LIBRARY
NAVAL paSTGRADUATE SC
MONTEREY, CALIFORNIA 9
NPS-61MD72121A
NAVAL POSTGRADUATE SCHOOL
It
Monterey, California
31 December 1972
NPS-61Md72121A
Coherent and Incoherent Components of Sound
Scattered at a Time Dependent Rough Surface
C.S. Clay
Physics & Chemistry
Department
Approved for public release; distribution unlimited,
FEDDOCS
D 208.14/2
NPS-61MD72121A
DUDLEY KNOX LIBRARY
NAVAL POSTGRA.DUATE SCHO'
MONTEREY CA 93943-5101
NAVAL POSTGRADUATE SCHOOL
Monterey, California
Rear Admiral M.B. Freeman M.U. Clauser
Superintendent Provost
TITLE: Coherent and Incoherent Components of Sound Scattered at a
Time Dependent Rough Surface
AUTHOR: C.S. Clay*
ABSTRACT :
Theoretical expressions are derived for the sound scattered
at a time-dependent rough surface. The calculations are made for
a Gaussian shaded source transducer and point receiver. The
Helmholtz theorem and Fresnel approximation are used. The rough
surface is assumed to be a traveling wave and to have a traveling
wave packet type of correlation function. The coherent component
of the signal is the product of the Fourier transformation of the
surface distribution function and the smooth surface reflection
signal. Comparison of theory and experiment shows the coherent
component to be sensitive to the non-Gaussian character of the
wind-blown water waves. The incoherent components and the temporal
correlation function of the scattered sound are given. For the
special case of a traveling cosine wave type of rough surface, the
spectrum of the scattered sound includes components which are mul-
tiples of the frequency of the surface wave. For surfaces describ-
able by a bivariate Gaussian distribution function, the temporal
correlation is a function of, but not the same as, the time de-
pendence of the rough surface. The scattered sound is insensitive
to the spatial correlation function of the surface at distances
larger than the dimensions of the transducer divided by the cosine
of the incident angle. The final expressions are complex error
integrals and can be used for all values of roughness. This task
was supported by Naval Ship Systems Command (Code PMS 388) .
TABLE OF CONTENTS
I. EXPERIMENT AND THE INVERSE PROBLEM 4
II. SCATTERING FUNCTION FROM THE HELMHOLTZ
EQUATION 6
III. EVALUATION OF THE SCATTERING INTEGRAL 18
IV. COHERENTLY SCATTERED SIGNAL 24
V. COSINE CORRUGATED SURFACE 28
VI. TOTAL SIGNAL SCATTERED AT A NON-GAUSSIAN
SURFACE 31
REFERENCES 33
INITIAL DISTRIBUTION LIST 34
FORM DD 1473 38
I. EXPERIMENT AND THE INVERSE PROBLEM
Sound scattered at a time dependent surface takes on some of the
time dependence of the surface. For example, the upward or downward motion
of the surface causes the frequency of the reflected signal to be Doppler
shifted. It is easy to demonstrate the effects in laboratory experiments
(Fig. 1.1). Small waves on the water cause the phases of the signals to
fluctuate. Experiments have shown the reflected signals are (crudely
stated) modulated by the surface and often the spectrum
of the surface can be identified as a component of the spectrum of the
reflected signal. With our growing technology in remote sensing, the
importance of being able to go from scattering measurements back to a
description of the time dependent surface is extremely important. It
may be more important to know when and how far one can go for a given
technique.
Doing the inverse problem requires a complete knowledge of the relation-
ship of the measurement to the quantity being sensed. Too often, the
result of having this knowledge is "I can't measure what I want by doing
the experiment I am doing." For example, in the plane wave approximation,
Eckart (1953) showed that mean square scattered signals are simply related
to the correlation function of the surface at small ifO
(f = k(cos 9,+ cos9„)/2, k is 2it/wavelength, a is rms roughness, and
and 9 are the incident and reflected angles). Unfortunately the
2 2
scattered signal is proportional to y a and correspondingly very small.
If the source and receiver are omnidirectional or broad beamed, the
scattered sound is usually identified as being a fluctuation of the signal
and is often inseparable from the noise.
4
In this paper I add another limitation (which is derived in the
Fresnel approximation. The scattered sound is insensitive to spatial
correlation function of the rough surface at correlation distances larger
than the dimensions of the transducer/cos9- .
I believe a bit of discussion of the theoretical problem and
approximations is needed. In seeking the source of a discrepancy between
my theory (1971) and some of our laboratory measurements, G. A. Sandness
identified the calculation of the incident sound at the surface as being
the difficulty for two reasons, (private communication). First, in using
the Helmholtz equation, the incident sound signal should satisfy the wave
equation. The combination of a point source and directional function (or
illumination function on the surface) does not satisfy the wave equation.
Second, the limitation of the expansion to second order terms may be a
very poor approximation for a diverging wave at the surface.
Another way of thinking about this is to regard the phase and amplitude
of the incident signal as being the hologram of the source. The hologram
of a finite object is different from that of a point source. The reader
may wish to read Melton and Horton (1970).
Having said that there are difficulties with the Fresnel approximation,
I have chosen to use it. Also, I use a Gaussian shaded source transducer
and a point receiver. This shading fits the measured response of our
transducers and facilitates repeated integrations. I have obtained ex-
pressions that can be used over a wide range of surface roughnesses and
correlation functions without resorting to separate expansions for the high
and low frequency limits. The end result is the mean square scattered sound
and its temporal correlation function.
II. SCATTERING FUNCTION FROM THE HELMHOLTZ EQUATION
The development of a theoretical relationship of a scattering function
to the properties of the surface requires analysis of the scattering of
acoustic signals by rough surfaces. This has been the subject of a number
of papers and is treated in several books, notably Beckmann and Spizzichino
(1963), Tolstoy and Clay (1966), Ol'Shevskii (1967), Fortuin (1970), and
Horton (1971). Although we will not discuss it here, in the electronic
and radio engineering journals an extensive literature exists on the
scattering of electromagnetic waves by various types of irregular and rough
surfaces. We are obliged to present the theory in some detail because the
relevant underwater acoustics theory has not been applied to these types
of problems. Most of the studies have been empirical.
The derivation of the scattering equation from the Helmholtz theorem
is usually based upon the assumption of local plane reflections. It is
suggested that the reader who is interested in the details refer to the
development given by Tolstoy and Clay (1966), who base their derivations
on those of Eckart (1953) and Beckmann and Spizzichino (1963). The sound
absorption will be ignored in the derivation.
The general assumptions are: 1) The source and receiver are far from
the illuminated area. 2) The dimensions of the source are small compared
to R-. and R . 3) The source is a Gaussian shaded transducer and is directed
along R^ , (Fig. 2.1). Also, the individual elements are in phase and
incident sound pressure at (x,y,^) is the integral over ds ' . 4) No shadows
are present.
SOURCE
Fig. 2.1 Geometry
ds is at the position x, y, ^ relative to the origin. The
plane of the source is perpendicular to R .
Because the surface is rough, we can expect that at grazing angles
some of the surface will be in shadow. The proportion of the area in
shadow to the total area scanned depends upon the shape of the surface
and on the grazing angle. Wagner (1967) discussed this for rays and a
randomly rough surfaces. In any case, the assumption of no shadowing
is violated. It is also likely that the radii of curvature of some
features on the surface are small compared to the acoustical wave-
length. In that case, the sound does not have a local plane reflection
and in order to describe the scattered sound we mist use higher order
expansions.
Eckart (1953) remarked that the boundary conditions are troublesome.
For example, one can set
P = Sp^ (2.1)
where p^ and p are the incident and reflected signals on the local surface
and ^/ha is the normal derivative. Using ER = -1 for a free surface, Eckart
suggested that in the smoother areas, the second condition might hold. In
deep shadows, he thought that
h,.b.
Sn = ^^ (2-«
might be reasonable. Horton and Muir (1967) assumed an average boundary
condition and combined the two equations involving the normal derivatives
to obta in
* P2C2COS9 - P C COS0
^ " P,c rose, + p^c^cose'' » reflection coefficient for plane waves,
2 2 1 11 r
here 1 and 2 refer to medium 1 and 2. is refracted direction
m medium 2 .
8
hp /^ = or hp/hn = (2.4)
They also used Eckart's small slope approximation and replaced the
normal derivative by b/bz. Horton et al. (1967) compared theoretical
computations and the experimental scattered sound and found the
agreement to be quite good. The slope correction can be included
by performing an integration by parts [Tolstoy and Clay, p. 196-199
(1966)]. In the specular direction, the result is the same as
obtained with h/bn. ~ ^/^z. The result is different for backscattered
s ound .
The most direct way to demonstrate the approximations is to
start with the integral expression for the scattered acoustical
pressure. With the aid of the Helmholtz theorem [Born and Wolf,
p. 375 (1965)] it is
r- /gMs^" ""t \P)
P = 47- / hi ^- U^ I, 1 ds (2.5)
i(kR - a)t^)A
where U = e^^'^" ""^Z'/r (2.6)
i(kR - cot )/ •
Pi - B e 1/R ^2.7)
B^ = npc {2iO'^
n = source power
I
and the normal is drawn toward the receiver. R and R are the source
and receiver distances to ds (at x, y, z) . In applying the Helmholtz
theorem, the surface is closed by a hemisphere at infinity. There
are no other sources and all waves are outgoing from our source.
Hence the contribution to the integral for the surface at infinity can
be ignored and the integral over the illuminated interface is used.
We choose to use the less conventional form of the normal toward
the receiver because that is the direction of the wave propagation
and it does not matter where the source is placed. The substitution
of (2.1) and (2.2) into (2.5) gives
^ ^' /.
"^ %r (Pi") I ds (2.8)
or the Hdrton-Muir condition, 2.4 gives
ds (2.9)
P = 4^ ^^ Pi
After expansion of the integrands for a moderately directive source,
the integrals will be the same and the kind of boundary condition
depends upon a constant factor in front of the integral.
There are several ways of writing the algebra for the expansion
and they are all messy. Since the transducer plane is perpendicular
'2 2
to R , R and R are
1 2 ' 2 ' 2 '
R = (-R sin + X cos 9 - K) + (y - y) + (R cos0- - x sinS.-^'
R = (R sine^cose„ - x)^ -f (R sinS^sine - y)^ + (R^cosQ. -
(2.10)
With the aid of the binomial expansion and retaining ^ and second
order terms, R and R are
'2, '2 x^cos^e,+ y^ x X cose,
R R I X + v 1 I . I_^
- 1 2R 2R R R
+ X sin0 - ^ cos© + . . .
2 1 \ v^ r 2 2^
1 - sin cos e ) + -J^ (^1 - sin ^2^^'^ ^3/
2 . . , ^ 2
- X sin0^cos6»^ - y sine^sin©^ - ^ cos0^ + . . . (2.11)
10
For small slopes, the normal derivative is approximately o/oZ (or ^/^
here). On making this approximation and also including the integral over
the transducer ds', one obtains
P ~
ik BOT e
-icut
2n R^ R^
r , , r ik (R • + R)
jgds je
ds
(2.12)
(cos© + cos9 )/2, boundary cond. (2.1 and 2), ^/^ ~ ^/hz
F =
1 + cos9i + cosS^ - sinS, sin0- cos9„ , , /« -.^x
1 2 1 2 3 , slope correction (2.13)
cos9 + cos0„
cos0_/2, boundary cond. (2.4)
(Tolstoy and Clay, 1966)
,2
S = 1^ ^^P
W
(2.14)
Integration over the source yields (after algebraic manipulation)
f-iQt -ik(R^ + S^l
ik B9F e -'
P ~ - T 7-
2n R^R^d - id^)^ (1 - id^)
// exp -ax -ay +2 icnx + 2 ipy + 2 iyM dydx
ik cos^e^R Fl - i (1 - R-/R ) d, 1
i x L 2 X L J
a =
X
2 R^R.
1 - id.
a = - ^
y 2 R^R2
ik R [l - i (1 - R^/R )d^]
1 - id
w
(2.15)
R s R^ + R^ (1 - sin^e2 sin^e^)
d^ = kL^/(2R^) d^ = kW^/(2R^)
(2.16)
2a = k(sin0 - sin0 cosS )
2p = -k(sine sine )
2y = -k(cos9 + cos0 )
11
Integration of 2.15 for ^ = yields the image solution. Although
the size of the transducer is included and the expressions are dependent
on L and W, I don't think the expressions are accurate for near
field computations because the higher powers of x and y were dropped
in 2.11.
Proceeding, the covariance of the sound pressure is
/p (t + T) p* (t)\ =
, 2„2_2_2 -iooT
= k B gj F e
2 2 2 2 i: 2 J-
hT^W (1 + dL^)^(l + d^ )^
(il/r^^'i'^x^-V" - V^- V
2 * ,2 2 * ,2
-ax'-^-ay -"'
XX y
+ 2ia(x-x') + 2iP(y-y') + 2i^'(^ -^') }
dy dx dy' dx'y (2.17)
I assume the surface is random and ^ has a zero mean. The correlation
function of t is X X. T
^ 2 2 2
X Y T
"2 2 2
where in functional notation ^ could be written as i^(|,ri, t,).
Assuming a Gaussian surface, the bivariate Gaussian probability density is
2-1 2-3-
W - (2na ) (1-t ) 2
exp
/ - [ 2 a-^^)o^] ~^ [ t^ + V'^ - 2^^'^3 J (2.19)
Since the only random quantities are t, and ^', the average in (2.17)
operates on them as follows:
•k
For a given roughness wave length A, the shorter wave length roughness
in an area having the dimensions of several A has the same statistical pro-
perties as all other similar size areas.
12
(e^irC^ -r)> = //we2i^(^r),^,^.
= S
/ \ = exp [ - ^/a^a-^)] (2.20)*
The following changes of variable permit integration over x" and y",
X = x" + 1/2 y - y"+ T,/2
x' = x" - 1/2 y'- y" - T,/2
iie substitution of 2.20 and 2.21 into 2.17 and evaluation of the
integral yields, after the usual algebra,
2„2 2 -iooT
(2.21 )
/p(t + t) p*(t)\ ^ B-F-g?-e
2nR-^ LW cose
r r expV- a^l^ -a t]^ -f- 2iQ ^ + 2ipTi - 4r^cr^ (l-i|r)l dl dTi (2.22]
-00
2 2 r~ 2 li
R COS0T 1 + d^ (1-R2/Ry)
where a, = ^^ 3_J= L ^ 4
5.22
2 R2 L
R ^ Tl + d ^(1-Ro/Rv)^]
a = y ^ , -^
1 2 2
2^2 " (2.23)
If '^ can be expressed as a sum of first and second order polynomials
in 5 , T\, and t, (2.22) can be integrated directly for all values
of rev.
The intensity of scattered sound is often expressed with the
aid of a scattering function S as follows:
g2
(p^) = — T S (2.24 )
* r 2 2 1
In Section 6, I replace exp [_- 4y a (l-i|f)J by the characteristic function
^2 and remove the restriction that the distribution be Gaussian.
13
where A = illuminated area
To change (2.22) into this form I need to express A in terms of the
transducer dimensions and R . Since transducers are often described
by their Fraunhofer "beam width", I do so here. For plane waves,
defining and X as shown on Figure 2.2, and ignoring time dependence,
,2 .2
/ fexp I - ^
nLW
|p| = -L- Tfexp I - ^- - K- - ikx'sin0 - iky'sinxj dx' dy' (2.25)
L2 2 2 2 2 2 I
- k L sin 0/4 - k W sin X/4J
The response is shown on Figure 2,3. I choose to use |p|= e to define
A0 and AX, as follows
at X = 0, (kLsin0)/2 = 1
and at 0=0, (kWsinX)/2 = 1
Thus sin0 ~ A0 = 2/ (kL)
sinX ~ AX = 2/ (kW) (2.26)
The substitution of (2.26) into (2.22) and casting into the form
(2.24) yields
/p (t + t) p*(t)\ = —^ — f- S (2.27)
S = k^F ^v^e"^ ^ r Texp -a^^^ - a T]^ + 2 10:^ + 2 i^i\
- 4/a^ d-!')] d^ dT] (2.28)
A = R^ A0 AX/ (cos0^) (2.29)
Eq. (2.22) is equivalent to Eq. (2.27) and (2.28). We can proceed in several
ways, direct evaluation at 2.28, by numerical integration, polynomial ex-
pression of ^ and integration, and special functions for ^ and integration.
14
Fig. 2.2 Transducer. The transducer is Gaussian shaded. L and W
indicate the dimensions for e amplitude shading.
15
Fig. 2.3 Response of the Gaussian shaded transducer. The far field
(Fraunhofer) response of the Gaussian shaded transducer,
g = exp [-x'2/l2]
16
Before we cast off, sail into the confused sea, and expand the
correlation functions, it will pay us to look at the convergence of
(2.22). The contributions to the integral are small for values of
i and T] larger than
^f > ^^ ^f ^ \ (2.30)
and for R^ = R ^1 " ^2
where ^l^ - L/cos0^
a ^ ~ W
Ti -
Since the contributions are small, we need not worry about the shape
of ^ at distances larger than i and t] in evaluating the integral.
Or, the dimensions of the transducer and 0, determine the sensitivity
of the scattered sound to ^. On the other hand^ accurate fits at small I and
T] are extremely important at large xo.
1 would like to close this section by giving my thoughts on the
physical significance of the I, t] integral for S. This Integral
ought not be confused with the first integral over the illuminated
area because the ^ , t) integral relates the phase of the scattered
signal at any point on the surface relative to a nearby point at
a displacement i and t]. The contributions to S are small for
i and i^ greater than i and t) . Paraphased, the integral is
in correlation space. For a given roughness, the constancy of the
phases of the scattered components of signal depends upon the dimen-
sions of the transducer.
17
m. EVALUATION OF THE SCATTERING INTEGRAL
Experimental measurements of the i, t dependence of wind blown
water waves have shown that ^ has the form of traveling wave packet;
Fig. 3.1. The envelope moves at the group velocity and the phases
35--
30--
25--
20--
V*
., _L » 10 ISTSEC
Fig. 3.1 i a, t)
On the graphs R(t) is our t
of the damped oscillation move at the phase velocity. Near the origin
(i > I , etc), the dependence on I and t can be approximated as
^ ~ ^if(^ - vt) (3.1)
for waves traveling in the + x direction. i|f is S3nnetric about
1=0, and T =
^ (t) = ^(-e)
^ (t) = ^(-t) (3.2)
18
Near ^ = 0, ^ can be approximated as a polynomial
t~l-ax -b|xl
or \1^ - 1 - a^^ + 2 ax&r - b|| - "D'tJ - a^v'^r^ (3.3)
and the expansion has cross terms ^t. Correspondingly, waves moving
in the y direction have rix terms and waves moving in an arbitrary
direction have both. This doesn't make the analysis more difficult
because t is a parameter.
In an earlier paper, I used polynomial fits to the correlation
function for a random surface (Clay, 1971). The procedure was to
divide the surface into sub areas and to integrate the scattering
function for each of the sub areas. I will do the ssime here except
that the t dimension is added to the problem. The I, -q and t
nomenclatures are shown on Figure 3.2. A polynominal expansion for the
ijkth sub region is
- f ...T] - g'. ., T^ - h' ... T - m'..,|T
i-jk ' ° ijk ijk ijk
,2 2 2 2
- Tl . ., T]T - r '. ., I T - S . ., Tl T
' ijk ' ijk ijk '
(3.5)
It isn't necessary to expand the t dependence as a polynomial in
some problems. For example, the correlation function can be expanded
as the product of polynomials in ^ and functions of t such as cos pr
and sin pr. If I had the polynomial form programed, I wouldn't bother.
The coefficients of the variables i and t) can be combined
19
§ L J '^^^
Fig. 3.2 sub areas
The I, -n map is at constant t. The I, t map is at constant t\
^1/ is assumed to be for regions beyond i^^^, r\ ^^ and t
max 'max max
20
2
C — c' - c' T - h ' T
ijk ijk ^ ijk ijk
2
a, ., = a' . ,, + r' . ., T
ijk ijk ijk
xjk = b' , ., + m . ,, T
ijk xjk
• , 2
e. ., = e '. ,, + s '. ., T
ijk ijk ijk
ijk " ^'ijk "^ '"'ijk^ (3.6)
2
t. ., = c. .. - a. .,1 - b, ., I
ijk ijk ijk xjk
2
" ^ijk^l " ^iik"! (3.7)
On designating S for the kth t region, we can write
E
-ijk ^ 8n ^ h""^ L~ H' - V
S..,. =k:C£:^-' ll^^p [•_,2 __2
^-1' ^J-1
+ 2ia| + 2ipT] - 4r^a^(l - ^^.j^) I dUr] (3.8)
The integral has the form of the product of two complex
error integrals. To use the tabulations, we transform the variables.
Since the integrals on |£ and t\ have the same form, we use an
integral on x and let it stand for either. Note that erf (z)
and erfc(z) are defined by Abramowitz and Stegun(1965) as follows:
erf(z) s — — / e dt
21
erfc(z) = _2
; /"•"
dt
-z
w(z) = e erfc( - iz)
(3.9)
We now cast S. into the form of (3,9) by defining the constant C
ijk ^ ^ ijk
and functions U , V ag follows
ijk ijk
2 2 2
U V
ijk ijk ijk
ijk 32
=ijk = exp[-4rV(l -c.j^)]
2 2
^i
ijk
exp
U-1
where with the aid of (2.23) and (3.7)
1
- B . .1
~7~ XX J k
xijk
(3.10)
(3.11)
d^ (3.12)
2 2
al + 4r cr a, .,
ijk
A 2. .,
X ijk
xijk
2 2
iO; + 4 r a b. .,
ijk
(3.13)
Similar expressions for the j) dependence are
ijk
exp
- -JL
^j-1 L ^ijk
yijk'
drj (3.14)
yijk
2 2
a + 4r cr e. .,
T] ijk
2 2
B . ., = - ip + 4 r cj f. .,
yxjk ijk
(3.15)
Eq . (3.13) and (3.14) have the same form so we drop subscripts and
cast it into the form of (3.9) by completing the square and changing
variables as follows:
22
t = X ,
'- + s
s = AB/2
g(i-l) =-1^1 + s
A (3.16)
I
g(i) = -1 + s
A
2 ^_/..x .2
A s /-gCi) -t ,^
U = Ae / e dt
/
g(i-l) (3.17)
2
U = A c^ erfc[g (i-1)] - erfc [ g(i) ] (3.18)
If g is real, (3.18) can be used for the evaluation of U and V.
This would be expected for the specular direction when 0! and P
are zero. For complex g, we change the form of erfc(g) to w(z)
in (3.9).
Let z = i g (i-1)
2
then e^ erfc [g (i-1)] = exp [s^ - g^(i-l)] w[ig(i-l)] (3.19)
Eq. 3.18 with the aid of (3.19) is
2
U = A e^ ^ e
2 2 "I
"^ ^^"^^ w[ig(i-l)] - e ' S ^^^w[ig(i)] J
(3.20)
V has the same form as U.
The substitution of A , ., , B , ., , etc. into (3.16) is simple and
xijk xijk
the evaluation (3.20) for U. ,, and V... follows directly. I don't
ijk ijk
see much reason to write the expression because there is a chain of
substitutions through to (3.5) for the t dependence. The analytic
evaluation of (3.8) is complete. From all of this mess, I would
not expect the time dependence of the scattered signals to have a
simple relationship to the time dependence of the surface.
23
IV. COHERENTLY SCATTERED SIGNAL
For my estimate of the coherently scattered signal, I will use the
Eckart procedure and calculate the ensemble average of p. This method is
very general and is applicable to a much wider range of conditions than is
superfically apparent. By considering the more general problem, it is
possible to use the dependence of the coherently reflected signal on y
to determine the probability density function of the surface W.
I begin by assuming the coherent signal is an ensemble average of
many transmissions \P/ . The sound pressure is given by Helmholtz
integral and the expansions of R and R' , Kq. (2.11 and 2.12). For a
moderately directive source and ^ very small relative to R and R' , I can
write
R + R' ~ R(^ = 0) + R' (^ = 0) + 2r^ (4.1)
I use [space] to represent all of the non-random part of x, y, and t
dependence of the Helmholtz integral,
p = /[space] e ^ds (4.2)
>-'-00
and r.00
p = / [space] ds
o /_
for ^ =
(4.3)
The latter expression is exact for all configurations and (4.2) is only
restricted by the approximation, (4.1). The ensemble average of (4.2) is
<^p^ = ^' [space] ds (e^^O ^^'^^
24
The average in (4.4) is given by
00
(e^^^; =/w^|i^^d^ (4.5)
-00
and this is the characteristic function of W . Hence, we can write the
following relations
[<»/Po] = /^W^e^^^d^ (4.6)
"^ —00
00
-00
Since \pVp is a function of positive yo, Eq. 4.7 should be regarded
as the sine and cosine integrals from to oo. The normalized coherent
reflection is the Fourier transformation of the pdf of the surface.
For a test, I use the experimental data of Mayo, Wright, and Medwin(70)
[T 9 9 9
/p\ /p versus ^y a = g,
where g is often referred to as the roughness parameter. If W is Gaussian
then
k'pN/p ~ exp(-g) for Gaussian W (4.8)
howevei^ their data did not fit (4.8). The data agreed quite well with
(4.8) for g less than 4 (or ya < 1). At larger yo, the coherent com-
ponent is orders of magnitude larger than predicted by (4.8). All aspects
of the measurements were painstakingly re-examined, including finite
illuminated area, but no "mistakes" were identified. The non-fit became
the motivation of this research. For simplicity in a numerical test, I
approximated their measured distribution function by the linear segments
shown on Figure 4.1. The substitution and evaluation of (4.6) for the
piece-wise linear function is routine and omitted here. A comparison of my
calculation of \P//p and their data are shown on Figure 4.2. The phase
of the coherently reflected signal probably changes as ya increases. The
phase is needed in the evaluation of Eq. 4.7. Since the phase is needed
for the inverse transform, I have not attempted to do it.
25
The procedure we have described, measure the reflected signal at a =
and then the coherently reflected signal for the roughened surface is easy
to do in the laboratory. Doing this for a rough sea surface would require
highly accurate measurements and calculations. I suggest that an alter-
native procedure might be to measure the cross correlation of signals
at a pair of separated hydrophones.
Fig. 4.1. Experimental probability density function of the model sea surface
a = .45 cm. The solid line is the piece-wise linear
approximation to the distribution function. Data furnished by
Mayo, Wright and Medwin.
26
Fig. 4.2. Coherently reflected signal. The data points are from Mayo,
Wright and Medwin (1970). The solid line is calculated for
the probability density function shown on Figure 4.1. The data
are for angle of incidence =45° and a = .45 cm,
27
V. COSINE CORRUGATED SURFACE
The surface of water waves often looks like a cosine corrugated surface
for relatively long times and over fairly large areas. If one watches
carefully the phases and amplitudes of the wave change randomly. It is easy
to make a cosine surface random analytically by letting the phase change
randomly between each observation (or signal transmission). In addition,
an ensemble can be formed of waves having different amplitudes. I assume
the m'-" wave surface is given
f = f cos K (x - vt - X )
^ ^m m
(5.1)
Where v is the wave velocity, t is the amplitude, and x is the phase.
^m m
The procedure for calculating the scattered sound signal is to substitute
(5.1) into (2.16) and to expand the result with the aid of the following
expansion in Bessel functions
^ia cos bx ^ J ^^ _^ 2 \ (i)"j (a) cos (nbx)
L
1
(5.2)
The result of integration and manipulation is
2 r 2
p = p e"P /^y )j (2r^ )e"" /^x +V(i)'' J (2rC ) exp
o ^ o ^m / n ^m
( 2a + nK) . ,^, , ,
^ — ; - inK(vt + x )
4a m
X
+ ia)\(2nj exp
1
( 2a - nK) , . ^, ^ , X
— ; + inK(vt + X )
4a m
X
(5.3)
where
A S B e
_ _q
*o ~ R^ + R^
i[cDt - k(R3^ + R2)]
(5.4)
28
A =
o
' '^^'l
- 1/2
1 - ^Vw
y.' _l
- 1/2
(5.5)
Before I wipe out most of the terms in (5.3) I should discuss them. The
traveling water wave introduced an infinite series of time-dependent terms
having frequencies nKv. These terms contribute a modulation of
the reflected signal. The higher harmonics are more important at larger
roughness, i.e. y"^ • If the reflected signal were processed by means of a
spectrum analyzer, the spectrum would be like that shown on Figure 5.1.
The powers in the components depend on the bistatic geometry.
L
9 e 9
1
frequency
Fig. 5.1. Spectrum of the signal reflected at a traveling water wave.
O) is the frequency of the sound signal and Kv is the frequency
of the water wave.
29
The simplest ensemble average is obtained by letting all phases of the
traveling wave be equally likely. An average over Kx from to 2it eliminates
all of the terms in the summation in Eq. (5.3). In the specular direction
(5.3) then reduces to
The distribution function of a cosine wave is
f = f cos K X
^ ^m
(5.7)
W = C > t
c ^ ^m
The circle can be completed by averaging (5.6) over a random set of
amplitudes. For example, the envelope of a Gaussian random function has
a Rayleigh distribution function:
W^ = o'^ ^ exp [- ^^/(2a^)j (5.8)
r J^ (2ri ) Wj^d^ = exp [-2r^a^J (5.9)
"^ o
This would be expected on the basis of the central limit theorem. I suggest
that fair approximations to non-Gaussian functions can be made by averaging
(5.6) for several values of ^ .
30
VI. TOTAL SIGNAL SCATTERED AT A NON-GAUSSIAN SURFACE
This is my last section and lest the reader has forgotten, my purpose
is to do the inverse problem. The basic formulation of the scattering
problem is given in Sections 2 and 3. There, the emphasis was on going
from a known surface to calculated scattered signal. This analysis
started with a bivarate Gaussian surface and its correlation function.
I could have used the characteristic function and removed the Gaussian
restriction as follows:
S
.<e^^^^^-?'*> (6.1)
2 2 1
The only change is to replace exp^ -4y a (1-^1^ )j by C in Eq. (2.22).
Following my procedure in Section 4, I evaluate p and express Eq, (2.22)
as follows:
/ A / 2 - x% -1 -icoT
//d exp [2iQ^ + 2i3Ti] C^d^d (6.2)
L2 2 "I
- a.^ - a T) J (6.3)
Since a. and a are functions of k, 9, , 9^, and 9., they can also be
expressed as a function of o; and 3. I make a Fourier transform operation
-1 ,
on assuming that D and D are slowly varying functions of s and t].
Integration of CC and P yields 6(^ - ^') and 5(t] - t) ' ) type functions and
the expression for C„ is the following
00
1 rrP/ ^-k-l vpp*) icjOT r -I
2n^jjJ ^^\ 2 ^ ^''P [-^(^'^ + 20;^ + 23ti)J dadpdw' = C^ (6.4)
-00 Pq
31
The q: and 3 dependences of a, and a are greatly simplified by letting
9 = -Q-^> Qo ~ 0> ^'^d holding R^ and R constant for a set of measurements,
Obviously, one's ability to do the inverse transformation depends upon
having accurate values of ^pp*^ . I assume the user would do the trans-
forms numerically. Although the integral has infinite limits, actual
measurements of /pp'=^ give both upper and lower limits. There is another
consideration. Since D operates as a spatial high pass filter in (6.2),
I doubt if C would have much accuracy at I and t] much larger than i and
32
References
Abramowitz, M. , and I. A. Stegun, Handbook of Mathematical Functions,
N.B.S. Applied Mathematics Series 55, U.S. Government Printing
Office, Washington, D.C. (1964).
Beckmann, P., and A. Spizzichino, The Scattering of Electromagnetic Waves
from Rough Surfaces , The MacMillon Co., New York, (1963).
Born, M. , and E. Wolf, Principles of Optics , Pergamon Press, (1965).
Clay, C. S., Notes on Ocean Acoustics, Univ. of Wis. Geophysical and
Polar Research Center Research Report, No. 71-2, August 1971.
Eckart, C, J. Acoust. Soc. Amer. 25 , p. 195 (1953)-
Fortuin, L. , J. Acoust. Soc. Amer. 47 , p. 1209-28 (1970).
Horton, C. W. , Sr., J. Acoust. Soc. Amer. 51 , p. 1049-61 (1972).
Horton, C. W., Sr., and T. G. Muir, J. Acoust. Soc. Amer. 41 , p. 627-34 (1967).
Mayo, N., H. Medwin, and W. M. Wright, J. Acoust. Soc . Amer. 47, p. 112(A),
(1970).
Mel
ton, D. R. , and C. W. Horton, Sr., J. Acoust. Soc. Amer. 47 , p. 290-98 (1970)
Ol'shevskii, V. V., Characteristics of Sea Reverberation , Consultants
Bureau, New York (1967).
Tolstoy, I., and C. S. Clay, Ocean Acoustics: T heory and Experiment in
Underwater Sound , McGraw-Hill Book Co., New York, (1966).
Wagner, R. J., Shadowing of randomly rough surfaces, J. Acoust. Soc. Amer. 41,
p. 138-47 (1967).
33
INITIAL DISTRIBUTION LIST
No. of Copies
1. Conunander, Naval Ship Systems Command 2
Attn: Code PMS 388
Navy Department
Washington, D.C. 20350
2. Commander, Naval Ship Systems Command 1
Attn: Code PMS 302
Navy Department
Washington, D.C. 20360
3. Office of Naval Research (Code 468) 2
800 N. Quincy Street
Arlington, Virginia 22217
4. Director, Naval Research Laboratory 1
Technical Information Division
Department of the Navy
Washington, D.C. 20390
5. Director, Advanced Res Projects Agency 1
Technical Library, The Pentagon
Washington, D.C. 20301
6. Defense Documentation Center 12
Cameron Station
Alexandria, Virginia 22314
7. Commanding Officer 1
Office of Naval Research Branch Office
219 S. Dearborn Street
Chicago, Illinois 60604
8. Commanding Officer 1
Office of Naval Research Branch Office
1030 East Green Street
Pasadena, California 91101
9. Commanding Officer 1
Office of Naval Research Branch Office
495 Summer Street
Boston, Massachusetts 02210
10. San Francisco Area Office 1
Office of Naval Research
1076 Mission Street
San Francisco, California 94109
34
11. New York Area Office
Office of Naval Research
207 W. 24th Street
New York, New York 10011
12. Director
Naval Research Laboratory
Library, Code 2029 (ONRL)
Washington, D.C. 20390
13. Commander
Naval Ordnance Laboratory
Acoustics Division
V?hite Oak, Silver Spring, Maryland 20910
14. Commander
Navy Undersea Res. & Develop. Center
Technical Library
San Diego, California 92152
15. Of ficer-in-Charge
Technical Library
Naval Underwater Systems Center
Newport, Rhode Island 02840
16. Commanding Officer
Technical Library
Naval Underwater Systems Center
Fort Trumbull, New London, Conn. 06321
17. Naval Ship Res. & Develop. Center
Central Library
Washington, D.C. 20034
18. Commanding Officer
Naval Air Development Center
Johnsville, Warminster, Penn. 18974
19. Commanding Officer
Navy Mine Defense Laboratory
Panama City, Florida 32402
20. Naval Research Laboratory
Underwater Sound Reference Division
Technical Library
P.O. Box 8337
Orlando, Florida 32806
21. Naval Weapons Center
Technical Library
China Lake, California 93555
22. Naval Undersea Warfare Center
Technical Library
3202 E. Foothill Boulevard
Pasadena, California 91107
35
23. Office of the Director of Defense
Research and Engineering
Information Office Library Branch
The Pentagon
Washington, D.C. 20301
24. Technical Library-
Naval Air Development Center
Warminster, Penn. 18974
25. Director, NOAA
National Ocean Data Center
Rockville, Maryland 20852
26. Commander
Naval Oceanographic Office
Washington, D.C. 20390
Attn: Code 1540 (Library)
27. Naval Postgraduate School
Technical Library
Monterey, California 93940
28. Naval Academy
Technical Library
Annapolis, Maryland 21401
29. Research & Technology Directorate
Naval Electronics Systems Command
Department of the Navy
Washington, D.C. 20360
30. RDT&E Planning Division
Naval Ship Systems Command
Department of the Navy
Washington, D.C. 20360
31. Research and Technology
Naval Air Systems Command
Department of the Navy
Washington, D.C. 20360
32. Research & Technology Directorate
Naval Ordnance Systems Command
Department of the Navy
Washington, D.C. 20360
33. Office of Naval Research
Code 461
Department of the Navy
Arlington, Virginia 22217
36
34. Office of Naval Research
Ocean Science & Technology Div. (Code 480)
Department of the Navy-
Arlington, Virginia 22217
35. Dr. Paul A. Crowther
Naval Division
Marconi
Chobham Road
Frimley, Camberly, Surrey
England
36. Dr. H. Medwin, Code 61 Md
Naval Postgraduate School
Monterey, California 93940
37. Mr. R. Uranicar (Code 370E)
Naval Air Systems Command
Department of the Navy
Washington, D.C. 20360
38. Sonar Directorate (Ships 901)
Naval Ship Systems Command
Department of the Navy
Washington, D.C. 20360
39. Librarian
Applied Research Lab
University of Texas
Austin, Texas 78712
40. Prof. C.S. Clay
Dept. of Geology & Geophysics
Geophysical & Polar Research Center
University of Wisconsin
Middleton, Wisconsin 53562
41. Dean of Research
Naval Postgraduate School
Monterey, California 93940
42. Dr. Jerald W. Caruthers
Department of Oceanography
Texas A & M University
College Station, Texas 77843
43. Dr. R.G. Williams
NOAA
Rockville, Maryland 20852
37
Unclassified
security Classification
DOCUMENT CONTROL DATA -R&D
iSiTuniy classification ol tlllu, body of ahslraci and indexing annolaliun mu.sl he entered when the overall report is classified)
1 OHIGINATING ACTIVITY (Corporate aultior)
Naval Postgraduate School
Monterey, California 93940
2a. REPORT SECURITY CLASSIFICATION
Unclassified
26, GROUP
3 REPORT TITLE
Coherent and Incoherent Components of Sound Scattered at a Time
Dependent Rough Surface
4 DESCRIPTIVE NO T E5 (Type of report and, inclus i ve dales)
Technical Report, NPS-61Md72121A 31 December 1972
5 AUTHOR(S) (First name, middle initial, last name)
Clarence S. Clay
6 REPOR T DATE
December 1972
Sa. CONTRACT OR GRANT NO.
b. PROJEC T NO.
7a. TOTAL NO OF PAGES
39
76. NO OF RE FS
13
9a. ORIGINATOR'S REPORT NUMBERIS)
9b. OTHER REPORT NO(S) (Any other numbers that may be assigned
this report)
10. DISTRIBUTION STATEMENT
Approved for public release; distribution unlimited,
11. SUPPLEMENTARY NOTES
12. SPONSORING MILITARY ACTIVITY
Naval Postgraduate School
Monterey, California 93940
13 ABSTRACT
Theoretical expressions are derived for the sound scattered at a time-
dependent rough surface. The calculations are made for a Gaussian
shaded source transducer and point receiver. The Helmholtz theorem and
Fresnel approximation are used. The rough surface is assumed to be a
traveling wave and to have a traveling wave packet type of correlation
function. The coherent component of the signal is the product of the
Fourier trans foarmation of the surface distribution function and the
smooth surface reflection signal. Comparison of theory and experiment
shows the coherent component to be sensitive to the non-Gaussian char-
acter of the wind-blown water waves. The incoherent components and the
temporal correlation function of the scattered sound are given. For the
special case of a traveling cosine wave type of rough surface, spectrum
of the scattered sound includes components which are multiples of the
frequency of the surface wave. For surfaces describable by a bivariate
Gaussian distribution function, the temporal correlation is a function
of, but not the same as, the time dependence of the rough surface. The
scattered sound is insensitive to the spatial correlation function of
the surface at distances larger than the dimensions of the transducer
divided by the cosine of the incident angle. The final expressions are
complex error integrals and can be used for all values of roughness.
This task was supported by Naval Ship Systems Command (Code PMS 388) .
DD
""" 1473
I NO V fiS I ^ / *J
S/N 01 01 -807-681 1
(PAGE 1 )
38
Unclassified
Security Classification
A-3t408
Unclassified
Security Classification
KEY wo R DS
Rough-surface transmission
Sound scattering
Coherent reflection
DD ,rr..1473 BACK)
ROLE W T
39
Unclassified
S/N OIOt-807-6821
Security Classification
A- 3 I 409
DUDLEY KNOX LIBRARY
3 2768 00391402 9