuits is a range having a width of 2,500 cycles, extending from
about 250 cycles to about 2,750 cycles. In determining the frequency
range of such a circuit, the cutoff points are taken as those at which
the attenuation reaches a value 10 db greater than that at 1,000 cycles.
This frequency range for design is taken in general as applying to
the overall transmission characteristic of the circuit between the ter-
minal central offices of a connection. Where such offices are connected
by a direct trunk this frequency range applies to the individual trunk.
Where two or more trunks are used in tandem the frequency range
of the overall connection will tend to be somewhat less than that of
the component parts. For this reason then, to meet the specified range
for an overall multi-switch connection, it will be necessary to have the
frequency ranges of the trunks which are used as parts of built-up con-
nections, somewhat greater than the specified range for the overall
circuit.
In view of the relatively lower cost of toll switching trunks and other
similar trunks from toll offices to local central offices, it is desirable that
these terminal circuits have a broader frequency range than that speci-
fied above, so as to avoid their narrowing the range transmitted by the
toll trunks with which they are connected.
It may be stated that the general purpose in working to the trans-
mitted frequency range given above is that each individual trunk
should have a frequency range at least as great as that specified and
that the frequency ranges of those trunks which are frequently used
as parts of built-up connections should be somewhat greater than the
specified range.
In setting up the requirements for the various transmission char-
acteristics of telephone message circuits, the aim is to arrive at the
combination of requirements which will give the most economical tele-
phone system for furnishing the desired grade of transmission service.
Since the effects of many of the factors entering into the determina-
483
484
BELL SYSTEM TECHNICAL JOURNAL
tion of this ideal are not susceptible to definite evaluation, the selection
of a requirement such as that for the transmitted frequency range, is
necessarily a matter of judgment, taking into account the various
factors involved.
In Figure 1 are given the results of recent tests showing the effect
upon articulation of varying the upper and lower cutoff frequencies
of a circuit similar to the Master Reference System for Telephone
Transmission. These results apply to the condition of no noise and
the received volume at the optimum value. It will be noted from the
curve for the variation of the upper cutoff frequency, that the rate
100
90-
i, 80
o
s: 70
2 60
O
< 50
o
40
<
uj 30
m
<
■^ 20
>-
in
10
~-
-
V
•s
y
y
-
- ■
HIGH PASS
FILTERS
•s
s
\
/
"^OW PASS
FILTERS
/
\
\
\
\
/
/
/
/
/
2 34568 2 34566
100 1,000 10,000
CUTOFF FREQUENCY OF FILTERS IN CYCLES PER SECOND
Fig. 1 — Syllable articulation of circuit similar to master reference system at optimum
received volume under quiet conditions.
of growth is relatively slower above 2,500 cycles than below, and that
the total gain in going from this point to infinity is relatively small.
Figure 2 shows on a somewhat different basis the upper part of this
curve and also, for comparison, corresponding data for circuits having
commercial terminal apparatus of the types used in the Bell System.
The ordinates for these curves are the ratios of the increase in articu-
lation in going from an upper cutoff frequency of 2,000 cycles to some
higher point, to the total change in articulation in going from 2,000
cycles to infinity. For example, referring to the curve for the effect
of upper cutoff frequency on articulation of the Master Reference
System, it is seen that the articulation for the 2,000-cycle point is 70
TRANSMITTER FREQUENCY RANGE
485
per cent, for the 3,000-cycIe point is 87 per cent and for infinity is 97
per cent. Increasing the cutoff from 2,000 to 3,000 cycles gives a
growth in articulation which is 17/27, or .63, of the total increase in
articulation which would be obtained in going to a cutoff of infinity.
The values for the other curves of Figure 2 are obtained in a corre-
sponding manner, it being appreciated that the articulation values with
commercial instruments are lower than those for the Master Reference
Circuit. This method of plotting the results has the advantage of
showing the rate of growth of articulation for the three kinds of circuits
on a comparable basis.
1.0
09
OB
07
0.6
Ob
0.4
0.3
02
0.1
2000 2400 2800 3200 3600 4000
'.UTOFF FREQUENCY OF LOW PASS FILTER IN CYCLES PER SECOND
DESKSTAND SETS^
.^
-'liANDSETS
.^
y
^
/
^
/*
^
l-^IRCUIT SIMILIAR TO
^/(ASTER REFERENCE SYSTEM
f
//
7
/
f
/
/
Ordinate is:
Af - A:
2000
Am - Aa
Where Af = the syllable ar-
ticulation with a low pass
filter of cutoff frequency f.
A2000 = the syllable articula-
tion with the 2000 cycle
low pass filter.
Am = the syllable articulation
obtained with no filters.
Fig. 2 — Syllable articulation of telephone systems at optimum received volume
under quiet conditions.
It is seen from the curves of Figure 2 that raising the upper fre-
quency limit from 2,000 to 2,500 cycles gives about one-half of the
total increase which would be obtained in going to an infinite cutoff
and raising to 2,750 cycles gives for the commercial instruments about
two-thirds of the increase in articulation which would be obtained in
going to an infinite cutoff. These curves do not indicate any particular
cutoff frequency as a stopping point for commercial circuits but it is
considered that going as far as about 2,750 cycles is justified. While
there is some articulation advantage in going further, observations of
the number of repetitions occurring in conversations over circuits
having different cutoff frequencies have indicated but little reduction
in repetitions by going beyond about 2,750 cycles with commercial
types of terminal sets.
For the lower end of the range, the lower cutoff frequency curve of
Figure 1 shows little effect on articulation of cutoffs below 400 cycles.
486 BELL SYSTEM TECHNICAL JOURNAL
The selection of the 250-cycle point for the specified frequency range
is on the basis of maintaining reasonable naturalness.
It has been found that with present commercial station sets little is
gained either in intelligibility or naturalness by extensions of the trans-
mitted frequency range beyond the limits which have been set. This
range, moreover, permits effective utilization, particularly from the
standpoint of intelligibility, of the capabilities of much better station
instruments even if this improved apparatus should approach the ideal
in performance. With such terminal apparatus, major extensions
beyond the upper frequency limits give improvements from the stand-
point of naturalness largely as the result of better reproduction of the
fricative consonants and of some of the incidental sounds which accom-
pany speech. The^extension necessary to effect a material improve-
ment in this respect is a matter of a thousand cycles or more, rather
than hundreds of cycles. It has been considered that such an extension
for message circuits is not now justified.
Further in this connection, it must be borne in mind that an exten-
sion of the transmission range will in general increase the amount of
noise on the circuit and magnify the crosstalk problem. For trans-
mission systems such as carrier and radio where the noise may be
assumed to be uniformly distributed over the sideband range, the added
noise may be particularly important. Also a widening of the range
increases the difficulties of securing proper impedance balances and of
equalizing amplitude and phase distortion.
On the basis of these considerations, it has been decided that new
designs of telephone message circuits for the Bell System should have
an effective transmission band width of at least 2,500 cycles, extending
from about 250 to 2,750 cycles. Furthermore, this band width will be
made greater in those cases where this can be accomplished without
material increase in costs.
Some Recent Developments in Long Distance Cables
in the United States of America
By A. B. CLARK
THE transmission history of long distance circuits, and particularly
long distance cable circuits, has been one of continually improving
standards. It has also been one of continual reduction of circuit costs.
These have resulted largely from new developments to which have
been added economies resulting from heavy growth and improved
engineering.
To put it another way, present-day circuits are capable of trans-
mitting a kilocycle of frequency range more cheaply than those of
earlier days. As the cost per kilocycle of band width has been reduced,
part of the cost reduction has naturally been used in furnishing tele-
phone customers wider-band and generally better telephone circuits.
The accompanying chart is of interest in comparing the transmission
frequency characteristics of what were considered good telephone cir-
cuits some time ago with what are considered good telephone circuits
today and what are proposed for the future. At the left of the chart
are shown various types of circuits which have been in use or proposed
for New York-San Francisco service, a distance of a little over 3,000
miles. The original loaded transcontinental line, which remained in
service from January 25, 1915, until April, 1920, when it was unloaded,
gave a band width of only about 900 cycles.* The non-loaded circuit
was better, giving about 1,800 cycles. The modern carrier telephone
circuit is better still, giving about 2,700 cycles. The extra-light loaded
type of cable circuit (H-44, which has been the standard loading system
for long distance use for some time) will give a band even wider,
extending up to at least 3,000 cycles.
At the right of the chart are shown typical characteristics for New
York-Washington (about 225 miles) two-wire cable circuits with
various loadings. The now obsolete heavy-loaded system, H-245, gave
an effective range of 1,400 cycles, the medium-heavy loaded or H-174
gave 1,900 cycles while a new system which is being considered, called
B-88, will give about 2,700 cycles. (At the present time H-174 two-
* The limiting frequencies are taken as those at which the loss is 10 db greater
than the loss at 1,000 cycles.
32 487
488
BELL SYSTEM TECHNICAL JOURNAL
wire circuits are restricted to shorter lengths, the curve being given
simply for comparative purposes.)
In addition to this matter of frequency band width, there has been
improvement in the 1,000-cycle efficiency of long distance circuits and
also improvement with respect to noise and crosstalk. The original
loaded transcontinental circuit, for example, gave, during good weather.
NEW YORK TO
SAN FRANCISCO
165- MIL I OPEN-WIRE
NON-LOADED CARRIER
2 -WIRE LOADED CABLE — 19 GA.
NEW YORK-WASHINGTON
10
-I400CYCLES-
1
H-245
i
H-174
/
B-86
1
f
1900 CYCLES
07nn r*V("i re
/
J
J
/
7
/
/
T
5
,
/
\
\
1
1
\
i
1
1
1
1
1
\
^
J.
J
/
•
/
-5
1000 2000
CYCLES PER SECOND
3000
1000 2000 3000
CYCLES PER SECOND
Fig. 1 — Transmission-frequency characteiistics of representative types of
telephone circuits.
a 1,000-cycle transmission loss of about 20 db with a variation from
this of at least 10 db during bad weather. The non-loaded circuit gave
about 12 db during good weather with smaller variations. With both
of these circuits the noise was somewhat in excess of 1,000 noise units.
The carrier and cable systems compare very favorably with non-loaded
voice- frequency circuits in the matter of transmission loss and are
much quieter.
With the two-wire cable circuits shown, the H-245 circuit gave about
12 db loss at 1,000 cycles, the H-174 circuit 10 db loss and it is ex-
pected that the B-88 circuit will give about 9 db loss at 1,000 cycles.
All of the cable circuits are strikingly quiet as compared with older
type voice-frequency open-wire circuits. The cable circuits are also
considerably better from the standpoint of crosstalk.
It is of interest to consider the effect on service of the change in
standards of toll circuits as illustrated by the characteristics of the
RECENT DEVELOPMENTS IN LONG DISTANCE CABLES 489
above circuits. One way of indicating this is by the repetitions occur-
ring per unit time in commercial conversations. Assuming present
commercial telephone instruments, typical terminal circuit and room
noise conditions, following are some estimates on this basis:
Repetitions per
Circuit 100 Seconds
Loaded New York-San Francisco circuit 3
Non-loaded New York-San Francisco circuit 2
Carrier circuit. New York-San Francisco 1
H-44 cable circuit. New York-San Francisco 1
H-245 cable circuit. New York- Washington lyi
H-174 cable circuit, New York-Washington \%
B-88 cable circuit, New York- Washington 1
Short Cable Circuits
Consideration is now being given to giving up the H- 172-63 two-wire
circuits in favor of B-88-50 and H-88-50 two-wire circuits. H-1 72-63
four-wire circuits were given up some time ago. With the new two-
wire circuits the important line constants and circuit characteristics
are given in the following table.
H-88-50 loading is being considered particularly for those repeater
sections less than about 40 miles in length while B-88-50 loading is
being considered particularly for repeater sections whose lengths are
greater than about 45 miles. For intermediate repeater section lengths
the choice of loading will be dictated by various considerations appli-
cable to the particular circuit layout involved.
With either of the above two-wire circuits, the following transmission
results are anticipated :
Circuits for terminal business up to about 250 miles in
length to have a working net loss at 1,000 cycles of about
9 ± 2 db. The frequency range to extend from about 250
cycles to some frequency between 2,750 cycles and 3,000
cycles. Crosstalk between circuits to exceed 1,000 units in
only about 1 per cent of the combinations. Noise measured
at the receiving end of the circuit, including "babble," * less
than 200 units.
Circuits for "via" business to be limited to lengths in the
neighborhood of 100 miles so that adding a circuit link of this
type to a built-up connection will not, in general, add more
than about 2 or 3 db to the overall loss.
* Babble is the name given to the effect produced by a number of different circuits
crosstalking into a particular circuit at a given time and producing an unintelligible
murmur.
490
BELL SYSTEM TECHNICAL JOURNAL
Long Cable Circuits
Present plans are to retain H-44-25 four-wire circuits for the inter-
mediate and longer distances. This type of circuit is well known so
it is unnecessary to go into its characteristics. With the idea of trans-
mitting frequencies up to about 3,000 cycles with very little evidence
of phase distortion, phase correctors are being considered for very
long circuits of this type, say, circuits exceeding 1,000 miles in length.
Characteristic
77-88-50
5-88-50
Conductor gauge
No. 19 B. & S.
No. 19 B. & S.
Side circuit cable capaci-
tance per mile
0.062 microfarad
0.062 microfarad
Phantom circuit cable ca-
pacitance per mile
0.1 microfarad
0.1 microfarad
Inductance of loading coils
on side circuits
88 milhenries
88 milhenries
Inductance of loading coils
on phantom circuits
50 milhenries
50 milhenries
Spacing of loading coils
6,000 feet
3,000 feet
Nominal cutoff frequency
of side circuit
4,000 cycles
5,600 cycles
Nominal cutoff frequency
of phantom circuit
4,200 cycles
5,900 cycles
Nominal velocity of prop-
agation of side circuit
14,000 mi. per sec.
10,000 mi. per sec.
Nominal velocity of prop-
agation of phantom cir-
cuit
15,000 mi. per sec.
10,600 mi. per sec.
Nominal impedance of
side circuit
1,110 ohms
1,560 ohms
Nominal impedance of
phantom circuit
670 ohms
940 ohms
1000-Cycle attenuation of
side circuit at 55° F.
0.36 db per mile
0.28 db per mile
1000-Cycle attenuation of
phantom circuit at 55^
F.
0.30 db per mile
0.24 db per mile
RECENT DEVELOPMENTS IN LONG DISTANCE CABLES 491
Importance of High Velocity Circuits in Cable
Echo suppressors have proven quite effective in reducing the echo
effects on long distance circuits. For very long distance cable circuits,
however, echo is still a matter for concern, particularly with losses held
down to figures such as those given in the paper by Mr. H. S. Osborne
entitled "A General Switching Plan for Telephone Toll Service."
When cable circuit lengths become very great the actual delay suf-
fered by the speech waves in traveling from end to end of the circuit
becomes important quite apart from echo. We must look forward to
the time when a subscriber in San Francisco will talk by cable across
the United States to New York, then by cable and open wire to New-
foundland, by submarine cable to England and then by a long cable
circuit, let us say, to Constantinople; in other words, a 10,000-mile
circuit length. The highest velocity long distance cable circuits in use
today will give, for conversations over such a circuit, about >^-second
delay in going from one end to the other so that when one subscriber
speaks the other's reply cannot possibly reach him in less time than
one second. This is quite a long time interval. By utilizing speaking
tube delay circuits, connections have been set up involving delays as
great'as this. Conversations are possible over circuits with such delays
but the delay is a serious interference particularly when voice-operated
devices are added which tend to lock out portions of the conversations.
It is thus evidently important to seek higher velocity circuits.
Telephone Carrier in Cable
In seeking ways for obtaining high velocity circuits in cable in an
economical manner, consideration has been given to the proposition of
applying telephone carrier to long distance cables. For large groups of
long distance circuits it appears likely that a carrier-frequency range
can be advantageously used in cables, as wide or wider than the fre-
quency range which has been exploited on open-wire lines. Experi-
mental work on systems of this kind is actively under way at the
present time.
The higher frequencies involved together with the accompanying
attenuations and increased coupling between circuits introduce some
very interesting and unusual noise and crosstalk problems, as well as
problems of equalization and maintenance of transmission stability.
Also, there are interesting economic problems of conductor size and
type, loading versus no loading, repeater spacing, etc.
It is interesting to note that if non-loaded circuits are utilized, a
velocity of transmission of about 100,000 miles per second would be
492 BELL SYSTEM TECHNICAL JOURNAL
obtained while with loaded circuits a velocity perhaps half as great.
The non-loaded setup in particular would, therefore, provide circuits
whose velocity, like open-wire circuits, would be great enough to leave
no question of obtaining satisfactory conversations over any world-
wide telephone network. It is too early, however, to predict just
what the outcome of this development may be.
Phase Distortion in Telephone Apparatus *
By C. E. LANE
This paper shows that if, over its transmitting range, the phase shift, B, in
radians, of a four-terminal network may be written S = Oo + Oico (w = 2-wx
frequency in cycles per second), there is no phase distortion if ao = N, N
being any integer. However, there is a delay, for any signal, given by dB/dw
= fli (seconds). If N is not an integer, there is a delay, Oi, and in addition a
distortion, which distortion, generally for speech and music, may be neg-
lected. Typical phase characteristics for lines, filters and all-pass networks
are shown. In general over their transmitting range, such phase charac-
teristics which usually are curved, may be regarded as the sum of two
characteristics, a straight line having a slope corresponding to the minimum
slope of the original and which introduces a delay without distortion and a
curved portion to which all of the distortion of the signal may be ascribed.
Oscillograms are given showing the distortion for a loaded line and for band
filters for a signal which is of the form y sin (wo^ + 0) between ^ = and
t = T and zero for all other time. A description is given of the means em-
ployed for reducing the amount of phase distortion in telephone cable and in
low-pass filters in circuits used for program transmission and regular tele-
phone service. Also, phase distortion in repeaters and transformers is
described. Brief reference is made to the problem of phase distortion in
telegraph, picture transmission, and television circuits.
'^"^HE effects of amplitude distortion in the transmission of signals
-L has been taken into consideration in the design of telephone
systems for some time. Recently ^ increasing attention is being given
to the phase changes which waves undergo in the process of their
transmission. The necessity of this is, on the one hand, due chiefly to
the use of long distance telephone systems involving greater lengths of
loaded cable and numerous filters and repeaters in tandem, and, on the
other hand, due to the demand for improved performance. One place
where better quality has become particularly desirable is in circuits
for interconnecting broadcasting stations.
This paper will present some general considerations of the relation
between the phase characteristics ^ of telephone apparatus and signal
distortion,'' show the types of phase characteristics that most fre-
quently require consideration and discuss the manner in which the
amount of phase distortion is controlled. Brief reference will also be
* Presented at New York Section, A. I. E. E., May 1930.
1 At the end of this paper, a bibliography is given containing references to previous
publications on this subject.
^ For a definition of phase characteristic see Appendix I. The "insertion " phase
characteristic and "image transfer" phase characteristic are defined there.
^ A companion paper by J. C. Steinberg deals specifically with the effects of phase
distortion on the quality of speech and music. Another companion paper by H.
Nyquist and S. Brand treats of the measurement of phase distortion.
493
494 BELL SYSTEM TECHNICAL JOURNAL
made to phase distortion in systems for transmitting other than tele-
phone signals.
Interpretation of Phase Distortion
Telephone systems must be so designed that the received signal
approximates in wave form the sent signal within limits found by exper-
ience to be tolerable. We are here concerned primarily with the de-
parture of the received signal wave from the sent signal * which may be
attributed to the phase characteristic * of such networks as lines,
filters, repeaters, etc. which go to make up the complete system. Such
distortion is called phase or delay distortion. The reason for the term,
delay, will appear later. We shall summarize here some of the more
general conclusions of the effect on signals of certain phase characteris-
tics and discuss the validity of these conclusions in Appendix II.
If the phase characteristic of any network is of the type shown by any
of the dotted lines in Fig. 1 the received wave will be an exact copy of
the sent or reference wave (assuming no amplitude distortion). In
the case of Fig. 2 the received wave differs from the reference wave
only in that it is reversed in sign which is equivalent simply to reversing
the terminals of the load. In both cases the received wave is delayed
ivith respect to the reference wave by a time interval that is given by the
slope of the phase characteristic or dBjdw. If B is in radians, and co
= 2irf, where/ is the frequency in cycles per second, the delay will be
in seconds. There is no distinction in effect between the phase char-
acteristics for any of the dotted lines and furthermore they are identical
with any such solid broken line as that shown. Since this is true we
may completely represent any of the phase characteristics of Fig. 1
if we choose by the single line passing through zero in which case the
delay is 5/co.
If the phase characteristics are straight lines and intersect the verti-
cal axis at odd multiples of 7r/2 the received wave may be obtained from
the reference wave first by delaying it by dB/dco and then shifting the
phases of all its steady state sinusoidal components as obtained by
Fourier Integral or Series analysis by 7r/2.
If the straight line phase characteristics intersect the ordinate at
intermediate values the received wave may be said to be the sum of
* See Appendix I. If one desires to be specific the sent or reference wave in the
case of considering image transfer phase shift may look upon as the current entering
the network and in the case of insertion phase shift the current through the load with
the network omitted; and the received wave the current through the load in either
case with the network in place.
* In actual apparatus certain general types of phase characteristics are associated
with certain attenuation characteristics. For this reason it is difficult to separate
to the extent one might sometimes desire the effects of the two types of distortion,
attenuation and phase.
PHASE DISTORTION IN TELEPHONE APPARATUS
495
two parts as follows: The first part is an exact copy of the original wave
modified in amplitude by a factor cos a^ and delayed by dB/do:. The
second part is obtained from the original wave by shifting all of the
(720O)
(540°)
(360»)
a: (I80">J
O
Z (0°}
bJ
< (-180°)
a.
(-360°)
(-540°)
(-720°)
4TT
I
O^ -
' ^
'
b^"''
,/-
3lT
-
^
--
r""^^
c ,---
---
^-
^•'
2tT
.^--'
^'
"''
d--
^-^
-^'"
TT
-
^^ ^
^^
-'
"
e
,-
^'-^
,
,^^^^ ,
-^
^^
i-""
f=
^^,^^
-IT
-
^ ^ '
^
--'
2tt
'-
^^^
"^
3tt
-
-- "^
4TT
,-^
j = 2irf
Fig. 1 — ^Phase characteristics which introduce no distortion.
(720°; 4 IT -
(-540°J -3tt
fV20") -4tT -
(.i=2TT-r
Fig. 2 — Phase characteristics which introduce no distortion equivalent to those of
Fig. 1 with connections to the load reversed.
components of the original by 7r/2, multiplying the result by sin ao
and then delaying by dBldoi. ao is the value of the angle at the point
of intersection of the vertical axis.
496
BELL SYSTEM TECHNICAL JOURNAL
An important point to note here is that if a given signal has all its
important frequency components falling in a region between /i and /o
either because of the nature of the signal or as a result of attenuation
in the system we are only interested in the phase characteristic in that
region. Thus for such a signal a sufficient condition for negligible phase
distortion is that the phase characteristics be like those in Fig. 1 and Fig. 2
between /i and fx only.
The phase characteristics actually found in telephone apparatus
that frequently must be considered may for convenience be classified
as follows: (1) those typical of the low pass filter and the loaded line,
180
160
<n
uj 140
u
cr
u 120
C
? 100
I-
- 80
X
en
60
u
<
I 40
a.
20
1
\
/
7
y
/
/
-~^
y
y
/
1
'^
y
/
/
1 1
li
^-^
y
y
fv
^'
V
y /
/
/
^
/
/^
^
(A)
^
•
^
y^
:^
—
.^-■
, ^
u.^^^
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Ff
"REQUENCvf— j
Fig. 3 — Image transfer phase characteristics of typical low-pass filter sections.
Same as insertion phase characteristic except in neighborhood of fe. Curve C
resembles closely the phase characteristic of a loaded line.
(2) those of band pass filters, (3) those of high pass filters, and (4) those
of all pass networks.^ The attenuation of all of these networks is
fairly constant and small in the transmitting range.
The solid curves of Fig. 3 show the image transfer phase characteris-
tics of typical low pass filter sections having a cut-off frequency fc.
The significance of these different type sections will be discussed later.
The curve, C, resembles closely the phase characteristic of a loaded
* An interesting type of insertion phase is that obtained when these four types of
apparatus are terminated over a considerable portion of their transmitting range in
impedances that differ radically from their image impedances. The wa\-y phase
curves obtained may be said to be chiefly responsible for the so-called reflection ef-
fects thus obtained though here too the attenuation plays an important part. If the
attenuation in the network is large as in a line this waviness due to miss-match tends
to disappear and the phase characteristic resembles closely in general shape that which
would be obtained if no miss-match occurred.
PHASE DISTORTION IN TELEPHONE APPARATUS
497
lineJ These curves as well as those in the following three figures are
for non-dissipative networks terminated in their image impedances.
However, the insertion phase characteristic with dissipation and the
usual terminations, i.e. a resistance of fixed value, would not noticeably
180
160
140
120
100
80
60
40
20
20
40
-60
-80
-lOO
-120
-140
-160
-160
/
^ = IS
'/
^c,
/
/
i
^m = f
2
/
7
'
//
/
/
/
/
/
/
A, /
r f
(A)/ y
V
/
/
V
u
►,
/
^
/^
y^ ^
1 \
1
1
^c,
^
^m
/
'^c.
^
/jT
/
,,
^ /
(A2)
It
/
/
/}
/ 1
/
^,
O '
/
/
/
V
/
'
7
0.6
0.9
1.0
1.2
1.3
FREQUENCY f^^
Fig. 4 — Image transfer phase characteristics of typical band-pass filter sections
be different in most of the transmitting range. For filters in the latter
case in the neighborhood of jc the slope would be modified so as to
remain finite. This will be discussed further in the next section.
The curves of Figs. 4 and 5 show respectively the image transfer
' It will be noted that the second derivative of these low pass filter phase curves
are positive at all frequencies. However, special sections exist for which this is not
true and such sections are occasionally used as discussed later where it is desirable to
keep the characteristic as a whole nearly straight over a wider frequency range but
low pass filters when considered as a whole generally have phase characteristics of the
type shown.
498
BELL SYSTEM TECHNICAL JOURNAL
phase characteristics for typical band pass and high pass filter sections.
The curves of Fig. 6 are for all-pass lattice type network sections.
The frequency fr is the resonant or anti-resonant frequency of the
series arms and cross-arms of the network.
The above four figures show that the phase characteristics are
curved in every case over a considerable portion of the transmitted
frequency range.** A curved phase characteristic like that shown in
Fig. 3, curve B, for example, may be represented as the sum of a
distortionless phase characteristic, Bi, of the type shown in Fig. 1
and another curved one, B2, which is the difference between it and the
0.5
1.0
1.5
FREQUENCY ^fH
2.0 2.5
3.0
3.5
4.C
4.0
00
JA}_
*•
'^
-20
UJ
1
^
^
-
'
W
^
UJ -40
(T
LU -60
Q
Z-80
1-
= -100
I
<fi
.j-'20
U1
^-140
Q.
-160
/
/
^
^
—
/,
/
y
^
"^
/
/
/
/
p.
^
^^
*-
/
/
,,^
^
^
^
^
4
/
A
-—
^
-^
'^
1
Fig. 5 — Image transfer phase characteristics of typical high-pass filter sections.
original,^ i.e., B = By + Bi. The slope of the straight line charac-
teristic is the minimum slope of the original, i.e. the slope at very low
frequencies. B\_ introduces at all frequencies a definite delay without
distortion given by its slope. Bo, to which no delay as a whole may be
ascribed may be called the phase distortion characteristic of the net-
work, and its derivative {dBjdo))/ — (dB/d(jo)min. the delay distortion
characteristic or simply the delay distortion. This procedure is equiva-
lent to regarding the low pass filter as consisting of two parts in tan-
dem the first part introducing a delay without distortion and the second
part a distortion. If, after subtracting such straight line portions
from low pass filters the remaining curves are the same, the phase
* In discussing the phase characteristics of filters only the characteristics m the
transmitting range are considered, since in general the frequency components in this
range only contribute noticeably to the received signal.
' See Appendix II.
PHASE DISTORTION IN TELEPHONE APPARATUS
499
distortion would be the same even though the delays might be different
due to different slopes of the subtracted portion.^"
In the case of the band pass filter of Fig. 4 a distortionless portion
may be subtracted having the slope of the original at the mid-frequency
but only for special cases will it be tangent to the curve. Fig. 4 shows
curve A broken into two parts a distortionless part Ai giving delay
and a part Ai responsible for the distortion, i.e., A = Ai -\- Ao. It
is doubtful if effects of ^2 on speech and music would be noticeably
different than any other curve obtained by adding a constant angle
to it at all frequencies. This type of variation occurs to the same ex-
360
£ 320
y 280
UJ
'='240
Z
,_ 200
I 160
in
^ 120
<
I 80
40
fF)
r ■
^
- —
/
^^
■
/.
A
—
\a
^
1
1
^^
^
=^
• —
(A)
/
<^
^^
7,
/,
/.
^/
7
/'
//
/
/
k
/
^
^
/
'
/
/
t
/
0.5
1.0
1.5
3.0
2.0 2.5
FREQUENCY {^^
3.5
4.0
4.5 4.5
Fig. 6 — Image transfer phase characteristics of typical all-pass lattice type network
sections.
tent in apparatus observed to produce negligible phase distortion as
where distortion is apparent. It is for this reason that the slope of the
phase curve rather than the curve itself is generally taken as the
criterion for determining the amount 0/ phase or delay distortion in a
network. In other words if two phase curves give the same dBjdoi or
delay characteristics (sometimes called "envelope" delay characteris-
tics) they are regarded as introducing the same phase distortion par-
ticularly for speech and music though such an assumption is not
rigorously true.
In the case of the high pass filters and all pass networks of Figs. 5 and
^'^ This procedure of breaking the phase characteristic into two parts one part
introducing no distortion but a delay To given by the minimum slope in the transmit-
ting range and another part responsible for the distortion is perfectly rigorous, but it
must not be interpreted to mean that for a signal starting at < = absolutely nothing
will be received before t = To. In the first place there is generally a small amount
of energy outside the transmitting range that will come through earlier and in the
second place, unless the residual phase characteristic is of a kind that can be produced
by physical apparatus the distortion characteristic alone will cause the received wave
to differ from zero prior to / = T^.
500
BELL SYSTEM TECHNICAL JOURNAL
6 the minimum slopes are zero hence no delay can be subtracted which
applies to the signal as a whole. '^
In order to observe the effects of phase distortion in some simple
cases ^2 we shall show oscillographs of some sent and received non-
periodic waves. These waves are of the type that are zero up to time
/ = 0, take the form y sin (wo^ + Q) between / = and / = T and are
zero for all future time.^^ A Fourier Integral analysis of these waves
would show they contained energy over the entire frequency range,
though most of it is confined to frequencies in the neighborhood of /o
where /o = ooq/Itt.
79,200
72,000
0.220
800 1200 1,600 2,000
frequency' in cycles per second
2.400
2.800
Fig. 7 — Insertion phase and delay characteristics of a 600 mile length of medium
heavy loaded cable (including repeaters).
Such waves as these are elementary waves which can readily be
produced in the laboratory and the effects of distortion on them ob-
served or in special cases the effect may be calculated.
" This assumes of course that energy falls in the frequency range where the slope
approximates zero. If such were not the case for a particular signal a definite delay
could be ascribed given by the minimum slope in the range where the frequency com-
ponents of consequence fall.
1- The effect of phase distortion on speech and music signals is discussed in the
paper by J. C. Steinberg already mentioned.
'2 It is of interest to note that any complex wave which is zero at all times prior to
/ = and also at all times after t = T may be regarded as the sum of such finite
components as these. Analyze it by means of Fouriers Series as though it repeated it-
self as a steady state wave for all time. Then multiply all of the steady state sinusoidal
components by zero for all time prior to / = and after t = T retaining only the por-
tion for the interval of time T. The resultant simple components will add up to giv-e
the original complex wave. After distortion the distorted components will add up
to give the distorted wave as a whole.
PHASE DISTORTION IN TELEPHONE APPARATUS
501
Fig. 7 is the insertion phase and delay characteristic of a 600 mile
length of medium heavy loaded cable including repeaters. This cable
has a theoretical cut-off of about 2500 cycles. Fig. 8 shows the dis-
tortion for two simple waves ^^ of the above type, for one /o = 1000
cycles and for the other /o = 1500 cycles. The oscillographs show as,
predicted, that practically nothing in either case is received until
after the time given by the minimum value of dBjdo}, i.e. .0654 seconds.
After this time a distorted form of the sent wave occurs. Since for
/o = 1000 most of the energy of the wave as analyzed by the Fourier
Integral method of analysis falls in the neighborhood of 1000 cycles
SENT
1000 CYCLES
Fig. 8 — Distortion resulting from 600 mile length of cable of Fig. 7 for signals of the
form y sin uut + 6 between t = and t = T and zero for all other time.
where the delay characteristic is reasonably constant this wave is not
distorted much. In the case for/o = 1500 cycles a larger portion of the
energy falls in the neighborhood of 1500 cycles where the delay char-
acteristic is changing more rapidly and it is therefore distorted more.
{dBjdio)/^ — {dBldw)m\n. may be taken as a fair measure of the dis-
tortion of such simple waves although higher derivatives are also
involved.
Fig. 9 shows the insertion delay characteristic for a system consisting
of four band filters in tandem. The attenuation characteristic is also
" Reproduced from the paper by Sallie P. Mead, loc. cit.
502
BELL SYSTEM TECHNICAL JOURNAL
shown. Fig. 10 shows oscillographs for waves of the above type for
/o = 260, 300, 480 and 680 cycles per second. Notice that the distor-
tion is much greater where /o falls near the edges of the transmitting
band, although in every case the wave starts noticeably building up at
about .0109 seconds after / = for the sent wave. This is the value of
{dBld(xi)m\-n. in the transmitting range of the filters. In both Figs. 8
48
44
40
36
32
Si
Z 28
Z
o
P 24
<
D
^ 20
<
16
12
1
1
1
/
\
//
\
\
/
/
^
^^
.HJIE
NUATl
0^^^
^/
\
. DELAY
^
200
300
400 500 600
FREaUENCY IN CYCLES PER SECOND
700
0.060
0.055
0.050
0.045
0.040
<0
Q
0.035 Z
O
U
UJ
«
0.030
z
0.025
<
UJ
o
0.020
0.015
0.010
0.005
800
Fig. 9 — Insertion delay and attenuation characteristics of 4 band-pass filters in
tandem.
and 10 some of the distortion may be ascribed to attenuation although
the elongation effect is primarily due to phase distortion. It is this
elongation effect that is noticeable to the ear in speech and music.
Phase Distortion and Its Correction
This section of the paper will contain a more specific discussion of
the phase characteristics of apparatus and the means employed for
keeping phase distortion within desirable limits.
PHASE DISTORTION IN TELEPHONE APPARATUS
503
260 CYCLES
RECEIVED
300 CYCLES
RECEIVED
Mwv^M^
00109
"SECONDS
480 CYCLES
680 CYCLES
1-ig. 10 — Distortion resulting from four band filters of Fig. 9 for signals of the form
y sin coo/ + 6 between / = and t = T and zero for all other time.
504
BELL SYSTEM TECHNICAL JOURNAL
1. Telephone Cable. — Two of the cases where phase distortion in
telephone cable has been considered objectionable will be discussed:
(1) In the newly developed high quality cable circuit for transmitting
programs to and between broadcasting stations '^ which circuit is
M,i-nju
//
10,000
/
/
y
9,000
//
//
/
8,000
/
/
/
A
7,000
10
UJ
UJ
/
M
//
/
/M'
a.
o
UJ
C 6,000
z
/
V
Z^
/
w
W
J
1-
u.
S 5,000
Ui
(0
^y^
-
/)
' / <
<<y
i-
/
//
/
/ /
^ -
■V
<
I
"■ 4,000
/
y.o
/
/
,/
/
/ /
V?
3,000
//
r
/)
'&
■^
/
/
A
^
2,000
/
/
'/
4?
-^
^^
)p
^
A
V
^
ov
1,000
//
/
^^
■^v^^
/
y
r
^
^
^
1,000
2,000 3,000 4,000 5,000 6,000 7,000
FREQUENCY IN CYCLES PER SECOND
8,000 9P00
Fig. 1 1 — Insertion phase characteristic of a 50 mile length of cable for program trans-
mission. Also, phase characteristic of a phase corrector for this cable and for the sum
of the two
designed to transmit with negligible amplitude and phase distortion
all frequencies between 50 and 8000 cycles and (2) a proposed cable
i"* Long Distance Cable Circuit for Program Transmission, A. B. Clark and C. W.
Green, to be presented at Summer Convention A. L E. E., Toronto, June, 1930.
This cable consists of No. 16 gauge copper wire witli 22 millihenry loading coils
spaced e\-ery 3000 feet.
PHASE DISTORTION IN TELEPHONE APPARATUS
505
circuit ^® for regular long distance telephone message service designed
to transmit with negligible distortion frequencies between 200 and
3000 cycles.
Fig. 11 shows the insertion phase characteristic of a 50 mile length
of the cable for program transmission (exclusive of repeater). The
broken line is drawn to bring out more clearly the curvature in the
I C| C| C3 c^ (.5 e., I-!, C5 <-5 ti
KflfiJ loortvoJ UiyT-soJ 'jKL/riaJ Imr^oJ Ul/|^!kJ lfifl7T^_aJ UiL'TafflJ UsL/tviiiJ Ua^jiflJ l&tL^W
my
i
t" T" T" T- T'- T" T" T" T" T" T
Fig. 12 — Schematic of the jjhase corrector for a 50 mile length of cable for program
transmission.
phase characteristic. Fig. 12 shows a schematic of the phase corrector
used to correct for a 50 mile length of this cable. The insertion phase
characteristic of this corrector as well as that of both the cable and
corrector combined are also shown in Fig. 11. It will be noticed that a
0.005
0.004
in
o
z
O 0.003
UJ
in
>- 0.002
<
_j
UJ
o
0.001
1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000
FREQUENCY IN CYCLES PER SECOND
Fig. 13 — Insertion delay characteristics corresponding to phase characteristics of
Fig. 11.
TOTAL DELAY ,
:^
L-^
STH
OF
CAB
_E F
OR
= B0(
iBA^
t C^F
DEL
AY
OF 5
0-M
ILE
LEN
^Ric
:tor
DEL
AY C
^F PHASP /-^
■~~-
-^
single phase corrector consists of eleven sections, seven of one kind,
three of another, and one of a third. Each section contains one two-
terminal inductance coil and one three-terminal coil with mutual
between the two windings and also two condensers. Fig. 13 gives
'^ This cable consists of No. 19 gauge copper wire with 44 millihenry loading coils
spaced every 6000 feet.
506
BELL SYSTEM TECHNICAL JOURNAL
dBjdoi for the cable, the phase corrector, and the two combined. A
50 mile length of corrected cable gives a delay of .00375 seconds. A
3000 mile length " of this cable such as would extend from coast to
coast would give a delay of .225 seconds, with a difference between the
36,000
32,000
28,000
24,000
(0
lU
Ul
a.
o
Q 20,000
I
W 16,000
tu
<
I
Q.
12,000
8;000
4,000
//
/
/
k
/y
//
w
K-
/'J
w
i-
//
9
^
>p
' 4
v^
•v^
:>
/y
''/.
//
%
w
J
A
r
/
/
A/
A
>"
pHA'
iS-^
-i\rr_
OF
„ — y~-r/-iD
GORP
f- ":
■ '
^^
400
800 1,200 1,600 2,000 2,400
FREQUENCY IN CYCLES PER SECOND
2,800
3,200
Hg. 14 — Insertion phase characteristics of a 500 mile length of cable for telephone
message service.
minimum and maximum value of .006 seconds in the corrected range.
Without correction this difference would have been .055 seconds.
Figs. 14, 15 and 16 correspond respectively to the previous three but
are for a 500 mile length of the cable for regular telephone message
service. The total delay for 500 mile length of this cable after correc-
'' In designing apparatus going in long distance circuits in general the parts are so
designed that if a circuit 3 or 4 thousand miles long is used the total accumulated
distortion due to either amplitude or phase will be within tolerable limits.
PHASE DISTORTION IN TELEPHONE APPARATUS
507
tion is .029 seconds. A 3000 mile length would give .174 seconds delay.
The difference after correction for 3000 miles between minimum and
maximum value of dB/dco is .007 seconds and before correction .035
seconds. This phase corrector (for 500 miles) consists of 12 sections,
8 of one kind and 4 of another. Each section contains four condensers
and two four-terminal inductance coils with mutual between windings.
Both this phase corrector and the previous one are formed by connect-
ing together such all-pass network sections as to give the phase char-
acteristic desired. In the first bridge T-sections are used and in the
second lattice type.^~* The former are more economical when unbal-
anced apparatus may be used, though similar phase characteristics
may generally be obtained with either.
Fig. 15 — Schematic of the phase corrector for 500 mile length of cable for telephone
message service.
2. Filters. — The following factors influence the phase distortion in
filters: (1) The width of the frequency band transmitted, (2) the amount
of discrimination between transmitted and attenuated regions (cor-
responds to number of filter sections), (3) the rate at which the atten-
uation rises at the edges of the transmitting band, (4) the types of filter
sections used, (5) the number of filters in tandem, (6) the amount of
reflection due to impedance mis-match near the edges of the trans-
mitting bands, and (7) the amount of dissipation in the filter
elements.
The insertion phase characteristics of Fig. 17 and the insertion delay
characteristics of Figs. 18 and 19 are for two low pass filters ^^ of the
usual type. As will be seen from their attenuation characteristics
(insertion loss) each gives a discrimination of about 35 db, although
the second requires an additional section in order to provide the rapidly
'^ Nyquist, U. S. patents Nos. 1,675,460 and 1,735,052 and Zobel patent Xo.
1,701,552; Maximum Output Network for Telephone Substation and Repeater Cir-
cuits, by G. A. Campbell and R. M. Foster, Trans. A. I. E. E., Vol. 39, pp. 231-280.
^^ This note explains symbols used in these three figures and also the following
three. Zj is the image impedance. Zo for a low pass filter is the value of Zj at zero
frequency and for a high pass filter at infinite frequency. Q is the ratio of the coil
reactance to its effective resistance. Dissipation in the condensers is considered
negligible. For a filter section having an attenuation peak at frequency, /«,, and a
cut-off at, fc, "a " is the ratio /„//c for a low pass filter and/e//„ for a high pass filter.
508
BELL SYSTEM TECHNICAL JOURNAL
rising attenuation at the edge of the band. These curves show that
the delay distortion in the transmitting band is increased by increasing
the slope of the attenuation curve at the cut-off although the minimum
value — i.e. the delay which applies to the signal as a whole, does not
increase appreciably. The effective band width transmitted depends
upon both the delay and attenuation characteristics since especially
0.032
0.030
0.028
0.026
0.024
0.022
^ 0.020
o
z
O 0.018
o
UJ
0.016
^ 0.014
_I
UJ
° 0.012
0.010
0.006
0.006
0.004
0.002
TOTAL DELAY
rE
b^
^OM^
^''■
,p.g£
,-1^11-'
: LEN
GTH C
3F CABUE f
OB'f
EUEP^
De
.AY C
^.0.
^
^y
P PH>
i£2o/
=?
400
800
1,200 1,600 2,000 2,400
FREQUENCY IN CYCLES PER SECOND
2,800 3,200
Fig. 16 — Insertion delay characteristics corresponding to the phase characteristics of
Fig. 14.
for a number of filters like these in tandem the delay of the frequency
components of the wave near the cut-off may be so great that these
components contribute little to articulation. Therefore in the design
of filters a proper balance must be determined between the rate of
attenuation and the delay distortion. A more complete discussion
of the relation between delay, attenuation, and the effective cut-oft' is
given in the paper by J. C. Steinberg.-" When low pass filters are to
be designed with sharper cut-offs from the standpoint of both delay
and attenuation, there are two ways in which this is usually done, one
20 Loc. cit.
PHASE DISTORTION IN TELEPHONE APPARATUS
509
600
550
500
450
in
(ij
iiJ 400
cc
o
UJ
Q 350
I- 300
I
^ 250
UJ
200
150
100
50
r
L|
3M1
L2
cm
^
1-3
L|
00(
n
Q =
200
O- l.25f-
F^
■^000 H
- I,25f^
— 1
(X
0.=
NO
DISSIPATION
1
SCHEMATIC (A) CONSISTING OF ONE
1^
(X SECTION, ONE a = l.25 SECTION
AND ONE a = 1.05 SECTION
,/^
\
>-l L|
/
/-
^^
=^5 --
//
7
/
/
SCHEMATIC (B) SAME AS (A)
WITH 1.05 SECTION OMITTED
/
>
/
TERMINATED
IN Zi
^
^
o/^
/
--
_A
vN^
l
>^
^
^
"^^K
a^^
0.2
0.4
0.6
0.8
1.0
FREQUENCY (-fj
1.2
1.4
1.6
F"ig. 17 — -Insertion phase characteristics and schematics of low-pass filters. (/I)
For a filter consisting of 3 full sections, one section having no attenuation peak, one
with a peak at 1.25/c and another at 1.05/c. (5) Same as (/I) with the 1.05 section
omitted.
10
7 -
«i|3
>-
<
_i
UJ
Q
fl
J
1
\
r200
-80
) DISSIPATION
r
/
\
■s^
NC
V
/
1
\,
\
\
''
z
o
1- ~
<
D
3-
re
DEC.
\
UJ
h-
y
1
1
o£^
■^
y
I
'
J
0.2
0.4 06 0.8
FREQUENCY (-^)
1.0
1.2
50
45
40
Z
30 -
Z
o
<
D
20 Z
UJ
\-
10
1.6
Fig. IS — Insertion delay and attenuation characteristics for filter .1 of Fig. 17
510
BELL SYSTEM TECHNICAL JOURNAL
is the use of a few filter sections of a type different from the usual
types above having phase characteristics with a negative second
derivative rather than positive so that the combination will postpone
the occurrence of phase distortion until very near the theoretical cut-
off and the other is the addition of all pass network sections which
accomplish the same general result.
O
^
= 200
) DISSIPATION
/
S
^,
NC
/
/
/
/
/
o
'
~ i
'
I
/
2
* V
/
'
1
DEL£1J„-
^
/
0.2
0.4
0.6 0.8
FREQUE
1.0
1.2
1.4
ncy(^)
50
45
40
35 -°
30?
25
Z
o
I-
<
20 i
I-
15 5
10
5
1.6
Fig. 19 — Insertion delay and attenuation characteristics for filter B of Fig. 17.
One other point should be noted here. The shape of the insertion
phase curve as shown at the cut-off frequency owes its departure from
the image transfer phase shift without dissipation shown by the dotted
line more because of the reflection than dissipation. The value of the
Q so long as it is within the usual range makes little difference.
Figs. 20, 21 and 22 correspond to those of 17, 18 and 19 but are for
high pass filters. High pass filters introduce no initial delay to signals
as a whole. The distortion of the signal is dependent upon sharpness
of cut-off, Q etc. just as for low pass filters.
Band pass filters give an initial delay defined by the shape of the
phase curve. Other factors remaining the same this delay as well as
the amount of phase distortion is inversely proportional to /o — /i
the band width in cycles and is independent of the position of the band
on the frequency scale. The effect of reflection, dissipation, sharpness
of cut-off, etc., are about the same at the lower cut-off as for a high pass
filter and at the upper cut-off as for a low pass. As already noted one
distinguishing feature of the phase characteristic of a band pass filter
PHASE DISTORTION IN TELEPHONE APPARATUS
511
FREQUENCY
1.2 1.4
Fig. 20 — Insertion phase characteristics and schematics for high-pass filters. {A)
For a filter consisting of 3 full sections, one section having no attenuation peak, one
with a peak at 1/1.25/c and another at 1/1.05/c. (S) Same as {A) with the 1/1.05/c
section omitted.
10
<3. 3 6
■01 T>
u
-^ 5
>-
<
-I 4
0.6
0.8
/
\
■*
X
Y
\
'
\
t
Q = 200
a = 8o
NO DISSIPATION
1
/-
5
1
-I
m
z
-c -
>
\
\
V
\
\
\
•^^
W
'o^.
—
1.2 1.4
FREQUENCY
(^c)
1.6
1.8
2.0
50
45
40
35
30
25
20
15
10
2.2
Fig. 21 — Insertion delay and attenuation characteristic for filter .4 of Fig. 20.
512
BELL SYSTEM TECHNICAL JOURNAL
is that the straight portion of the phase curve may if extended intersect
the vertical axis at any point and does not like the low pass filter pass
always through Ntv.
As an example of a condition where it has been found necessary to
take phase into consideration in designing low pass filters let us con-
sider the case of line filters used in open wire circuits transmitting
simultaneously both programs for broadcasting and carrier telephony.
Here as many as 40 or 50 filters may be used in tandem.
Fig. 23 shows the measured delay and attenuation characteristics
i
3
"olx
>-
<
u
9
8
7
/
\
^
/
\>,
= aoo
3 DISSIPATION
6
5
4
3
2
'
N(
\
ft
O
V
V
\
^
\
\
%
\
^^vL^
()
50
45
40
35
30
25
20
15
10
5
z
o
1-
<
D
Z
UJ
I-
I-
<
0.6
0.8
1.0
1.2 1.4
FREQUENCY
1^^)
1.6
1.8
2.0
2.2
Fig. 22 — Insertion delay and attenuation characteristic for filter B of Fig. 20.
for 25 of the present 5000 cycle quality line filters now in use, these
filters being connected in tandem. The circuit is so designed that it
practically equalizes up to 5000 cycles for the attenuation distortion.
When a number of these filters are used in tandem as in the longest
"hook ups" the phase distortion of these filters is somewhat noticeable
but not seriously so and their effective cut-off is appreciably lowered
because of this phase distortion.'^
Fig. 24 gives the calculated delay and attenuation of twenty-
five 8000 cycle low pass line filters connected in tandem. This
filter is being considered for use in place of the 5000 cycle line
filter of Fig. 23. The delay for 25 of these filters in tandem is ap-
proximately constant up to 7500 cycles (within .001 seconds). The
attenuation is also constant up to this frequency. The attenuation
-' J. C. Steinberg, loc. cit.
PHASE DISTORTION IN TELEPHONE APPARATUS
513
distortion between 7500 and 8000 cycles is purposely left uncorrected
so that if these filters are used when a number of them are connected
in tandem on a line, the effective cut-off will be lowered therefore
eliminating the effects of any accumulating delay over this range.
Thus for short distances 8000 cycle quality will be realized, and for
very long circuits 7500 cycle quality, and for intermediate lengths
something in between. In any case the effects of delay distortion will
be negligible. Special filter sections are used in order to meet the
unusual delay and attenuation requirements. The same results could
L|
I
-2
SCHEMATIC
1-3
OF
FILTER
■-4
L5
R|
lO— ^005
Cu
■ vv
c
l_
-
§1-6 |l7
C2^ C3^ C4:
§1-6 k^i
T V V V V ^~>
OLq
?-
g^S
125
Ll 1-2 "-3 L4
L5
■i— ^AA/v-04
R|
1
100
/
/
75
/
/
/
50
^
/
/
1
y'
/
-
"
/
25
1 ATI
ON
y
"AT
...
0.020
0.016
m
a
z
0.012 8
m
0.008 >
<
-I
UJ
a
0.004
1,000
2p00 3,000 4,000 5,000 6,000 7,000
FREQUENCY IN CYCLES PER SECOND
8p00
9,000
Fig. 23 — Schematic, insertion delay and attenuation characteristics of twenty-five
5000 cycle low-pass line filters connected in tandem.
have been obtained using the usual type sections like those used in the
5000 cycle filter and then correcting for phase by an all-pass structure.
Such a method would have resulted in a somewhat more expensive
filter and one giving more overall delay.
3. Repeaters. — The chief sources of delay in telephone repeaters are
the transformers. However, some additional low frequency delay is
caused by shunt retard coils and series condensers. This is kept within
negligible limits by using large values of both. Conversely inductance
in series and capacitance in shunt cause high frequency delay but this
effect can easily be made negligible in any frequency range in which
one is interested. Fig. 25 is a schematic of the telephone repeater
514
BELL SYSTEM TECHNICAL JOURNAL
used in the cable circuit previously referred to for program transmis-
sion. A repeating coil and input transformer at its input and the out-
put transformer at its output are shown. These will be discussed in
the next paragraph.
4. Transformers. — As an example of phase distortion in transformers
we shall consider those shown in the telephone repeater of Fig. 25.
Here, the phase shift of a small number of transformers would be of
little importance but where a large number are connected together as
in long toll circuits their effect cannot be overlooked. The delay
caused by a transformer is similar to that of a high-pass filter. The
140
120
100
80
2
O
<
Z
u
I-
60
40
20
h
/
1
—
.
DELAY
■^
y
^^
i
f
kTTE
NUA
TlO^
i
J
0.028
0.024
0,020
If)
Q
Z
o
0.016 o
0.012
>-
<
0.006
0.004
UJ
O
1000
2000 3000 4000 5000 6000 7000
FREQUENCY IN CYCLES PER SECOND
8000
9000
Fig. 24 — Insertion delay and attenuation characteristics of twenty-five 8000 cycle
low-pass line filters connected in tandem.
insertion phase characteristics of these three transformers between
impedances for which they were designed are shown in Fig. 26 and their
insertion delay characteristics, dB/dco in Fig. 27. It will be noticed
that practically all of the delay occurs below 100 cycles. The three
together give at 40 cycles a value of dBjdi^ of .0008 seconds. 25 sets
would give .020 seconds delay. Experience has shown that this amount
of delay for speech at low frequencies is negligible whereas at high
frequencies such would not be the case.-'^
5. Attenuation Equalizers. — Attenuation equalizers introduce some
phase distortion but the amount is so small that it can generally be
-- At high frequencies {dB/dw)aiax. — {dB/dui)^^. must generally be kept under
.005 to .010 seconds if its effect can be neglected for speech.
PHASE DISTORTION IN TELEPHONE APPARATUS
515
UJ
a.
Hii'
^im^
P — r"i_iw.A^^
c
o
in
C
a
^=" rt
o
o
o
u
'o
T3
in
a
o
o
CD
516
BELL SYSTEM TECHNICAL JOURNAL
neglected. The presence of such equalizers in the circuit for program
transmission did not influence the design of the phase correctors.
However in considering the design of the equalizers the particular
structure used was chosen on the basis of its giving a minimum amount
of phase distortion.
Phase Distortion in Other Communication Systems
Phase distortion has for some time been considered a real problem
in submarine cable telegraphy. If the highest reversal frequency of a
telegraph signal is /„ it has been found expedient to correct for both
14
12
10
in
UJ
LU
a.
o
LU
Q
Z
I
IT)
in -2
<
I
a -4
-10
1
.__
—
=--=
/
^^"^ 
