Underwater Acoustic Systems

Macmillan New Electronics Series

Series Editor: Paul A. Lynn

Rodney F.W. Coates, Underwater Acoustic Systems

Paul A. Lynn, Radar Systems
A.F. Murray and H.M. Reekie, Integrated Circuit Design
Dennis N. Pim, Television and Teletext
Martin S. Smith, Introduction to Antennas
P.M. Taylor, Robotic Control

Series Standing Order

If you would like to receive future titles in this series as they are published,
you can make use of our standing order facility. To place a standing order
please contact your bookseller or, in case of difficulty, write to us at the
address below with your name and address and the name of the series.
Please state with which title you wish to begin your standing order. (If you
live outside the United Kingdom we may not have the rights for your area,
in which case we will forward your order to the publisher concerned.)

Customer Services Department, Macmillan Distribution Ltd

Houndmills, Basingstoke, Hampshire, RG21 2XS, England.
Underwater Acoustic
Systems

Rodney F. W. Coates

Professor of Electronics
School of Information Systems
University of East Anglia
Norwich

Macmillan New Electronics

Introductions to Advanced Topics

M
MACMILLAN
© Rodney F.W. Coates 1990

All rights reserved. No reproduction, copy or transmission

of this publication may be made without written permission.

No paragraph of this publication may be reproduced, copied

or transmitted save with written permission or in accordance
with the provisions of the Copyright Act 1956 (as amended),
or under the terms of any licence permitting limited copying
issued by the Copyright Licensing Agency, 33-4 Alfred Place,
London WC1E 7DP.

Any person who does any unauthorised act in relation to

this publication may be liable to criminal prosecution and
civil claims for damages.

First published 1990

Published by
MACMILLAN EDUCATION LTD
Houndmills, Basingstoke, Hampshire RG21 2XS
and London
Companies and representatives
throughout the world

Typeset by
Comind (UK), Cambridge

British Library Cataloguing in Publication Data

Coates, R.F.W. (Rodney F.W.)
Underwater acoustic systems.
1. Acoustics. Underwater
I. Title
620.2'5

ISBN 978-0-333-42542-8 ISBN 978-1-349-20508-0 (eBook)

DOI 10.1007/978-1-349-20508-0
Contents
Series Editor's Foreword viii
Preface ix
1 Sound Transmission Fundamentals 1
1.1 Introduction 1
1.2 Sound Speed 3
1.3 The Propagation Equation 6
1.4 Logarithmic Power Measurement 8
1.5 The "New" Reference Unit 8
1.6 Sound Reflection and Refraction 10
1. 7 Pressure Reflection and Transmission Coefficients 11
1.8 Behaviour at Normal Incidence 12
1.9 Reflection and Transmission with Varying Incidence 13
1.10 Reflection and Transmission: the "Fast" Bottom 14
1.11 Reflection and Transmission: the "Slow" Bottom 15

2 The Sonar Equations 16

2.1 Introduction 16
2.2 Source Intensity Calculation 16
2.3 Source Directivity 17
2.4 Transmission Loss 18
2.5 Target Strength 22
2.6 Reflection Intensity Loss Coefficient 24
2. 7 Sea-floor Loss 25
2.8 Sea-surface Loss 26
2.9 Noise 28
2.10 Reverberation 29
2.11 Calculating the Signal Excess 30

3 Characteristics and Analysis of Sonar Waveforms 32

3.1 Introduction 32
3.2 Swept Frequency (Heterodyne) Spectrum Analysers 34
3.3 Filter-bank Spectrum Analysers 36
3.4 Fast Fourier Transform Analysers 36
3.5 Prony Analysis 40
3.6 Further Model-building Techniques for Spectral Estimation 41
3.7 Four-dimensional Space-Time Waveform Analysis 43

v
vi Contents

4 Ray Trace Modelling of Sonar Propagation 52

4.1 Introduction 52
4.2 Ray Tracing Sonar Models 53
4.3 Ray Trace Calculations 56
4.4 Some Examples of Ray Modelling 58
4.5 Modelling Transmission in the Shelf-seas 65
4.6 The Lloyd Mirror Effect 70

5 Normal Mode Modelling of Sonar Propagation 73

(co-authored by P.A. Willison)
5.1 Introduction 73
5.2 A Heuristic Treatment of Normal Modes in an
Acoustic Waveguide 74
5.3 Normal Mode Solution for Long Ranges 80
5.4 Normal Modes as Interfering Plane Waves 83
5.5 The Normal Mode Solution Formalised 84
5.6 Normal Mode Solution for All Ranges 87
5.7 The Horizontally Stratified Channel 88

6 Noise and Reverberation 90

6.1 Introduction 90
6.2 Deep Sea Ambient Noise Level 91
6.3 The Variability of Ambient Noise with Time 93
6.4 The Variability of Ambient Noise Level with Depth 94
6.5 The Angular Distribution of the Ambient Noise Field 95
6.6 Ship-generated Noise 100
6.7 Reverberation 104
6.8 Scattering 106

7 Acoustic Transduction 112

7.1 Introduction 112
7.2 The Basic Principles of Acoustic Transduction 113
7.3 Piezo-electric Transduction 116
7.4 The Langevin Projector 117
7.5 Ring and Tube Transducer Designs 122
7.6 Resonance Behaviour of Transducers 123
7.7 Multiple Matching Layer Transducers 127
7.8 Polar Response Measurements on Transducers 129
7.9 Admittance Measurements of Terminal Response 130
7.10 Hydrophones 132
 Contents vii

8 Transducer Arrays 136

8.1 Introduction 136
8.2 The Linear Hydrophone Array 137
8.3 The Fourier Transform Approach to Pattern Synthesis 143
8.4 Array Beamsteering 146
8.5 Directivity Index 147
8.6 The Parametric Source 147
8.7 Synthetic Aperture Sonar 149

9 Sonar Engineering and Applications 153

9.1 Introduction 153
9.2 The Basic Echo Sounder 154
9.3 Sub-bottom Profiling 159
9.4 Fishing Sonars 160
9.5 Side-scan Terrain-mapping Sonars 162
9.6 Seismic Survey 165
9.7 Acoustic Positioning and Navigation 166
9.8 Doppler Measurements 168

10 Acoustic Communications 171

10.1 Introduction 171
10.2 The Gross Attributes of the Received Signal 172
10.3 The Channel Transfer Function 175
10.4 Combating Multipath 178
10.5 Diversity Reception 178
10.6 Equalisation 181
10.7 Communication using Parametric Transmission 183

Index 186
Series Editor's Foreword

The rapid development of electronics and its engineering applications

ensure that new topics are always competing for a place in university and
polytechnic courses. But it is often difficult for lecturers to find suitable
books for recommendation to students, particularly when a topic is covered
by a short lecture module, or as an 'option'.
Macmillan New Electronics offers introductions to advanced topics. The
level is generally that of second and subsequent years of undergraduate
courses in electronic and electrical engineering, computer science and
physics. Some of the authors will paint with a broad brush; others will
concentrate on a narrower topic, and cover it in greater detail. But in all
cases the titles in the Series will provide a sound basis for further reading
of the specialist literature, and an up-to-date appreciation of practical
applications and likely trends.
The level, scope and approach of the Series should also appeal to
practising engineers and scientists encountering an area of electronics for
the first time, or needing a rapid and authoritative update.

Paul A. Lynn

viii
Preface

This text is the result of a period of some fifteen years spent both researching
and teaching, primarily at Master's degree level, aspects of underwater
acoustics. Its content is aimed primarily at a professional engineering or
advanced undergraduate and postgraduate student audience. It is the author's
intention that the book should provide a brisk, comprehensive tutorial
treatment adequately referenced so that the reader may readily delve further
into its subject matter. Its first two chapters are concerned with the basics
of the propagation of sound in the sea and with the preliminary assessment,
via the Sonar Equations, of system performance. Often, both in these first
chapters and, indeed, in the remainder of the text, the treatment of practical
problems- for example, in modelling propagation behaviour- is handled
empirically rather than theoretically. This is because the practising engineer,
or the scientist seeking to utilise underwater acoustics in, for example,
oceanographic investigations, frequently needs only a first-cut visualisation
as to the scope or implication of a particular task, rather than a detailed
mathematical dissection of the problem.
In contrast, in Chapter 3, some detailed consideration has been given to
the need for and problems associated with waveform analysis, since this
is often a most important way of gaining insight into the nature of both
propagation and of system performance. Recent years have witnessed a
dramatic increase in the power and availability of signal processing hardware
together with a decrease in its cost. Additionally, there has been a considerable
broadening of the scope of algorithmic techniques available for application
to signal processing tasks, to the confusion of many who might benefit from
the use of such methods.
Similarly, Chapter 5 delves into the complexities of Normal Mode
modelling of sound propagation in the sea. In contrast to Ray Trace modelling
(covered in Chapter 4) the Normal Mode approach is far from easy to
appreciate. It is, however, of profound importance in describing propagation
in shallow water, or at low frequencies, as well as in sound-channels or
waveguides. Chapter 5 thus purports to present a map of the territory to
assist the reader in venturing further into this intricate area of computer
modelling. The author would wish to express his grateful thanks to Peter
Willison, also of the School of Information Systems at the University of
East Anglia, who played a major part in the writing of Chapter 5.

ix
X Preface

Chapter 6 deals, again often empirically, with the subjects of noise and
reverberation in the sea. Some novel material is introduced here, in discussing
the angular variability of ambient noise.
The subject of acoustic transduction is covered in Chapter 7, and is
followed in Chapter 8 by a treatment of the formation of groups of transducers
into arrays with preferred pattern propagation characteristics. It is
unfortunately the case that the subject of acoustic transducer design is
almost always but poorly treated- if treated at all- in texts on underwater
acoustics. Indeed, the subject as a whole is inadequately covered in the
scientific and technical literature and is almost invariably described by its
proponents as being 90% black art and 10% science. This, of course, is to
be regretted and is, in part at least, a consequence of a policy of need-to-
know on the part of Naval Laboratories and Contractors and in part the result
of a lack of commitment to fundamental research in this area during the
past two or three decades. Regrettably also, the subject can be done but
scant justice, in a single chapter, in a text such as this.
Chapter 9 reviews the utilisation of the various techniques discussed in
the preceding chapters in the construction of a range of underwater acoustic
equipments. In Chapter 10, the burgeoning new field of underwater acoustic
communication, which is assuming considerable importance in scientific
data gathering, communication with autonomous vehicles and sub-sea oil-
field control and telemetry, is also treated in some detail.

Rodney Coates
1 Sound Transmission Fundamentals

1.1 Introduction

The science of acoustics involves the study and practical application of

sound transmission in solid and fluid media. Although the subject is one
of considerable scope, as figure 1.1 illustrates, our interest will lie particularly
in applications involving sound transmission in the sea and in its underlying
sediment layers and rock strata. Sound transmission is the single most
effective means of directing energy transfer over long distances in sea-
water. Neither radio-wave nor optical propagation is effective for this
purpose, since the former, at all but the lowest usable frequencies, attenuates
rapidly in the conducting salt water and the latter is subject to scattering
by suspended material in the sea. Underwater acoustics is thus a topic of
extreme importance in military and commercial applications.

Sound is a longitudinal wave motion which can exist in any compressible

transport medium. Imagine the transport medium to resemble a three-
dimensional lattice of elastically interconnected "particles". Suppose that
one particle is displaced from its rest position and then released. The elastic
interconnection between the particles will allow a disturbance to propagate
outwards from the location of the initial displacement.

The key factors which describe the propagation, physically, are particle
velocity and the local pressure which is responsible for creating particle
displacement. A mathematical study of the physics of sound leads to the
formulation of "wave equations" which are differential equations inter-
relating particle velocity and pressure. These equations, which we shall
examine in greater detail later, incorporate as a "constant of proportionality"
a quantity which determines the rate at which a disturbance propagates
through the medium. This quantity is the speed of sound and is an important
characteristic of all physical media which sustain sound propagation and
of all engineering materials used in equipments for the generation or
detection of sound.

1
 N

FREQUENCY
10mHz 100mHz 1Hz 1 ~ Hz 100Hz 1kHz 10 j Hz 100 r Hz 1MHz 10MHz 100MHz 1 GHz 10GHz
I I I I I I I I I I I

SEISMICS SAWTECHNOLOGY ~
KXXXXXXXXXXXXXXXM

SUB-BOTIOM PROFILING - - SIDESCAN SONAR

POOOOOOOUOOOOOOOOO

ATMOSPHERIC ECHO S O U N D E R S - - NON-DESTRUCTIVE TESTING

- FATHOMETER ECHO SOUNDERS ~ -
ACOUSTICAL MICROSCOPY """""""' ~
INDUSTRIAL ACOUSTICS ~ ACOUSTIC IMAGING ~
MAJIUUUUOOUU
~
~~ BIOMEDICAL DIAGNOSTICS l:l
...
O<:RXICOX:X:XX XO[XX"'X""X"'XXICOX:X:XX""X""XJ ~~~~~n~~RS
iii
..,
1
PASSIVE DETECTION
- FISHFINDING SONARS
)...
....
<::)
$::

""
~-
-
~
10km 1 km 100m 10m 1m 10an 1cm 1mm 100""' IO )IIl 111"'
100km iii
""
3
ACOUSTICAL WAVELENGTH IN WATER
""

Figure 1.1 The scope of acoustical engineering

 Sound Transmission Fundamentals 3

1.2 Sound Speed

Sound speed, conventionally denoted by the letter "c" is, itself, determined
(through the wave equations) by three other physical properties of the
medium, namely its specific beat at constant pressure, y, its density, p,
and its isothermal bulk modulus of elasticity, B. The inter-relationship
between these various quantities is given by the following equation, which
is attributed to Newton

c = (yB/p)ll2
In distilled water at 20"C and at standard atmospheric pressure, the physicist
measures y as 1.004, p as 998 kg m-3 and B as 2.18 x109 N m-2 • We thus
calculate sound speed as 1481 m s-1.

In practice, constraints in measurement accuracy of y, B and p limit the

value of this equation in providing a prediction of sound speed. It is, perhaps,
sufficient to note that all three variables are quantities which depend upon
temperature, T, pressure P and, for real sea-water, chemical composition.
Chemical composition is classically expressed in terms of salinity, S, and
more recently in terms of electrical conductivity, G. Consequently, it has
been argued that sound speed may be expressed, to adequate accuracy, as
some suitable function of temperature, pressure (or depth, Z) and salinity
(or conductivity):

c = f(T,P,S)
Many polynomial approximations have been formulated to yield suitable
expressions to satisfy such a relationship. These polynomials have been
derived from the results of experimental measurements of sound velocity
by curve fitting. The measurements are performed under carefully controlled
laboratory conditions using precision "primary standard" sound velocimeters.
Less accurate, but smaller and more robust "secondary standard" velocimeters
are used for field measurements.

Early measurements on distilled water at atmospheric pressure performed

at the US Bureau of Standards were used to produce the fourth-order
polynomial

c = 1402.736 + T(5.03358 + T(-0.0579506 + T(3.31636E-4 +

T(-1.45262E--6 + 3.0449E-9))))
4 Underwater Acoustic Systems

Later work on distilled water at elevated pressures and on carefully prepared

"standard" sea-water samples led to the formulation of a bewildering array
of sound speed equations. At the present time, the benchmark equation
should be taken as the Lovett [1.1] expression:

c = 1402.394 + T(K1 + S(K2 + K3SP) + K4P2 + T((K5 + K6 S) + ~ T)) +

K8S + P(K9 + P(K10 + P (K11 T + K12S)))

where

K1 = 5.01132 K2 = -1.266383E-2 K3 = 2.062107E-8

K4 = -1.052396E-8 K5 = -5.513036E-2 K6 = 9.543664E-5
~ = 2.221008E-4 K8 = 1.332947 K9 = 1.605336E-2
KID = 2.12448E-7 K11 = 2.183988E-13 K12 = -2.253828E-13

and T is measured in degrees Centigrade, S in parts per thousand (%o or

ppt) and P in decibars. To assist in computer coding this and other key
formulae in this text, set-point test values are provided. Check, forT= 14 ·c.
S = 35%o and P = 100 decibars that c = 1505.100285. In using the above
formula, to obtain highest accuracy, it is necessary to correct the inter-
relation between depth and pressure for geographical latitude e. A useful
"universal" formula for the inter-relation of pressure, P, in decibars and
depth, z, in metres is given [1.2] by

P = 1.0052405(1 + (5.28E-3)sin29)z + (2.36E--6)z2

An alternative formula permits calculation of pressure in kg cm-2 rather than

decibars:

P = 1.04 + 0.102506(1 + 0.00528 sin2 9) z + (2.524E-7)z 2

and check that P = 104.0690158 for a = x/4 and z =1000 m.

Finally, we note that recent developments in oceanographic measurement
practice demand the measurement of salinity via conductivity, G, which
is measured in Siemens per metre (S m-1), this parameter being more
amenable to direct, rapid in-situ observation. A useful but perhaps not
definitive relationship is given by the "Collias" equation

S = -0.505 + 11.15294G + 0.36800676 2 - 0.35412GT- 0.0120291TG2

+ 0.0086GT2 + 0.0000048G2'f3 - 0.00011GTl

and check that S = 33.215408%o for G = 4 S m-1 and T = 14•c

(1 S m-1 = 10 mmho cm-1)
 Sound Transmission Fundamentals 5

Practical estimation of sea-water sound speed may thus be made by measuring

temperature, depth and salinity. For military purposes, which involve primarily
the measurement of sound-speed versus depth profiles, and which are
important in ray-trace sonar prediction programs, only temperature and
depth need usually be measured. Salinity would be taken at the "typical"
value of 35%o. The commonest way of acquiring this information is by means
of a bathythermograph, particularly the "expendable" kind. This latter
equipment, shown in figure 1.2, consists of a bomb-shaped projectile
containing a thermistor connected to a surface electronics package by a
spool of fine wire. The wire plays out as the bomb falls through the water
column and snaps at maximum extension. The bomb rate of fall is predictable,
so that depth can be inferred at any known time after launch. The resistance
of the thermistor measured via the connecting wire, varies with, and thus
measures sea-water temperature.

Figure 1.2 The Expendable Bathythermograph: XBT

For more accurate measurements, such as might be required in the offshore

survey industry, or for definitive measurements in oceanographic acoustics,
either a "CTD" or a field sound velocimeter would be employed. The term
"CTD" is used to describe a nest of sensors, often employed in oceanographic
surveys, which measure conductivity temperature and depth. Confusingly,
the "C" in "CTD" stands not for sound speed but for conductivity for which,
as we have seen, the internationally agreed abbreviation is actually G. Field
measurement accuracy on conductivity is typically ±0.03 S m-1 {±0.3
mmho cm-1), being equivalent to an accuracy of 1% of full scale on a salinity
scale extending to 35%o and thus equivalent to a full-scale conductivity of
3 S m-1 • A field measurement accuracy for temperature is taken to be ±o.1·c
and for pressure 1% of stated depth. The accuracy of estimation of sound
speed by conversion is thus depth dependent. At the surface, and at depths
to a few hundred metres the accuracy on estimation is of the order of ±0.6
6 Underwater Acoustic Systems

m s-1 (±400 parts per million) rising modestly to ±0.8 m s-1 (530 parts per
million) at a depth of 4000 m. The field sound velocimeter [1.3], [1.4]
illustrated in figure 1.3 operates on an acoustic pulse "sing-around" principle.
A high-frequency acoustic pulse is launched into the water sample by a
transmit/receive transducer. It traverses a folded 10 em path, returning to
retrigger a new pulse. The pulse repetition frequency is thus ten times the
speed of sound in metres per second. The folded path is used to minimise
the effect of errors which could be caused by water flow in the vicinity
of the instrument. Typically, a sound velocimeter should be able to measure
sound speed to an accuracy of the order of one part in 105• Sound velocimeters
have the disadvantage of being expensive and extremely difficult to calibrate
and maintain, despite their apparent simplicity.

ill{}
Adjustable reflector

J:f----~~
lnvar space!.
rods

generator

Output repetition frequency

= 10 x speed of sound in
metres per second
Figure 1.3 The sing-around secondary standard or
"field" velocimeter

1.3 The Propagation Equation

Sound speed is the critical parameter which inter-relates transmission

frequency and wavelength. Sound is a travelling wave. This means that,
to an observer at a fixed location, sound pressure fluctuates as a function
of time. However, to an observer with a global view of the pressure field,
the propagating wave is also a function of spatial displacement from the
sound source. For a pulsed sinusoidally-varying sound wave emanating as
a plane wave-front in the horizontal, we write

p(t,x) = A(t + kx/ro)sin(rot + kx)

 Sound Transmission Fundamentals 7

where A(t) is the envelope of the pulse shape, c.o is the radian frequency
and k is the (horizontal) wave number. The following equations apply:

c.o = 2xf; c = D..; T = 1/f; k = c.o/c = 2xf/c

where f is the frequency measured in Hertz (Hz) or "cycles per second".

T is the temporal period and A. is the wavelength. At a fixed location,
for example at x = 0, the signal has a temporal dependence

p(t) = A(t)sin rot

and at a fixed instant of time, for example at t =0, a spatial dependence
p(x) = A{kx/c.o) sin kx
A plausible envelope for a sonar "ping" is the gaussian pulse shape

A(t + kx/c.o) = exp(-0.72 (1tB(t + kx/c.o))2')

where B is the signal bandwidth, determined primarily by the transducer(s)

used to convert acoustic signals to electric signals, or vice versa. Typically
B ""O.lf. The gaussian envelope and the resulting time and spatial functions
corresponding to the sonar "ping" are shown in figure 1.4.
Ping envelope

Pulse "width" lllV8fS81y

-==:::::::::::.__-1---===-
proportional to pulse bandwidth

Ping waveform

Perlod·T-1/1

Figure 1.4 The sonar "ping" waveform

8 Underwater Acoustic Systems

1.4 Logarithmic Power Measurement

Acoustic power levels discernible by the human ear span an extremely wide
range. For example, the intensity of the sound made by the proverbial "drop
of a pin" is about one millionth of a millionth {lQ-12) of the intensity you
might hear if you listen to a jet aircraft taking off. In order to draw this
range within a more handleable compass, a logarithmic measure of power,
known as the decibel is widely used. The decibel, a dimensionless unit,
for which the abbreviation "dB" is used, is defined in terms of the ratio
of a measured power to a reference power, thus

dB ratio = 10log10 (measured power/reference power)

Thus, if measured and reference powers are equal, their ratio is unity and
the dB ratio is zero. If the measured power is ten times the reference power,
then the power ratio is 10 and the dB ratio also 10. The table given below
extends this example

power ratio: 1/100 l/10 1 10 100 1000

dB notation: -20 -10 0 +10 +20 +30

When the decibel unit is used, the scaling of power, to account for example
for loss during transmission, is achieved by addition, rather than by
multiplication.

1.5 The ''New" Reference Unit

Underwater acoustics has always been bedevilled by a plethora of confusing

systems of units. This problem has been exacerbated by the frequent
intrusion of Naval terminology so that, in any one document, horizontal
range may be found measured in kiloyards whilst depth is referred to in
feet, kilofeet or fathoms, speed of sound in feet per second and target speed
in knots. We shall adopt SI units in this text.

Acoustic waves in the sea are pressure waves. Pressure is force per unit
area. In the SI system, force is measured in Newtons, with one Newton being
(about) the force exerted on the palm of your hand by an apple placed
thereon. Pressure would thus be measured in Newtons per square metre and
the unit of pressure is referred to as the Pascal, for which the abbreviation
"Pa" is used. A pressure of 1 Pa is far larger than one would ever expect
to encounter in normal underwater acoustic operations. Consequently,
pressure is typically measured in units of one millionth of a Pascal, which
is referred to as one microPascal, and abbreviated to 1 ~a.
 Sound Transmission Fundamentals 9

By way of providing an illustration to give a "feel" for the practical

significance of such fluctuations. a typical underwater hydrophone, sensing
a sound field with pressure fluctuation of 1 J.l.Pa, could be expected to
generate at its terminals a waveform with an amplitude of 100 picovolts
(1Q-10 volts). Such a hydrophone would be said to exhibit a sensitivity of
-200 dB relative to a standard of 1 volt per JlPa.

The pressure wave discussed above, p(t,x). describes the instantaneous

fluctuation of pressure as a function of time and spatial displacement. As
it happens, the sonar "ping" depicted above is a waveform of a type referred
to as a finite energy waveform. Waveforms which are repetitive. such as
the "ping ...ping ...ping ... " of an echo sounder, or which are continuous in
nature, such as bio-acoustic emanations or sea noise, are referred to as finite
power waveforms. For waveforms in this latter class, it is preferable to think
not in terms of instantaneous pressure fluctuation, but in terms of an
averaged pressure fluctuation. We use the mean square pressure, since this
quantity is proportional to the time averaged power transferred by the wave.
In fact the most useful measurement, derived from the average power
transfer. is the power transferred per unit area normal to the direction of
propagation. This quantity. measured in watts per square metre, is the
acoustic intensity.

The acoustic intensity of any pressure wave is measured in decibels relative

to the intensity of a reference wave. The reference wave is taken to be a
plane wave, of root mean square pressure equal to one microPascal. The
student of acoustics will frequently encounter. in the literature, the terminology
"N dB re 1 JlPa". This means that the intensity of the measured pressure
wave is greater by N decibels than the intensity of the reference pressure
wave. Notice that the terminology misses out, for brevity. an important part
of the definition. It should more correctly read "N dB re (the intensity of
a plane waveform ofrms pressure equal to) 1 JlPa". The part of the definition
in the brackets should always be borne in mind.

The actual acoustic intensity of the reference wave is calculated from the
formula

-2-
1 = p (t)/pc
where the product pc is referred to as the acoustic impedance of the
transport medium. is denoted CJ and is ascribed the unit of the Rayl. Thus,
for the plane wave of rms pressure 1 JlPa. we calculate, knowing that p
= 1000 kg m-3 and c = 1500 m s-1, or CJ = 1.5 MRayl, an intensity of 0.67E
- 18 wm-2 •
10 Underwater Acoustic Systems

By way of example, suppose that an acoustic source is supposed to exhibit

an output intensity of +120 dB re 1 JlPa. We may calculate the power ratio
as

power ratio = 1Q<dB ratio/10)

and in this case, the power ratio is 1()<120110> = 1012• The actual intensity of
the source is thus 0.67E-6 watts per square metre.

1.6 Sound Reflection and Refraction

Sound reflection, that is, specular ("mirror-like") reflection, obeys the

same law as in geometric optics, figure 1.5, with

Sound refraction obeys Snell's law, with

Transmission will always take place from lower to higher acoustic impedance.
For example, sound will always penetrate from air to water, irrespective

Incident
wave Reflected
wave

Boundary
p c
2 2

Refracted
wave

Figure 1.5 Reflection and refraction at the boundary between

acoustically dissimilar media
 Sound Transmission Fundamentals 11

of angle of incidence. Total internal reflection can occur (if the angle of
incidence is inclined sufficiently far from the normal) if transmission is
attempted from a medium of higher to one of lower acoustic impedance.
The reader should recall, from geometric optics, that the critical angle of
incidence, ec, occurs when, as the incident ray swings away from the normal,
the angle 92 made between the emergent ray and its normal increases to
graze along the interface, so that 92 = 90°. This marks the onset of internal
reflection. Then, sin9 2 = 1 and 91 = ec = sin-1 (c/c 2).

1.7 Pressure Reflection and Transmission Coefficients

The acoustic impedances of the materials on either side of a boundary

determine the degree of reflection or transmission across the boundary. Such
properties are important in designing transducers, in determining sea-floor
sediment properties acoustically, in sonar modelling and in assessing target
strength. In this section, the materials on either side of the boundary
are assumed lossless. In section 2.7 an empirical formula which is not
restricted by this presumption is presented to describe, for practical modelling
purposes, loss in acoustic intensity following reflection from the sea-floor.
For a detailed analysis of the derivation of the following results, the reader
is referred to Brekhovskikh [1.5] and to Clay and Medwin [1.6]. Inclusion
of sea-floor attenuation, particularly, in predicting the pressure reflection
and refraction coefficients, is due to Mackenzie [1.7].

The pressure reflection coefficient at a boundary is given in terms of the

incident angle el by
R 12 = P/P 1 = (A- B)/(A + B)

and the pressure transmission coefficient is given by

T 12 = P/P1 = 2A/(A + B)
where

Notice that if c1 and c2 have the same value (but could none the less be
properties of materials of differing density and thus differing acoustic
impedance - a phenomenon observed in, for example, some sea-floor
sediments) - then el and e2 will have the same value and both reflection
and transmission coefficients will exhibit values which will be independent
of angle of incidence. These values will be R12 = (p 2 - P1)/(p 2 + P1) and
T1z = 2p.j(Pz + PJ
12 Underwater Acoustic Systems

1.8 Behaviour at Normal Incidence

If 9 1= 90• then the reflection and transmission coefficients become,

respectively

The fact that T12 = 1 - R12 reflects a condition inherent in the derivation
of the formulae, that the interface itself sustains no excess pressure.

Two extreme conditions are of immediate interest. For an upward-going

soundwave, normally incident on the (inner) sea-surface, <r2 becomes the
acoustic impedance of air and <r1 that of water. We note that <r2 « <rl' so
that R 12 = -1 and T12 = 0. We refer to the sea-surface as being a "pressure-
release" boundary and note that phase-inversion of a reflected wave is to
be expected.

By contrast, a sound-wave, normally incident on, say, a granite sea-wall

or a thick steel plate so that <r2 » <rl' will exhibit R12 = +1 and T12 = 0.
We note that phase inversion on reflection does not now occur but that
reflection is again virtually total, with no transmission into the second
medium from the water.

Next consider the case when 0'2 = 0'1 • This condition might be considered
to represent the circumstance of sound incident upon a "rho-cee" rubber
membrane such as is used to protect some underwater transducers ("rho-
cee" rubber is a rubber designed specifically to have the same acoustic
impedance as water). For this case, R12 = 0 and T12 = +1; transmission is
total and no sound is reflected.

At the sea-floor, sediment acoustic impedance ranges from about 4 MRayl,

for coarse sand, to about 2 MRayl, for fine abyssal clays. The corresponding
range of reflection coefficients lies between +0.45 and +0.14 and the range
of transmission coefficients therefore lies between +0.55 and +0.86.

It is of some interest at this point to consider the power or energy reflected

from or passing across the sediment interface; that is, we wish to determine
an intensity reflection or transmission coefficient. Because the reflected
wave remains within medium 1, the intensity reflection coefficient, J.lr, is
simply the square of R12 and will be given as
 Sound Transmission Fundamentals 13

and, because the sum of reflected and transmitted powers must be unity,
the transmission coefficient will be

It thus follows that, for 0'1 = 0'1 , power flow is total, as we should hope.
For our previous examples, we find power flow into the sediment of between
80% (for coarse sand) and 98% (for fine abyssal silt).

1.9 Reflection and Transmission with Varying Incidence

We note first that, for angles of incidence less than critical, total internal
reflection will not occur and transmission and reflection will both (in
general) be present. In our previous expressions, B will be real and of value

B = cr1(1 - (c'J/1
'c )1 sin1 91
)111·
.1 9 < 9c

Both the reflection and transmission coefficients will thus be real and will
vary in value with changing angle of incidence. No phase shift will occur.

For angles of incidence equal to the critical angle, 91 = 90° so that

cos9 1 = 0 and thus T 11 is zero. This remains true for angles greater than
critical since then total internal reflection occurs and no transmission across
the interface between the two media can take place. At critical incidence,
B becomes zero.

Above critical incidence, B becomes imaginary but may for convenience

be re-written as

so that, in turn, the reflection coefficient may be expressed as the ratio of

complex conjugates

R11 = (A - jB')/{A + jB')

It then follows that the magnitude of the reflectivity of the surface is

=
invariant with angle of incidence: IR111 1; 91 > 9c"

The reflected wave exhibits, however, a phase lag determined by the ratio
of the imaginary to the real part of the numerator or denominator of the
reflection coefficient
14 Underwater Acoustic Systems

cp = -2tan-1{(sin291 - (c/c2)2)112/{{cr.jcr1) cos91)}; 91 > 9e

= 0; 91 < 9e
The reflection coefficient may then be written in polar form, as

R 12 = exp(cp); 91 > 9e
= (A - B)/(A + B); 91 < 9e

If we return to the problem of reflection from the sea-surface from within,

we note that cr1 » cr2 and that p1 » p2. Our expression for cp thus reduces
to tan( cp/2) :::::>~.or cp :::::> 1so·. That is, irrespective of the angle of incidence,
a surface-reflected wave will always suffer phase inversion.

1.10 Reflection and Transmission: the "Fast" Bottom

Many realistic sea-bed conditions are represented by the so-called "fast"

bottom, where both sound speed and density significantly exceed those of
the overlying water. Figure 1.6 illustrates the amplitude and phase functions
for the pressure transmission and reflection coefficients for such a sea-bed.

Here, the bottom is presumed to have a high (for a sedimentary deposit,

at least) sound speed of 1800 m s-1. The higher density values depicted
correspond loosely to the coarser sands. As can be seen, the critical angle
does not change, being dependent only on the ratio of sound speeds in the
two media. For angles greater than critical, total reflection occurs. For
angles less than critical, sound passes into the sediment. The greatest
transmission occurs at normal incidence, when 91 is zero. The reflected and
transmitted waves exhibit zero phase shift for angles of incidence which
are less than critical, and a phase shift rising to 1so· at grazing incidence.

1.0 0.0 180

0.8 0.2 150f.\

12().,.;D
1\)
I R,~ o.6 0.4

0.4 0.6

0.8
IT121
:f Ill
0.2 30 ~
0.0 1.0 0
0 20 40 60 80 0 20 40 60 80
Incident angle Incident angle

Figure 1.6 Reflection and transmission pressure coefficients versus

incident angle 8r· the "fast" bottom
 Sound Transmission Fundamentals 15

1.11 Reflection and Transmission: the "Slow" Bottom

Under some circumstances, even though the sediment density will exceed
the density of the overlying sea-water, the sound speed may yet be less.
This is the condition referred to as a "slow" bottom. Now it is possible for
"intromission" to occur. Figure 1.7 illustrates this phenomenon. We see that,
at an angle

perfect transmission into the sediment may occur, the reflection coefficient
being zero.

1.0 0.0 180

0.8 0.2 1sot\

P2= 1.4 120...:0
N
0.4
I ~21o.6 IT121 90 "tJ
0.4 0.6 P2•1.8
i
60 Ill
0.8 30 "'
0.2

1.0
P2=1.0, 1.4, 1.8
I
I
I
1J
0
~

20 40 60 80 0 20 40 60 80
Incident angle lnadent angle

Figure 1.7 Reflection and transmission pressure coefficients versus

incident angle 91 : The "slow" bottom

References
[1.1] J.R. Lovett, Merged Sea-Water Sound Speed Equations, J. Acowst. Soc. Am., Vol. 63,
No. 6, 1978, pp. 1713-1718

[1.2] C.C. Leroy, Development of Simple Equations for Accurate and More Realistic Calculation
of the Speed of Sound in Sea Water, J. Acowst. Soc. Am., Vol. 46, No. 1, 1969, pp. 216-
226

[1.3] R.L. Williamson, G. Hodges and E. Eady, A New Sound Velocity Meter, The Radio
and Electronic Engineer, June 1967, pp. 387-393

[1.4] K.V. Mackenzie, A Decade of Experience with Velocimeters, J. Acowst. Soc. Am., Vol.
50, No. 5, Pt 2, pp. 1321-1333

[1.5] L.N. Brekhovskikh, Waves in Layered Media, Academic Press, New York, 1960

[1.6] C.S. Clay and H. Medwin, Acoustical Oceanography, Wiley, New York, 1977

[1.7] K.V. Mackenzie, Bottom Reverberation for 530 and 1030-cps Sound in Deep Water,
J. Acowst. Soc. Am., Vol. 33, No. 11, 1961, pp.1498-1504
2 The Sonar Equations

2.1 Introduction

Underwater acoustic systems inevitably involve the detection of signals.

The fundamental criterion which determines the effectiveness of all detection
processes has to do with determining the extent to which the received signal
exceeds, or is swamped by, such corrupting influences as may exist. In
electromagnetic detection equipments- radio and radar systems -antenna-
borne signals are of such small size that the significant corrupting influence
may well be the similarly small, random, gaussian noise waveforms deriving
from charge transport processes or molecular agitation in the antenna and
front-end receiving circuits.

Whilst it is true that sonar systems also can be subject to a predominating

gaussian noise corruption, the fact that the sea is bounded by excellent
reflectors - its surface and floor - means that reverberation (multiple
delayed echoes of the transmitted signal) may in some instances present
a far greater problem. The sonar equations provide a variety of statements
of sonar system efficacy. They are established by combining information
pertaining to source power output and directivity, transmission loss and
noise or reverberation corruption. The sonar equations produce - let it be
said- guidelines, rather than exact results. The sea is a complex transmission
medium and the designer, in applying the sonar equations, would do well
to err in the direction of safety, in setting signal levels, and in the direction
of pessimism, in predicting loss or corruption level.

2.2 Source Intensity Calculation

The source intensity is a measure, in dB re 1 IJ.Pa, of the power flux

[W m-2] delivered into the water by a source and is always referred to
standard range from a presumed acoustic centre of the source. Standard
range is 1 metre. The acoustic centre is a convenient fiction which provides
a starting location for loss calculations. At 1 m, the acoustic centre is
surrounded by a spherical envelope of area 4xr2 = 12.6 m2 • If the source
16
 The Sonar Equations 17

power output is P watts, and the source radiates equally strongly in all
directions, then the source intensity at standard range is P/12.6 W m-2 • The
relative acoustic intensity, denoted SL and measured in dB re 1 ~a. is
calculated as

SL = 10 log10((P/12.6)/reference wave intensity)

Recalling the result for reference wave intensity from section 1.5, we find
that

SL = 10 log 10 ((P/12.6)/0.67E-18)

= 10 log10 P + 167

Notice that source intensity can usually only be measured remotely from
the source and in deep water, to avoid reverberation problems. Scaling of
the remote measurement back to 1 m reference range must be accomplished
by taking into account spreading and loss laws within the water.

2.3 Source Directivity

Our calculation of source intensity assumes omnidirectional spreading of

sound from its acoustic centre. Often, like the light beam produced by the
focusing mirror of an electric torch, the sound does not spread
omnidirectionally. If the beam sub tends cjl steradians of solid angle, then
the power concentration, for given source power output, in the beam and
relative to the acoustic intensity which would have been generated, had
spreading indeed been omnidirectional, is scaled by a factor

41t/cjl

This scaling factor is the most elementary description of a source parameter

known as the directivity index. It is customary to express directivity index
in decibels, since it represents an intensity ratio. We write

The beam solid angle depends upon the dimensions of the source, relative
to a wavelength, at the frequency of interest. It, and consequently directivity
index, may thus be quite difficult to envisage, let alone assess for sources,
such as ships, which emit broadband radiation and are of complex mechanical
structure and surface geometry. We shall address the problem of estimating
18 Underwater Acoustic Systems

beam angle for transducers in chapter 7 and arrays of transducers in chapter

8. For some sonar projectors, which are usually narrowband and of well
defined acoustic aperture, beam angle may be quite easily defined. For
example, for the circular piston transducer of diameter D, the directivity
index is calculated as

so that increasing the number of wavelengths encompassed by the diameter

increases the directionality and thus improves the directivity index.

Finally, calculation of source intensity, if modified to reflect the concentrating

effect of limited solid beam angle, must be amended to

SL = 10log10 P + 167 + Dl

2.4 Transmission Loss

It is frequently of value to be able to assess the accumulated decrease in

acoustic intensity, as a pressure wave propagates outwards from a source.
The parameter which describes the decrease of intensity with distance is
known as the transmission loss, and is denoted TL. The transmission loss
is the sum of a spreading loss and an attenuation, the latter caused by
the unavoidable frictional conversion of sound into heat during propagation.

The most fundamental spreading-loss law is that which describes spherical

or free-field spreading. Free-field conditions will be approximated only
when all reflecting boundaries are so far from the source and receiver that
no channelling of acoustic energy can occur. At low frequencies this will
typically be the case only in deep water. At very high frequencies, because
attenuation per unit distance rises with increasing frequency the effect may
also be evident in shallow water. The basic loss law for spherical spreading,
the "inverse square law", giving the intensity I(R) at range R, relative to
intensity at 1 m standard reference range, is

I(R) = R-2
or, expressed in dB

I(R) = 20 log10 R
If reflections from sea-surface and sea-floor result in the sea behaving like
an acoustic waveguide, free-field conditions will not pertain. Propagation
may then take place with a cylindrical spreading law, for which
 The Sonar Equations 19

I(R) = R-1
or, in dB

I(R) = -10 log 10 R

Since boundary reflections vary with sea-state and bottom material properties,
and since sound penetration into the sea-bed at low frequencies is quite
good, the cylindrical law tends to under-estimate loss. A "practical" law,
intermediate between the spherical and cylindrical laws, is thus often
invoked for "first-cut" calculations in sonar system design. The practical
law is defined by the expression

I{R) = -15 log10 R

Attenuation in the sea is caused mainly by viscous friction and at frequencies
in excess of (about) 1 MHz, loss is as measured for distilled water [2.1].

102

101

10°

_,
]"
itl
10 ..
::;;M~~~;~~~;;;·;~I~ate
..
§
OJ
::J
10
-2 OVerall attenuation
exh 1bited by sea-water
at 35 ppt sahmty and at
/// relaxanon

~ -3 14'C /:,..._ D1stilled water

< 10 ,/ attenuation

/ ,/
..
... ...
-4 ·······;/·······:;;'················--······(k;~~--~~;d·;~j~~~t~~n
10

.. ..
-5 ,/'' ,,/
10
/,' //

10 6 ~--~~--~...-----.------.------.------r
100Hz 1kHz 10kHz 100kHz 1 MHz 10MHz

Transmission frequency
Figure 2.1 Attenuation in distilled water and in sea-water,
showing the increase in attenuation brought about by molecular
resonance effects
20 Underwater Acoustic Systems

As frequency is decreased, molecular resonance effects conspire to worsen

the attenuation figure for sea-water, by comparison with the distilled water
value. At frequencies below (about) 500kHz, the presence of magnesium
sulphate, in solution in sea-water, begins to intrude an excess attenuation
over the distilled water loss, ultimately increasing the attenuation uniformly,
by a factor of about eighteen for frequencies below (about) 70 kHz [2.2].
At frequencies below (about) 700 Hz, boric acid - despite its small
concentration in sea-water - adds in a further uniform sixteeen-fold
increase in loss [2.3].

The graphs presented in figure 2.1 illustrate the variability of sea-water

attenuation factor, a., measured in dB m-1 • Final calculation of transmission
loss, TL, is effected by applying the formula

TL = k log r + a. r

where k is chosen as 20 for free-field, 10 for cylindrical or 15 for "practical"

spreading. For purposes of computer calculation, with fin kHz, the following
formulae apply to the calculation of a., to an accuracy acceptable for sonar
calculations

(Freshwater attenuation, [2.2])

(MgS04 relaxation, [2.3])

(Boric acid relaxation, [2.4])

where, below, S is salinity (o/oo) and T is temperature (° C). Also

a = 2.1 X I0-10 (T - 38)2 + 1.3 X I0-7

b = 2S X I0-5

f0 = 50(T+1)

C = 1.2 X 1Q-4

fl = 10(T-4)/100

At 35o/oo and 14° C check values are a = 2.5 x I0-7 , b = 7.0 x I0-4,
f0 = 750 kHz and f1 = 1.26 kHz.
 The Sonar Equations 21

Planktonic marine life, suspended material and entrained gas bubbles can
all lead to "anomalously" higher attenuation than might be predicted by
the equations discussed above. These various entities contribute to attenuation,
if present in sufficiently great densities, by a combination of scattering and
resonance effects. The latter phenomenon is most likely to be evident in
bubbly water, or where plankton which may contain or which may respire
gas bubbles is encountered. All such effects are difficult to quantify. In
general if field experiments suggest a level of attenuation greater than can
be ascribed by invoking the standard transmission loss equations, experimental
error should be the first assumption, not anomalous attenuation. An excellent
review of the subject of attenuation of sound in sea-water is provided in
reference [2.4].

Attenuation in marine sediments [2.5] is another area of potential importance

in some sonar situations. The attenuation of sound in marine sediments has
been found to vary approximately exponentially with both distance and
frequency. It is specified by an attenuation coefficient, a.5 , measured in
dB m-1 which is, itself, a function of frequency given by

a.. = PfY

Experimental measurements indicate that y is close to unity for marine

sediments. The total attenuation, in decibels, for sound which has travelled
a distance r within a single sediment layer and which has a transmission
frequency f, is thus

The attenuation coefficient a.. is, itself, a function of bottom porosity, n,

which is defined as

n = VjV
where V,. is the volume of water in a sample of sediment and V is the volume
of the sediment.

Alternatively, moisture content, m, may be quoted and is taken to be the

weight of water in a sample, divided by the weight of solids, so that

n = (P/P.)(m/(m + 1))

where P. is the sediment density and P.. is the density of water.

22 Underwater Acoustic Systems

Table 2.1
Marine sediment properties

n P. m c 0' p
(kg m-3 ) (m s-1) (MRayl)

Coarse sand 0.4 2000 0.25 1800 3.6 0.3-0.6

Fine sand 0.5 1900 0.35 1700 3.2 0.4-0.7
Silt 0.6 1800 0.50 1600 2.9 0.1-0.5
Clay 0.8 1400 1.33 1600 2.2 0.05-0.2

Marine sediments may become heavily saturated with hydrocarbon gases,

as a result of discharges from deeper gas and petroleum reservoirs, or as
a result of biological decomposition. Saturation with gas will cause the
sediment to become highly reflective to sound waves. Normal attenuation
levels will then be far exceeded.

2.5 Target Strength

Target strength, TS, is the decibel measure of intensity returned from a

target, referred to 1 m standard range from the notional "acoustic centre"
of the target, relative to the incident intensity. Suppose for example, that
an active sonar is used to gauge the target strength of a midwater target,
such as a "trials" submarine, located vertically below the source vessel,
figure 2.2. Imagine the depth of the submarine to be 120 m, the sonar
projector electrical power input to be 500 W and the echo return intensity
to be 143 dB re 1 JJ.Pa. The calculation proceeds as follows. The source
level is evaluated, as we have seen in section 2.2, as

SL = lOlog P + 167 = 194 dB re 1 JlPa

To calculate transmission loss, we assume free-field conditions in deep

water and hence spherical spreading, so that

TL = 20log r = 20log 120 = 42 dB

The intensity arriving at the submarine is

SL - TL = 194 - 42 :: 152 dB re 1 JJ.Pa

 The Sonar Equations 23

I
SL = 194dB Intensity received
= 143 dB

Transmit path TL Return path TL

=42dB =42dB

Intensity at target Re-radiated intensity

= 152 dB = 152 + TS

( ______ ~

Figure 2.2 An example of the use of the sonar equations in assessing

target strength

We presume the submarine to have a target strength TS. The signal strength
at one metre notional range is then 152 + TS. We insert the return transmission
loss, to find the intensity at the surface vessel. This we have measured at
143 dB re 1 J.I.Pa. Simplifying and solving for target strength, we find that

TS = 33 dB re 1 J.I.Pa

The example quoted above gives an indication of the way in which target
strength may be measured. For military targets, such as submarines, mines
and torpedoes, such measurement is common practice. Optical assessment
of target highlights used to be carried out by photographing a model of a
particular type of submarine, painted matt black. Only particularly good
reflective planes "stood out" against a matt black background when the
model was illuminated with bright light. A more modern approach utilises
images of the submarine created by computer graphics techniques, to
achieve the same end. A theoretical assessment of target strength can be
made for some of the simpler solid forms, some of the commoner of which
are listed in table 2.2.
24 Underwater Acoustic Systems

Table 2.2
Target strength of simple forms

a= radius; range = r; L = length; k = wavenumber

Solid sphere 10 log(a 2/4) r>>a; ka >> 1 (sphere "large")

Long cylinder 10 log(ar/2) e.g. oil-pipeline
Cylinder 10 log(aU/2) r>>a; ka>> 1 e.g. side view of torpedo

The target strength of biological targets can also be assessed by measurement

or by crude approximation. The problem is complicated by the fact that
many marine animals which are of scientific or commercial interest exhibit
shoaling behaviour. Social behaviour within the shoal may invalidate statistical
preconceptions regarding member distribution in space and may thus cause
further confusion.

The acoustic impedance of marine animal flesh is about 1 MRayl. Since

water has an acoustic impedance of 1.5 MRayl, we may assess the absolute
value of reflection coefficient at normal incidence (using our earlier formula,
section 1.7, with 9 1 = 0) as

The attenuation of sound in animal tissue is high, by comparison with

attenuation in water, being of the order of decibels per centimetre, rather
than per metre. We see that marine organisms are thus only moderate
reflectors of sound.

Some planktonic organisms, and all teleost fish, have one characteristic
which can lead to unusually high echo returns, however. This feature is the
presence of a gas "bubble" within the animal (the "swim-bladder" in the
case offish). Gas bubbles, unlike rigid reflecting spheres, exhibit high target
strength at frequencies corresponding to their mechanical "resonance"
frequency. Because shoaling organisms do not return a "coherent" reflection
(as might the sea-bed, were it suitably smooth) the shoal acoustic characteristics
are better described in terms of scattering rather than reflection coefficients.

2.6 Reflection Intensity Loss Coefficient

When sound reflects in a specular manner (that is, mirror-like, without

scattering) from a smooth surface, the intensity reflection coefficient may
be defined as
 The Sonar Equations 25

where 13 and 11 are the intensities following and preceding reflection, as

in our previous figure 1.5. We have that

So that the intensity _ _ loss coefficient in
decibels is Jlr = 10log(p/p1)
2 2
=
20log(IR12 0

where R12 is the pressure reflection coefficient defined in section 1.7.

2.7 Sea-floor Loss

Those parts of the sea-floor which are sedimentary deposits are often
capable of being regarded as a sensibly plane, smooth, reflector of sound.
Acoustic impedance is then dominant in determining acoustic properties.
Combining the above results with those presented in section 1.9 we establish
the Rayleigh bottom-loss reflection model for reflection at the interface
between media themselves, presumed lossless.

Of course, marine sediments must be, to some extent, lossy because the
passage of pressure waves through them causes particle motion and, through
friction, the dissipation of acoustic energy. To take account of the fact that
sediments are lossy, we have the NUC (Naval Underwater Center) model
[2.6], which is empirical and specified in terms of bottom porosity, 0 < n
< 1, which gives bottom-loss directly in decibels

-(17.5n-1.025)f1!3[tanh((6.55 - 0.07249 1)n)<UJn> +

(0.08-0.296n)(l - 0.01179) 2]

Sample curves are shown in figure 2.3. Here f (in bold) is the transmission
frequency in kHz, 9 1 is the angle of incidence measured in degrees, relative
to the normal, as before, and the bottom porosity is as defined in section
2.4 above.

A useful and adequately accurate approximation for the tanh function in

the NUC model is

tanh~ = ~(1.09 - ~(0.395 - 0.0472~)) for ~ < 3

= 0.995 for ~ > 3
26 Underwater Acoustic Systems

Angle of Incidence - degrees

---...
0 10 20 30 40 50 60 70 80 90
0
-2 ~ t-..-
\."'-' ~" '........._, n=0.2
!8 ""- .~

""\\. '" """- "

' -4
.i -6
\. ~" .~ n=0.3
.Q
= \\ ~ n=0.4

"
-8
8c: -10
' -"""'
~

,g -12 \ ....__
g n=0.9
.;::: -14 .......
£ -16
-18
-20
(a)

Frequency - kHz
0.01 0.1 1.0 10.0
0
-2

........... :::::
- r- ~--~
r-
10'

20')>
~ ~
CD"
~
a
i"' [
s
i
r--..1\
1\
30'
90'

(b)

Figure 2.3 The NUC bottom loss model: (a) versus angle of incidence
and bottom porosity at a transmission frequency of 1 kHz and (b)
versus frequency and angle of incidence at a bottom porosity of 0.5

2.8 Sea-surface Loss

The sea-surface acts both as a reflector and as a scatterer. When calm,

reflection with no scattering occurs. Then, as we have already seen in section
1.8, because the impedance of air is vanishingly small, by comparison with
that of water, 0'2 = 0 so that A= 0 and R12 = -1. Also T12 = 0 for all angles
of incidence. Total internal reflection therefore always occurs.
 The Sonar Equations 27

When the sea-surface is rough, however, scattering as well as specular

reflection takes place and any of several empirical models may be used to
predict reflectivity. The criterion governing the action of the sea-surface
as a reflector is thus its surface roughness, which is determined by the
Rayleigh parameter

Q = kcrcos9

where k is the sound wave-number (k = ro/c), cr (written in bold to distinguish

it from specific acoustic impedance) is the root-mean-square crest-to-trough
wave-height (the standard deviation ofthe wave-height distribution function)
and 9 the angle of incidence, measured to the normal to the sea-surface.
The surface pressure reflection coefficient is given as R = -exp(-Q) and
this result is considered to apply when Q << 1. If Q >> 1, then the surface
acts primarily as a scatterer, dissipating the incident radiation in a non-
coherent manner.

The following equation inter-relates wave height cr and windspeed, w, for

a fully arisen sea
cr = 0.005wz.s

The Beckmann-Spizzichino surface reflection loss model [2. 7] permits

calculation of the reduction of acoustic intensity following reflection,
internally within the ocean, from the sea-surface. The surface reflection
coefficient is a function of angle of incidence to the horizontal, 9 (herein
measured in degrees), windspeed, w (knots) and frequency, f (kHz).

The geometry of the problem is defined as follows. Sound of acoustic

intensity 11 is incident upon the sea surface from below, at an angle 0 to
the normal to the surface. That sound is in part specularly reflected, with
intensity 12 , and in part scattered and absorbed. The surface loss coefficient
is defined, in decibels, to be the logarithmic ratio

A convenient algorithm for calculating the Beckmann-Spizzichino loss

coefficient in decibels is presented here as

where f1 = --/10f2 and f2 = 378w-2• "Spot check" parameter and loss values
are w = 20, 9 = 60, f = 7, loss= -11.518.
28 Underwater Acoustic Systems

Knee frequency I 1
proportional to windspeed, w

+
r--------------------~----~·---·
log!

-r-----~--~-~--~.- ----------: -----t- ---

'•, I
'•.... I
1 Loss proportional to

····..
••

10 dB 1 windspeed, w, and the square

l- -
of the angle of incidence

'•· ........ I

- - - - - - - - - - - - - - - - - - - -·.-···.....- - - - - - - -
'.1

Figure 2.4 The Beckmann-Spizzichino loss coefficient

Presenting the loss algorithm in this form has the advantage that the
interaction of the various parameters governing loss may be easily visualised.
On log-frequency,log-attentuation (decibel) scales, figure 2.4, the "flattened
Z" of the loss curve is fixed in shape for all conditions of parameters. It
will slide leftwards with increasing windspeed, indicating increasing
attentuation at yet lower frequencies as the sea-surface becomes more
agitated. Under this same condition, the entire curve will also shift downwards,
reflecting a gross overall increase in attenuation. Finally, the curve will
exhibit a gross downwards shift for increasing angle of incidence, again
as might be expected from a heuristic argument.

2.9 Noise

We shall consider the extensive subject of noise in the sea more fully in
chapter 6. For the purpose of illustrating the necessary manipulations, in
the context of the sonar equations we note two principal classifications of
acoustic noise which may corrupt a received signal: ambient noise, generated
by a variety of natural and man-made mechanisms, and self-noise, produced
by hull friction, the wake and bow-wave, and machinery on board a vessel
carrying a listening sonar. For the moment we concentrate upon the more
general principles of noise analysis and commence by observing that noise
must be specified as a spectral density.
 The Sonar Equations 29

Output
cc lo

------------ -----------
·----------- ---------- Bandpass Squaring Averaging
analysing circuit filter
·-----------
····-···-··· ..........
---··---··. filter
·----------- -----------
·----------- ·----------
-------·---- -----------
----------
···--------- Bandwidth
·---------- BHz
---------- ----------
---------
...._.__ Receive hydrophone sensing a noise intensity 10
within the frequency band B at centre frequency f
·-----------------------
-----------------------
·-----------------------
Figure 2.5 Analysis equipment for measurement of noise spectral
density

In order to measure a noise spectrum, we utilise some form of analyser such

as is shown in figure 2.5. The tunable measuring filter will exhibit a
bandwidth of B Hz at a frequency, f, and the entire system will be so
calibrated as to yield a measure of received intensity within this bandwidth,
again in decibels relative to our reference unit. To plot the spectrum, we
plot dB re 1 IJ.Pa per Hz of measurement filter bandwidth. That is, if the
analyser output at frequency f is I0 dB re 1 jlPa, then we plot N(t) =
Io- 10 IogB dB re 1 jlPa per Hz. To utilise such a spectral plot, we need
to know the useful bandwidth - W, say - of our sonar, as well as its
transmission frequency. The calculation may then be reversed to yield a
corrupting noise level

NL = N(t) + 10logW

2.10 Reverberation

As with noise, we cover this topic more fully in chapter 6. We note however
that reverberation, like noise, is a corrupting influence which may mask
sonar returns. We note also that, unlike noise, reverberation is caused by
the operation of the sonar, being the result of the reflection of transmitted
signals back towards the receiver by adjacent scatterers or surfaces. Whereas
increasing signal power will assist in improving signal detectability against
noise, this will not be the case with reverberation. In this case, increasing
30 Underwater Acoustic Systems

the transmitted power level will increase the received reverberation level
in like proportion. In stating that the cause of reverberation is the return
of transmitted signals by adjacent reflectors, we recognise two principal
causes of such corruption: surface reverberation, wherein the sound is
reflected or scattered back to the receive transducer by the sea-surface or
sea-floor, and volume reverberation, which is caused by reflection from
particulate matter, marine organisms, bubbles and so on, in suspension in
the path of the transmitted beam.

We may anticipate dominance of one or other of these two conditions in

the vast majority of operational situations, and note that the reverberation
level, RL, will be a function of source level, pulse duration and range to
the scattering location.

2.11 Calculating the Signal Excess

The ultimate objective in applying the sonar equations is to determine the

level of detectability of received signals. Many sonars are monostatic: the
transmit and receive transducers are at essentially the same location. Bistatic
sonars would utilise transmit and receive transducers at widely differing
locations. For echo-ranging purposes such systems are uncommon, although
the bistatic geometry is inherent in communication, navigation and positioning
systems. For a monostatic, echo-ranging system, we should seek to determine
a signal excess as

SL + 201 - 2TL + TS - NL

If the sonar is reverberation-limited, rather than noise-limited, replace NL

by RL. As with the distinction between the two possible causes of reverberation,
it is usually the case that either noise or reverberation will present the
dominating corrupting influence.

For a bistatic sonar, such as an acoustic communication link, we need only

consider the single-way transmission loss, and target strength is, of course,
irrelevant. We evaluate the signal excess (essentially the signal to noise
ratio) as

SL + DIt + DIr - TL - NL

where Dlt and DI. are the transmitter and receiver transducer (or array)
directivity indexes, respectively. Most often, for reasonably efficient
communication, the signal excess would need to be of the order of 10 dB.
For echo-ranging, and some positioning systems, a much worse signal
 The Sonar Equations 31

excess can be tolerated, either because signal repetition can be used to

integrate the signal out of the noise or because the display format itself may
offer substantial visual integration, as with the XYt plots common in sub-
bottom profiling and sidescan sonar work.

References
[2.1] L. Hall, The Origin of Ultrasonic Absorption in Water, Phys. Rev., Vol. 73, 1948, p.
775

[2.2] L.N. Liebermann, Origin of Sound Absorption in Water and in Sea Water, J. Acoust.
Soc. Am., Vol. 20, 1946, p. 868

[2.3] E. Yeager et al., Origin ofthe Low Frequency Sound Absorption in Sea Water, J. Acoust.
Soc. Am., Vol. 53, 1973, p. 1705

[2.4] R. Urick, Sound Propagation in the Sea, Peninsula Publishing, Los Altos, Calif., 1982,
pp.S.l-5.17

[2.5] E.L. Hamilton, The Elastic Propenies of Marine Sediments, J. Geophys. Res., Vol. 76,
1971, pp. 579-604

[2.6] H.R. Hall and W.H. Watson, An Empirical Bottom Reflection Expression for Use in
Sonar Range Prediction, NUC Technical Note 10, July 1967

[2.7] R. Coates, An Empirical Formula for Computing the Beckmann-Spizzichino Surface

Reflection Loss Coefficient, IEEE Trans. Ferroelectrics Frequency Control, Vol. 35, No.
4, July 1988, pp.522-3
3 Characteristics and Analysis of Sonar
Waveforms

3.1 Introduction

Sonar emissions and the noises which corrupt them are pressure waves
travelling in a four-dimensional space-time continuum. In the past, it has
often been adequate to restrict consideration of such processes only to the
time-domain signals emerging from the outputs of each of the hydrophones
used for signal detection. Increasing computing power, together with a rapid
evolution in appreciation of the mathematics of signal analysis and its
application to particular physical problems must now lead the underwater
acoustician towards a more all-embracing comprehension of the spatia-
temporal nature of the processes he is called upon to handle.

In this chapter, it is not intended to cover each possible method of

characterisation or analysis in extreme detail. Copious references are provided
to allow the reader to delve deeper into any particular method which may
commend itself as being particularly appropriate for a given task. Rather,
the objective is to provide a chart of a complex and often confusing territory
which will, hopefully, provide some insights into method and applicability
for the wide range of interesting and powerful analysis techniques and
equipments which have become readily available to the marine acoustician.

Although popularly described as "the silent world", the sea is, as we might
by now suspect, an extremely noisy environment. Sound propagates well,
at sonic frequencies, and derives from sources which are many and varied.
Thus far, our characterisation of the sound has been in terms of source level
alone. That is, we have been interested only in average power output.
Common sense would suggest that we should seek a deeper insight into the
temporal, spectral, spatial and directional characteristics of the man-made
signals received by sonar equipments and also of the marine sounds which
may corrupt or mask them.

Temporal behaviour has to do with the way in which source level varies,
as a function of time, in the short term. We might expect that some types
32
 Characteristics and Analysis of Sonar Waveforms 33

of sound source, for example noise caused by distant storms, would maintain
some constancy of power level over quite long periods (longer at least than
our inspection period). On the other hand, "vocalisation" by marine animals
would exhibit relatively rapid changes in power level. Sound sources of
the first kind would be referred to as stationary, because their average power
level is constant on a time-scale which is relatively long by comparison
with the "period" of their fastest amplitude fluctuations. Those in the second
category are non-stationary and would be expected to exhibit relatively
fast fluctuations of short-term averaged power.

Both sorts of waveform, as exemplified above, would be further classified

as finite power processes. This allows us to distinguish them from burst-
like emanations, such as explosive detonations which, although admittedly
non-stationary, differ from the continuous, if fluctuating, sounds produced
by marine animals in that they are oflimited duration. They are finite energy
processes. It should be stressed that the boundaries between many of these
classifications are blurred when an attempt is made to use them to characterise
perceived marine sounds. For example, the crackle produced by myriad
snapping shrimps is a sound source produced by marine animals, and is
almost always best described as a stationary, finite power process.

Spectral behaviour is concerned with the distribution in frequency of the

power or energy content of a waveform. For stationary, finite power processes,
we examine a long-term averaged power spectral density, which plots
power per Hertz of measuring system bandwidth versus frequency. For non-
stationary, finite energy processes, we attempt to establish an energy
spectral density. Finally, for non-stationary, finite power processes, we
have recourse to an energy time-frequency plot. Clearly these various
forms of representation require further explanation.

We shall also consider the temporal correlation attributes of signals. That

is, any two signals may share some similarity of waveform as, for example,
when a signal is subject to transport delay (a common enough occurrence
with sonar waveforms). The source and received signals may well not be
exactly the same because, given sufficient transport delay, noise will have
been added, and attenuation and other corrupting influences will have been
at work. The temporal correlation function, the cross-correlation, will
indicate both the extent of the delay and the degree of comparability between
the two waveforms. Indeed, the method also works with a single waveform,
to unravel, for example, contained echoes. The correlation function is then
referred to as an auto-correlation.

The estimation of a temporal correlation function may be described as a

transformation from the time-domain into the delay-domain. It is not the
34 Underwater Acoustic Systems

only such transformation. As we shall see, evaluation of the correlation

function of a signal can be effected by (amongst other techniques) Fourier
Transformation of the power or energy spectral density (as appropriate) for
that signal. If, instead, we Fourier Transform the logarithm of the power
spectrum, we obtain a new delay domain characteristic, referred to as the
cepstrum. The term cepstrum is simply an anagram ofthe word "spectrum".
The word has, of itself, no deeper significance than that.

The reader will probably be aware that, in evaluating a power spectral

density or an energy spectral density, phase information that would be
present in, for example, the Fourier Transform (the voltage spectral density)
will be lost. However if, in evaluating the logarithm of the power spectrum,
phase information from the calculation of the voltage spectral density is
retained, and used to establish a complex log power spectrum, then
transformation into the delay domain yields a complex cepstrum, which may
be processed to eliminate or suppress chosen bands of delay wherein, for
example, reverberant information might lie. By reversing the procedure and
transforming back to the time domain it is possible to establish signal
waveforms free from contamination by echoes. This process is referred to
as homomorphic deconvolution.

In establishing a temporal correlation function we compare two waveforms

subject to a relative time displacement between them. Alternatively, we may
assume both processes to be recorded with, as it were, zero time slippage,
but with a variable spatial displacement. This allows us to compute a spatial
correlation function. The spatial correlation function is related by a transform
identity, to the angular distribution of acoustic intensity of the sound
field.

3.2 Swept Frequency (Heterodyne) Spectrum Analysers

The most basic of the commonly available spectrum analysers operates by

shifting the baseband spectrum of the waveform being analysed to a higher
frequency, by mixing with a local oscillator, the output of which is swept
in frequency. The baseband spectrum, thus translated, is slid across a
narrow-band bandpass analysing filter. The filter is followed by a precision
envelope detector to yield an oscilloscope display Y signal. The X signal
is provided by the ramp waveform inducing the frequency sweep. The
analysis procedure, illustrated in figure 3.1, is analogous to the
superheterodyne principle used in radio receivers, to obtain high selectivity
following radio-frequency tuning which is a coarse selection process with
poor rejection of out-of-band noise and interference. Superheterodyne
detection involves using a multiple stage, intermediate frequency amplifier,
 Characteristics and Analysis of Sonar Waveforms 35

with each stage a double-tuned. critically-coupled radio-frequency

transformer. Attempting to sweep the centre-frequency tuning on such an
amplifier would be quite difficult. Consequently. for practical swept-frequency
analysers. sweeping the centre-frequency of the analysing filter would only
be contemplated in relatively low-frequency (10 kHz or less) specialised
applications. Much useful advice on the practical design of swept spectrum
analysers. including selection of filter characteristics. sweep rate and averaging
times for both deterministic and random signals is to be found in reference
[3.1].
Input Gain-Set Precision Output Buffer

[0--
Amplifier R.F. Mixer I.F. Amplifier Rectifier

[> ~ ~

Ramp Voltage Controlled

Generator Oscillator

Figure 3.1 The heterodyned frequency sweeping spectrum analyser

AI

INPUT

BAND SELECT AVERAGE

ANALYSING
FILTER

P(l)
AREA • AVERAGE POWER AT
[~ OUTPUT OF ANALYSING
FILTER

FREQUENCY I [Hz]

Figure 3.2 The filter-bank analyser and the interpretation of analyser

output power spectrum
36 Underwater Acoustic Systems

3.3 Filter-bank Spectrum Analysers

A different approach to spectrum analyser design is to be found in the filter-

bank analyser, which is illustrated in figure 3.2. Here, the need to sweep
is eliminated. A virtually instantaneous appraisal of the entirety of the
spectrum is obtained and the possibility exists for effecting a display which
will adequately mirror spectral change when examining a quasi-stationary
process. However, it is arguable that resolution will be potentially inferior
to that of the swept analyser, simply because of the cost of implementing
a large bank of fine resolution filters.

It is usually the case that either full-octave or, more probably, 1/3 octave
filters would be employed in the filter bank. A full-octave filter is defined
such that its upper and lower -3 dB cutoff frequencies, fu and f1, are related
as fu = 2f1• The filter centre frequency is defined to be the geometric mean
of the cutoff frequencies: fc = (fi1) 112 • Such a filter has a proportional
bandwidth (fu- f1)/fc which is 70% of its centre frequency. This means that,
to cover three decades of frequency, some ten filters will be needed in the
filter-bank. The 1/3 octave filter is defined such that fu = 2113 f1• Such a filter
offers a 23% proportional bandwidth. Thirty such filters would be required
to span three decades of frequency.

3.4 Fast Fourier Transform Analysers

The most flexible and sophisticated of spectrum analysers currently available,

the "FFT" analysers, make use of hardware implementations of an algorithm
known as the Fast Fourier Transform [3.2] which is, itself, a particularly
efficient method of computing a result known as the Discrete Fourier
Transform {DFT). An introduction to the properties of the DFT and the
nature and use of the FFT is to be found in reference [3.3].

The DFT calculates Fourier Series components from a sample sequence

segmented from the process to be analysed. The segment is presumed to
represent one period of a hypothetically periodic waveform. Because the
input waveform will not in general be periodic, figure 3.3(a), averaging on
a segment by segment basis allows a stable, smoothed spectrum of an
aperiodic fuction to be built up.

The DFT algorithm computes N harmonically related spectral coefficients

X(n) as the sum
=~L
N-1
X(n) x(k)exp{-27tjkn/N) n =0, 1 ... N-l
k=O
 Characteristics and Analysis of Sonar Waveforms 37

where- as figure 3.3(b) illustrates- x(k) represents the sampled, periodic

input vector derived from a bandlimited time function v(t) containing a
highest frequency f0 , by sampling at equal time intervals T = l/f0 , so that

x(k) = v(kT); k = 0, 1 ... N-1

An inversion transformation also exists and can also be computed using

the FFT algorithm. This is the Inverse Discrete Fourier Transform {lOFT).
The IDFT has the form
N-t
x(k) = L X(n) exp(21tjkn/N) k= 0, 1 ... N-1
n=O
The inter-relation between the (line) spectrum and the input data sequence
is shown in figure 3.3(a) and (b).
v(t)

(a)

x(k)

0 T-,
-+-------=-+~----~- k
r+ N-1
(b)

Figure 3.3 Illustrating the waveform sampling and spectral periodicity

inherent in the Discrete Fourier Transform, applied to a segmented
aperiodic, bandlimited waveform sampled in accordance with the
Sampling Theorem
38 Underwater Acoustic Systems

t4- Line spacing = 1/NT

Amplitude response
of n'th and adjacent
-- --
./\ \
effective line ffiters

.
\!
\i
\\ \:
.. .t""-..l ...- -...._ ..-

L Centre frequency of n'th

effective line ffiter

Figure 3.4 The representation of the DFT as being equivalent to a

"sincx" transfer-function parallel filter-bank

x(k)

x(k) w(k)

Figure 3.5 Windowing to reduce terminal discontinuities and reduce

leakage between spectral lines by modification of the "sinxlx" filter
transfer function induced by rectangular windowing
 Characteristics and Analysis of Sonar Waveforms 39

A time-series such as defines each DFT spectral coefficient can be taken

to represent, also, a Finite Impulse Response {FIR) digital filter. The FFT
analyser is thus, in this respect, directly analogous to the filter-bank analyser,
except that its operation is numerical, rather than _analog-electronic. By
employing the methods of Laplace or z-transform calculus, it is possible
to elucidate the effective transfer function of each "filter" in the DFT "filter-
bank". The filters, if the DFT is used in its most primitive formulation (that
is, literally as stated above) are poorly selective by comparison with well-
designed analog octave or l/3 octave filters such as would be used in an
analog filter-bank analyser.

Each DFT "line filter" has a "sin x/x" response against frequency, figure
3.4. This means that, if a large spectral line lies partway between adjacent
spectral line locations, leakage into the adjacent lines (and, indeed, other
close-to lines) will occur. This problem can be ameliorated by impressing
a windowing function [3 .4] upon the abstracted segment as figure 3.5 shows.

Because the windowing process "loses" some information, the segmentation

process will probably now be engineered to include some degree of overlap,
thus potentially reducing data throughput and hence curtailing effective
analyser input bandwidth.

Because the FFT analyser relies on a hardware implementation which is

firmware-controlled and software-evolved, great flexibility in processing
options can be made available to the user. In particular and quite unlike
the swept-filter and filter-bank analysers, the FFT analysers can compute
(subject to sampling limitations, of course) literal Fourier Transformations
of input waveforms. This means that they may display not only power or
energy spectra, but the amplitude and phase functions of a suitably
synchronised input waveform. Yet more sophisticated processing is possible.
The analyser, having computed a power spectrum, may transform again,
to display a correlation function. The majority of FFT analysers will allow
input of two waveforms, simultaneously, making possible the calculation
of cross-spectra and cross-correlation functions.

Finally, the ability to synchronise FFT analysers, in "grabbing" a waveform

segment, means that the analyser may be triggered by, for example, an
internal combustion engine firing cycle. This makes possible most
sophisticated displays of phenomena which are only short-term stationary,
such as would derive from the detonations which take place within the
engine.
40 Underwater Acoustic Systems

3.5 Prony Analysis [3.5]

The OFT approach to spectral analysis remains the workhorse of waveform

analysis techniques. However, it is possible to call upon other procedures
which may, under some circumstances have advantages. The OFT equation,
in one sense, presumes a structure to the waveform being analysed. For
example, suppose we were to employ the technique in analysing a sound
from a musical instrument which was rich in harmonics. Suppose also that
we could synchronise the sampling period to that of the fundamental of the
musical instrument. Then the OFT "filter bins" would co-locate with the
harmonics of the sound, and great precision in identification of amplitude
and phasing of the harmonics would result.

The generation and need for identification oftonal components in a waveform

is by no means unusual in underwater acoustics. For example, the self-noise
of a ship will contain readily identifiable line spectral components. They
will derive from the whines, whistles and hums associated with reciprocating
and rotating machinery. There is no prior reason to assume that such tonal
components will exhibit harmonic inter-relationships although such a feature
may from time to time be present. For example, blade-rate radiation from
a three-bladed propellor will occur at the third overtone of shaft vibrations.
By contrast, where there is gearing between an engine and a shaft, the
gearing ratio will not, in all probability, be a simple integer relationship
and tonal components initially deriving from a single source will not exhibit
harmonic relationship.

Furthermore, in using the OFT, it is common to employ a large number

(typically 512 or 1024) of spectral samples. There may be far fewer tonal
components of profound interest in, say, a ship's self-noise spectrum. An
alternative analysis option due, with much subsequent modification, to
Prony and thus bearing his name involves modelling the signal (or perhaps
more accurately the signal generating mechanism) as a suite of anharmonically
related sinusoidal generators, the amplitudes, phases and frequencies of
which are to be adjusted by a least squares procedure to minimise an error
vector between the model time series, xm(k) and the signal time series x(k).
Thus we might write

L
N-1
xm(k) = A(n) cos(ro(n) t + <J>(n)) k = 0, 1 ... K-1
n=O

and attempt to minimise {xm(k) - x(k))Z. The result of the minimisation

is an array of values of the vector [A, ro, «1>1 for theN values of n. Drawing
the spectrum then means simply drawing in zero-width lines which purport
to locate, and specify amplitude and phase of, the anharmonic components
of x(k).
 Characteristics and Analysis of Sonar Waveforms 41

Of course, the method will work only poorly if x(k) is rich in atonal spectral
components; that is, if the spectrum is dominated by noise-like sounds. This
could be the case (at least in part) when ship sounds are being analysed,
because the bow-wave and propellor cavitation give rise to such processes.
Clearly, the choice, use and interpretation ofthese various analysis techniques
require care and experience. Finally, it should be noted that the technique
described above is predominantly employed in an off-line, software-dominated
context although, with the ready availability of highly portable computing
machinery of great power and flexibility, no major difficulty would be
thought to attend a hardware, real-time implementation.

3.6 Further Model-building Techniques for Spectral Estimation

[3.6, 3.7]

The Prony method is a subset of a more general class of model-building

methods of analysis which have assumed great popularity during the past
decade. This wider class of techniques is based upon the assumption that
many sound-generating mechanisms can be modelled as, effectively, noise-
stimulated digital filters. The general form of a digital filter, specified in
the frequency domain, is as a z-operator [3.8] transfer function

H(z) = B(z)/A{z)
where z = exp(-jroT) and T is the sampling interval. A(z) and B(z) are
polynomials capable of representation in the factored form A(z) = (z + a1)
(z + a 2) ••• (z + am); B(z) = (z + b 1)(z + b2) ••• (z + bn). This amounts to
supposing that a sound generating mechanism may be at least identified
by a z-plane pole-zero filter model and that even its very physical structure
may yield to such an interpretation. In some ways, such a line of thought
should not, perhaps, be suiprising. Many physical structures capable of
vibrating are representable as interacting collections of lumped energy
storage and dissipative elements. Thus the practical application of the theory
of differential equations and the use of transform methods in their solution
should be recalled as a commonplace, in the investigation of such phenomena.

Whether the sampled-data or difference equation form is truly likely to be

innately representative of underlying physical structure is, perhaps, to be
doubted. Its utility in the context of digital computation is clearly evident
although it should be remembered that, with the current rapid increase in
computing power, the rather simple integration model implied by the z-
transform operator may yet be superseded by digital processing which more
closely parallels the end-effect of the analytical process of integration and
thus permits a return to a more direct attack upon modelling the actual
physical structure of the generating mechanism.
42 Underwater Acoustic Systems

Use of the model presented above in a practical spectral analysis context

can take on any of several different forms. If the A(z) = 1, then the model
is essentially that of a moving-average (MA) filter and the analysis is
referred to in that way. If the B(z) = 1, the model is an all-pole digital filter
and is referred to as being auto-regressive (AR). If both A(z) and B(z) are
used to model the process, then it is referred to as an auto-regressive,
moving-average (ARMA) analysis.

The objective of the analysis procedure is the moving about of the pole/
zero locations on the z-plane so as to minimise the mean-square error
between some appropriate attribute of the filter output and some chosen
equivalent attribute of the waveform being matched. This could involve,
paralleling in a sense the Prony method described in the previous section,
the minimisation of the error between the filter output vector and a vector
containing the sampled signal being analysed. In fact, the analysis takes
as its input a (short) Blackmann-Tukey [3.9] time-domain derived correlation
function and uses the ARMA, or MA or AR technique, to establish a
matching correlation function in the filter output. Since the correlation
function is a statistical attribute of the digital filter, rather than a deterministic
output, a white noise excitation of the filter input (numerically) would be
entirely satisfactory.

It may seem strange at first sight, particularly to those who are familiar
with the speed tradeoff in computing correlation functions, using an FFT
algorithm by first calculating a power spectrum, then re-transforming.
However, it is appropriate to recall several features of the application of
ARMA-type modelling. First the model size or order will be modest by
comparison with the size of an FFT. Then the method may yield attractive
improvements in the quality of a spectral estimate, by comparison with the
FFT approach, particularly if record lengths are short.

Finally, we should ask how does the method produce a spectral estimate,
if all that has happened is the jiggling of pole/zero positions, to match up
target waveform and model output waveform correlation functions? Notice
that, once the jiggling process has been completed to within some required
level of error magnitude, we are left with a transfer function H(z) defined
in terms of pole and zero (or pole, or zero) locations. By re-writing H(z)
with z replaced by exp(-jroT), we obtain a function in ro, which may then
be readily reduced to the form of a power spectrum, since

P(ro) = IH(exp(-jroT))I2
 Characteristics and Analysis of Sonar Waveforms 43

3.7 Four-dimensional Space-Time Waveform Analysis

Imagine, as figure 3.6 suggests, a model of the ocean, wherein exists a

spatially distributed pressure field which is the result of the combined effect
of a multiplicity of sound sources. In this illustration one might be tempted
to suppose that a single source existed somewhere down in the bottom rear
left-hand corner. This could be the case, but such a visualisation actually
simplifies the reality, where wavefronts from many directions would
interweave to create a locally fluctuating pressure at a hydrophone location.
What possible observables and statistical measures might interest us?

Figure 3 .6 A randomly fluctuating sound-field with two sensing

hydrophones H 1 and H 2

At hydrophone H1 at co-ordinates (x 1,y 1,z 1) we might observe pressure p 1(t)

= p(xl'yl'zl't). In fact, the hydrophone output would be a voltage v1(t) =
kh p1 (t), where ~ is the hydrophone constant, measured in volts per J.l.Pa.
Since it is probably easier for most readers to think in terms of the hydrophone
output voltage and its various attributes, when manipulating units, than the
actual pressure field itself, we shall continue by utilising this form of the
sound field variable. We can, of course, think immediately in terms of a
Fourier Transform of v1 (t): V1 (f) ¢:::> v 1(t). However, it is probably better
to skip this stage of thinking and to anticipate other frequency-domain, time-
frequency and delay-domain attributes of the process.
44 Underwater Acoustic Systems

We thus observe v1(t), the hydrophone output voltage which is proportional

to the time-domain instantaneous pressure fluctuation and which is measured
in volts [V]. We may also seek to determine, by calculation or direct
measurement, one of the following spectral quantities, because each allows
us to form an appreciation of the distribution of power or energy (as is
deemed appropriate) within the waveform as a function of frequency, or
of both time and frequency . An electronic mechanism for achieving these
objectives is shown in figure 3.7.

I
'
X r-• % ___.. Output
oc P(f)

·----------- ---------·· I 4 Averaging

Bandpass filter
Squaring
analysing circuit
filter
---------··· --------···
·----------- ----------
• f~
___.. Output
Bandwidth
oc E(f)
B Hz

'
.· : : : : : : : : : : <-..,; · : : : : : : : : : .

·: ::::::::: ~ ·::::::::: SetT 4

·::::::::::. ~ :::.·:::::: Integrator
·········-·-------------
····· · ·······-------·--
·:::::::::::::::::::::::
Receive hydrophone sensing a noise intensity I 0
within the frequency band B at centre frequency f

Figure 3.7 The measurement of power and of energy density estimates

by means of a narrow-band, band-pass filter

P(f)

Figure 3.8 Integration of power spectral density to yield total power

 Characteristics and Analysis of Sonar Waveforms 45

P1 (t): the power spectral density, measured in [V2 Hz-1] or, preferably,
[V2 s], assuming p1(t) to have been a stationary, finite power process,
or
E 1(t): the energy spectral density, measured in [V2 s Hz-1] or, preferably,
[V 2 s2], assuming p1(t) to have been a non-stationary, finite energy
process, or
e1(t,t): the short-term energy spectral density, measured in [V2 s2], assuming
p 1(t) to have been a quasi-stationary, finite energy or finite power
process.

Notice, figure 3.8, that the integral of P1(t) with respect to frequency yields
power- the average power in a continuously transmitted broadband signal,
perhaps. The integral of E1(t) with respect to frequency yields energy - the
total energy in an explosive detonation, for example. The integral of e1 (t,t)
versus time and frequency (the volume beneath the e1(t,t) surface over the
t,f plane) also yields total energy, the function itself depicting, figure 3.9,
energy observable within a short epoch L1t by an analysing filter of width
M.

Let us pursue the quest for observables deriving from v1 (t) further. We may
attempt to observe a correlation function R/t) = Jv 1(t) v1(t + 't) dt which
seeks to measure the internal temporal similarity within v1(t). That is, if
we skip over a time interval 't, do we observe some measure of similarly
between the delayed version of v1 , namely v1 (t + 't) and v1 (t) itself? For
example, if v1 were a periodic function, of period T, then we would expect
to encounter strong correlation at delay intervals 't = nT, n = 0, 1, 2 ....
Notice that correlation may be observed in both finite power and finite
energy signals. It is not uncommon, for example, to observe strong correlation
in the (finite energy) waveform following an explosive detonation because,
particularly in shallow water, a high level of reverberation will be present,
and delayed and attenuated replicas of the explosive signature will be
contained within the hydrophone output waveform. We note that the correlation
function, as expressed above, actually pertains only to finite energy processes.
For finite power processes, the integral would necessarily become infinite,
in the limit. We should write, more correctly

f
+T/2

Ru(t) = ~~ T
. 1
-T/2
for finite power processes and

J
+T/2

R 11 ('t) = v1(t) v1(t + 't) dt

-T/2
46 Underwater Acoustic Systems

for finite energy processes. The "11" subscript denotes auto-correlation

estimation and is introduced here in anticipation of the need to define a
delay-domain "cross-correlation", as between the outputs of hydrophones
H1 and H2 • Note that the forms of correlation coefficient for the finite power
and finite energy processes are intrinsically different, dimensionally. The
reader will probably be familiar with the Wiener-Kinchine relationships

for finite energy processes and

for finite power processes, with the appropriate correlation integral being
used to define the relationship between R 11 and v1 in each case. We further
note that it is often the case that a normalised and dimensionless correlation
function will be employed, by computing a quantity Rl 1(t)/R 11 (0). This
quantity has extreme values of ±1.

Thus far we have concentrated on observables associated with a single point

of measurement (xl'y l'zJ At two different points (xl'y l'z 1) and (x 2,y 2,z2)
we might expect to encounter some usefully interpretable element of delay-
domain correlation as, for example, when the second point of observation
lies further back along a line drawn from an acoustic source through the
first point of observation. This is just the same as using two hydrophones

Figure 3.9 The e(t,f) spectrum, showing that the volume beneath the
surface yields a measure of the total energy in the signal within an
epoch Lit and over a frequency band of width Lif
 Characteristics and Analysis of Sonar Waveforms 47

to form a crude array, which might be pointed at a source to determine

azimuth and bearing angle. The important quantity needed in interpreting
the hydrophone outputs is the baseline distance between them which,
together with sound-speed, allows the differential delay to be determined.
This delay will be greatest when the two hydrophones are co-linear with
the source. We thus seek to measure or calculate a cross-correlation R 12{t)
where +T/2

R12('t) = ~~o ~ J
-T/2

for finite power processes and

+T/2

R 12('t) = J v 1(t)v2(t+'t)dt
-T/2

for finite energy processes. These quantities are related to cross-power,

P 1it) and cross-energy E 12(t) spectral densities, respectively. They describe
a similarity of power or energy spectral components between the two
locations as, for example, when directional hydrophones observe similar
attributes of a signal, at different spatial locations, when looking at a source,
but may observe different attributes of a directionally variable sound or
reverberation field corrupting the source signal. The relationships (again
chosen to be appropriate for the type of signal being investigated) are R12{t)
<=> P 12{t) and R12(t) <=> E 12(t).

It was mentioned in section 3.1 that the cepstrum [3.10] provides another
delay-domain method of examining a waveform. In order to appreciate
better the computational steps involved in cepstral processing we shall
concentrate upon finite power functions so that, given a hydrophone output
v(t) we may hope to obtain a power spectrum P(t), from which we establish
the power cepstrum as c(t) <=> logP(t). Why should this apparently trivial
processing step be of any value in waveform analysis? The reason is perhaps
most easily understood and, from our standpoint as underwater acousticians,
is best exemplified by considering the problem of multipath propagation
(or equivalently, reverberation caused by echoes from the sea-surface and
sea-floor). For this purpose, we imagine that the cepstrum is to be used in
a diagnostic sense, to tell us something about the reverberation delay: the
difference in arrival time between a main-path signal and one or more
echoes.

If we represent the main path signal, as registered at the hydrophone output,

as v(t) then a multipath signal may be represented as k(v(t + t 0 )), where
48 Underwater Acoustic Systems

k is a loss-factor caused by the increased path length of a reflected signal

as well as loss on reflection, and t 0 represents the differential delay between
the two paths, which we wish to be able to estimate. The hydrophone output
is then

V 0 (t) = v(t) + k(v(t + tJ) <=> V(f) + k(V(f) exp(21tjft0 ))

= V(f)(l + k exp(2pjft0 ) )

Manipulating the right-hand side of the transformation, we note that the

spectrum at the hydrophone output is that of the transmitted signal, but with
an impressed, sinusoidal modulation so that

IV0 (f)l 2 = IV(f)l2 { (1 + k cos(21tft0 ))2 + k 2 sin 2(21tft)}

oc IV(f)l 2 { 1 + k cos(21tft0 )}

Now, frequently, the source signal V(f) will be of"low-pass" form, exhibiting
decreasing spectral components with increasing frequency. This means that
the power spectrum of V0 (t) will have the form shown in figure 3.10(a).
Transformation into the delay-domain, as a correlation function, allows the
sinusoidal ripple to transform as a more or less well identified spike located
at delay magnitude t 0 , as figure 3.10(b) shows. Of course, if the delay
magnitude is relatively small, the spectral ripple period will be large and

_PD:Q~~!---------~--
...: ·.. -----L k
..

f
(a)

R(<t)JR(O)
-1

(b)
Figure 3.10 The power spectrum contaminated with a ripple
modulation in frequency caused by multipath, and the corresponding
correlation function, showing the delay identifying spikes
 Characteristics and Analysis of Sonar Waveforms 49

very few cycles of ripple will be available for the lOFT analysis to work
on. Identification and use of the delay-spike will then become difficult. If,
however, the logarithm of the power spectrum is calculated, then the ripple
amplitude remains large over the entire frequency range, even at low delay
magnitudes. Consequently, the cepstrum may be thought to provide a
preferable mechanism for delay-attribute identification and measurement
than, say, a correlation function.

In fact, because the non-linear operation implied by the logarithmic function

adds no further information to the signal being processed, it is without doubt
the case that further clever processing would cause a correlation function
to yield as good information concerning delay. The cepstrum should, then,
really reside as a convenient, rather than a unique weapon in our armoury
in analysing signals which have passed through a reverberant channel.
Indeed, because the logarithmic operation on the spectrum distorts a sinusoidal
ripple-shape into a cusp ripple-shape, transformation produces not a single
delay-domain spike, but a periodic multiplicity of such spikes. With more
than one delay path present, this can make the cepstrum at least as difficult
to interpret as an auto-correlation function.

As it happens, an ingenious variant of the cepstrum, known as the meta-

cepstrum [3.11], may get around this difficulty. Instead of applying a
logarithmic function generation, imagine the numerical equivalent of a
variable-gain amplifier being passed across the spectrum, almost to establish
spectral whitening by automatic gain control in the frequency domain. Then
- considerably more closely, at least - the ripple will retain its sinusoidal
form as well as maintaining constant amplitude across the full span of the
spectrum.

It was also mentioned in section 3.1 that a delay-domain suppression of

echoes within v(t) may be possible. To do this we form a complex cepstrum.
At this point we must make special note of a small point of philosophy,
concerning the use of the cepstrum method. We would normally imagine
that sampled data computations, to evaluate such attributes as spectra and
correlation functions, are but convenient or, indeed, in the sphere of digital
computation, necessary approximations to otherwise impractical "analog-
world" transformations which would be presumed to act upon continuous
signals.

In the case of the cepstrum - and particularly in the case of the complex
cepstrum - this is not really true. The complex cepstrum is essentially an
attribute of the sampled signal, rather than of the signal itself. This is
because, in order to form the complex cepstrum, we have no alternative
but to segment our time-series representing v(t) (::) v(k); k = 0, 1 ... N-1
50 Underwater Acoustic Systems

and, using a OFT, transform to compute an unsmoothed spectrum estimate

V(n); n =0, 1 ... N-1. We note that, as calculated by the OFT, the transform
V(n) will consist of a real and imaginary part: V(n) = X(n) + jY(n) from
which amplitude is calculable as A(n) = {X2{n) + Y2{n))ll2 and phase is
calculable as cjl(n) = tan-1{Y(n)/X{n)). We then form the complex quantity
ln(V(n)) = ln(A(n) exp(jcjl(n)) = ln(A(n)) + jcjl(n) and apply the lOFT tore-
transform this quantity into the delay-domain.

Thus far, we have investigated the temporal, spectral and delay-domain

properties of the sound field. Although, in calculating cross-correlation
functions, we happen to have investigated a spatially variable geometry,
we have not explicitly sought a statistical measure of similarity which was
actually a function of, let us suggest, a displacement vector

That is, a correlation function which would seek to establish a statistical

average such as

By way of example, let us restrict the problem to one wherein we seek to

determine the spatial correlation resulting from measuring the cross-power
spectral density

where V1 and V2 are Fourier Transforms of the hydrophone outputs. That

is, for a single separation, ~. we move the vertical pair to depth z and measure
the hydrophone cross-power spectral density. We move to a new depth and
repeat, averaging in order to assemble by this means the function C for fixed
separation. By repeating again for different separations, we build up a
picture of the spatial function itself. The spatial cross-power spectral
density may be Fourier Transformed to yield a new function, referred to
as a wavenumber spectrum, Q(p) [3.12]. We write

and calculate

-too

Q(p) =- 1
21t
JC(~)exp(-,ip~) d~
-oo
 Characteristics and Analysis of Sonar Waveforms 51

In this case, because the vertical hydrophone pair cannot respond to variability
in the horizontal plane, we must presume that (as for example in an open
ocean situation) no innate horizontal directivity is to be encountered. The
reader interested in pursuing further the question of angular directivity and
spatial correlation is directed towards reference [3.13]

References

[3.1] R.B. Randall, Freqwency Analysis, Bruel & Kjaer, Naerum, DK-2850 Denmark, 1987
[ISBN 87 87355 07 8]

[3.2] J.W. Cooley and J.W. Tukey, An Algorithm for the Machine Computation of Complex
Fourier Series, Math. Comp., Vol. 19, 1965, pp. 297-301

[3.3] R. Coates, Fourier Transform Methods, in An lntrodwction to Digital Filtering (Bogner,

R.E. and A.G. Constantinides, ed.), Wiley, 1975

[3.4] F.J. Harris, On the Use of Windows for Harmonic Analysis with the Discrete Fourier
Transfor", Proc. IEEE, Vol. 66, Jan. 1978, pp. 51-83

[3.5] S.L. Marple, Spectral Line Analysis by Pisarenko and Prony Methods, IEEE Conf.
Acowstics, Speech and Signal Processing, 1979, pp. 159-161

[3.6] S.M. Kay and S.L. Marple, Spectrum Analysis - A Modem Perspective, Proc IEEE,
Vol. 69, No. 11, 1981, pp.1380-1419

[3.7] O.L. Frost, Power Spectrum Estimation, in Aspects of Signal Processing (G. Tacconi,
ed.), Reidel Publishing, Dordrecht, The Netherlands, 1977, pp. 125-162

[3.8] A.V. Oppenheim and R.W. Schafer, Digital Signal Processing, Prentice-Hall, New
Jersey, 1975

[3.9] R.B. Blackmann and J.W. Tukey, The Measwrement of Power Spectra from the Point
of View of Commwnications Engineering, Dover, New York, 1959

[3.10] D.G. Childers, D.P. Skinner and R.C. Kemrait, The Cepstrum: A Guide to Processing,
Proc. IEEE, Vol. 65, No. 10, 1977, pp. 1428-1443

[3.11] P. Hirsch, The Metacepstrum, /. Acowst. Soc. Am., Vol. 69, No.3, 1981, pp. 863-865

[3.12] H. Cox, Spatial Correlation in Arbitrary Noise Fields With Application to Ambient
Sea Noise, I. Acowst. Soc. Am, Vol. 54, 1973, pp. 1289-1301

[3.13] S.M. Morley and C.L. Baxter, Angwlar Distribwtion Analysis in Acowstics, Springer-
Verlag, Berlin, 1986 [Series: Lecture Notes in Engineering, No. 17]
4 Ray Trace Modelling of Sonar
Propagation

4.1 Introduction

Sonar modelling has to do with predicting sound intensity at some point

in the sea remote from a source. It provides a more detailed way of predicting
performance than do the sonar equations, which would usually be used as
a "first cut" and preferably "worst-case" approach to system design. Sonar
modelling is of great importance in deducing the path traversed by sound
as, for example, in seismics- where it is required to determine the thickness
and acoustic characteristics of sea-bed sediment layers - or in military
applications where range and bearing to an underwater sound source, such
as an enemy submarine, must be found. A comprehensive review of modelling
software currently in use is to be found in reference [4.1].

Most frequently, the modelling problem is defined by assuming a known

vertical sound velocity profile as characterising a channel of fixed depth.
Horizontal variability of sound velocity is often discounted, on the grounds
that it is variable only over distances substantially greater than typical ocean
sonar ranges. This assumption should be regarded as dubious in shelf-sea,
or estuarine conditions or when operating at extremely low frequencies. The
structure of, and nomenclature associated with the world seas and oceans
is shown in figure 4.1. Table 4.1 provides an indication of the dimensions
associated with the various submarine features illustrated in this diagram.

Abyssal plain Cant inental rise

Figure 4.1 The structure and nomenclature of the ocean

52
 Ray Trace Modelling of Sonar Propagation 53

Table 4.1
Typical dimensions of various ocean features

Av. Max. Area Sound speed characteristics

depth depth x10 6
(m) (m) (mz)

Atlantic Ocean 3900 9100 82 Ocean: Fig. 4.10

Pacific Ocean 4300 11000 165 Ocean: Fig. 4.10
Arctic Ocean 1200 4600 14 Isothermal, O"C
North Sea 94 700 0.6 Isothermal, 2-10"C (seasonal)
Mediterranean Sea 1400 5100 3 Ocean: Fig. 4.10; S:38 ppt (av.)

Topography of continental margins

Width (km) Gradient

Shelf 75 (av.) 0.0002 (av.)

Slope 20-100 0.005 (av.)
Rise 0-600 0.0001-0.001

Modelling takes as its most fundamental basis, some attempt to solve the
wave equations, approximately and typically numerically. Two principal
methods exist, with several variants which we shall not discuss. The first
method parallels the processes of geometric optics, assumes a horizontally
stratified medium and is referred to as Ray Tracing. The second method,
known as the Mode Theory approach, develops particular solutions to the
wave equations, analytically, which describe the ability of the channel to
enter preferred "resonant states". Numerical methods are then used to
compute pressure at a specified depth, and range as a function of time and
frequency. The use of Mode Theory thus avoids the necessity, incumbent
upon the Ray Tracing method, of establishing the entire ensemble of rays
at all points between source and receiver. Ray Tracing is suitable for
applications where sound wavelength is small by comparison with range
and water depth. Mode Theory is often considered complementary to Ray
Tracing, being suitable for application to shallow water channels. It is a
subject that we shall examine in greater depth in Chapter 5.

4.2 Ray Tracing Sonar Models [4.2-4.5]

Ray tracing involves the application of Snell's law to a horizontally stratified

medium. The most important concept in ray tracing is that of the ray
54 Underwater Acoustic Systems

Figure 4.2 Refraction through water layers of differing sound speed

coefficient "a". If a ray is launched at angle e(z) from a source at depth

z, figure 4.2, then on intersecting a boundary between media of differing
acoustic impedance

If many layers are present, figure 4.3, then this result may be generalised
in the following way:

sin(9(z))/c(z) = sin(e(z 1))/c(z 1) = sin(9(z))/c(z = a

0)

and the parameter a is seen to depend on the initial launch angle and the
initial sound speed, and to remain a constant for that ray thereafter. By
applying this principle of constancy of ray coefficient along any given ray,
computer programs may be written to plot the courses of typical rays
launched from a source. The programs may also compute distance travelled
along a ray, and may be used to infer intensity at points remote from the
source.

Figure 4.3 Refraction through many layers : constancy of the ray

parameter a
 Ray Trace Modelling of Sonar Propagation 55

Figure 4.4 Propagation in isove/ocity (constant velocity with depth)

conditions

In order to be able to understand how ray tracing software evolves, we look

next at the way in which a ray will pass through a layer of constant sound
speed. That is, isovelocity conditions pertain, with c(z) = 1500 m s-1 for
all depths. Then propagation takes place in straight line paths, as figure
4.4 illustrates. Suppose, then, that we return to our stratified model, and
imagine that sound speed is constant within each layer. Then, as figure 4.5
shows, we imagine the sound speed profile with depth to be a histogram-
like approximation to what, in reality, would be a continuous and smoothly
changing sound speed profile. Now. if a ray is launched at depth z0 in the
first layer, at an angle 90 to the horizontal, one may compute first the
distance to the second layer, and thus obtain end-point co-ordinates for
plotting the first part of the ray, and then proceed on downwards through
the remaining layers to compute the entire ray trajectory. Clearly, the finer
the gradation of the stratification, the smoother will become the computed
ray trajectory. Equally clearly, so also will computation time increase.

Figure 4.5 Stepwise , stratified approximation to a sound speed profile

decreasing linearly with increasing depth
56 Underwater Acoustic Systems

Figure 4.6 Using linear interpolation of sound speed within the

stratified layers

An obvious step in improving matters might involve considering an alternative

approximation to the sound speed versus depth profile. Consider, for example,
a situation in which we assume sound speed to vary linearly, across a layer
after the manner shown in figure 4.6. It can then be shown that sound rays
travel in circular paths of specifiable radius and centre co-ordinates. Now
the computations become more complex, but the curves so generated are
smoother and fewer layer approximations become necessary. Whilst one
might contemplate going to even higher orders of approximation - making
for example, parabolic fits to the sound speed profile - this has not been
considered worthwhile in the past.

4.3 Ray Trace Calculations

For the linear sound-speed fit within a layer, the important results governing
ray path calculations are summarised here. It should be stressed that much
greater detail can be built into a ray-trace program, which can become an
entity of quite considerable complexity. The following equations will allow
the interested student to begin a process of development of modelling
routines suitable for use on any of a wide range of fast, graphics-oriented
modern microcomputers.

Figure 4. 7 depicts the geometry of the problem. For a ray, of ray constant
a, moving within a layer of water with a sound speed profile

c(z) = z0 + bz
the sound will travel in a circular locus [4.6] of radius, r, given by

r = (ab)-1
 Ray Trace Modelling of Sonar Propagation 57

z
Figure 4.7 Defining the geometry of circular ray propagation within
the depth layer z1 to z2

If the starting co-ordinates for the ray are (z 1,r1) then the centre of the
upward-curving circular locus will have co-ordinates

Notice that the centre of the downward curving locus, encountered when
the sound speed gradient in a stratum is negative rather than positive, will
have co-ordinates

In either event, the equation of the circle so defined is

(z - zc) 2 + (x - xy =r 2
so that, if the approached bound of the stratified layer in which the ray is
moving is at depth z2, then the terminating co-ordinates for the arc will be
(z2 ,x2) where x2 will be given by

x2 = xc + 0 .5 {2x c
2 - (z2 - zc)2+ r 2 } 112
58 Underwater Acoustic Systems

Since many modern microcomputers will access graphics software (or even,
in some instances, hardware) capable of effecting circular arc plots directly
and with great speed, the equations given above, plus some control logic,
afford the basis for a simple ray tracing program. Care is needed to trap
the ray turning situation and to cope with the isovelocity condition b=O,
which corresponds to straight line propagation and by implication an infinite
radius for the circular locus. Logic is also needed to determine the intercept
with the sea-floor and initiate the reflected ray.

Further equations which allow computation of distance and approximate

travel time along a ray segment are given below. ~ is taken to be the time
at the point of entry of the ray to the layer, at location (zl'x 1). ~is the time
of arrival at location (z2,x2). The length along the arc is determined by
evaluating first the angle to the vertical, 92 at (z2,x 2).

The included angle at the centre of the circle is then 9c = 92 - 91' and the
length of the arc is simply ~s = r9c. The average sound speed within the
stratum is given by

Consequently the approximate travel time is given as

More general expressions are to be found in the literature [4.2] for travel
time along a ray. However, since starting data for sound speed profiles, for
example, is often but poorly known, some considerable insight into sound
propagation may yet be acquired with models based even upon these simple
rules.

4.4 Some Examples of Ray Modelling

In this section, we shall assume a simplified statement of sound speed as

a function of depth, temperature and salinity which recognises approximate
coefficients of sound speed for these variables of 0.0016 m s-1 per metre
increase in depth, 4.6 m s-1 per Centigrade degree increase in temperature
and 1.3 m s-1 per part per thousand increase in salinity, giving

c = 1450 + 4.6T + 0.0016z + 1.3(S - 35)

 Ray Trace Modelling of Sonar Propagation 59

~t'UNO OI'C CO Ul !118

j" ' I ' IG 1' 11

•e
....

..,.
Figure 4.8 Sound rays propagating in upward circular arcs in
isothermal Arctic water

Isothermal water, such as might be found in the Arctic Ocean or in any

deep ocean below the main thermocline, leads to a sound speed profile which
rises approximately linearly with depth, because it is the increase of sound
speed with increasing pressure that becomes the dominant effect. Thus in
Arctic water with T = 0 ·c and S taken as 35 ppt:

c = 1450 + 0.0016z
The effect of projecting rays from a source is shown in figure 4.8. The
upward circular ray paths are readily identified. It will also be noticed that
some rays cannot reach the ocean-floor. The turning depth for a ray launched
at depth z1 and at a downward angle 9(z 1) to the vertical is readily calculated.
The ray constant for such a ray is

where z, is the turning depth. But at the turning depth sin 9(z,) must be unity
since 9(z,) will be 90•. Therefore we find that

and since c(z) is expressible, for isothermal water, as an equation of the

form

where c0 is the surface sound speed and b the pressure coefficient with depth,
60 Underwater Acoustic Systems

it follows immediately that

Z1 = b-1(c(z 1)/sin 9(z1) - cJ

Water in the main thermocline decreases in temperature, linearly with

depth. In the deep ocean, the thermocline might extend downwards to a
depth of, perhaps 500 m exhibiting a temperature fall of 10°C through this
interval. The temperature coefficient with depth is thus -0.02oC per metre
increase in depth. In the thermocline, with S again taken as 35 ppt, the
temperature-induced fall in sound speed with increase in depth is (about)
-0.02 x 4.6 = 0.092 "" 0.1 m s· 1 per metre increase in depth, which is nearly
two orders of magnitude larger than the pressure coefficient of 0.0016 m s-1
per metre. The sound speed equation thus becomes

c = 1450- 0.1z; 0 < z < 500

This sound speed profile gives rise to downward circular ray paths within
the thermocline, as figure 4.9 shows. Below the thermocline will exist
isothermal deep-ocean water at, perhaps, 4 oC. Consequently, from the base
of the thermocline at 500 m depth, to the ocean-floor at 2500 m, the sound
speed profile is (as with our first example) dominated by the pressure

IOUIIDIP!LD 1.!1 "" lt*K 1111 M

1U 0 lt..O 1M0
0
··~------~------~"------~~----~

,. .I
!

Figure 4 .9 Sound rays propagating in downward circular arcs within

the main thermocline
 Ray Trace Modelling of Sonar Propagation 61

Figure 4.10 Idealised deep ocean sound speed profile

IIIIWUN: Ill M

... f-' -------o--!~----..!;":.__----"+-"-

110 ...
" "
...
~ ~
~ ~

Figure 4 .11 Sound trapped in an acoustic waveguide : projector at

lower boundary of main thermocline
62 Underwater Acoustic Systems

coefficient and
c = 1400 + 0.0016z; 500 < z < 2500

Consequently we see the rays, as they pass out of the thermocline enter
an upward-curving circular locus, but now (because of the much smaller
gradient of sound-speed with depth) with far less pronounced curvature.
Also to be noted on figure 4.9 is the formation of a caustic on the lower
boundary of the pattern of rays. In the vicinity of a caustic, unusually high
sound intensities may be encountered.

A typical ocean sound speed profile, such as is shown in figure 4.10, adds
to these various possibilities a surface layer some tens of metres deep, in
which wave action induces mixing and establishes roughly isovelocity
conditions, for which z = 0, S = 35 ppt and T = 14• C, so that c = 1514
m s-1 • The peculiar nature of the deep ocean sound speed profile results in
some interesting transmission phenomena. As the ray trace diagram shown
in figure 4.11 illustrates, a sound source on the lower surface of the main
thermocline radiates rays which are trapped in its vicinity by the increase
in sound speed which occurs with increasing displacement from it. Such
sound channels are believed to be used by whale pods in establishing trans-
oceanic communication. This is because sound trapped in this way is subject
only to cylindrical spreading loss and, at sonic frequencies, suffers a
relatively small attenuation loss. It is interesting to note, also, that the Blue
Whale has been discovered to emit sounds at a source level equivalent to
that of a modem warship. This makes it not only the largest but also the
loudest animal species.

The ray diagram shown in figure 4.12 also illustrates the formation of the
sound channel, when the source is located at the base of the thermocline,
where the sound speed gradient changes sign. All operating and propagation
conditions remain as in figure 4.11, except that the angular spread of rays
from the source has been increased significantly. The plotted area is also
increased, to show reflections from the sea-surface. The strong channelling
along the base of the thermocline is clearly in evidence.

Finally, some operating and propagation conditions may introduce divergences

such that shadow zones of unusually low insonification are formed, figure
4.13. It is not true that no sound at all exists within shadow zones.
Diffraction effects will still produce weak insonification. Also, regions of
high sound intensity, called focii, may be observed. Notice that the only
difference between the operating and propagation conditions between this
figure and figure 4.12 is that, in the former, the sound source was located
at the base of the thermocline at a depth of 500 m. Here it is only 50 m
below the surface. The effect on the ray-plot of this change in depth is,
however, profound.
 30UNO 9rEED IN M/9
1500 1505 1510 1515 1520 1525
0

100 l ~ ~
Q
"C
200
200 + / ~
-=~
~ ':i~~{~~=~~~~::::;~Jt~~~-~~-/~: ::~:·:~~~~-~~~;~~~¥~~~%~~~ ~
900 " (")
~

•co ii::
•oo + ~
" ., sao f}
" ~
::::
~ ::s
600
Oo
i!: ~J ...~"
-·
~ 700 ~
C'-1
eoo +
\ \ eoo ~
::s
Q
900 t ~
...
~""'"--"'-...~
1000 + \ 1000 t 7 ./~
..."'tt
~''''""-'"-...~~~_J? ~Q
1100 t Oo
'''''""-'"-...-.........~"-.~<«:::<::----~ Q

1200 ... 1200 1 ........

,, ,, ...........:-.........~,.............~........__ ~
• --·
::s

Figure 4.12 . Showing the formation of the sound channel when the source is
located at the base of the thermocline ~
 ~

Kn
SOUND SPEED IN n/S
1500 1505 1510 1516 1620 1525 D
D

100

200 + / 2001~ c::::

~~ ;:s
SOD f 1»'/~//// / / / / / / // $ "#,~
......
!}
~'-~~'-''-''.''""'- '- '- '- ....
~
400 400 t 1:)
+ ~~~~=:,r;><.S rr/~ ///// f f ; J / / 7-
" ....
-"'
,..,)>..
600 c
;:
; 600 t ; ~, ...,
\ t~ 700
....,..,
-
800 + \ 800 ~
...,
900 ;!
t ~~~ ~~
-"'
...,
1000 + \ 1000

1100
12001 \ 1200

Figure 4.13. Showing the formation offocii and shadow zones

 Ray Trace Modelling of Sonar Propagation 65

4.5 Modelling Transmission in the Shelf-seas

The shelf-seas are those shallow seas which surround the continents but
which form the continental shelf. Typically, they will be only some tens
of metres deep. The North Sea, for example, is a shallow shelf sea with
an average depth of 94 m but which over large parts of its area is less than
half that depth. During the major part of the year, weather conditions and
tidal currents establish good mixing and lead to isovelocity conditions and
an absence of a marked thermocline. Water temperature, however, will vary
markedly both over the region and throughout the year.

To place these observations in context, a sequence of North Sea velocity

profiles was analysed to establish an average variation of sound speed,
which indicated a slight increase with depth, caused primarily by increase
in temperature, and at the level of 0.05 m s-1 per metre. This figure, although
small, still significantly exceeds the pressure coefficient of sound speed,
as we have assumed it above, of 0.0016 m s-1 per metre. Its significance
may be further appreciated if we note that the radius of curvature of sound
rays will be calculated as at least 30,000 m, this figure applying when a
ray is launched horizontally, thereby producing greatest curvature. It is clear
that, over kilometre ranges, propagation will be essentially in straight lines.

Propagation at ranges which considerably exceed water depth places some

significant computational demands upon the modelling procedure. In essence,
the sound channel then consists of an isovelocity medium separated by
plane, parallel reflecting surfaces. The reflection coefficent of these surfaces
will, of course, depend upon environmental conditions, frequency and angle
of incidence. For the moment, however, let us assume perfect reflection
and proceed by considering an optical analogy.

Imagine, figure 4.14(a), that a object, 0, and a point of inspection, P, are

defined, within the region separating plane, parallel mirror surfaces. There
will be a direct path between the object and point of inspection. There will
also, figure 4.14(b), be a pair of reflections, providing first images I11 and
I12 in mirrors 1 and 2 respectively. The notation Im1 and Im2 is used to signify
the m 'th multiple reflection in these mirrors.

Outside of mirrors 1 and 2 there will exist multiply reflected images of these
mirrors and figure 4 .14( c) illustrates the first such pair of images of mirrors
1 and 2. As seen from point P, there will now appear image I21 , the result
of apparent reflection in the first image of mirror 2. Similarly, there will
also appear an image I22 , because of reflection of image I12 in the first image
of mirror 1.
66 Underwater Acoustic Systems

Image 2 of Mirror 1
~~~--------------~~

Image 1 of Mirror 2 I21

~~~---------------~~
Ill

0--------- 0
Mirror 1

p
p

Mirror 2

~~~---------------~
Image 1 of Mirror 1

I
22 ~---------------~~
Image 2 of Mirror 2
(a) (b) (c)

Figure 4.14 Illustrating the formation of multiple images of a source

by two plane parallel reflecting surfaces

Clearly, this pattern will repeat, ad infinitum, with 121 reflecting in the
second image of mirror 1, 122 reflecting in the second image of mirror 2
and so on. In acoustical terms, the result of this process of sustained multiple
reflections will be from the observer's viewpoint at P, an effective array
of periodically placed sources. In the acoustic case, mirror 1 will be the
sea-surface, offering phase inversion on reflection. Mirror 2 will be the sea-
floor, which we shall presume to be a hard reflecting boundary offering
no phase inversion. We can build the phase functions into general pressure
reflection coefficients, the choice of which is largely at the discretion of
the investigator.

For example, we might under some circumstances choose to employ the

sea-bed reflection models giving a bottom reflection coefficient Rb(9) = R 12 ,
with R 12 as specified in sections 1.9 to 1.11. R12 is, of course, a function
of incident angle, e, as well as the sea-floor physical properties. Similarly
a simple model for the sea-surface reflection coefficient might just be R.(e)
= -1; all e. Alternatively, we might introduce more sophisticated models,
developed from, perhaps, the sea-floor and sea-surface intensity reflection
 Ray Trace Modelling of Sonar Propagation 67

models presented in sections 2.7 and 2.8 remembering, of course, that such
models do not inherently build in phase information and present intensity
rather than pressure coefficients which are directly calculated in decibels
rather than as a numerical ratio. Our procedure will follow the geometry
developed in figure 4.15.

In this figure we see, on the left-hand side, the first few reflection-pairs
in the sea-surface, the sea-floor and the multiple reflections of both. The
source is placed high in the water, at depth z1, so that its first reflection
in the sea-surface is relatively close to it, and its first reflection in the sea-
floor is more remote. The pairing of reflections is then quite marked. If
the reader examines the paths taken by rays passing from the image sources
to the receiver location, each will be seen to pass through a certain number
of surface and bottom reflection layers. Passage through a surface reflection
layer (each is marked "S" on the diagram) will correspond to multiplication
of the acoustic pressure at the layer by R,, the surface reflection coefficient.
Similarly, passage through a bottom reflection layer (marked "B") will
cause the pressure to be multiplied by the bottom reflection coefficient Rb.

The reader should also note that reflection image pairs are identified by
number. That is, we identify pairs m = 1, 2, 3 ... above and below the true
water layer. The "m = 1" pair "above" actually includes the source itself
and is thus partially within the true water layer, of course.

Notice also, that we denote the ray-length, in a given ray pair, for that ray
which is closer to the horizontal axis of the true water layer, as rm1 and the
ray-length corresponding to that which is further away, as rm2. Finally, we
prime and double-prime the r' s to denote ray-pairs which start, respectively,
"above" or "below" the true water channel. There are thus, for any one value
of m, four ray-lengths to consider in estimating spreading loss and transport
delay between the image sources and the receiver location: r'm1 and r'm2 as
well as r"m1 and r"m2·

If we inspect the right-hand diagram, which shows the m 'th ray pair below
the true water channel, we observe that
r~l = (((2mh -~) -z1)2 + R2)1/2

r~ = (((2nit-~) + ~) 2 + R) 112
If we examine the m'th ray pair above the true water channel, we find that

= (((2mh + ~) -z/ + R2) 112

= (((2mh + z2 ) + z1) 2 + R2) 1/2
68 Underwater Acoustic Systems

r
B
h

s ~

T
s

Figure 4.15 The geometry of multiple reflection in an isospeed

shallow sea

Next, we examine the number of surface and sea-floor bounces, for each
m'th ray (of four). For the ray r'mt' being the lower m'th ray above the true
channel, we develop a bounce sequence by examining the left-hand part
of the figure, which is: 0 (direct path, m = 1); S,B (for m = 2); S,B,S,B
(for m = 3); .... This observation we generalise, so that we may write, for
the lower m'th ray above the true channel, that the overall reflection loss
is Jl'mi' given by

where e· ml is the angle from the lower m'th ray to the normal to the reflecting
surfaces. Proceeding in like manner for the other three m 'th rays, we find
for the upper m 'th ray above the channel, an overall reflection coefficient,
J.1'm2, given by

rn-lre· )
J..l, ~ rn2 llsrn(e'm2)
 Ray Trace Modelling of Sonar Propagation 69

For the two rays below the channel, there will be overall reflection coefficients,
~"m1 and ~"m2• given by

m~e"
~b\mV
\ ',m-l(f1' )
1\ ml
and

respectively. We are now in a position to express the sound pressure field

at the receiver location. We write, for the source signal, that it should be
(the real part of) the unit phasor exp(jrot). The received pressure field will
then be (the real part of)

Figure 4.16 Location of source image-pairs on a circle with its centre

at the intersection point between a sloping sea-floor and the
sea-surface
70 Underwater Acoustic Systems

Here we see the four m'th paths combining additively, with inverse square
(power) spreading and thus inverse (first-order) pressure decrease with
range. We see, also, the effect of loss caused by multiple bounces from the
sea-surface and sea-bed. Finally, in the argument of the "exp" terms contained
within the summation, we see the effect of transport delay between image
sources and receiver. Numerically, the evaluation of this equation is
straightforward. Such is the power of the modern computing workstation
that the direct application of this result can yield useful insight into propagation
in the shallow seas. Because of the isovelocity propagation conditions, we
note that calculation of the pressure field at a point remote from the source
can be achieved without recourse to conventional ray-trace methods. In this
respect, the method anticipates some of the computational economies which
result from the application of normal mode theory, which we study in
Chapter 5.

The method discussed above may be extended to include the situation in

which the sea-floor is sloping. Clearly, this circumstance is of interest in
shelf-sea and continental-slope working. In this event, the image sources
form themselves about a circular locus, whose centre is the (perhaps
hypothetical) point of intersection of the sea-floor and sea-surface, figure
4.16. For further details, the reader is referred to [4.7].

4.6 The Lloyd Mirror Effect

One final subset of the modelling procedures discussed in the previous

section is worth alluding to. If sound reflects from the surface of the sea,
at such range and in such water depth, that no significant bottom reflections
are encountered, then the geometry illustrated in figure 4.17 will be
encountered.

If the signal transmitted is a unit sinusoid cos(rot), then at the receiver

location, both the source and a single image in the surface will be observed.
The received signal will thus be

where r2 =r1 + c't, where the negative sign between the terms indicates phase
inversion on reflection from the sea-surface, and where 't is the excess
propagation delay on the reflection path, by comparison with the direct path.
Expanding and manipulating, we find that the mean intensity at the receiver
is given by

I = 2r2 (1 - cos OO't)

 Ray Trace Modelling of Sonar Propagation 71

If next we turn our attention to the excess delay then, by inspecting figure
4.17, we see, by applying Pythagoras's Theorem to find the difference
between the direct and image path ranges and simplifying by means of the
Binomial Theorem, for horizontal displacements, r, significantly greater
than source and receiver depths, that 't "' 2z 1z.jrc. Consequently

r--------.;
''
------------------------------------------- '
~-------- -·

Zt+Z2

_......__t__J_
Figure 4.17 The Lloyd Mirror effect: geometry of a single
sea-surface reflection

This equation shows that the channel transfer function will exhibit constructive
and destructive interference which will decrease in spatial frequency with
increasing horizontal displacement, as figure 4.18 illustrates. This phenomenon
is known as the Lloyd Mirror effect and can sometimes be observed on
sidescan sonar records, because of a sea-bed, rather than a sea-surface image
interference. We shall also see, in Chapter 10, that it may prove deleterious
in the operation of some sub-sea acoustic communication systems.

Nonnalised range r/(41t z 1 z:z P.. )

0.1 1.0 10.0
---
·--
-----~ -2 ------~-~~-
---- __
------ '

- - - -=-"<,,,:~:<:--=~~:::~:::~:_

Figure 4.18 The Lloyd Mirror effect: attenuation as a function of

normalised range
72 Underwater Acoustic Systems

Note also that if r becomes very large, the argument of the cosine term will
become small. The cosine term may then be expanded as a power series,
of which only the first term will be significant: cos ~ = 1 + ~2/21 It follows
that I oc r-4; the intensity falls as the fourth power of range.

References
[4.1] R.I. Urick, Sound Propagation in the Sea, Peninsula Publishing, Los Altos, Calif.,
1982, pp. 3.1-3.8

[4.2] C.S. Clay and H. Medwin, Acoustical Oceanography, Wiley, New York, 1977

[4.3] Physics of Sound in the Sea, Vol. 1: Transmission, Reprinted and distributed by the
Research Analysis Group, US National Research Council

[4.4] I. Tolstoy and C.S. Clay, Ocean Acoustics, McGraw-Hill, New York, 1966, pp. 33-
36.

[4.5] C.B. Officer, Introduction to the Theory of Sound Transmission with Application to
the Ocean, McGraw-Hill, New York, 1958

[4.6] R.I. Urick, Sound Propagation in the Sea, Peninsula Publishing, Los Altos, Calif., 1982,
p.4.11

[4.7] J.D. Macpherson and M.J. Daintith, Practical Model of Shallow Water Acoustic
Propagation, J. Acoust. Soc. Am., Vol. 41, No. 4, 1967, pp. 850-854
5 Normal Mode Modelling of Sonar
Propagation
co-authored by P.A. Willison

5.1 Introduction

As we saw in the previous chapter, sound propagation in shallow water leads

us, via the ray tracing approach, to a model of propagation wherein there
will exist a multiplicity of reflected image sources. Given iso-speed conditions,
the ray paths from these sources will be straight lines. In water whose depth
is moderately shallow with respect to range, there may be sufficiently few
bounces for the problem of estimating the summed sound intensity developed
by each ray to be computationally viable. For channels which are extremely
long by comparison with water depth, the problem rapidly becomes intractable,
although with the increasing power of modern scientific workstations, this
difficulty is less significant than it once was.

Normal mode modelling provides an alternative to the ray trace approach.

It relies upon particular solutions to the wave equation and yields an
expression for the pressure field at a point remote from the sound source.
We shall discuss the interpretation and development of the normal mode
method in the context of an iso-speed channel over a horizontal reflecting
bottom. The reader is advised that the method may be extended in a number
of ways, to include propagation conditions which vary with depth and range,
as well as variable bottom geometry.

We begin by establishing an understanding of the nature of normal modes

in the sound channel. To emphasise an important point, imagine that a guitar
string is plucked. The string will vibrate, making a half-cycle sinusoidal
displacement in space, with nodal points at the bridge and nut, and a
maximum amplitude of displacement at its centre. This is the fundamental
mode of vibration of a stretched, thin string. A skilful guitarist can pluck
the string, whilst simultaneously damping very briefly the centre of the
string with the little finger of the right (plucking) hand. This will cause

73
74 Underwater Acoustic Systems

a node to form at the centre of the string. The string will then vibrate with
a full-cycle sinusoidal spatial displacement, with nodes at bridge, nut and
centre. The sound thus produced will be the first overtone, an octave up
on the fundamental frequency. Second and even third overtones can be
achieved by damping the string one third and one quarter of the way from
the bridge.

Clearly, the stretched string may be persuaded to enter any of a number

of vibrational states. These allowed states may be likened to normal modes
of vibration in a waveguide in underwater acoustics. The essential condition,
which dictates possible frequencies of vibration (apart that is, from distance
between the bridge and nut, string density and tension) is the requirement
of the existence of nodal points at the ends of the string. It is, quite simply,
impossible to conceive of a vibrational state wherein the ends of a thin string
are flapping up and down, and the nodal points are somewhere other than
at the ends of the string. Such a situation makes no physical sense.

5.2 A Heuristic Treatment of Normal Modes in an Acoustic

Waveguide

In figure 5.1, we imagine a "billiard ball" of high-pressure (the isolated

dark "blob") to be incident upon the surface of the sea. Like a real billiard
ball, its horizontal component of velocity will remain unchanged by the
impact but its vertical component will be reversed, so that it will appear
to "reflect" from the water surface. Curiously, another effect will also occur.
On reflection the "billiard ball" of high-pressure will become one of low-
pressure, relative to ambient (or atmospheric) pressure. This is because,
being a "free surface", the pressure at the sea-surface must remain at
ambient. Were the wave ipcident on a rigid surface which could sustain
compression of the adjacent water, then the incident high-pressure "billiard
ball" would reflect as a high-pressure "billiard ball". This effect occurs on
reflection from the sea-floor.

In fact, the isolated "billiard ball" of high pressure has no physical parallel;
fluid flow outwards from it would dissipate its pressure towards the ambient.
However, a plane wavefront, such as is also shown on figure 5.1, may be
likened, at least in horizontal section, to a row of "billiard balls" of high
pressure. As each "ball" reaches the free surface it will reflect and pressure-
invert, to form a row of low pressure "balls". That is, our high-pressure
(relative to ambient) plane wavefront will reflect to become a low pressure
(relative to ambient) plane wavefront. Both before and after reflection the
wavefront will be travelling obliquely to the sea-surface, at the speed of
sound in the sea.
 Normal Mode Modelling of Sonar Propagation 75

Figure 5.1 Illustrating the reflection of a section of plane wavefront

from the sea-surface

In figure 5.2, we see the high-pressure incident plane wave (the darker band
moving upwards towards the surface, and upon contact, developing the low-
pressure reflected wave). Although the direction of travel is, in both cases,
normal to the wavefront, the appearance within the frame of the picture
is of an inverted "vee" running (in this example) rightwards. The apparent
horizontal speed of the "vee", if the angle of incidence of the ray to the
normal to the sea-surface is 9, will be c sin e.

Figure 5.2 The plane wavefront, in section, moving rightwards and

continuously reflecting from the sea-surface
76 Underwater Acoustic Systems

Next, in figure 5.3(a), we imagine not a single ("shock-wave") plane-

wavefront but a periodically excited plane-wave pressure field. Now the
inverted "vees" must have a separation (normal to the wavefronts themselves)
equal to one wavelength at the excitation frequency. In this figure, the bold
lines represent high-pressure and the dashed lines low pressure, being
presumed to be the maximum and minimum pressure values for a sinusoidal
spatial pressure variation.

(a)

(b)

Figure 5.3 The periodic excitation of the channel, forming a

rightwards moving pressure pattern

Clearly, the pressure maxima will reinforce where two bold lines cross, to
produce a region of even higher pressure. A similar but reversed effect,
producing increased rarefaction, will occur where two dashed lines cross.
Where dashed and bold lines cross, the pressures will cancel to ambient.
These effects are shown as a shaded pressure density map in figure 5.3(b).
The entire pressure pattern will appear to propagate rightwards with speed
c sine, withe again measured between the normal to the wavefront and the
normal to the sea-surface.
 Normal Mode Modelling of Sonar Propagation 77

Now, like the guitar string which cannot have free, flapping ends, the
pressure field, if it is constrained also by a rigid lower boundary, can only
exist at a given frequency and angle of incidence if water depth is such
as to engender the correct bottom reflection conditions so that phase reversal
does not occur. Thus figure 5.4 shows the allowable depths which satisfy
this constraint. Also superimposed upon this picture are sketches showing
the variation of pressure amplitude with depth. This latter function will vary
sinusoidally with time, at the excitation frequency.

Figure 5.4 The allowable normal modes for channels of varying

water depth excited by a harmonic source of fixed frequency.
In the inset figures, pressure is plotted horizontally, as
a function of depth

Of course, we cannot choose channel depth, h, to suit our own convenience,

so in practice we find that an appropriate normal mode of vibration can
develop, at a given drive frequency, only at a particular angle of incidence.
Reference to figure 5.5 shows that this angle, e, is given by

e = cos- (A/4h)
1

for the first mode and, by extension, is given for higher modes by the
expression

en = cos-1((A./2h) (n - 1/2)); n = 1, 2, 3 .....

This equation, which is known as the characteristic equation, is a variant

of the Bragg equation which governs the interference pattern of a diffraction
grating. Weston [5.1] based a series of ripple tank experiments on this
relation. We note, once more, that once the channel depth and the transmission
frequency are fixed, only certain angles of incidence are allowed for
propagation along the waveguide.
78 Underwater Acoustic Systems

~ 2h/tan 9 = ( ).J2)/sin 9 _.. j

I :

Direction of propagation

Figure 5.5 Illustrating the derivation of the inter-relationship

between angle of incidence, 8, wavelength, A. and water depth, h,
for the first normal mode

We also note that, for each mode, because of the orientation of the wavefronts
to the channel boundaries, the speed of pressure interference pattern
propagation down the channel, which is known as the group velocity, will
be given for each mode by the expression

If en tends to 90•, the plane wavefronts approach vertical alignment and

pass down the channel at a speed close to the speed of sound in free space.
This condition is approached only as the excitation frequency becomes very
high.

When the excitation frequency for the first mode falls to such a value that
the channel depth is spanned by one quarter wavelength, we encounter a
situation where the plane wavefronts form only standing waves in the
vertical, reflecting continuously backwards and forwards between the sea-
surface and sea-floor. Then the group velocity falls to zero, en ~ o·, and
no forward propagation can occur.

More generally, this effect will occur when the excitation wavelength for
the n'th mode has fallen to a value satisfied by the relationship

h = (A..J2) (n - 1/2);

that is, when the frequency f"" = c/A.cn is given by

fen = (c/2h) (n - 1/2)

 Normal Mode Modelling of Sonar Propagation 79

C-

0+---~-----+------+-----~------~------~----_..
0 f

Figure 5.6 The variability of group velocity with frequency

This frequency is referred to as the cutoff frequency of the n'th mode. For
frequencies above the cutoff, propagation can occur and will take place at
an incidence angle to the surface normal of en and group velocity un. Since
the geometry of figure 5.5 allows us to express en in terms of the ratio of
wavelengths at excitation frequency and cutoff frequency, A and "-en
respectively, as

J..? -1.?
en
=----
)._2
en

we may derive a new expression for group velocity in terms of mode

excitation and cutoff frequencies which is

un = c[l - (fc jf)2]-112

This expression is shown plotted in figure 5.6. If the excitation frequency

were, say, half-way between the fourth and fifth mode cutoff frequencies,
we should expect to experience the lowest four modes propagating in the
duct. Whilst this would be true if the sea-surface and sea-floor were perfect
reflectors, practical conditions will tend to reduce the number of modes
which can propagate effectively.

We thus anticipate a situation in which our harmonic source will excite a

certain number of modes within the acoustic waveguide. These several
modes will simultaneously co-exist. They will propagate, each at its own
group velocity, thus establishing a pressure at some remote point which will
80 Underwater Acoustic Systems

be the superposition of the pressures at that point due to each mode

considered separately.

It is also worth noting that the depth of the source will have an important
effect in establishing the strength, or even the existence, of a given mode.
For example, if the source depth is such as to locate the source at a nodal
point, for which for a given mode p(z) =0, then that mode cannot be excited.
For maximum excitation, the source should be at a depth corresponding to
one of the anti-nodes. The inset curves in figure 5.4 show the nodal and
antinodal points. The former correspond to zero-crossings on the depth axis
of the pressure versus depth curves. The latter correspond to pressure
maxima and minima.

5.3 Normal Mode Solution for Long Ranges

The reason for developing the normal mode solution was that the method
of images involves, in principle at least, an infinite summation of rays.
Because of this, application ofthe method, particularly in an inhomogeneous
medium and at long ranges, can be computationally cumbersome. We have
yet to show that the mode solution improves on the image approach. Indeed,
we have deduced that, to describe accurately the pressure field at a distant
point, we require an infinite sum of modes so that, at first sight, matters
might not be thought to have improved. We will now show that, assuming
that we are interested in long range propagation, the pressure at a point
can be found by the summation of a finite number of normal modes.

The model to be employed retains, for simplicity only, the homogeneous

water layer and free surface condition, as in section 5.2, but now employs
a "non-rigid" bottom. If a wave is incident on the bottom boundary with
an angle less than some critical angle, the wave is partially reflected back
into the water layer and partially transmitted into the lower layer. The
critical angle is given by

where c and C10d are the sound speeds in the water and in the lower,
sedimentary sea-floor respectively. We assume that, for angles of incidence
greater than the critical angle the wave undergoes total internal reflection
and no energy is lost to the lower layer. This is an idealisation of the
reflection characteristic for the "fast bottom" which is illustrated in figure
1.6. To avoid the problem of energy being transmitted back into the surface
layer we assume that the lower layer is infinitely thick.
 Normal Mode Modelling of Sonar Propagation 81

The modes for which (because of angle of incidence) energy is partially

transmitted into the lower layer will be quickly attenuated as a result of
this energy leakage, and thus will have no significant bearing on the acoustic
pressure field at long ranges. To calculate the pressure at long range we
need only be interested in the modes that represent the guided or trapped
waves.

To determine the highest mode that we need to consider, we rearrange the

characteristic equation to give

n = (2hf/c) cos ec + 1/2

where ec is the critical angle measured from the normal. The highest mode
of interest, m, is then given by rounding n to the next lower integer value.
To calculate the pressure at a distant point we therefore sum from 1 to m,
assuming the transmit frequency is higher than the cutoff frequency of the
first mode. The attraction of the Normal Mode method is now readily
apparent. The solution of the long range propagation problem is simply
given by summing a finite, often small, number of modes.

Let us consider an example. Assume a channel depth of 7.5 m. The channel

cutoff will then be 50 Hz for the first mode, 150Hz for the second, 250
Hz for the third, and so on. This calculation applies strictly to the channel
with a rigid bottom. Here, as figure 5.7(a) illustrates, we see the pressure
fluctuation amplitude (about ambient) decrease immediately to zero, below
the perfect reflecting rigid sea-floor. The modal patterns of pressure amplitude
with depth are as we have been led to expect from our previous discussions.

In contrast, if the sea-floor boundary is not perfectly reflecting, then we

should expect sound to penetrate and for there to be a reasonably rapid
decrease of amplitude with increasing depth into the sediment, because of
the high attenuation within the sediment. This situation is shown in figure
5.7(b). The pressure amplitude will be the same in the sea just above the
sediment as it will be just within the sea-floor. This reflects a requirement
for continuity of pressure at this interface. If there is still some sensible
difference between the density of the sea-water and the density of the
sediment then, although an "ideal" waveguide will not exist, and the sub-
bottom half-space will undoubtedly influence the modal cutoff frequencies,
there will yet remain at least a loose correspondence between the cutoff
frequencies for the real and for the ideal channel. For a fuller treatment
of this topic, the reader is referred to any of the advanced texts which
specifically treat sound propagation in the sea, of which several are listed
at the end of this chapter [5.4 - 5.14].
82 Underwater Acoustic Systems

For our tutorial purposes, we accept that the mode cutoff frequencies, as
indicated by considering behaviour with a rigid bottom, are at least adequate
in predicting the approximate maximum number of modes which could
propagate, at a given excitation frequency.

(o) (b)

Figure 5 .7 Showing pressure variation with depth for the first two
modal states: (a) for a rigid sea-floor and (b) for a sea-floor, such
as an unconsolidated sediment, which is not perfectly reflecting and
into which sound may penetrate

For example, suppose we attempted to excite the channel with a 2000 Hz

source. Then we should expect, with a perfectly rigid bottom, that some
twenty modes would propagate, albeit with decreasing sound speed as mode
number increased. We now need to estimate how many of the higher of those
modes might safely be ignored, by virtue of a systematic loss of energy
into the sediment, for incidence angles greater than critical.

Suppose for example, that sound speed in the water layer is taken as 1500
m/s and in the sediment layer is arbitrarily chosen as 1600 m/s.This is the
condition of the so-called "fast bottom" which might well correspond to
our unconsolidated sea-bed composed of fine sand. We can now calculate
the critical angle, e. as

sin-1(1500/1600) = 70°; cose. = 0.34

The number of modes of interest is then

m=next integer value below (((2 x 7.5 x 2000 x 0.34)/1500) + 1/2}=7

We thus need only calculate the pressure field by summing the effects of
the first seven modes, rather than all of the twenty which could in principle
propagate with greater than zero group velocity, since energy leakage into
the sea-bed will mean that the thirteen highest modes will not propagate
to long ranges at 2000 Hz.
 Normal Mode Modelling of Sonar Propagation 83

The advantage of the normal mode solution is now obvious, with the
reduction of an infinite summation to that of a relatively few terms. It should
be stressed that this is only an approximate solution and is only valid for
ranges in excess of a few water depths. At short ranges the influence of
the rapidly attenuated modes must be included. Additionally, we should
include the effect of a non-rigid bottom more comprehensively than is
possible here. These situations make for a far more complicated analysis
and the necessary extensions are reviewed briefly in section 5.6.

Another factor to be borne in mind when considering propagation to long

ranges is that the channel will behave in a markedly dispersive manner,
because of the variability of group delay with frequency. Sound from an
explosive source, such as might be used in marine seismic survey, will
contain a rich distribution of spectral components in the audio range
spanning, in the main, the 200Hz to 2kHz regime used as an example in
the previous paragraph.

Heard close-to, the detonation will sound like a deeply pitched "crunch".
When detected remotely in water which, near the coast, may have shallowed
over many hundreds of metres to a depth of only a few metres, the detonation
will sound remarkably like a down-chirping sinusoidal sonar pulse of a
second or so duration, with a start-frequency of about 2kHz and an end-
frequency in the low hundreds of Hz. The reason why this is so is because,
at the higher frequencies, the group velocity of most of the (9 or so, by
our previous calculation) normal modes will be clustered towards the
(asymptotic) free-field sound speed, c. The lower frequencies will sustain
fewer and fewer modes and will exhibit group delay values significantly
less than the free-field value. The higher frequency components will thus
arrive first and with lower loss, producing the down-chirp effect.

5.4 Normal Modes as Interfering Plane Waves

To give a more mathematical interpretation of the propagation of sound as

normal modes, we shall show that the modes can be generated by the
summation of up-going and down-going plane waves. We will then be able
to show the connection with the normal modes and the method of images
discussed earlier. We can represent an up-going and a down-going plane
wave by the first and second terms, respectively, in the expression

p = A1 exp(j(rot - 10' - yz)) + A,. exp(j(rot - 10' + yz))

84 Underwater Acoustic Systems

where K is the horizontal component and y the vertical component of wave

number k = ro/c, measured along the propagation direction e, such that k2
= x:2 +f. At the surface, when z =0, the contributions from the plane waves
must sum to zero:

A1 exp(j(rot - x:r)) + A2 exp(j(rot - x:r)) =0

and it follows that A1 = -Az = A. At the bottom, when z = h, the pressure
must be allowed to be a maximum, so that dp/dz = 0.

Manipulating, we find that

dp/dz = 2jAy exp(j(rot- x:r)) cosyh =0

which can only occur when yh = n/2, 3n/2, 5n/2 ... or

'Y = }i(n
1t
-1/l); n = 1, 2, 3...

Now, since

we find, again, that

en = cos-1((A/2h) (n - 1/2)); n = 1, 2, 3 ...

As the frequency of the excitation is decreased, the angle of incidence

(between the surface or sea-floor and the normal to the wavefronts) becomes
larger. The frequency at which the angle is 90" defines, again, the mode
cutoff. When in this condition, we see once more that we have a purely
standing wave in the z direction (upwards and downwards) with no forward
propagation component. Consequently, it is obvious that for propagation
we require the transmission frequency to be greater than the cutoff frequency
of the first mode.

5.5 The Normal Mode Solution Formalised

We continue by retaining our assumption of the iso-speed channel model

with an ideal pressure release surface and a rigid bottom. At locations which
do not include any sources, the acoustic pressure field p(x,y,z,t) in a water
layer can be shown [5 .2], to obey the wave equation
2
v2p = _.!._a
ac
P
c2
 Normal Mode Modelling of Sonar Propagation 85

We shall show that the acoustic pressure field can be represented by a linear
superposition of travelling normal modes. In a physical sense, as we have
seen, the normal modes describe the way in which the fluid is vibrating,
akin to the vibrations of a taut string when plucked. The wave equation
is most easily solved by the method of separation of variables. For our model
we have the 2-dimensional wave equation

The pressure field extending from a source between plane, parallel surfaces
will exhibit circular symmetry in the horizontal. This suggests that it will
be mathematically more convenient to solve the equation in cylindrical co-
ordinates giving
2 2 2
ap +_!_~ +4=-A~
a? r ar az c at

To simplify the solution we assume that p(r ,z,t) has a time dependence of
the form expUwt). This allows us to re-write the wave equation in cylindrical
co-ordinates as

where k = ro/c. This result is known as the time-independent Helmholtz

equation. Using the method of separation of variables, we assume a solution
of the form
p(r,z) = R(r)Z(z)
Substitution gives
ZR"+_!_ZR'
r
+Z"R+k2RZ=O

Dividing by ZR and re-arranging, we find that

R" R' Z" 2

R + R =- Z -k
The left-hand side of this equation is dependent only upon the range, r,
whilst the right is dependent only upon the depth, z. This can only be the
case when both sides are equal to a constant, referred to as the "separation
constant". We write the separation constant as -'A.2 and thus obtain two
ordinary linear differential equations
86 Underwater Acoustic Systems

The general solution to the first of these equations is

Z(z) = A sin ('lfz) + Bcos ('lfZ)

where

[ ol 2 ]112
c - A.
"' = -:'2

We know that at z = 0 we must have Z(z) = 0, therefore B = 0. The second

boundary condition requires the pressure to be a maximum at z = h and
therefore
Isin ('lfh)l= 1

This requires that 'lfh = 7t/2, 37t/2, 57t/2, ... or that vn = (7t/h)(2n - 1);
n = l, 2, 3 ... vn is termed an eigenvalue and the solution

Za (z) = Au sin ('Jin z)

is an eigenfunction of the problem. We see that there is an infinity of

eigenvalues and thus infinitely many eigenfunctions as solutions to the
boundary value problem. By solving this equation as a Sturm-Liouville
problem, see [5.3], we see that the eigenfunctions so obtained have the
property that they form an orthogonal set with respect to some weight
function, that is
h

PJ Zm(z) Zn(z) dz = 3mn

0

where d is the Kronecker delta function. Thus any single eigenfunction or

any linear superposition of eigenfunctions forms a solution to the wave
equation. The eigenfunctions are known as the normal modes of propagation
in our perfect waveguide model. The range-dependent differential equation
derived above is a form of Bessel's equation, the method of solution of which
will depend upon the accuracy and range involved in a particular problem.
 Normal Mode Modelling of Sonar Propagation 87

Exploiting the orthogonality of the eigenfunctions, we may expand the

acoustic field as a sum of the normal modes to yield the pressure field
equation. A typical form establishing a particular solution to the wave
equation which defines pressure as a function of time, horizontal range and
depth, and vertical sound velocity profile is

where A is an amplitude factor, describing source level, the function Zn

defines the nature of allowed vertical standing waves and kn is the
horizontal component of wave number, k. The vertical component of wave
number, 'Yn• must be such as to ensure that a pressure node exists at
the surface and, if we assume a perfectly rigid reflecting bottom, a pressure
anti-node at the sea-bed. Note that the eigenfunctions essentially plot the
depth dependence of pressure. For given water depth and transmission
frequency, only a finite number of modes may exist. The sum of the allowed
modes, as determined by the defining equation given above, establishes the
pressure field. Time dependence, for a harmonic source, is determined by
the leading "exp" function. Range dependence is built into the denominator
function, and the contained "exp" function.

Local waveshape in the vicinity (r,z) may be obtained for non-harmonic

sources, by Fourier transforming p(r,z,t) to obtain an effective channel
transfer function P(r,z,f), multiplying by the source spectrum and re-
transforming. Such manipulation requires obvious care, considerable skill
in numerical analysis and computing machinery of some power. The
method is of importance, because the channel, at low frequencies and
shallow water depths, is a "dispersive" waveguide. That is, as with light
split by a glass prism into its spectral components, sounds of differing
frequencies travel with different speeds in the guide, a phenomenon not
observable in free-field propagation.

5.6 Normal Mode Solution for All Ranges

In section 5.3, we saw that the Normal Mode approach exploited the fact
that only those modes that represent the trapped energy are of any great
significance at long ranges and therefore the modes that represent the energy
which leaks out of the channel could be ignored. For an exact solution,
applicable to all ranges, we must include those modes that were ignored.
This is a more difficult problem, but one which has received a good deal
of interest over the last forty years. Pekeris [5.4] is generally regarded as
having pioneered the waveguide approach to the solution of acoustic
propagation in the shallow water channel.
88 Underwater Acoustic Systems

We have seen that the eigenvalues for the perfect waveguide are quantised.
The associated modes are called discrete modes. If we allow water depth,
h, to tend to infinity, then the difference between successive eigenvalues
will tend to zero. The eigenvalues will form a continuous set and the
eigenfunctions, the normal modes, are called continuous modes. For an in-
depth discussion of continuous and discrete modes the reader is referred
to Tolstoy and Clay [5.5]. It is generally accepted that at short ranges, where
continuous modes must be taken into account, the Ray Tracing approach
is more favourable.

5.7 The Horizontally Stratified Channel

We now wish to consider the solution ofthe more general problem in which
the sound speed and density are allowed to vary with depth. This problem
is usually solved by extending the model introduced in section 5.3 to allow
for many horizontal layers each with its own constant sound speed and
density, or density as a known function of depth. As the width of the
stratifications is reduced the model tends to that of a continuously horizontally
varying medium.

The shallow water channel that is of interest to us may not contain appreciable
differences in sound speed and density in the water layer. It is, however,
possible that the parameters of a number of the sub-bottom layers may be
known and can then be easily included in the model. Waveguide propagation
will only take place when there exists a layer with a sound speed minimum
relative to the other layers.
Early studies by Pekeris [5.4] and by Officer [5.6] dealt with a maximum
of three layers. Tolstoy extended the theory to many layers using physical
and geometric reasoning [5.7, 5.8] to generate the characteristic equation
for the normal modes. Other authors, such as Budden [5.9], Brekhovskikh
and Lysanov [5.10], Brekhovskikh [5.11] and Clay and Medwin [5.12] have
employed similar approaches. More recent mathematical treatments, such
as those given by Stickler [5.13] and by Boyles [5.14], all solve the wave
equation with varying degrees of accuracy.
 Normal Mode Modelling of Sonar Propagation 89

References

[5.1] D.E. Weston, A Moire Fringe Analog of Sound Propagation in Shallow Water, J. Acoust.
Soc. Am., Vol. 32, No. 6, 1960, pp. 647-654

[5.2] L.E. Kinsler, A.R. Frey, A.B. Coppens and J.V. Sanders, Fundamentals of Acoustics,
Wiley, 3rd edition, New York, 1982

[5.3] E. Kreyszig, Advanced Engineering Mathematics, Wiley, 5th edition, New York,l983

[5.4] C.L. Pekeris, Theory of Propagation of Explosive Sound in Shallow Water, in Propagation
of Sound in the Oceans, Geol. Soc. Am. Memoir 27, 1948

[5.5] I. Tolstoy and C.S. Clay, Ocean Acoustics, McGraw-Hill, New York, 1966, pp. 33-
36

[5.6] C.B. Officer, Introduction to the Theory of Sound Transmission with Application to
the Ocean, McGraw-Hill, New York, 1958

[5.7] I. Tolstoy, Shallow Water Test of the Theory of Layered Wave Guides, J. Acoust. Soc.
Am., Vol. 30, No. 4, 1958, pp. 348-361

[5.8] I. Tolstoy, Note on the Propagation of Normal Modes in Inhomogeneous Media, J.

Acoust. Soc. Am., Vol. 27, No. 2, 1955, pp. 274-277

[5.9] K.G. Budden, The Wave-Guide Mode Theory of Wave Propagation, Logos, London,
1961

[5.10] L. Brekhovskikh andY. Lysanov, Fundamentals of Ocean Acoustics, Springer-Verlag,

Berlin, 1982

[5.11] L.M. Brekhovskikh, Waves In Layered Media (R.T. Beyer, transl.), 2nd edition,
Academic Press, London, 1980

[5.12] C.S. Clay and H. Medwin, Acoustical Oceanography, Wiley, New York, 1977

[5.13] D.C. Stickler, Normal-mode Program with Both the Discrete and Branch Line
Contributions, J. Acoust. Soc. Am., Vol. 57, No. 4, 1975, pp. 856-861

[5.14] C.A. Boyles, Acoustic Waveguides, Wiley-lnterscience, New York, 1984

6 Noise and Reverberation

6.1 Introduction

The study of acoustic noise and the reverberation of acoustic signals is of

importance because one or other of these phenomena will set the limit to
sonar system performance. Whilst it is true that both sources of corruption
can co-exist, it is most common to find that one of the two will predominate.
Notice that in a noise-limited sonar, increasing the signal power will
improve the signal-to-noise ratio and hence the system performance. The
same need not be true for a reverberation limited sonar, since the reverberation
is, itself, directly a function of the output signal level.

Noise in the sea may derive from many sources: seismic events, wave action,
shipping, thermal agitation, rainfall, sounds made by marine animals and
so on. The ambient noise field is considered isotropic, in the sense that,
wherever a sensing hydrophone is placed in the sea, the observed intensity
of the noise will remain the same - subject of course to a similarity of
relevant environmental conditions. This does not mean, however, that the
noise is not directional. It is easily possible to envisage a directional
hydrophone sensing a higher angular intensity density of noise, when
viewing along the axis of a sound channel such as would form at the base
of the main thermocline (see section 6.3, for example), than it would when
looking vertically upwards or downwards.

Many noise sources would present a continuous spectrum together with

gaussian statistics. This would certainly be true of thermal noise at high
frequencies but would also hold for wave and rain-induced noise. By
contrast, shipping noise would present a mixed spectrum with both continuous
and discrete components, the former deriving largely from the bow-wave
and propeller cavitation, the latter from machinery noise. Whilst the bow-
wave might well present gaussian and stationary statistics, it is quite likely
that propeller cavitation, although also a noise process, would exhibit short-
term non-stationary behaviour producing a measurable pulsing of acoustic
intensity at a frequency proportional to the blade-rate.

90
 Noise and Reverberation 91

6.2 Deep Sea Ambient Noise Level [6.1]

In using the term "ambient", we discount such local effects as the "self-
noise" of a moving vessel which might be carrying a sonar. Self-noise is
caused by the passage of the vessel through the water and by vibrations
induced by its machinery and propellers. Note that we shall not need to
correct ambient noise level for range in sonar calculations because it is an
all-pervading quantity, much as heat in a greenhouse, though having the
sun as a "localised" source, is all-pervading.

/ Turbulence Noise
1

::c
N

~ 60 Surface Agitation
~
,....; 50
w=15
w=10
e w=5

l
I:El 40 w=O
"0

z~ 30

2
Thermal
Noise
10

0
0.001 O.Ql 0.1 1.0 10.0 100.0 1000.0
Frequency - kHz

Figure 6.1. Deep-water ambient noise spectrum level

We commence by examining, figure 6.1, the gross spectral behaviour of

acoustic noise observed in the open sea. The lowest frequency decade,
between 1 Hz and 10Hz, is dominated by noise originating from oceanic
turbulence. Over the next frequency decade, shipping is the dominant cause.
Noise in the next three decades, between 100Hz and 100kHz, is primarily
caused by surface agitation. Finally, at frequencies in excess of about 100
kHz, it is thermal noise, originating in molecular motion in the sea which
92 Underwater Acoustic Systems

is the principal effect. The overall noise spectrum level, NSL, or the
intensity attributable to a spectrum measurement bandwidth of 1 Hz, is the
power sum of the noise spectrum levels NSLl to NSL4 attributable to these
four predominant sources of noise. We may empirically calculate the
individual noise spectrum levels as

NSLl = 17 - 30(log f) Turbulence Noise

NSL 2 = 40 + 20(0 - 0.5) + 26(log f) - 60(log (f + 0.03))

Shipping Noise

NSL3 = 50 + 7.5 w 112 + 20(logf) - 40(log(f + 0.4))

Surface Agitation Noise

NSL4 = -15 + 20(log f) Thermal Noise

Table 6.1
Nautical Beaufort Wind Scale

Beaufort Wind l!IDd Sas:s:d Description of Sea Surfac1 Sea M~ID !!1!1 b&
Number Name knoll m/s state ft m

0 calm <1 <0.5 sea minor-like 0 0 0

1 light air 1-3 0.5-2 scale-like ripples, no 0 0 0
foam-crests
2 light breeze 4-6 2-3 small wavelets, crests 1 0-1 0-0.3
glassy but not breaking
3 gentle breeze 7-10 3-5 large wavelets, crests 2 1-2 0.3-0.6
begin to break
4 moderate 11-16 5-8 small wavea, fairly 3 2-4 0.6-1.2
breeze frequent white horses
5 fresh breeze 17-21 8-11 moderate waves, many 4 4-8 1.2-2.4
white horses
6 strong breeze 22-27 11-13 large waves, white foam 5 8-13 2.4-4
crests, some spray
7 moderate gale 28-33 14-16 sea heaping; foam begins 6 13-20 4-6
to be blown in streaks
8 fresh gale 34-40 17-20 moderately high wavea, 6 13-20 4-6
marked foam streaking
9 strong gale 41-47 21-24 high wavea, rolling sea, 6 13-20 4-6
spray reduces visibility
10 whole gale 48-55 24-27 v. high wavea, overhanging 7 20-30 6-9
crests, sea white
11 storm 56-65 28-33 except. high waves, ships 8 30-45 9-14
lost to visibility
12+ humcane 65+ 34+ air filled wlth foam; 9 45+ 14+
visibility v. poor

where f is frequency in kHz, Dis shipping density on a scale 0 (very light)

to 1 (heavy) and w is windspeed, in metres per second (1 m s-1 ... 2 knots
 Noise and Reverberation 93

= 2 mph). Table 6.1 summarises the nautical Beaufort scale of wind. We

then calculate the overall noise spectrum level as

Finally, to allow for an actual transmission bandwidth of B Hz, we calculate

the noise level as
NL = NSL + lOlog B

6.3 The Variability of Ambient Noise with Time

In contemplating the general utility of the curves presented in figure 6.1,

we are drawn immediately to question the stationarity of the various
causative processes. Clearly, over some identifiable time-scale, all the
underlying causative processes, except for thermal noise, must be considered
exceedingly variable. It is, in general, considered that noise due to shipping
exhibits the shortest-term variability. Wind-induced noise, by contrast,
exhibits statistics which vary relatively slowly because ofthe inertia imparted
by the time required to build up, or dissipate, a "fully-arisen" sea. The
following broad generalisations are worthy of note.

1. The ambient noise spectrum has not, thus far, been observed to yield
evidence of seasonal variability.

2. Short-term variability of wind-induced noise has been observed to follow

the loose pattern indicated by figure 6.2. Typically the noise samples used
to estimate standard deviation, as depicted here, were of the order of 100
seconds' duration and were taken at intervals of one hour [6.2].

3. In shallow water, ambient noise exhibits wider, more rapid variability,

particularly in the vicinity of harbours and shipping lanes. This variability
also exhibits longer period, diurnal fluctuations because of the reduction
in traffic during the night.

4. Coastal locations where breaking surf is present also exhibit a significant

increase of, perhaps, some 10 dB over deeper water predictions. In the main,
however, the general form of the spectrum noise level curves mirror those
shown in figure 6.1.

5. In all these various aspects, available data concerning variability is poorly

co-ordinated and often inadequately documented in so far as precise description
of experimental method is concerned. Appropriate statistical descriptors for
both magnitude variability and duration remain yet to be properly developed.
94 Underwater Acoustic Systems

Vertical width measures standard

+
~
90 deviation of NSL
~
8. 80
&!
.....::1.. 70
~
IXl
"0 60
...l
z
tl)
50

40

30
20
0.001 0.01 0.1 1.0 10.0 100.0 1000.0
Frequency - kHz

Figure 6.2. The gross short-term temporal variability of noise

spectrum level

6.4 The Variability of Ambient Noise Level with Depth

In general, ambient noise level decreases with increasing depth . This is

because the noise is generated at the sea-surface by wind or shipping, and
because, as depth increases, so also does range-dependent attenuation
caused by sound absorption. However at frequencies in the band between
10 and 100 Hz, where shipping noise may predominate, there is in deep
ocean water only a small decrease with depth, because of duct-like propagation
within the deep ocean sound channel.

The deep ocean sound channel has its origin in the nature of the deep ocean
sound speed profile, which we examined in section 4.3 . Figure 6.3 shows
a simplified sound speed profile, together with a ray launched at that depth,
zl' at which the sound speed is a minimum. The launch angle, el' has been
so chosen that the ray just grazes the surface. Were the angle (upwards,
towards the vertical) only marginally smaller, we should expect reflection,
rather than grazing, at the ocean-surface, and thus (relatively) high loss.
Note that the characteristic ray parameter, a, is given by sin 9/c(z 1). The
 Noise and Reverberation 95

same ray, passing below the depth z1 will curve circularly and become
horizontal at a new depth z2 , for which - because of Snell's law - c(z 2)
must equal c(O). The ray launched at an angle 0 1 downwards will also behave
in similar manner, by virtue of the geometry of the ray construction. All
rays between these two extremes of angle will also snake forward along
the deep sound channel axis at depth z1 and all will exhibit only cylindrical
spreading loss.

The previous paragraph describes the way in which sound may propagate
along the deep sound channel if launched on the axis, at depth z1 • However,
surface generated noise can also pass into the deep sound channel and
become trapped there, propagating onwards with low loss. Only rays so
angled that they pass below the turning depth z2 will return to intercept the
ocean surface in a reflection, thus engendering loss. It is of course possible
for rays to become incident reflectively on the ocean floor where, again,
loss will occur. Because of this, it is observed that at depths down to z2 ,
which is referred to as the critical depth, negligible decrease in spectrum
noise level will occur. Below this depth a more rapid decrease will take
place, as figure 6.3 also illustrates.

NSL
OdD .........

-Sto -IOdB
IJPcall1

--------~- --- ~~~~~~~~

Figure 6.3. Sound trapping in the deep sound channel and the
quietening of ambient noise below the critical depth, z2

6.5 The Angular Distribution of the Ambient Noise Field

The sea-surface may be considered for most purposes to be the originating

region for ambient noise over the normal operating frequency regime of
the vast majority of sonar equipment. The sea-surface is frequently thought
of as consisting of densely packed dipole random sources, for each of which
will exist the characteristic dipole (power) radiation pattern

1(0) = 10 sin:ZO
96 Underwater Acoustic Systems

where 10 is the intensity encountered in a downward vertical direction. figure

6.4. Dipole radiation is considered more fully in section 8.2.

To determine the angular distribution of intensity in a deep ocean. we

consider the geometry illustrated in figure 6.5. In this figure. it is assumed
that no horizontal directionality will be encountered. so we seek to sum
contributions from circular annuli at the hydrophone location. We also
assume that. because the ocean is deep. reflections from the ocean floor
may be neglected. This is because attenuation will greatly reduce the noise
contribution induced by ocean floor reflection. by comparison with that
emanating from the ocean surface. at least for a shallowly immersed
hydrophone. The intensity at the hydrophone will then be

dl = 1(9) 21trl-2 dr
where r is the radius of the annulus and 1 the slant range to the hydrophone.
The equation thus expressed includes both the radiating area and the
spreading loss along the transmission path.

Reflective sea-surface

/ C?'
'

Intensity distribution
pattern

Figure 6.4. The angular intensity of a dipole source in the sea-surface

 Noise and Reverberation 97

Figure 6.5. View from below of a circular annulus in the sea-surface,

acting as a distributed noise source as sensed by a directional
hydrophone with an upwards viewing angle 9

The annular radius and slant range, and hence the incremental intensity dl,
may now be expressed in terms of hydrophone viewing angle e and hydrophone
depth zh. We then find that

We may recast this result to provide the intensity per unit solid angle 'I'
by writing

dl/d'lf = dl/(21t sine d9) = 10 cose

This remarkably simple result tells us that the intensity is greatest when
the hydrophone views the sea-surface directly from below. It further tells
us that viewed horizontally, we expect to encounter no ambient noise
contribution. Because the sea-floor was presumed infinitely far removed
from the surface and the hydrophone, no noise will be received from below
the vertical. Finally, we note that neither the pattern of directional characteristic
of the noise, nor its magnitude, varies with depth. We thus write, for the
noise intensity per unit solid angle

N(9) = 10 cose 0 ~ q ~ 1t/2

=0 1tl2 < e ~ 1t
This curve is illustrated in figure 6.6 and has been found to accord reasonably
well, at frequencies in excess of several hundreds of Hertz, with experimentally
acquired results.
98 Underwater Acoustic Systems

We turn next to the question of vertical directionality in shallow water. Here,

the problem is complicated by the possibility of a multiplicity of reflections
from the sea-floor and sea-surface. For the moment, we assume the sea-
surface to be an ideal reflector but the sea-floor we assume to be a lossy
reflector with a loss coefficient Jl(9) (see, for example, sections 1.10, 1.11
and 2.7). We may anticipate that any ambient noise incident within, say,
a cone of inspection of given solid angle, will derive from a surface "source
patch" such as "source patch 1",shown in figure 6. 7. However, the hydrophone
angle of acceptance will also sense noise generated from the further multiplicity
of source patches 2, 3 ... and so on. The folded cone of acceptance allows
us to write the summed intensity of all such source patches as

N(9) = 10 cos9(1 + 1.1(9) + Jl2(9) + ....)

Vertical
aspect

N(e) =cos e
~
Horizontal
aspect

Figure 6.6. The vertical directivity of ambient noise in a deep ocean,

where ocean-floor reflections are greatly attenuated by the relatively
large distance between the ocean surface and floor by comparison with
the distance between the surface and hydrophone
 Noise and Reverberation 99

source patch 4

Figure 6.7. The folded cone of inspection for a directional hydrophone

looking upwards towards the sea-surface in shallow water, where
multiply reflected surface noise patches contribute to the overall
acoustic intensity at the receiver location

which may be re-written as

Strictly, this result holds for upward-looking cones of inspection. If the cone
is directed downwards, the same calculation applies, but with an extra
bottom bounce, so that the computation of vertical noise directivity may
be re-written in the more complete form, as
o :::; e :::; 1t/2

N(9) = 10 cose Jl(9) ( 1 - Jl(9) }-1 ; 1t/2 < 9 :::; 1t

Vertical Vertical Vertical
aspect aspect aspect
20 2

Horizontal Horizontal
aspect aspect

Silt sea-floor Coarse sand sea-floor Soft sedimentary rock

Figure 6.8. Vertical directionality of ambient noise in a shallow sea,

calculated in decibels relative to dipole field source vertical
axis intensity 10
100 Underwater Acoustic Systems

Again we note that the vertical directionality is independent of depth.

Clearly, calculation of N(O) depends upon the model chosen for sea-floor
reflectivity. Chapman [6.3] has identified six typical sea-bed types and
computed both bottom reflection loss and relative noise intensity in decibels,
as a function of the direction angle 0 for the cone of inspection. His results
show considerable variability in directionality with sea-bed type. Figure
6.8 summarises his conclusions for three bottom types: silt, coarse sand
and soft sedimentary rock; these materials span Chapman's range of observable
ambient noise directionality characteristics.

6.6 Ship-generated Noise [6.4]

Whereas ambient noise has spectral characteristics which, over the operating
bandwidth of a typical sonar, might well be described as uniform or "white",
the same cannot necessarily be said of the acoustic emanations originating
in machinery. In particular, our concern will be with the various propulsion
machinery: propellers, motors, gearing and drive shafts associated with sea-
going vessels. Other mechanical noise sources exist, however, and due
attention may need to be addressed to them, in particular circumstances.
Examples might include the high intensity shock waves encountered during
piling operations for rig emplacement or other marine civil engineering
purposes, or the variety of clanks, bangs, squeals and rattles associated with
a wide range ofloosely maritime activities, particularly in harbour and near-
shore locations.

It must be stated that, whilst some general information concerning such

noise sources is to be found [6.4, 6.5], an extensive (or at least publically
accessible) database concerning source level, spectral content and spectral
stability is not at this time available.

The propellers of a ship serve to generate noise in several ways. Firstly,

propeller cavitation - a commonly encountered phenomenon manifest by
the entrainment of air-bubbles into the wake- is the major cause of noise
generated by surface ships. Propeller cavitation results from pressure
fluctuations in the water in the vicinity (usually) of the blade tips. When
these fluctuations fall below ambient pressure (effectively below atmospheric
pressure, for a surface ship) dissolved air is caused to leave solution,
forming the cavitation wake. Cavitation will exhibit a broadband, continuous
(noiselike) spectrum peaking in the high tens of Hertz and falling at some
20 dB/decade, with increasing frequency. The spectrum may, however, be
further confused by the presence of modulations at the propeller blade-rate.
The magnitude of the spectrum level will in general increase markedly with
 Noise and Reverberation 101

increased speed. The propeller cavitation noise spectrum level of a surface

ship, referred to 1 m standard range, may be estimated, in dB re 1 Jl.Pa per
Hz, as a function of speed, v, in knots, frequency, f, in kHz and displacement
tonnage T by using the empirical formula

SL = 10log(3.103 v6 Tf-2); f > 1 kHz

Secondly, a poorly designed propeller may exhibit "singing" when, typically,

one blade-edge enters a high-frequency state of vibration whilst running.
Such an effect may be most pronounced and will militate against quiet
running. The problem may also be extremely difficult to cure, without re-
design of the propeller itself. Finally, there may be expected a blade-rate
tonal, with a frequency which is the product of the shaft speed and number
of blades on the propeller itself. Virtually all modem merchant ships may
be expected to generate shaft frequency tonals in the region 1-3 Hz and
blade rate tonals with a fundamental in the range 6-10 Hz. The blade rate
line series is the dominant feature of the low-frequency spectrum of ship-
generated sounds. It does not follow that these tonals will increase in
frequency with increasing speed, since thrust is usually controlled by using
variable pitch propellers, with engine speed and thus shaft speed, maintained
at an economical but sensibly constant level.

Noise deriving from reciprocating piston-engines, rotating (turbine)

propulsion units and the various drive equipments associated with them is
characterised by both a line and a continuous spectral content. In the medium
and high-speed marine diesel engines which power many smaller vessels,
piston-slap is the major source of noise. Piston-slap is the impact of the
piston against the cylinder and is caused by the sideways movement induced
by direction changes of the piston during the firing cycle. The reader may
well be familiar with the observation that any diesel engine tends to run
substantially more noisily when first started. This effect is primarily caused
by excessive slap when the engine is cold; quietening occurs because the
running tolerances tighten as the engine block warms up. Piston slap may
occur several times in a single cylinder, during a single rotation of the
crankshaft, and will be occurring also on all other cylinders in the engine.
The result will be a rich harmonic line spectrum with a fundamental at the
crankshaft rotation frequency.

Large vessels which incorporate large, low speed (<250 rpm) diesel engines
will, however, not exhibit piston-slap as a significant noise-generation
mechanism. This is a consequence of the design of articulated connecting
rods, which virtually eliminate transverse motion of the piston.
102 Underwater Acoustic Systems

Gears are also important sources of noise. In this case the tones generated
will be at multiples of the tooth contact frequency and thus related to the
product of drive shaft speed and the number of teeth on the driving gear.
The drive shaft, by flexing or whipping during rotation, may also act as
a source of mechanical noise with a line component, as may the shaft
bearings themselves. Although ball-race bearings are in general noisier than
friction (block) bearings, it is only when such bearings are poorly installed
or approaching the end of their useful life that excessive noise will be
generated.

Taking all these various effects into account, figure 6.9 illustrates the
general form of the spectrum to be anticipated from a surface vessel,
excluding tonals at the lower frequencies. It should be noted that tonals
cannot be usefully incorporated on a graph such as this, since the vertical
axis measures a spectral density in units of dB re 1 JJ.Pa per Hz of measurement
analyser bandwidth. The tonals, being a spectral representation of a pressure
sinusoid, are measured in dB re 1 J.I.Pa. Thus typical tonals for a large
merchant vessel, the Chevron London, were identified at multiples of 6.8
Hz, with strongest components in the region 40-70Hz, which corresponded
to source levels (for each tone) of up to 190 dB re 1 J.I.Pa. The significance
of this distinction lies in, for example, the ability of spectrum analysis
equipments to discriminate tonals from cavitation noise. Taking the spectrum
level in the 10 to 100Hz decade, as indicated in figure 6.9 as being (about)
160 dB re 1JJ.Pa per Hz and guessing that we should need an analyser
bandwidth of significantly less than the tonal spacing of 6.8 Hz - say 2
Hz - then the noise level due to cavitation noise will be

NL = NSL + 101ogB
where B is the effective analyser bandwidth, so that NL = 160 + 10log2
= 163 dB re 1 J.I.Pa, which compared against the tonal strength of 190 dB
re 1 J.I.Pa indicates a signal-to-noise ratio of the order of 27 dB, which would
clearly facilitate line identification. Notice that averaging times in building
the spectral record would need to be significantly longer than the reciprocal
of the analyser bandwidth- which is 0.5 second. Averaging over at least
5 seconds would appear to be necessary. In practice, it is probable that a
digital (FFT) spectrum analysis on recorded time-series data would be used
to accomplish such an analysis (see Chapter 3).

Because of the need to acquire and record frequency components down to

1 Hz, digital or FM tape recording of frequencies from (notionally) zero
to perhaps 1 kHz would be taken on one channel of a multichannel recorder.
Analysis could later proceed on this part of the spectral record using a 512
point real FFT, with suitably selected overlap, window weighting and
 Noise and Reverberation 103

-- ... ------ .. ---- .. ---:--· ...

I
I

I
I
I
I '
: 20 dB/decade ' '
: roUoff ____.:,
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I

10 100 lk lOk lOOk

Frequency -Hz
Figure 6.9. The noise spectral density for a large (30,000 tonne)
merchant vessel

averaging, to yield the required 2Hz of analyser bandwidth. Simultaneously,

a full-width direct (analog) signal might be recorded on a second tape-
recorder channel, holding frequency components to some high tens of kHz,
but losing (because of the ac coupling of the direct recording format) the
lower hundred Hz or so of the incoming signal. The second channel could
then be analysed, but with coarser frequency resolution, a much higher
sampling rate and the presumption that the effective analyser filter bandwidth
was only some lOO's of Hz, rather than the 2Hz required to resolve the
lower frequency spectral lines.

The various modulations on a surface vessel's acoustic output make its noise
signature exceptionally distinctive, to the advantage of the experienced
passive sonar operator intent upon target classification and identification.
However, the reader should be aware that the measurement of spectral
characteristics is at best difficult, if only because of the degrading influence
of the acoustic channel separating a source vessel and a listener. The
problem is, of course, most severe in shallow water. It should also be noted
that, in such a context, "shallow" pertains to acoustic wavelength as much
as geographical circumstance. At the tonal frequencies (1-30 Hz)
corresponding to shaft and blade rate radiation, wavelengths are ofthe order
of 100-1000 m. Under such circumstances, in the shallow shelf-seas,
interference and waveguide effects will present experimental problems.
104 Underwater Acoustic Systems

6.7 Reverberation [6.6]

All sonars, whether active or passive, are subject to the corrupting influence
of noise. Active sonars give rise to another source of corruption known as
reverberation, which is innately associated with several interlinking physical
effects. These effects are:

1. Multipath propagation caused by boundary (sea-floor and sea-surface)

effects.

2. Multipath propagation caused by a possible multiplicity of refractive

("mainpath multipath") transmission paths.

3. Volume scattering caused by suspended reflective and diffractive objects

such as plankton and nekton.

4. Surface scattering, caused by sea-surface and sea-floor roughness or

entrained air bubbles in the immediate surface layer.

The relative significance of these various causative phenomena will depend

upon the circumstances under which an active sonar is required to operate.
Their gross effect will be to cause a time-spreading of received signal
energy. Furthermore, the reverberation intensity will increase, with increasing
transmitter power. Both noise and reverberation can act to obscure a
received signal. However, it is usually the case that one corruptive process
will dominate. The sonar will tend to be either noise-limited or reverberation-
limited. If the former is the case, then increasing the signal power will have
the effect of improving the signal-to-noise ratio. If the sonar is reverberation

Sea-surface

- - - - - -r
150m
omnidirectional
hydrophone

llbTNT
detonation
l 300m
~: ~-----
c = 1500 m s-l

lOOOm

Deep scattering
layer

Figure 6.10. An experiment demonstrating the properties of surface

and volume reverberation
 Noise and Reverberation 105

limited, then no advantage will accrue from adopting such a strategy. In

the main, the corruption induced by volume and surface scattering will,
because of the large number of scattering entities, be largely incoherent.
The same may well be true of"mainpath multi path". However, reverberation
caused by multiple specular reflections from the sea-surface and sea-floor
will cause attenuated and delayed replicas of the transmitted signal to be
returned to the receiver.

To illustrate some of these various phenomena, consider the scenario

depicted in figure 6.10. Here, we imagine a 1 lb charge of TNT to be
detonated at a depth of 450 m, and thus 300m vertically below a hydrophone
at a depth of 150 m. The water depth is presumed to be (about) 1000 m.

Some 200 ms after detonation, a first pressure shock-wave will arrive at

the hydrophone. This wave will have the temporal characteristics of an
explosive detonation, exhibiting a rapid rise to a peak pressure of the order
of 230 dB re 1 JJ.Pa, followed by a slower, noisy, envelope decay of intensity
caused by bubble formation and collapse. The time-constant of this decay
will be about 200 ~s for an explosive charge of the size envisaged here.
The next event experienced by the hydrophone is the passage of the surface
reflection and the associated surface reverberation. The specular surface
reflection will dominate, in magnitude, at the outset. The reverberation will
persist longer, however, than the nominally three time-constants taken for
the specular component to decay to a negligibly small level. Notice that,
as the reflection passes the hydrophone, the upper rim of the reflected
spherical region of high pressure will be travelling circularly outwards
within the surface layer of the sea, inducing a reverberant return which will
be substantially omni-directional, if not exactly isotropic - or of equal
intensity in all directions- as figure 6.11 suggests. We shall thus expect
the hydrophone to respond first to the reflected pressure-wave, then to the
reverberant surface scattering return.

p(t)

Figure 6.11 . Tlze expanding high-pressure shell following a

submarine detonation
106 Underwater Acoustic Systems

Some 200 ms after the detonation, the expanding pressure shell will intersect
with the deep scattering layer. This layer contains a profusion of planktonic
animals and, feeding upon them, a range of nekton often particularly
characterised by species of squid. In any event, a strong volume reverberation
is to be anticipated, arriving at the hydrophone (as a backscatter signal)
about 600 ms after the detonation.

The shock-wave will progress outwards until it reflects from the ocean-
floor. A specular reflection will follow, which will initiate a response from
the hydrophone approximately one second after detonation. Associated with
this reflection, but following it, will be a further bottom reverberation. Yet
further multiple surface and bottom reflections, all with associated volume
and surface reverberation, may be expected, decaying - of course - as
thunder from a lightning-strike rolls away between the hills surrounding
a valley. All these various events are depicted in figure 6.12, which provides
an indication of both the time-scale and magnitude of the various events
contributing to produce a reverberant response to a pulse of high energy
but short duration.

6.8 Scattering [6.7]

Scattering, as we have seen, takes place in such a manner that sound,

incident upon a scattering surface or within a scattering volume, is re-
radiated in directions other than that which would correspond to specular
reflection. Because active sonars are often monostatic (the transmit and

230 I Direct response

to detonation
/Surface reflection
220
If
::1. 210
. _ _ _ Deep scattering lay...
1!200 volume reverberation
!1l Sea-floor reflection
of Surface reverberation
caused by sea-floor

130
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5
Time after detonation - seconds

Figure 6.12. Response at the hydrophone to an explosive detonation.

The explosion time constant (here about 200 ms) and peak amplitude
(230 dB re 1 f.l.Pa) both increase with increasing charge size
 Noise and Reverberation 107

receive transducers are at the same location) much attention has been given
to the problem of backscattering, where the intensity of sound returning
on the same path as the transmission is the quantity which is required to
be estimated. However, there has been growing interest in bistatic geometries,
in particular for communication purposes where, necessarily, the transmitter
and receiver will be at different geographical locations, and also in passive
sonar applications. This latter class of problem stems from a military
objective: the covert detection and recognition of surface or submarine
targets. In both these examples, the receiver is required to isolate a main-
path received signal from reverberation and noise. The reverberation may
now contain a significant forward-scattered component, and backscattering
is likely to be quite unimportant. Since most available data pertain to
backscatter, some care will be needed in acquiring modelling information
for system performance prediction.

We turn our attention first to the problems of volume and surface backscatter
prediction. We note that the performance criteria we shall investigate
pertain specifically to pulsed sonars and would not be accurate for continuous
wave transmissions, which have received relatively little study in the
scattering context. Because we deal with pulsed sonar, let us examine, for
a moment, the temporal and spatial characteristics of the pulse. A sonar
pulse of duration T seconds will clearly correspond to a pressure perturbation
in the water of cT metres spatial extent. Imagine such a pulse to enter a
region of scatterers, such as the deep scattering layer. Imagine also the layer
boundary to be reasonably precisely defined. The pulse front edge will
become incident upon the layer boundary and will begin to produce backscatter
which will proceed backwards towards the transmitter. The front edge will
penetrate the scattering volume, initiating backscatter as it travels inwards.
Some time later (depending upon how far the layer is from the transmit-
receive transducer location) backscatter will begin to be detected by the
receive transducer. The actual acoustic contribution at any instant will be
the sum of backscatter initiated by all parts of the pressure perturbation
which is the travelling pulse, which are at that instant equally time-spaced
from the receiver.

Put in another way, the front end of the pulse will have stimulated a scatterer
to return an energy contribution towards the receiver. When the pulse is
"half-way into" the layer, it will stimulate a backwards travelling pulse
which will have got back to the layer boundary at exactly the same time
that the pulse rear edge will also have got to the boundary, travelling
inwards. The pulse rear edge will thus, at that instant, be initiating an
incoherent backscatter contribution which will be power additive with the
backward travelling contribution from the front edge of the pulse. Of course,
the same effect will be occurring for all other parts of the pulse. It is as
108 Underwater Acoustic Systems

if, at that instant, a layer depth of one half the spatial length of the pulse
were contributing, power additively, to form the backscatter wavefront
about to proceed back to the receive transducer.

Having thus established the spatial extent of the scattering volume as

cT/2, we may now consider the geometry of a typical volume backscatter
situation, figure 6.13 . For simplicity, let us assume a transmit-receive
transducer with an axially symmetric beam-pattern, and a beam-width of
«1> radians. We discuss the evaluation of «1> in terms of transmit frequency
and transducer aperture in Chapter 8. At range R, the subtended circular
radius of the scattering volume will be c!>R, and the scattering volume itself
will be 1tci>2R2cT/2. If the transmitted intensity is 10 , then the intensity
incident on the scattering volume will be I 0R-2 • We assume a backscatter
constant Kv, determining the proportion of incident energy per unit scattering
volume, returned in the direction of the transmitter, and calculate, subject
again to a further inverse square-law spreading, a received intensity given
by

scattering
layer

• •
transmit-receive
transducer
cT/2

Figure 6.13. Volume backscattering: the geometry of the problem

 Noise and Reverberation 109

This relation may be used to predict the backscatter strength, if Kv is known

for a given ocean environment. Alternatively, it may be re-cast to allow
Kv to be estimated from measurements of 10 and 11•

We further note that the backscatter intensity increases with the source
intensity, decreases as the square of range to the scatterer location, increases
as beam width increases and also increases with increasing pulse length. As
was mentioned in an earlier paragraph, the formula derived above is good
for "short-pulse" sonars. Such sonars would encompass some tens or hundreds
of cycles of carrier within the pulse envelope.

A good example of volume backscatter is to be found in the deep scattering

layer, referred to above in the context of observations of reverberation
produced by a small explosive detonation. The depth and population dispersion
of the deep scattering layer is subject to wide diurnal variation. It might
typically be found at depths of less than 200 m during the night, making
a rapid movement to and from depths in excess of 1000 m at sunrise and
sunset. The population concentration at the surface at night is such as to
produce a value of 10log Kv of about -70 dB. During the day, the value
of 10log Kv falls to between -90 and -100 dB. Finally, the backscatter
coefficient shows little observed frequency variability at frequencies in
excess of 20kHz. The use of backscatter as a method of determining biomass
concentration is now a matter of considerable scientific interest, and is a
subject to which we return, in Chapter 9.

Another example of volume backscatter is to be encountered when a sonar

is made to impinge on the wake of a vessel. Because the wake is heavily
entrained with air-bubbles which persist long after the passage of the vessel
itself, a strong backscattered return may be observed for many minutes. The
wake strength, measured in decibels per metre of wake length insonified,
replaces the scattering strength coefficient 10log Kv. Typical values of the
order of -10 to -30 dB at frequencies of some tens of kHz have been
observed, higher values corresponding to larger vessels. The wake strength
decay rate is of the order of 1 dB per minute. Much information on the
subject of wakes and their acoustic properties is to be found in reference
[6.8].

A similar argument to that presented above for volume scattering may be

applied to show that, for surface scattering, the backscatter expression is

where 'I' is the beam width of the transmit-receive transducer, in the horizontal
110 Underwater Acoustic Systems

plane, and K is a surface reverberation constant. The major causes of sea-

1

surface backscatter are surface roughness induced by wave action and

backscatter produced by air bubbles in the immediate surface layer.

Urick [6.9] discusses various theoretical and empirical formulae which have
been suggested to allow calculation of the surface scattering coefficient K 1•

However, the unification of the available data to provide anything approaching

a general formula K (9,f, v), which is of widespread applicability, yet remains
1

wanting. Here, 9 is incidence angle, v is windspeed and f the transmission

frequency. Indeed, the more general problem of quantifying the scattering
coefficient for arbitrary incident and observation angles, 9 and 'If respectively:
K {9,'Jf,f,v), is even more elusive.
1

Surface scattering at the sea-floor could likewise well afford some further
study and unification of results. The general problem has been likened to
that of the diffuse reflection of light from a matt surface, which is described
by Lambert's Law. The acoustic analog of Lambert's Law may be stated
thus

where 11 is the intensity at unit distance from a unit area of diffuse scattering
surface of reflection coefficient Jl, insonified by plane waves of intensity
10 • Note that 9 and 'I'· the incidence and observation angles respectively,
need not be co-planar. The backscatter coefficient (for which 'I'= 9) is thus

K1 = Jl sin20
Lambert's Law fits, by observation, reasonably well to some types of sea-
floor, particularly those that are rough by comparison with the wavelength
of the sound incident upon them. The value of Jl may be bracketed between
Jl = 0.32 and Jl = 0.002. The former value is the maximum reflectivity if
the scattering is omnidirectional and no sound loss into the sea-floor takes
place. The smaller value corresponds to experimental observations in the
deep ocean [6.10, 6.11].
 Noise and Reverberation 111

References

[6.1] G.M. Wenz, Acoustic Ambient Noise in the Ocean: Spectra and Sources, I. Acoust.
Soc. Am., Vol. 34, No. 12, 1962, 1936-1956

[6.2] R.J. Urick, Ambient Noise in the Sea, Peninsula Publishing, Los Altos, Calif., 1984,
pp. 3-5

[6.3] D.M.F. Chapman, Surface Generated Noise in Shallow Water: A Model, Proc. lnst.
Acoustics (London), Vol. 9, Pt 4, December 1987, pp.l-11

[6.4] D. Ross, Mechanics of Underwater Sound, Pergamon Press, New York, 1976

[6.5] R.J. Urick, Principles of Underwater Sound for Engineers, McGraw Hill, New York,
2nd edition, 1975, pp.298-342

[6.6] Physics of Sound in the Sea - Part II: Reverberation, US National Research Council
(Originally issued as Division 6, Volume 8 NRDC Summary Technical Reports)

[6.7] R.P. Chapman, Sound Scattering in the Ocean, in Underwater Acoustics, Vol. 2. (V.M.
Albers, ed.), Plenum Press, New York, 1967, pp. 161-183

[6.8] Physics of Sound in the Sea · Part W: Acoustic Properties of Waves, US National
Research Council (Originally issued as Division 6, Volume 8 NRDC Summary Technical
Reports)

[6.9] R.J. Urick, Principles of Underwater Sound for Engineers, McGraw Hill, New York,
2nd edition, 1975, pp. 240-243

[6.10] K.V. Mackenzie, Bottom Reverberation for 530 and 1030-cps Sound in Deep Water,
I. Acoust. Soc. Am., Vol. 33, No. 11, 1961, pp.1498-1504

[6.11] P.B. Schmidt, Monostatic and Bistatic Backscattering Measurements from the Deep
Ocean Bottom, J. Acoust. Soc. Am., Vol. 50, No. 1, Pt 2, 1971, pp. 326-331
7 Acoustic Transduction

7.1 Introduction

Underwater sound transducers convert electrical energy into or from

mechanical energy, the latter quantity being perceived as longitudinal
pressure waves in water. We thus impress a voltage waveform v(t) on the
terminals of a transmit transducer, or projector, and generate a sympathetic
pressure fluctuation p(t) in the water. The reverse occurs with a receive
transducer, or hydrophone.

For acoustic projectors we seek, in partial characterisation, a projector

constant kP measured in terms of the logarithmic ratio of generated intensity
per volt rms applied to the projector terminals. That is, kp will have "units"
of [dB re 1 J!Pa v-1]. More specifically, the intensity so used to characterise
the projector will be the on-axis intensity, reduced to a standard range of
1 m.

In making these further stipulations, there is a presumption that any projector

will have some polar response (about which we shall concern ourselves
shortly) and that calibration measurements (by whatever means) will be
executed "in the far field" at some suitably large range R. Yet further
presumptions would require that, not only was the calibration measurement
far-field, but that it was also "free-field", so that reflections of sound to,
for example, a calibrated test hydrophone from the sea-surface or sea-floor,
or the walls of an acoustic test-tank, would not render the calibration
inaccurate. If our calibration measurement thus yielded an on-axis intensity
I at range R, then we should expect an intensity IR 2 at the reduced range
of 1 m. In estimating the extent of the near-field and thus the range at which
we might expect far-field conditions to start, we use the empirical result
Rc = A(A where A is the projector surface area and A. is wavelength in water
(recall that c = fA where f is the transmission frequency in Hz).

For the hydrophone, by contrast, the calibration constant kh represents rms

terminal voltage generated by immersing the hydrophone in a pressure field
of given rms pressure. ~ thus has units of [V J,JPa-1]. The same comments
112
 Acoustic Transduction 113

regarding the practicalities of calibration remain, of course. Commonly, a

reference hydrophone (a "secondary standard" carefully calibrated by a
manufacturer with access to good calibration facilities) will be used as the
basis of a "comparison method" of further calibrating both acoustic projectors
and equipment hydrophones. For the former, rms terminal voltage at range
R from the projector will be used to infer rms pressure in the water at the
hydrophone location caused by driving the projector with a known rms
terminal voltage. This pressure will then be used to calculate acoustic
intensity at the hydrophone. Finally the intensity at rangeR will be reduced
to the intensity at 1 m standard range, and the calibration will be complete.

Main Axis

1(0)

Figure 7.1. Projector polar response

For the hydrophone calibration, straightforward substitution of equipment

hydrophone for reference hydrophone in a pressure field remote from any
suitable projector will allow relative sensitivity at a given transmission
frequency to be evaluated. A wider range of calibration methods is discussed
in some detail in Urick [7.1].

In discussing projector calibration, the projector polar response was referred

to. We write b(9) as the ratio of off-axis to main-axis intensity, figure 7.1,
so that b(9) = 1(9)/1(0). The "3 dB" half-beamwidth (9 measured away from
the main axis, until 1(9)/1(0) = 0.5 or 10 log 10{1(9)/1(0)) = - 3 dB) will often
be of particular interest in estimating the suitability of a projector for a
particular task. For a circular transducer of diameter D, the half-beam width
measured in degrees will be approximately 30A/D where, again, A is the
wavelength of sound in water. Similar polar response considerations apply
to the hydrophone.

7.2 The Basic Principles of Acoustic Transduction

The objective of any transduction operation is to bring about the conversion

of energy from one physical form to another. Often, because of the wide
use of electronic methods for data gathering and message transmission, the
114 Underwater Acoustic Systems

transduction will be to or from the electrical state. Thus for underwater

acoustics applications we often seek physical phenomena which will translate
between the electrical state and some form of mechanical displacement,
since the latter will then engender pressure fluctuations in the water.

The most common form of transduction, and the one upon which we shall
primarily focus our attention, makes use of the piezo-electric effect. However,
it should be noted that magnetostriction is of great importance in the design
of some forms of transducer, particularly for low-frequency applications.
An example of a scroll-type flooded ring transducer is shown in figure 7.2.
The scroll is made of nickel-iron alloy and the application, via the surrounding
coil, of a magnetic field, induces circumferential length changes and thus
a hoop mode of displacement of the surrounding water.

Although it is possible to envisage applications involving other transduction

mechanisms, such as piezo-resistivity or displacement capacitance (perhaps
for hydrophone applications), the piezo-electric and magnetostrictive methods
have the advantage of extreme robustness and relative immunity to pressure
in deep water applications.

/ Magnetising Coil

Figure 7.2. Free-flooding, low-frequency magnetostrictive transducer;

diameter - 1 m

It is often interesting and sometimes helpful, in the context of instrumentation

transduction, to envisage the transduction process as being composed of
several separable stages. Thus for example, although the strain gauge
diaphragm might be thought of just as "a transducer", it should more
correctly be seen as a diaphragm transducer, converting pressure fluctuations
 Acoustic Transduction 115

to mechanical displacement, followed by the strain gauge, converting

displacement to resistance by a body-distortional mechanism. Such a
visualisation allows greater insight into the role played by the various
components involved in the transduction, and how they may individually
affect the sensitivity, linearity or inherent accuracy of the process itself.

Usually, acoustic transduction is a rather more "rough and ready" job than
instrumentation transduction. However, there is at least one relatively new
transducer which closely parallels the diaphragm displacement pressure
sensor described above, albeit in reverse- to generate pressure fluctuations
rather than detect them. That device is the flextensional transducer, figure
7 .3, wherein an elliptical containment provides a diaphragm-like displacement
across the minor axis, when the piezo-electric drive stack produces an
extension or contraction along the major axis. The elliptical form also
confers a leverage action (a circular container would not), increasing the
face displacement and thus the amplitude of the pressure wave launched
into the water, thereby assisting matching and providing a relatively broadband
transduction operation.

Finally, it is interesting to note that, whereas piezo-electric transduction

and, to a lesser extent magnetostrictive transduction remain the workhorses
in so far as transducer design is concerned, there yet remains a wide range
of mechanisms primarily, but not exclusively, utilised for marine seismic
survey, which include explosive and air-gun detonations and electric spark
discharges.

-1-3 em

Figure 7.3 . The f/extensional transducer, showing use in a short stave

of three f/extensional units. Also shown are the drive stack of piezo-
electric elements, metal endpieces and an aluminium or fibreglass
shell. Operating frequency is in the range 500 Hz to 3 kHz depending
upon wall thickness and material
116 Underwater Acoustic Systems

These mechanisms are used to launch high-intensity, low-frequency shock

waves into the water and, particularly, into the underlying geological strata.

7.3 Piezo-electric Transduction

The phenomenon of piezo-electricity lies at the heart of the acoustic

transduction process. Of the thirty-two classes of crystal structure, about
two-thirds exhibit the property that charge separation, and thus the generation
of a transverse electric field, occurs with mechanical displacement along
appropriate crystal axes. Quartz is well-known to exhibit the effect, and
its various physical properties make it most suitable as a stable, frequency-
defining oscillator element. In the past, quartz and, indeed, a range of other
materials were employed in the construction of underwater transducers. At
the present time, most transducers make use of piezo-electric ceramic
materials and of these, the lead zirconate titanate (PZT) ceramics are by
far the most commonly encountered. By analogy with the phenomenon of
ferromagnetism, the piezo-electric ceramics are often referred to as
"ferroelectric" materials, despite the fact that iron is not a constituent of
their molecular structure.

Although, of the possible piezoelectric transducer materials, ferroelectric

ceramics are now used exclusively for projector transduction, a small
proportion of receiver applications do make use of a plastic film known
as poly-vinylidene fluoride (PVF). PVF is sensitive as a receiver, but
exhibits high capacitance and is a poor projector material for practical as
well as physical reasons. Unlike the ceramic materials, it is well-matched
to water and is thus an efficient converter of acoustic to electrical energy.

PZT ceramic is produced as a powder which is compressed and fired to form

brittle and (by comparison with other ceramics) soft solids, which typically
take the form of rings, discs, tubes, plates and half-spheres. Further working
by cutting, using a diamond saw, and by grinding and lapping to produce
special shapes with preferred acoustical properties, is also possible. As
manufactured, the fired ceramic is on-poled: its crystal structure is on-
orientated and no piezo-electric action is discernible. By way of example,
a disc of the material will have its opposing flat faces silvered, with a heat-
fired paint, to provide plane-parallel circular conductive electrodes. The
crystal will be placed in a high electric field and heated to a temperature
in excess of some 3oo· C. This is the "Curie temperature" above which
piezo-electricity is destroyed, the crystal structure having been "loosened
up" because of the heating. The high electric field has the effect of pulling
the electric dipoles within the crystal, to make them lie with their dielectric
 Acoustic Transduction 117

axes along the lines of electric flux. As the crystal is cooled down, still
with the high electric field applied, the crystal dipole orientation becomes
"frozen" and, reduced to room temperature, strong piezo-electricity will be
evident in the disc. Pressure applied across the silvered flat faces of the
disc will result in the generation of possibly sizeable electric fields and,
even, kilovolts of terminal voltage. By contrast, connecting an ac signal
generator to the disc faces will, at audio frequencies and modest voltages,
produce an audible whistle.

Ferroelectric ceramics are materials which have a high acoustic impedance,

of the order of 35 MRayl, and low internal mechanical loss. The transmission
coefficient between the material and water is given by T = p2c,j(p 1c1 + p 2c2)
which thus has a value of about 0.04. Consequently, if an element is caused
to vibrate, by means of an applied alternating voltage, and thus establish
an internal acoustic wave motion, very little of the sound emanates into
the water. This has two consequences. Firstly, in order to get useful amounts
of acoustic energy into the water, high drive voltages are required. Secondly,
if excited with an impulse, the transducer (like a struck cymbal) "rings"
sustainedly. The time domain "envelope" is thus of long duration , and
consequently the transducer pass-band is small. The transducer element,
if impulsively excited, will ring at a frequency determined by its width,
h, and by the speed of sound within it.

In order to improve the transfer of sound into the water, it is possible to

provide a matching layer of acoustic impedance which is intermediate
between that of the ceramic and that of the water. The optimum layer
impedance is about 13 MRayl. No engineering materials quite provide such
an impedance. However, aluminium, titanium and loaded epoxy resins come
usefully close. In the following sections we consider in greater detail the
structure and design of a variety of composite transducers which use
ferroelectric ceramics as the driving elements.

7.4 The Langevin Projector

Any of the naturally occurring piezo-electric materials and even, from the
viewpoint of economy, the PZT ceramics, are not readily to be had in objects
of sufficiently large physical dimension to produce resonant behaviour in
the audible range. Many sonars are required to operate at such frequencies
and this was particularly the case in the early days of practical electronic
sonar, after World War I. In 1920 Paul Langevin published [7.2] a patent
disclosure where he described how mechanical structures sandwiching
quartz driving elements might be employed for the resonant generation of
sound. The simplest Langevin resonator is shown in figure 7.4.
118 Underwater Acoustic Systems

~·~-------------L

Figure 7.4. The Langevin resonator

Here, a PZT disc is imagined to be glued to cylindrical end-bars of, perhaps,

a metallic material with a similar acoustic impedance and sound speed as
that of the PZT itself. For example, the various PZT -type ceramics exhibit
acoustic impedance in the range 25-35 MRayl. Brass has an acoustic
impedance of about 30 MRayl. Metals such as brass, steel and the various
aluminium magnesium and titanium alloys have high quality factors (even
by comparison with PZT) and thus make potentially good resonator materials
with low internal conversion of acoustic energy to heat. The entire mechanical
structure is "driven" by the central crystal and will resonate at a frequency
f dictated predominantly by the length, L, of the structure according to a
law L = 'A'/2 = c'/2f where 'A' is the wavelength of sound in a material of
sound speed c'. The graph in figure 7.4 shows "particle" velocity, u, versus
axial displacement, x. The end-faces make the greatest velocity and spatial
excursions.
 Acoustic Transduction 119

Even designing a simple resonating structure such as is shown in figure

7.4 involves consideration of some less obvious properties of sound
propagation in solids. The speed of sound in a thin wire, often referred to
as "bar velocity", vbar, is higher than the velocity in bulk materials, vbutk"
The effect is a consequence of the bulk displacement nature of vibrations,
wherein a lateral contraction gives rise to a thickness increase in, for
example, a stubby bar. This effect is described by Poisson's ratio, which
for most metallic and piezo-electric ceramics of interest is about 0.3. Figure
7.5 illustrates the effect [7.3, 7.4], which is most marked for "squat"
transducers.

1.0

...
! 0.8
]
>

0.6

•
0 1 2 3

+ D/L

[0 ([)
Figure 7.5. Normalised phase velocity for extensional waves in round
bars; Poisson's ratio: 0.3

We may now take the design of the Langevin projector through a series
of stages which will allow us to develop it into a rather more practical
underwater projector transducer. The first three stages are of a practical
nature. If we use two or, indeed, any even number of driver PZT discs, then
interconnection and orientation may be made so as to add end displacement
and parallel the discs in terms of their electrical loading. The method is
illustrated in figure 7.6. Metal shims between the discs provide electrical
connection. The two metallic endpieces are commoned and "earthy". This
is particularly convenient when the discs are replaced by PZT rings and
a central "pre-stress" bolt is used to strengthen the whole structure, figure
7.6. The shims may be beryllium copper gauze or thin sheet or even 0.5
mm sheet stainless steel, without unduly reducing efficiency. Yet another
feature of the design may be appreciated by noting, in figure 7.4, that the
120 Underwater Acoustic Systems

Figure 7.6. Development of the Langevin resonator ( 1)

Figure 7.7. Development of the Langevin resonator (2)

....___ Neoprene Diaphragm

~+.?::.:::11---- Alloy radiating endpiece

~~~~~===:~~f-1--- Pre-stress bolt

Figure 7.8. Development of the Langevin resonator (3)

 Acoustic Transduction 121

vertical centre-line is a nodal point and thus, potentially at least, a good

place at which to attempt a mechanical mounting of the device, since the
mount will not then dampen vibration. This leads us through to the next
phase of design, where we note first that the coupling between PZT and
water via brass ends (assuming our first resonator to be so formed) is poor.
The coupling is best achieved between materials such as PZT and water,
with an endpiece material the specific acoustic impedance of which is the
geometric mean of the impedances of the materials which it separates. For
PZTand water (with acoustic impedances of 1.5 and 30 MRayl respectively),
we anticipate an endpiece material impedance of about 7 MRayl. Magnesium
alloy has an acoustic impedance of about 8 MRayl and so would do well,
at least in this one respect. Aluminium alloy exhibits an acoustic impedance
of about 14 MRayl.

Although this would produce a severer mismatch, such a transducer would

still work well. The design method is forgiving. If the tailpiece is actually
made of a material with a higher acoustic impedance than PZT, then
coupling into a fluid at that end will be yet poorer than from PZT alone.
We may thus evolve a design such as is illustrated in figure 7.8, where the
entire resonant structure is housed in an oil bath, the oil having much the
same acoustic impedance as water, and separated from the water by a tough
neoprene rubber membrane.

In fact, the design may be radically simplified for some applications, albeit
at the possible cost of transducer efficiency, by using a polyurethane rubber

Polyurethane Casting

Isolation pad

Figure 7.9 Use of the Langevin resonator in multiple transducer

arrays
122 Underwater Acoustic Systems

encapsulation technique, figure 7 .9. If an array of resonators is encapsulated

in this way, in an attempt to increase the effective aperture and reduce
beamwidth, then resonator/resonator and resonator/backplate interaction
may also prove troublesome.

Radial groove for compliant Mounting face

region between active face
and mounting and sealing 0-ring seal groove
faces

Figure 7.10. The Tonpilz resonator

The final stage of development of the Langevin resonator is into what is

often termed the "Tonpiltz" resonator, figure 7.10. Tonpiltz means "sound
mushroom" and would appear to be World War II terminology adopted from
German Naval parlance. In any event, the term is descriptive of the structure
itself, wherein the radiating face is flared outwards, again to improve the
matching into the water load.

7.5 Ring and Tube Transducer Designs [7.5]

Our objective thus far has been to develop transducer designs which allow
relatively little ceramic material to act as the driving element for a larger
mechanical body with a relatively low resonant frequency. Another method
of attaining the same end with possibly less technical complication is to
utilise a free- flooding ring transducer. This, figure 7.11, simply consists
of a short, large-diameter, thin-walled ceramic tube operating in a hoop
resonance mode.
 Acoustic Transduction 123

2r

t
.__ t

Figure 7.11. The free-flooding ring

Similar structures have been used fully encapsulated in polyurethane, rather

than dip-coated to be free-flooding. Smaller, and thus higher frequency,
tubes have also been discussed in the literature [7 .6] as transmitter devices
of simplicity and robustness. It is, perhaps, appropriate at this stage to
summarise the hoop, radial, length and thickness resonances for the tube
transducer, since this ceramic shape is readily available and extremely
useful for a wide range of transmission and reception applications.

The radial mode resonance is given as f = c/2TCr, the length mode resonance
by f = c/21 and the thickness mode by f = c'/2t where c is the velocity of
longitudinal waves, c' the velocity of thickness waves, r the mean radius,
1 the length and t the wall thickness of the tube.

7.6 Resonance Behaviour of Transducers

If the development of the various transducers described above has been

correctly carried out, some care will have been taken to ensure that the
transmission resonance is kept reasonably separated from other dominant
resonances. Then the transducer will act as a single resonator and will have
a terminal electrical response which will equate to that of a series tuned
LCR circuit. Because the drive elements are ceramic rings with fired plate
electrodes, and because ceramic is a dielectric material, there will be, placed
in parallel with the LCR equivalent of the mechanical resonator, a capacitor
of significant value, which models the static capacitance, C0 , of the drive
elements themselves. The equivalent circuit will then have the form shown
in figure 7.12. Here the resistive component in the series LCR circuit has
been split into two components: the radiation resistance R, and the loss
resistance R1•
124 Underwater Acoustic Systems

Figure 7.12. Equivalent circuit of the single resonance transducer

Frequency,f

Figure 7.13. Transmitting response of the untuned single

resonance transducer

Figure 7.14. The single resonance transducer with series tuning

 Acoustic Transduction 125

We are interested first in finding the overall voltage transfer function, v1

which, since v2 is proportional to the pressure field developed in the water,
gives the effective shape of the transmitter response, figure 7.13. It is easily
shown that

where ro0 = (LC)-112, Q0 = (ro0C(Rr + R1))-1 and the bandwidth, measured

in Hertz, is given by B = rorf27tQ0 if~» 1. This function is not a particularly
attractive transmitter response from, at least, a sub-sea communications
standpoint. By employing series tuning, figure 7.14, a double-humped
response can be obtained, figure 7.15, although tailoring the depth of ripple
may be tricky, in practical terms.

Note that the one remaining variable which is not, as it were, totally buried
within the transducer structure and thus difficult to gain access to without
a complete structural re-design, is the static capacitance, which may be
padded-up to a larger value. This will inevitably have the effect of drawing
together the two humps in the transmission characteristic, increasing the
overall "gain" and reducing the ripple, but at the expense of bandwidth.
In many sub-sea communications applications, where tailoring the
transmission response in this way might be desirable if data rate is not at
a premium, this might not be a problem.

There remains one further possible objective in performing series tuning.

Given a Langevin or Tonpilz-type transducer, the Q-factor may only be
modest and the damping on series tuning could significantly undermine any
Q-magnification or transformer action in "stepping up" the drive voltage
levels issued by semiconductor drive circuitry. Remember, also, that acoustic
transducers often present terminal resistance, when properly tuned to cancel
the quadrature current drain demanded by the static capacitance, of some
hundreds or even thousands of ohms. To drive powers of the order of tens
or hundreds of watts into such transducers may require drive voltages of
hundreds of volts or even kilovolts. For the Langevin and Tonpilz transducers,
this means that a transformer must be interposed between the source of the
drive waveform and the terminal input to the transducer, including the series
tuning inductor. In contrast, for the flooded ring or for any piezo-electric
crystal left deliberately unmatched mechanically, where the Q-factor is
intrinsically rather high, series tuning can lead to a worthwhile transformer
action, albeit with relatively small available transmitter bandwidth. In some
applications this may not matter.
126 Underwater Acoustic Systems

Frequency, f

Figure 7.15. Transmitting response of the tuned, single

resonance transducer

At series resonance, the inductive and capacitive reactances due to L and

C in figure 7.14 cancel, so that the total resistance Rr + R1 appears across
C 0 • The overall voltage gain when ro = ro 0 then becomes

where RL is the self-resistance of the tuning inductance, L•. If RL ~ 0 then

lv/v 11 ~ ro0 CR,. If R ~ oo then lv/v1 1 ~ (ro0 L,/RL)(R/(Rr + R1)). Both
extremes represent Q-factor gains which might thus be expected to offer
an effective transforming action with a voltage step-up of about ten. It is
also worth noting that, in this mode of tuning, the two resonance peaks which
inevitably occur when a series inductor is added are widely separated, not
close together forming a notionally flat, passband characteristic, such as
was shown in figure 7.15.

Finally, it is interesting to reflect on the design requirements imposed upon

the series tuning inductor and on any parallel, padding capacitance, required
to increase the value of the static capacitance, C0 • First, the tuning inductor
must not saturate, or even approach saturation. This can be quite a difficult
objective to meet, and it is as well to be prepared for difficulties, particularly
when tuning modest to low frequency transducers. If saturation is even
approached, the non-linear behaviour of ferrite materials leads to significant
"describing function" phase shifts and consequent departure from "inductor-
 Acoustic Transduction 127

like" behaviour. The result will be a dramatic and often mysterious de-
tuning and loss of performance. Curiously, choosing adequate capacitors
can be extremely difficult, also. Here the problem is that the terminal voltage
across the transducer, and hence across any additional padding capacitance,
may well be of the order of kilovolts. Typically, capacitors are specified
according to a de voltage rating, which may well be only one half or one
third the effective safe minimum ac rating. Since it is quite difficult to find
suppliers of capacitors rated for even de operation to 1 kilovolt, never mind
find a complete range of preferred values, the problem will be readily
appreciated.

An interesting further consequence to the series tuning problem has been

investigated by Coates and Mathams [7.7]. A moment's thought will show
that the transducer equivalent circuit of figure 7.13 is capable of being
thought of as the terminating end of aT-filter section. If, as figure 7.16
suggests, we increase the complexity of the matching network, we may hope
to fabricate transducer matching networks capable of realising some, if not
all, of the all-pole modem passive filter designs. It is even possible to treat
the mechanical end of the transducer equivalent circuit in a similar way,
as Dunn and Smith [7.8] have recently pointed out. Considerable potential
thus exists for tailoring a transmission response to desired passband criteria.

Figure 7.16. The matching circuit approach

7.7 Multiple Matching Layer Transducers

Transducers of the Langevin and Tonpiltz type are predominantly the most
popular designs for operation in the broad range of frequencies from (about)
1kHz to (about) 100kHz. For frequencies in excess of 100kHz, up to (for
reasons related only to the practicalities of fabrication of robust, well-
engineered transducers) perhaps a few MHz, another approach to design
becomes feasible. At such frequencies, it becomes relatively easy to obtain
128 Underwater Acoustic Systems

drive elements (typically round PZT discs) which are of large diameter by
comparison with their thickness: the so-called "thin disk". Such drive
elements, figure 7.17, if placed directly in contact with a water load, and
assuming an air backing, will exhibit a high Q-factor (typically about 15)
and a Lorentzian (single tuned circuit) transmission response into water,
of the form already illustrated in figure 7.13. The transmission band centre
frequency (the resonance frequency of the disc) will be f0 = c'/2d where
c' is sound speed in the disc (typically about 3000 m s-1} and d is the disc
thickness.
................................
............................ .
............................... ............................ ..
...............................

---------------·
............................

·::. Water load:::

Transducer thickness d

---------------
............................... ...................... ..
...............................

Figure 7.17. The thin disc transducer and its matching into water

If an acoustic quarter-wave matching layer, of thickness w = dc"/2c' where

c" is sound speed in the matching layer, and acoustic impedance equal to
the geometric mean of the acoustic impedances ofPZT and water respectively,
is interposed between the PZT disc and the water load, improved matching
will result. The bandwidth will broaden, and the transmission response will
exhibit, as in figure 7.15, a double humped bandpass characteristic. Some
difficulty attaches to finding a suitable matching layer material at the
required impedance of about 7 MRayl. Curiously, the author- following
a hint in Wells [7 .9] has successfully made up composite samples of
mammalian long-bone which could be turned into matching layer discs. The
acoustic impedance of mammalian long bone along, not transverse to, its
long axis, has almost exactly the right acoustic impedance. Fabrication is
tedious and unpleasant, however. A preferable solution is to use two
matching layers, such that each is quarter-wave resonant and each lies at
the geometric mean of the impedances of its neighbours. Thus if we take
the acoustic impedance of PZT as 0' 1 = 30 and that of water as 0'4 = 1.5
MRayl, the required layer impedances are

0' 0'
2 3

or 0'2 = 11 and 0'3 = 4. Goll [7.10], who provides a comprehensive analysis

of multiple matching layer transducers, suggests the use of glass and lucite
 Acoustic Transduction 129

to manufacture these layers. The author has developed other designs, more
reliable in their manufacture, which work equally well. If carefully crafted,
excellent octave-bandwidth transducers of high efficiency can be fabricated
using this approach.

7.8 Polar Response Measurements on Transducers

The measurement of polar response is an important guide to transducer

performance. A possible test-tank set-up is illustrated in figure 7 .18. The
equipment utilised may be as simple or as sophisticated as the user may
wish or be able to afford. The most expensive item is usually the tank itself,
together with its housing and installation cost. Ideally the tank dimensions
should encompass many tens of wavelengths, if only because this would
allow tone burst testing with range gating at the receive hydrophone to
isolate the first received, direct path signal, before any reverberation
(reflections from the tank walls and water surface) is received. The point
here is, that most transducers have Q-factors of the order of 10 and thus
take some ten cycles to complete their transient rise and fall times. If a useful
dwell at maximum (i.e. a flat and measurable pulse top) is also required,
then the need to be able to transmit perhaps fifty or so cycles per pulse
becomes obvious. However, when it is recalled that one wavelength at 1.5
kHz is 1 m in extent, this is not always easily achievable in any but the
largest tanks. Clearly the problem becomes less critical at higher frequencies.
At frequencies of the order of hundreds of kHz, tank testing can be done
in a water cistern in a laboratory at minimal installation cost.

Rotation of transmit transducer Fixed reference hydrophone

Figure 7.18. Polar response measurement

130 Underwater Acoustic Systems

Tank lining materials have been proposed as a method of reducing or

eliminating reflections from the walls, bottom and surface. For high-frequency
work, in excess of 100kHz, stipple rubber car mat works well. Rubberised
horse-hair bats have also been suggested but the author has had little success
with this material. At lower frequencies, marine ply baffle boards may be
disposed about a large tank to break up the ordered nature of reflective
returns. All such measures are, to an extent, cosmetic. Some value may be
had from employing them or from searching for new approaches, but too
much should not be hoped for. The moral of the story is: put your money
in the tank; make it as large as you can afford in the first place.

The drive transducer will be rotated by some means, in order that the relative
response at different angles of rotation may be measured by the fixed
hydrophone. Rotation may simply be by hand, with a protractor on the
rotating shaft used to measure angle. At the other extreme, a robust stepper
motor drive, with microcomputer control of drive angle, transmitter timing
and range gate timing may allow for automated measurement, data logging
and display of polar response. In terms of today's technology, such a
solution is neither difficult nor particularly costly.

7.9 Admittance Measurements of Terminal Response

The classic approach to measuring polar response, described well in Tucker

and Gazey [7.11], is to plot the terminal admittance Y = G + jB on the (G,
B) plane. The author [7.12] has described a microcomputer-based system
for making such measurements which avoids a commonly encountered axis-
warping phenomenon experienced with some of the cheaper such equipments.
However, the use of modulus of admittance IY(f)l plots is in some ways
to be preferred, partly because it makes for an easier identification of certain
key frequency and admittance modulus values required in the estimation
of equivalent circuit values and partly because the experimenter may
employ any convenient swept spectrum analyser to perform the measurement,
provided only that it has available a swept sinewave output.

The easiest way to establish an equivalent circuit is that shown in figure

7.19. The tracking oscillator output of the spectrum analyser is used to drive
the transducer, perhaps through a buffer amplifier although, being a
substitution method, this should not really be necessary. Most spectrum
analysers have a dual trace display with storage and if this facility is
available, it should be used. The current into the transducer (proportional
to the terminal admittance) is monitored by means of a current probe. A
small-value sensing resistor may alternatively be used; the current probe
will reflect into the line such resistance, anyway.
 Acoustic Transduction 131

In performing the experiment, sweep frequency up from zero to some

suitable frequency above the resonance frequency. Store the admittance
trace so obtained. Remove the transducer and replace it with a capacitance
box. Adjust the capacitance box value so that the straight line of IY(f)l =
21tC0 matches the slope of the stored transducer admittance at zero frequency,
figure 7.19(b). In this way, the value of static capacitance is obtained. Next,
figure 7.19(c}, measure the resonance and anti-resonance frequencies, f. and
fP respectively.

IY1

G£)
Spectrum
Analyser f

Analyser -:I------
_I _____
Input

(a) (b)

IY1 IY1 IY

f 5 fp
•
f f f

~~
T ----
~c~
T ~'0·
(c) (d) (e)

Figure 7.19. Making up a model of an acoustic transducer by

measurement and substitution

Using the measured value of C0 evaluate

and then evaluate the series inductance, L, by calculating

L = (41t2 rcr
s
1
132 Underwater Acoustic Systems

Again, using capacitance boxes and a tray of standard value inductors (or
an inductance box if the reader is inclined towards self-indulgence in the
fabrication of such an item; it can be well worth the trouble, for use in tuning
experiments using a spectrum analyser, as well), set up the physical model,
figure 7.19(d). Some "knob twiddling" may now be justified because the
equations given only approximate the C and L values. In any event, connected
as shown, a high-Q or "peaky" resonance should be obtained.

Lastly, add in a resistance box and adjust, figure 7.19(e), to reduce the Q-
factor and make the synthesised and measured traces coincide, and the
model is complete. This method will work, even for transducers with several
significant resonances within the sweep band. In fact it is probably desirable
to sweep upwards in frequency until no further significant resonances are
identifiable, if the lowest resonance is not the one of interest, and then
characterise the transducer in totality. Be warned, however, that any real
transducer will exhibit myriad small resonances and clearly it will not be
feasible to treat any but the most obvious. Some entertainment may be
afforded by trying to establish the modal nature of the most significant
resonances.

It is clearly possible, also, to improve the sophistication of the modelling

operation. One may envisage an experiment which reads admittance data
from the analyser, across to a microcomputer, which solves the equations
and overplots the measured with a computed admittance curve, and so on.
Coates and Maguire [7.13] have published equations for computing the
equivalent circuit component values (to an approximation) for multimode
transducers. They have also published the description of an iterative method
[7 .14] for refining the model so produced, so that the computed admittance
response converges upon the measured response.

7.10 Hydrophones [7.15, 7.16]

In the main, our attention thus far has centred upon the projector transducer.
Hydrophones, in some ways, offer less scope for variation of form to the
designer and, because they may often be required calibrated, tend less to
be the result of hand-crafting. Most hydrophones use PZT ceramic rings
as their sensitive elements. Some utilise pairs of PZT half-spheres in an
attempt to improve the omni-directionality of the device. A few utilise flat,
or rolled, bar-mounted PVF film material. We shall focus attention on the
first, largest class of devices.

Figure 7.20 shows the stages in the manufacture of a simple hydrophone.

In this case, a single, capped tube element is soldered to the cable conductors
 Acoustic Transduction 133

and encapsulated in a neoprene or polyurethane rubber. The design may

utilise a tube of any appropriate size, subject primarily to the constraint
that the hoop and length mode resonances occur at significantly higher
frequencies than the hydrophone is intended to detect. The larger the
diameter of the tube, the bigger will be its receiving sensitivity. It is also
desirable that the tube wall be as thin as is practicable. Indeed, there are
ratios of wall thickness to radius which can make the hydrophone virtually
useless as a sound detector, because of a subtractive interaction between
voltages generated within the crystal as a result of compression along
orthogonal axes [7.15] . The hydrophone may also, with great advantage,
be provided with an integral head amplifier, which will provide also a
buffering and cable-driving action. If possible, the hydrophone should also
be electrically shielded from external electric fields. This should be done
by connecting the shield to the cable screen and thus to the surface amplifier
input ground point. The shield should not be grounded through a sea-water
return; this can prove extremely noisy. Some hydrophone designs provide
also, a precision calibration resistor which is used forremote, field calibration
purposes. Figure 7.21 illustrates a typical configuration. The hydrophone
cable itself should exhibit low microphonics and great physical robustness.
Twisted-pair conductors should be individually screened. For some
applications, choice of a bitumen in-filling between the pairs may be

(a) (b) (c)

Figure 7.20. The basic hydrophone design using a PZT ceramic tube
134 Underwater Acoustic Systems

.....--....-..,.--....,-r----------------.;-..
~------------------------------------,
,...-;;-------r.:.._.;.~
.-------------------------...
"

Y-----------------------:..·
·-------------------------------------~
\....____v---""
3-Core, twisted pair, screened Calibration
oceanographic cable SOID'CC

Figure 7.21 Hydrophone head amplifer, power supply and

calibration circuitry

helpful in preventing the ingress of sea-water, should joints weaken or the

sheath be damaged in any way. If a heavy hydrophone assembly with,
perhaps, provision of a parabolic reflector and ballast weights is envisaged,
then kevlar fibre reinforced oceanographic cable may be used.

Figure 7.22 shows the system outline for a hydrophone design used by the
author for deep-ocean trials of a communication equipment. The underwater
housing provided electronics which both passed a raw, but amplified received,
modulated carrier signal to the surface, as well as a demodulated signal.
During the course of the experiment, a completely separate equipment
monitored and recorded both the raw signal and a demodulate, obtained in
a different way. On board ship, two different receivers, as well as yet further
recording capability were available. In fact, the direct microcomputer data
input option shown in figure 7.22 worked perfectly but, obeying Murphy's
Law, would certainly have failed had insufficient backup been provided.

Underw11ter Umt
Low No1se
Ampllf1ers 1
L - - - - - - - - - - - - - - - - - - - - - - ~
F1breg1ass Hydrophone
paraboloid actlve area
1111ed w1th
syntact1c
foam

Figure 7.22. A data communications receiver hydrophone unit

 Acoustic Transduction 135

References

[7.1] R.J. Urick, Principles of Underwater Sound, McGraw-Hill, New York, 1975

[7.2] P. Langevin, Brit. Pat. Specification, NS, 457, No. 145, 691, 1920

[7.3] J. Van Randeraat and R.E. Setterington, Piezoelectric Ceramics, Mullard Ltd, 1974
[ISBN 0 901232 75 0]

[7.4] L. Camp, Underwater Acoustics, Wiley, New York, 1970, p.136 ff.

[7.5] R.M. Davies, A Critical Study of the Hopkinson Pressure Bar, Phil. Trans. Roy. Soc.,
Series A, Vol. 240, 1948, pp. 375-457

[7.6] D. Church and D. Pincock, Predicting the Electrical Equivalent of Piezoceramic

Transducers for Small Acoustic Transmitters, IEEE Trans. Sanies and Ultrasonics, Vol. SU-
32, No. 1, 1985, pp. 61-64

[7 .7] R. Coates and R.F. Mathams, The Design of Matching Networks for Acoustic Transducers,
Ultrasonics, Vol. 26, March 1988, pp. 59-64

[7.8] J.R. Dunn and B.V. Smith, Problems in the Realisation of Transducers with Octave
Bandwidth, Proc. lnst. Acoustics., Vol. 9, Pt 2, 1987, pp.58-69

[7.9] P. Wells, Biomedical Ultrasound, Academic Press, London, 1977

[7 .10] J.H. Goll, The Design of Broadband Fluid Loaded Ultrasonic Transducers, IEEE Trans.
Sanies and Ultrasonics, Vol. SU-26, No 6, 1979

[7.11] D.G. Tucker and B.K. Gazey, Applied Underwater Acoustics, Pergamon Press,
London, 1966

[7 .12] R. Coates, A Microcomputer-based Measurement System for Displaying the Admittance

Characteristics of Acoustic Transducers, Trans. lnst. Meas. Control, Vol. 9, No. 4, 1987,
pp. 218-223

[7.13] R. Coates and P.T. Maguire, Multiple Mode Acoustic Transducer Calculations, IEEE
Trans. Ultrasonics, Ferroelectrics and Frequency Control, Vol. UFFC-36, No.4, July 1988.

[7.14] R. Coates and P.T. Maguire, An Iterative Method for the Determination of Acoustic
Transducer Lumped Equivalent Circuit Parameters, Report No. SYS! E87!3, School of Information
Systems, University of East Anglia, Norwich NR4 7TJ

[7.15] O.B. Wilson, An Introduction to the Theory and Design of Underwater Transducers,
Peninsula Publishing, Los Altos, Calif., 1988

[7.16] BS5653: Specification for Hydrophones for Calibration Purposes, British Standards
Institution, London, 1978
8 Transducer Arrays

8.1 Introduction

In the previous chapter we examined the process of acoustic transduction

and the way in which transducers might be assembled to meet particular
design objectives. The common requirements in specification are:

1. Selection of transmission frequency

2. Selection of transmission bandwidth (or Q-factor)
3. Selection of power drive capability
4. Transducer electrical to acoustic conversion efficiency

In order to meet a given specification, additional consideration may need

to be given to electrical termination and matching circuits. It may also prove
possible to incorporate into the transducer design certain criteria regarding
beamwidth and polar response pattern. However, in many situations it is
preferable, or even essential, to separate these latter objectives from those
of the design of the transducer elements themselves.

We should recall (section 7.1) that, for a circular aperture of diameter D

transmitting sound of wavelength A, the half-beamwidth of the polar response,
measured in degrees, will be 30A/D. It is commonly the case in the design
of a sonar system that beamwidth will, in fact, be a primary specification.
Thus we might find it convenient to construct a piston transducer for, say,
a 15kHz (A= 10 em) sonar, by utilising a sandwich construction employing
50 mm diameter ceramic rings, an aluminium head-mass and a steel tail-
mass. The overall diameter (and hence the effective aperture) could not be
much different from 50 mm, which would suggest a full beam width of 120°.

If the initial design specification were for, for example, a full beamwidth
of 20o, the implication would be that the transducer diameter should increase
by a factor of six, to 30 em. This is quite impractical, in a single, solid
transducer design. The cost of ceramic material would be prohibitive.

136
 Transducer Arrays 137

Casting large ceramic pieces is extremely difficult and presents the

manufacturer with a high failure rate and the customer with an extremely
high cost product. The solution is to assemble an array of the smaller (-50
mm diameter) transducer elements (perhaps twenty or thirty in number) to
provide the required aperture area. There are, needless to say, both advantages
and disadvantages in such a strategy.

Among the advantages, we note that control of the relative "gain" of the
peripheral elements in the array allows a measure of aperture "shading"
to take place. This can help reduce sidelobe levels in the polar response,
thereby minimising ambiguity in the discrimination between strongly
reflecting off-main-lobe targets and weaker on-main-lobe targets. Set against
this advantage, the array elements may well be prone to local interaction,
making perlormance prediction difficult and introducing a costly "loop"
into the array design process whilst such problems are ironed out. Yet a
further advantage in utilising an array of transducers is that, by adding
various electronic delay circuits to the individual transducers in the array,
the polar response main axis may be "steered" electronically in any desired
direction. This means that, to point the array in a given direction, we may
dispense with servo-motor drives and mechanical interconnections to a
soundhead and establish a virtually instantaneous shift of beam direction.
It should be stressed, however, that the electronics required to achieve this
is far more complex than that required for mechanical steering. The advantage
lies primarily in speed of response: "inertia-less" beam-steering.

8.2 The Linear Hydrophone Array

The simplest array is a line of equally spaced, equally weighted (unshaded),

omnidirectional elements. The elements are omnidirectional because they
are dimensionally small by comparison with a wavelength of sound at the
operating frequency. Such an array is illustrated in figure 8.1 in both an
idealised form and a practical implementation. It would typically be used
in marine seismic survey or in a military context, in the deployment of long
arrays for passive, covert detection of enemy vessels. If it is of interest to
detect frequencies of the order of 1.5 kHz or less (typical for both covert
passive sonar and marine seismics) then wavelength will be of the order
of 1 m or greater and thus much larger that typical array element dimensions.
The array elements could well be capped ceramic tubes of one inch (2.5
em) length and diameter.
138 Underwater Acoustic Systems

N-element array

• • • • • •
d

Aexible plastic ---,

tubing •

•'•'•'•'•'•'•'•'•'•'•'•'
•'•'•'•'•'•'•'•'•'•'•'•'
•'•'•'•'•'•'•'•'•'•'•'•'
•'•'•'•'•'•'•'•'•'•'•'•'

Individual hydrophone _J
PZT tube elements

Figure 8.1. An N-element hydrophone array. Typically (}. - 2 m.

N - 24 for seismic surveying; N - 100 - 300 for military arrays for
passive detection, which could be up to 1 km in length if towed

Figure 8.2 . A plane parallel acoustic pressure field approaches a line

array at inclination (}

Although the hydrophone array elements are presumed omnidirectional, the

array itself will exhibit strongly directional properties normal to its direction
of extension. To see why this is so, consider the array depicted in figure
8.2. A plane parallel wavefront, of wavelength A., approaches at angle e
to the array. The relative spatial separation parallel to the direction of
propagation, between the first and second array elements, is d sine. The
relative spatial separation between the first and n'th elements is thus
 Transducer Arrays 139

nd sin9. Following the approach outlined in section 1.3, we represent the

incoming perturbation as a travelling wave of unit amplitude cos( rot+ kx).
It follows that, for the n'th element, the signal output may be written in
the form cos(rot + (nrod/c) sin9). It is, at this point, manipulatively easier
to proceed by noting that

cos(rot + (2mtd!A.) sin9) = Re(exp(j(rot + (2mtd/A.) sin 9))}

= Re(exp(j(rot)) exp(j(2mtd/A) sin 9)}

The array output we take to be the simple arithmetic sum of the outputs
of the individual elements, scaled by a factor N-1 so that the polar response
maximum is unity, irrespective of the number of elements in the array. It
is possible, as we shall see later, to weight element outputs and, indeed,
to introduce other forms of array processing than a simple arithmetic sum.
Utilising this procedure, the array output becomes (the real part of)

exp~rot) ~ exp(j(2mtd/A.sin9)
n=O
Broodolde

(I) (b)

Broodolcle Broodolde

(c) (d)

Figure 8.3. The response (a) of an omnidirectional monopole in the 8

plane. Two monopoles in line form the simplest "interferometer" linear
array, shown in (b) with half-wavelength spacing, in (c) with a full
wavelength spacing and in (d) with 1110 wavelength spacing
140 Underwater Acoustic Systems

The time-dependence implied by the first term is of no interest to us. The

summation in the second term may be shown to equate to the function

R(O) = sin((N7td/A.) sinO)

Nsin((11d/A.) sinO)

which represents the directional sensitivity of the array. If N = 1, then R(9)

= 1 for all 9; a single element is omnidirectional, figure 8.3(a). If N = 2,
we obtain the simple two-element interferometer. It can further be shown,
for this case, that R(9) = cos((7td/A) sinO). Given an element separation of
one half-wavelength, we find that R(9) = cos((7t/2) sine), which produces
the directional pattern shown in figure 8.3(b), wherein a single ambiguity
in direction identification occurs. We cannot know where, in the vertical
plane (for a horizontally disposed array) a source will lie. If we increase
the spacing to (say) one wavelength, the pattern becomes that shown in
figure 8.3(c). Now the interferometer presents two ambiguities, in the sense
that there are two principal axes along which a strongly detected (and thus
presumably on-axis) source might be imagined to lie. Finally, if we reduce
the distance between the elements so that they lie close together, relative
to one wavelength then R(9) again tends to unity for all e, which is sensible,
since two additively associated elements which are acoustically close,
appear to merge as one, figure 8.3(d).

The reader should note that the two element interferometer is not the
classical dipole (although a single element is a monopole). The dipole
requires that the array element outputs are effectively differenced (either
on reception or transmission, depending upon circumstance), not added. If
we do this, we seek to establish a polar response

R(e) = (expU 1td/A.) sine - exp(-j 1td/A.) sine)/2

= j sin((7td/A.) sinO)
This equation we derive by imagining the dipole to be symmetrically
disposed about the origin, so that the spatial displacements to each element
are respectively +(d/2) sinO and -(d/2) sine.

The dipole is presumed to involve an element separation which is small

by comparison with a wavelength. Then the response modulus becomes

IR(9)1 = (7t5/ A.) sine

 Transducer Arrays 141

Broadside Broadside

End fire

(a) (b)

Figure 8.4. Polar response for the dipole element (small separation of
elements) (a) and for theN = 2 array with 180 o relative phase shift
between the elements, forming a )J2 separation end-fire array

which is the classical form of the dipole response already alluded to in

describing sea-surface noise sources in section 6.5, and which is depicted
in figure 8.4(a). Finally, we might enquire as to the effect on the polar
response of accepting a wider spacing. If we make the element spacing on
the (phase reversed) dipole elements one half-wavelength, the effect is to
produce a 90° rotation on the polar response of the two-element interferometer
pattern. We have essentially, by the implied phase shifting associated with
the dipole geometry, converted a very short broadside array into an equally
short endfire array, figure 8.4(b).

..

,..,
Figure 8 .5. Polar response plots for unshaded, uniformly spaced
arrays, with one half-wavelength interspacing between the elements.
The "zero angle" direction equates to broadside sensitivity.
142 Underwater Acoustic Systems

'""
Figure 8 .6. Polar response plots for unshaded, uniformly spaced 10-
element arrays , with variable interspacing between the elements. The
"zero angle" direction equates to broadside sensitivity.

Let us next investigate what happens as we increase the number of elements

in our array. (We revert to the assumption, for the moment, of simple
element output addition.) Figure 8.5 illustrates a progression through several
stages indicating that as the number of elements is increased, for fixed d
= A./2 interspacing , the directivity also increases.

If, next, as figure 8.6 shows, we select a modest array length of 10 elements
and then allow the spacing to vary, we see that for larger interspacing,
multiple main lobes arise, because the array is effectively spatially
undersampling the wavefield. For smaller interspacing, we worsen the
directivity.

Finally, figure 8. 7, we see the effect of fixing the array length, Nd = L,

to be constant, whilst causing d to approach zero. The bottom line in our
expression
R(a) = sin((N1td/A) sin a)
N sin((1td/A) sin a)

then tends to N1td/A. sine and the polar response assumes the familiar "sine"
or "sin(x)/x" form :

R(a) =sin((N1td/A.)sina) = _sin((1tL/A.)

___;.;._;.. . . .:;_sina)
___;. .;_
(N 1td/A.) sin a (1tL/A.)sina
This pattern corresponds to the uniform, linear array with even spacing
across its aperture.
 Transducer Arrays 143

Half beam-angle (-2 dB down)== 25 A. I L

Figure 8 .7. Directional response of a continuous line array of length L

8.3 The Fourier Transform Approach to Pattern Synthesis

In the previous section, in summing the contributions from individual

elements from a discrete array, we anticipated a more formal approach to
the determination of a far-field radiation pattern from the way in which a
hypothetical acoustic aperture was filled. In the most general sense, we may
assume the existence of some two-dimensional aperture, a function of x
and y, say. This aperture will produce a far-field response which will be
a function of the normalised sines of the beam angles in the horizontal and
vertical: a = (2rr/A.) sine and 'P = (2rr/A.) sin\jl. The inter-relation is the
two-dimensional complex Fourier Transform

W(x,y) ~ R(E>,'P)

The reader is referred elsewhere [8.1] for proofs. Here we note the one-
dimensional simplification
+oo

R(9) = J W(x)exp(-j9x)dx
-oo
144 Underwater Acoustic Systems

and, by way of example, reconsider the uniform line array referred to at

the end of the previous section. If we define the array to be of length L,
uniformly illuminated, so that

W(x) = L-1 ; lxl < L/2

= 0 elsewhere

then the Fourier Transform R(9) follows as a standard result from any
suitable Table of Transform pairs, as

R(8) = sin(eL/2)
8L!2
Upon making the necessary substitutions L = Nd and e = (2rt/A.) sine, our
previous result will be found to have been obtained. The radiation pattern
for the continuous line array has already been illustrated in figure 8.7.
Inspection of the pattern reveals that it has fallen to - 3dB of its main axis
value for an angular off-axis shift (the nominal "half beamwidth") of
25/../L degrees.

The method is a powerful one for investigating the behaviour of far-field

response. We may, for example, apply "shading" to the array aperture. In
the example referred to above, the one-dimensional aperture, the continuous
line array was unshaded. Equal weighting was given to all increments along
W(x) W(x)
Aperture
Weighting

X X

n n
R(9)

~""<---+-+----\,.----,~~. .
Directional
Response

-=-- ==>
¢ (9)

== .......... •
9 9

(a) (b)

Figure 8.8. The use of aperture shading to reduce the sidelobe level of
an aperture (a) with uniform illumination. To maintain the same main
lobe width, the shaded aperture (b) should be physically wider, but
with illumination tapering off at its extremes
 Transducer Arrays 145

its length. Had we so chosen, we could have applied any of a large family
of aperture weighting functions to modify the sensitivity, in the case of a
receiving array, or the power output, in the case of a transmitting array.
In particular, if we choose to smooth the transition at the end of the array
aperture, then, as figure 8.8 suggests, we may reduce the size of the sidelobes
in the radiation pattern and thus reduce the hazard offalse target identification,
without significantly losing angular resolution. That is, if the sidelobes are
prominent, there is a danger that an off-axis target of relatively large target
strength may be mistaken for an on-axis, albeit weaker target. Not surprisingly,
perhaps, the mathematics of weighting closely parallels the windowing
applied in numerical power spectrum evaluation. The reader is referred to
the common source of information on this topic [8.2] and to [8.3], where
several examples pertinent to aperture shading are quoted.

The method may also be applied to provide the evaluation of certain notable
results. The first of these pertains to the uniformly illuminated circular
aperture. This is equivalent, in effect, to a single circularly symmetric
acoustic projector and the result obtained for the far-field response is thus
of considerable importance as a first-cut design aid. The reader might care
to note that, the transducer being axially symmetric, the Fourier Transform
may be modified to yield the result
J1(8L/2)
R(8) =2 el./2

where e has the same meaning as before, L is the diameter of the transducer
and J 1 is the first-order Bessel function "of the first kind". The radiation
pattern is loosely of "sine" form, with an axially central main lobe nested
within a cone of sidelobes of decreasing amplitude, figure 8.9. In this case
the nominal half-beamwidth is 30A./L degrees.

Figure 8.9. The directional response of a circular aperture in an

infinite baffle, approximating to the response of, for example , an air-
backed circular ("piston" ) transducer element.
146 Underwater Acoustic Systems

8.4 Array Beamsteering [8.4-8.6]

The simple hydrophone array illustrated in figure 8.2 involved only the
summation of all hydrophone outputs, to "form" the beam. No shading was
used, nor was there any other form of processing following reception. The
result, for A./2 spacing and a reasonable number of elements within the array,
was an angularly selective main-lobe directed in the broad-side direction,
normal to the direction of extension of the array itself. In examining the
dipole, however, we noted that, with just two elements if, instead of simply
adding outputs, we inverted (phase shifted by 180.) the output of one of
the hydrophones before adding, then the broadside pattern would twist
through 90• and become an endfire pattern.

This principle may be extended for the general array to produce a beam-
steering effect which allows us to direct the narrow main lobe at any angle
we wish, to the line of extension of the array itself. Thus consider the array
implementation shown in figure 8.10. Here, the n'th element has, imposed
upon its output, a delay n't. The effect will be to produce a beam-swing
through an angle e given by

e= sin-1 (c't/d)

Of course, array shading may also be applied, in order to tailor the main-
lobe shape and minimise sidelobes. Indeed, quite complicated array processing
may be envisaged, which can allow the creation of steerable nulls in the
polar response, which can be used to eliminate a return from an unwanted,
off-axis target. Arrays constructed on this principle may take on a wide

n=O n=l n=2 n=3 n=N-2 n=N-1

2't 3't (N-l)'t

Beamformer Output

Figure 8.10. The "inertia-less" electronically steered

beamformer array
 Transducer Arrays 147

range of physical forms. Line and plane arrays have obvious benefits in
allowing transmitter or receiver beams to be formed and directed. However
cylindrical and part-spherical arrays are also used. It should also be mentioned
that, simple as the beamforming concept may appear, there are many
practical difficulties to be overcome in the development of effective
beamformers. In particular, the delay synthesis electronics is complex and
costly and there will almost inevitably be problems with element-to-element
interaction in the array itself.

8.5 Directivity Index

One consequence of improving the directional properties of a sound source

by forming an array is that the available power is directed within a smaller
solid angle than would have been the case if an omni-directional source
had been used. In this context, we have already seen, in section 2.3, that
if the array subtends a solid angle ell. then the on-axis acoustic intensity is
raised by a factor 41t/ell. Expressed in decibels this factor is the directivity
index, DI.

The quantity ell is calculated, if the array radiation pattern R{9,'1f) is known,
by evaluating the integral

ff
21t -Mt/2

ell= R(O, 'If) cos 9 dO d'lf

0 -1t/2

If the radiation pattern has rotational symmetry, the integral may be further
simplified thus

+1t/2
ell = 21t J R(9) cosO d9
-1t/2
Applying the result to our expressions for the continuous line of length L,
we find that ell = 2L/'A., and to the circular aperture of diameter L, we find
that ell = (1tL/'A.)2.

8.6 The Parametric Source [8.7-8.9]

In all our considerations thus far, in this chapter, we have dealt solely with
arrays of transducer elements so combined as to produce given transmission
characteristics. In the main, we find that desirable characteristics are: high
148 Underwater Acoustic Systems

sensitivity combined with directional angular response and freedom from

excessive sidelobes. In this and the next section, we investigate two other
methods of achieving high angular resolution, but with "non-physical"
apertures of large spatial extent. It should be stressed that, although both
methods to be discussed utilise transducers which are physically small by
comparison with the transmission wavelength, none the less the aperture
remains large, as indeed it must if effective beamwidth is to be small.

The first approach to creating a large non-physical aperture using a relatively

small transducer is known as "parametric" transmission. If a low-frequency
transducer is driven at adequately high power, a phenomenon known as
"cavitation" will be seen to occur. Cavitation is manifest as bubble formation
in the water in front of the transducer, coupled with a pronounced fluid
streaming, away from it. The effect occurs because, at high power drive
levels, the pressure fluctuations in the water can become so extreme as
literally to draw gas out of solution, tearing the water apart, as pressure
falls below ambient. Cavitation becomes more difficult to induce at depth,
because yet higher pressure amplitude fluctuations are needed. It is also
more difficult to induce if frequency is increased. Because cavitation
represents, in a sense, the approaching of an upper limit to possible pressure
amplitude fluctuations, studies of sound generation at such levels are
referred to as "finite amplitude acoustics". Fully developed cavitation, by
and large, is to be avoided in underwater acoustics because it can lead to
erosion damage on the face of high-power transducers. The effect finds
practical application in another context, in ultrasonic cleaning equipments.

The propagation of sound in water relies on the fact that the magnitude of
pressure fluctuations is directly related to the magnitude of particle velocity
fluctuations. At the levels of pressure fluctuation most frequently encountered
in underwater acoustics, the relationship between these quantities is linear
and described by the equation, analogous to Ohm's law in electrical circuit
theory

P = <m
Here p is rms pressure fluctuation, u is rms particle velocity and <J = pc
is, of course, specific acoustic impedance. If transducer face vibrations
become sufficiently great, then as cavitation is approached, this relationship
breaks down and the fluid medium begins to behave in a non-linear fashion.

If, then, we cause to be launched into the water by a high-power projector,

two frequency tones f1 and f2 , these tones will, as it were, spatially co-exist
in the water column in front of the projector. Since the water behaves non-
linearly, a power-series development is possible, wherein a multiplicative
 Transducer Arrays 149

mixing may be anticipated, producing along the length of the column in

front of the transducer, products of the form cos(ro 1t) cos(ro 2t). These
products may be re-written as sum and difference frequencies {cos(ro1 t -
ro2 t) + cos( ro1t + ro2t)} /2. It is only the difference frequency that is of interest
to us. Although the non-linearity produces only fairly weak mixing and
hence difference frequency generation at low power levels, relative to those
of the primary frequencies, the physical sites of difference frequency
generation are spatially distributed in front of the projector for a considerable
distance. This means that a spatially large, effectively endfire aperture at
the difference frequency can be obtained. This in turn implies high directivity
at the difference frequency.

In practice, we should normally choose primary frequencies in the region

of hundreds of kHz. Then we should have available some tens of kHz of
transducer bandwidth and an ability to synthesise a difference frequency
of up to half the projector bandwidth. This means that a very wide, albeit
relatively low-frequency, sweep of frequency is possible by the parametrically
generated secondary. Also, because the sidelobes at the two primary
frequencies do not exhibit any particular spatial correlation, there is virtually
no sidelobe structure at the difference frequency.

Parametric generation thus provides us with a method of generating low-

frequency sound, with extreme frequency agility, high angular resolution
and almost total freedom from sidelobes. The penalty is low efficiency of
conversion: in the region 1 to 10%. It has been suggested that an acoustically
transparent tube, containing fluid of lower vapour pressure than water, be
placed in front of the projector, in order to enhance the parametric generation.
The reader is referred to a paper by Muir [8.9] for further information on
the practical design of parametric sonars. The reader is also warned that
difference frequency generation can also occur within the drive power
amplifier, within any matching transformer or tuning inductor used with
the projector transducer, within the transducer itself and, indeed, within a
receive hydrophone or its associated electronic circuitry. This is because,
at high levels of drive, all these equipments may be expected to offer
significantly non-linear behaviour in some respect. Great care is needed,
therefore, in ensuring that the (weak) generation mechanism being sought
really is parametric in nature.

8.7 Synthetic Aperture Sonar [8.10-8.13]

If parametric sonar provides a virtual endfire array, then synthetic aperture

sonar establishes a virtual broadside array. The synthetic aperture principle
is widely used in terrain mapping radar [8.4]. It has, for example, been used
150 Underwater Acoustic Systems

Moving Antenna
or Transducer t Trajectory

~~---7--- signal intensity

flight time along trajectory

Figure 8.11. Non-coherent synthetic aperture sonar

for radar mapping both from aircraft and from earth and planetary
reconnaissance satellites. A major requirement with airborne or spaceborne
radars is that the antennae should not be overly large, for purely practical
aerodynamic reasons. This militates, at radio frequencies, against achieving
angular resolution which adequately complements range resolution capability.
The concept underlying the synthetic aperture principle is that, if an antenna
can be physically translated through space, it can act successively as the
individual elements on a long, linear array. The way in which this can occur
is illustrated in figure 8.11 in the context of "non-coherent" aperture
synthesis.

The non-coherent synthetic aperture is realised by recognising that the

crescent-shaped pattern of received signals on the graph of signal round-
trip delay time versus flight time of the antenna is a characteristic identifier
of the target. Indeed, cusp-shaped features are a commonplace on the output
of the dry-paper recorders such as are used in conjunction with many modern
echo-sounders. Whenever the echo-sounder records a strong but small-scale
feature such as a rock-scaur or a dense school of fish, the crescent form
 Transducer Arrays 151

may be noticed. Of course, the echo-sounder does not seek to make use
of such information. However, it is not difficult to imagine that some
numerical method could be devised to gather up the spatially distributed
information in the crescent and present it as a high-resolution point at the
target location. In fact, some airborne synthetic aperture radars utilise a film
recording process, whereby specially designed conical lenses re-focus such
features. Although it is relatively easy to appreciate the nature of the
synthetic aperture principle by considering the non-coherent case, it is
actually preferable to measure both the amplitude and the phase of the
incoming signals. Then the delay time versus flight time diagram becomes,
effectively, a zone-plate hologram and the reconstruction is, essentially,
holographic in nature, whether performed optically or numerically.

For many years, the possibility of creating a practical synthetic aperture

sonar has been held in question. Three objections to the possible efficacy
of the technique are commonly quoted. First, it was supposed that the
stability and control of any towfish could not be accurate enough to maintain
an adequately linear trajectory. The required trajectory should hold to a
straight line, to within a fraction of a wavelength, over many complete
synthesised apertures. The author recalls, for example, a photograph published
in the mid 1970's which purported to show an actively controlled US Navy
experimental towfish for synthetic aperture research. The nose-cone bore
the legend "Murphy's Law" attesting, we may presume, to the many
experimental difficulties encountered with the device. However, depending
upon context, there seems no reason why the problem of towing stability
should be insuperable, or that the tow geometry and trajectory should not
be sufficiently accurately monitored to make suitable correction possible.
For example, there is some current interest in the use of synthetic aperture
to enable low-frequency, long towed arrays to be reduced in size. Then
wavelengths are such as to suggest that towing errors would be handleable.

A second objection had to do with water path inhomogeneity. Again, whilst

this might be a problem in some localities or with some particular applications,
such as high-frequency, ultra-high resolution sidescan employed in estuarine
site survey situations, the difficulty does not, at lower frequencies and in
open ocean water, appear to be serious. The third reason had to do with
the fact that, because sound travels so much more slowly that light, the sonar
-unlike the radar- would exhibit an unacceptable spatial aliasing problem.
The use of frequency modulation encoded transmission appears to circumvent
this last difficulty [8.13]. Whereas it is reasonably certain that the synthetic
aperture principle can be made to work for active sonar, in terrain mapping
and other applications, it is less certain that the technique is applicable to
passive target detection. Finally, it is interesting to note that, whereas in
a conventional sonar, a large physical aperture is essential for good azimuth
152 Underwater Acoustic Systems

resolution, with a synthetic aperture, the moving antenna or transducer, the

physical aperture should be small. This is because the desirable mode of
operation is omnidirectional insonification of the field being mapped.

References

[8.1] R.N. Bracewell, The Fourier Transform and its Applications, 2nd edition, revised,
McGraw-Hill, New York, 1986

[8.2] F.J. Harris, On the Use of Windows for Harmonic Analysis with the Discrete Fourier
Transform, Proc. IEEE, Vol. 66, Jan. 1978, pp. 51-83

[8.3] C.S. Clay and H. Medwin, Acoustical Oceanography, Wiley, New York, 1977, pp. 138-
177

[8.4] M.L. Skolnik, Introduction to Radar Systems, McGraw-Hill, New York, 1980

[8.5] Special Issue on Beam Forming, IEEE. J. Oceanic Engineering, Vol. OE-10, No. 3,
July 1985

[8.6] Special Issue on Underwater Signal Processing, IEEE. J. Oceanic Engineering, Vol.
OE-12, No. 1, January 1987

[8. 7] H.O. Berktay, Possible Exploitation ofNon-Linear Acoustics in Underwater Transmitting

Applications, J. Sound Vibration, Vol. 2, p. 435, 1965

[8.8] H.O. Berktay, Some Finite Amplitude Effects in Underwater Acoustics, in Underwater
Acoustics, Vol. 2 (V.M. Albers, ed.), Plenum Press, New York, 1967, pp. 243-261

[8.9] T.G. Muir, Non-linear Acoustics and its Role in the Sedimentary Geophysics of the
Sea, Physics of Sound in Marine Sediments (L.L. Hampton, ed.), Plenum Press, New York,
1974, pp. 241-287

[8.10] L.J. Cutrona, Comparison of Sonar System Performance Achievable Using Synthetic
Aperture Techniques with the Performance Achievable by More Conventional Means, J.
Acoust. Soc. Am., Vol. 58, No. 2, 1975, pp. 336-348

[8.11] L.J. Cutrona, Additional Characteristics of Synthetic Aperture Sonar Systems and a
Further Comparison with Non-synthetic Aperture Sonar Systems, J. Acoust. Soc. Am., Vol.
61, No. 5, 1977, pp. 1213-1217

[8.12] P. de Heering, Alternate Schemes in Synthetic Aperture Sonar Processing, IEEE J.

Oceanic Engineering, Vol. OE-9, No. 4, October 1984, pp. 277-280

[8.13] P.T. Gough, A Synthetic Aperture Sonar System Capable of Operating at High Speed
and in Turbulent Media, IEEE J. Oceanic Engineering, Vol. OE-11, No. 2, Aprill986, pp.
333-339
9 Sonar Engineering and Applications

9.1 Introduction

In this chapter we examine the way in which electronics and acoustics

interact to provide practical solutions to a wide range of sonar engineering
problems. It is a matter of some regret that the history attaching to the
development of the subject of underwater acoustics is relatively poorly
documented. As history, the documentary material is of relatively recent
origin and is frequently difficult to gain access to because of the explicitly
military nature of much of the research which has been conducted during
the past several decades. An excellent general account, of a largely non-
technical nature, has been published by Haines [9.1]. This text views the
development of underwater acoustics from a British standpoint but with
copious reference to contributions made elsewhere. Urick's introduction
[9.2] provides a brief historical perspective, which is nicely complemented
by the first background chapter in a recent text by B urdic [9 .3]. A text edited
by Albers [9.4] provides a collection of benchmark papers of particular
interest to the underwater acoustician.

Reference is frequently made to what is possibly the earliest written reference

to the detection of sound at sea, which was made by Leonardo da Vinci
[9.5], writing in 1490: "If you cause your ship to stop, and place the head
of a long tube in the water and place the outer extremity to your ear, you
will hear ships at a great distance from you." Despite the obvious mismatches
in acoustic impedance involved in using such a principle, variants of this
technique (almost certainly without regard to Leonardo's original observation)
were employed during World War I, both at shore listening stations and
on surface vessels, for the primary purpose of submarine detection and
localisation. The simplest such device, which can easily be replicated and
tested, consisted of a pair of submerged air-filled rubber bulbs separated
by about two metres, connected by stiff-walled tubes to a pair of stethoscope
earpieces. The device has directional characteristics for frequencies in the
medium and high audible range.

153
154 Underwater Acoustic Systems

The first electrical sensors utilised carbon button microphones contained

in flexible rubber tubing some 10 metres in length, to form a towed array.
This device was introduced during the last months of World War I and was
a development which specifically recognised the need to remove the sensor
from the self-noise generated by the ship deploying it. Again, simple
microphonic detectors based upon this principle are easily made. Suitable
button microphone elements are still to be had in army surplus stores: they
form a part of the throat-microphones used by aviators in past years.

Although the early objective uses of sonar were certainly of a military

nature, it was the loss of the Titanic in 1912 that first stimulated thought
into ways of detecting objects in the sea - in this instance, icebergs rather
than submarines- by using active ultrasonics. In 1914, Fessenden was able
to demonstrate the detection of an iceberg at two miles range using a
moving-coil transducer [9.6]. Work by Langevin, Chilowsky, Boyle and
others lead to the development, in the years during and following World
War I, of quartz transducers, and metal-quartz-metal sandwich transducers
of the type discussed in section 7 .3. From this time but particularly during
the period of World War II, with the rapid development of electronics as
a valuable complementary technology, sonar engineering became firmly
established as a scientific and engineering discipline in its own right.

In the period following World War II, commercial developments connected

with marine civil engineering survey, hydrographic survey, oil-prospecting,
scientific investigations and sub-sea site operations, such as navigation and
re-location, have taken underwater acoustics out of its narrower military
framework, making it a major enabling technology for a wide range of
maritime activities. This burgeoning in application has gone hand in hand
with vital developments in transducer materials, in particular the discovery
of the lead titanates and their generic successors. Equally important has
been the rapid increase in complexity and decrease in cost of electronic
components and systems for signal generation and processing and, indeed,
the dramatic progress in development of appropriate algorithmic techniques
for a host of signal processing activities.

9.2 The Basic Echo Sounder

The simplest and most widely used of all sonar equipments is the Echo
Sounder, which is to be found on all Merchant and Military shipping and,
in any of a number of economical designs, on even the most modest of small
pleasure and sporting craft. In essence the device is quite simple, consisting
of a transmitter amplifier capable of generating a voltage pulse of sufficient
amplitude to launch, almost always by means of a piezo-electric transducer,
 Sonar Engineering and Applications 155

an acoustic sounding pulse vertically downwards into the water. Often,

although not exclusively, the same transducer will act to pick up the received
signal, which will be suitably amplified and then detected to allow estimation
of the travel time to the sea-floor and back. In pursuing the following
description, emphasis will be placed on the "unclever" parts of the circuitry.
The microprocessor-based, real-time signal processing is not usually the
greatest problem. Being sure that the electrical engineering is reliable and
the mechanical design well thought out requires great care.

Figure 9.1 provides a block diagram of the basic echo sounder. Operation
is as follows. The master clock generator issues timing pulses whose period
is equivalent to the round-trip delay for a pulse reflected from the sea-floor
at maximum depth. Thus, for an instrument required to operate in the range
0-150 m, the maximum round trip is 300m and the pulse repetition period
is 200 ms. Each master clock pulse initiates the generation of a fixed-width,
pulsed sinusoid. For a small-boat echo-sounder intended to operate in
shallow water with reasonably good depth resolution, choice of operating
frequency is often dominated by the availability of cheap 150kHz transducer
elements. If greater depth capability is required, a lower frequency - some
few tens of kHz - will be used.

The RF pulse waveform is then amplified. The amplifier will be designed

to generate peak power at a level appropriate for a particular application.
A power of the order of watts will be required for the high-frequency, small-
boat echo sounder, tens or even hundreds of watts for low-frequency deep

--411111~ PWse Waveform

Master
Clock
RF
PWse
~
Gcncn.tor Gcncn.tor

Power
AmpIiller

<J
Low Noise
---------
TxSignal
'
Variable Gain Bandpass
Amplifier Selective Amplifier Ltmiter
Amplifier

Figure 9.1. The basic echo sounder shown in schematic form

156 Underwater Acoustic Systems

water echo sounders. The objectives in designing a suitable semiconductor

power amplifier are: low output impedance, high efficiency and adequately
high current output capability. The last of these criteria is important because
with semiconductor circuitry, the power line voltages will be very much
less (some tens of volts) than the drive voltages (hundreds or even thousands
of volts) which will typically be needed to generate the design power levels
from the (relatively high impedance) transducer. Consequently, the power
amplifier will be required to supply high current to the primary of a
transformer which will have a substantial turns ratio.

The transformer itself will be required to be quite carefully designed. At

the drive frequencies envisaged for virtually all echo-sounder applications
(although not necessarily for sub-bottom profilers, which operate at
frequencies of the order of 1-5kHz) ungapped ferrite cored assemblies will
be mandatory. For the higher frequency applications, these may be of the
pot-core type. Otherwise, E-core assemblies will probably be found to be
more appropriate. If the transformer is properly designed, the core magnetic
volume will be relatively unimportant. Core size will most probably be
dictated by the volume of copper needed to accommodate primary current
with safe turn-to-turn voltage difference and by the fact that winding layers
in the secondary will require to be individually insulated, to cope with the
need to place hundreds or even thousands of volts across the transducer
element.

It is, of course, possible to wind the transformer and measure, using a

suitable meter or bridge, the secondary inductance. This inductance cannot
be used to tune the static capacitance of the transducer crystal. To see why
the transformer, if restricted to provide a high-efficiency voltage step-up,
should not be used to provide tuning, it is necessary to recall that the high
coupling coefficient of an ungapped ferrite core results in extremely low
leakage-flux and thus only very small values of primary and secondary
leakage inductance L and L •. Note that leakage inductance is not the same
as the primary and se~ondary winding inductance. The winding inductances
should be large. Indeed, they should be infinite for an ideal transformer.
Furthermore, they will also be effectively unseen by any terminating circuitry.

Any secondary tuning must derive from the leakage inductances, referred
to the secondary, where they will appear in series with the referred amplifier
output impedance. The amplifier output impedance should, with good
design, be negligible. Even the total referred leakage inductance for a
transformer with a l:N turns ratio, which will be L. + N2LP, will usually
be far too small to provide tuning. Of course the leakage inductance could,
in principle, be increased to a value which might tune the transducer. This
would, however, require air-gapping of the core. This, in turn, would
 Sonar Engineering and Applications 157

increase the core reluctance, to the detriment of effective power transfer.

The inevitable conclusion is that separation of function is advisable: let the
transformer act primarily as an impedance changer, and employ a series
inductor, specifically designed for the purpose, to cancel static capacitance.

Design of the inductor is not necessarily straightforward. In many respects,

it presents a significantly more difficult task than designing the output
transformer. The following point should be borne in mind. If the tuning
inductor begins to saturate, it will act as a non-linear circuit element. In
the following account the reader will be reminded of describing function
theory, in control engineering. Suppose, first, the drive voltage to be modest,
with the inductor behaving as it should, without evidence of core saturation.
Suppose also the drive frequency, the tuning inductance value and the static
capacitance to be such as to engender resonance. Now gradually increase
the drive voltage. The voltage across the inductor will increase proportionally.
So also will the current through it, and its core flux. At some point, the
flux will rise to such a magnitude that saturation will begin to occur. A
clipping of the current waveform will result. The saturating inductor will
then effectively enter a phase shift into its current waveform which, since
resonance is critically phase sensitive, will immediately de-tune the resonant
circuit with consequent loss of power.

There is a rather crude moral in this story. If the transducer you wish to
tune, hefts heavy in the hand, do not expect to tune it with a 15 mm pot
core! A considerable volume of ferrite will be required to hold adequate
energy within its magnetic field, per half cycle, to pass on to the electro-
mechanical resonator which is the transducer.

At this point, the amplified RF pulse has been delivered to the transducer,
from which it is emitted as a high-power acoustic pulse. The pulse travels
to the sea-floor, reflects with some loss, and returns to the same transducer,
which acts reciprocally to detect its presence and convert it into an electrical
signal. It is true that two transducers can be used, one to transmit and one
to receive. This approach increases cost in some measure. In some sonars,
it is imperative to use separate transmit and receive transducers, if different
insonification and inspection characteristics are required. This is often the
case with terrain-mapping sector-scanning sonars, for example. However,
it is when a single transmit/receive transducer is used that circuit design
complications result, so that is the situation we treat here. Clearly, because
of the extremely high drive voltages during transmission, it is inappropriate
merely to connect the receive amplifier directly to the transmit/receive
transducer. Instead a solid-state changeover network must be employed.
One such is shown in figure 9.1. During transmission, the parallel diodes
in series with the transformer secondary present a low impedance connection
158 Underwater Acoustic Systems

between transformer and transducer. On reception, when the voltage across

the transducer will be small, perhaps no more than millivolts in amplitude,
the diodes will present a relatively high slope impedance, effectively
isolating the transmission circuitry from the transducer. Again, during
transmission the shunt diodes across the amplifier input will conduct,
presenting a fraction of the transmit signal no more than the forward diode
knee voltage in amplitude, to the receiver. On reception, again because of
the relatively high slope resistance for input signal amplitudes with a
magnitude much less than the knee voltage, the shunt diodes will appear
as an open circuit. More complex, active switches can be devised but the
circuit shown in figure 9.1 will work well in many applications.

The first receive amplifier will almost always be chosen to be a solid-state

low-noise design, using either discrete components or an ac coupled,
integrated low-noise amplifier. Although the transducer will be to some
extent selective, it is most probable that it will respond to a number of
unwanted frequency bands, because of the presence of spurious resonances.
It may well be necessary to make the first low-noise amplifier at least
broadband selective at the transmit/receive frequency. If this is not done,
it is possible that out-of-band ambient noise may swamp the amplifier,
driving it into saturation and thus inhibiting the detection of the wanted
signal, despite any subsequent bandpass filtering.

Making the front-end amplifier broadband selective will not normally

provide enough bandpass filtering to eliminate all the out-of-band ambient
and front-end receiver noise. It is customary to introduce several stages of
selective, narrowband bandpass filtering prior to detection, in order to
achieve this.

The fact that the acoustic signal both spreads and is attenuated in passing
to the sea-floor and back, means that the received signal amplitude falls,
roughly as the square of water depth. In order to compensate for this effect,
a time variable gain (TVG) circuit is usually incorporated. This is simply
an amplifier, or cascade of amplifiers, whose gain is dynamically increased
by a ramp control voltage waveform triggered by the transmit pulse.

Finally, envelope detection will allow the received RF pulse reflected from
the sea-floor to be isolated and prepared for use as the round trip timing
pulse. The timing circuitry is relatively straightforward and in a modern
echo sounder might present its output on a numeric display, or on a monitor
screen in any of a variety of formats, or on some form of dry-paper chart
recorder capable of presenting a "facsimile" display.
 Sonar Engineering and Applications 159

9.3 Sub-bottom Profiling

As has been mentioned, the basic echo sounder finds use in ship and small-
boat navigation where, more often than not, its function is that of a warning
device, to indicate inadequate under-keel clearance. It also finds use, in
much the form described here, but with operation at only some few kHz,
in sub-bottom profiling. The use ofthese lower acoustic frequencies reduces
the potential resolution but allows penetration into marine sediments,
permitting geotechnical inspection of the sea-floor, as well as possible
location of buried objects such as wreck artifact, cables or oil-pipelines.
Figure 9.2 shows a sub-bottom profiler output. Features beneath the sea-
floor, such as sediment layers and rock outcrops, can be clearly identified.

Because high power is required to penetrate the sediment, a range of

alternatives to the more conventional piston-type piezo-electric transducer
(section 7 .3) has been devised. These include sparkers, where a bank of
high-value capacitors is discharged across a spark-gap under water, and a
range of air-gun and explosive detonation devices, synchronised in some
way to the master clock. Repetition rates for such applications are likely
to reflect, respectively, the spatial scale of the sub-bottom features being
I
I

' . "'·

Figure 9.2. A sub-bottom profi/er record, taken off the West African
Coast, showing a sediment layer above consolidated sediments
representing the channels of an ancient river delta. A gas emission can
also be seen on this record (courtesy of Gardline Surveys Ltd, Great
Yarmouth)
160 Underwater Acoustic Systems

investigated, and the survey speed. Because the acoustic aperture of reasonably
easily handleable low-frequency soundhead arrays will be relatively small,
and the beam-spread and resolution uncomfortably large, this is an area of
application where parametric sonar (see section 8.6) may yet prove valuable.
Yet further future improvements in sub-bottom profiling sonars are likely
to involve spread-spectrum encoding of the transmitted waveform to assist
in improving range resolution.

9.4 Fishing Sonars [9.7]

The simplest echo sounders can be used, with skill, in fish-finding applications,
since a clear return signal is often obtainable from fish swimming in shoals.
However, the application is one of commercial importance and a range of
improvements upon the basic echo sounder have been made to enhance its
value to the fishing community. Clearly, a "facsimile" type display, such
as produces a strip, dry-paper record of the water column beneath a boat,
as the boat progresses on it way, is of value as a diagnostic aid since, in
addition to providing a permanent record of past events, it allows a measure
of visual integration, making possible the detection of features which, on
a scan-to-scan basis, might not be apparent.

One shortcoming of the dry-paper record is that little information concerning

received signal strength is available. In order to improve interpretation yet
further, the detected output of a basic high-frequency (-100-200 kHz) echo
sounder may be processed to yield a gray-scale or, even better, a false-colour
display of received signal strength as a function of time. Yet a further
sophistication is to be found in the use of several possible different transmission
frequencies. The idea behind multi-frequency echo-sounding is that the
reflective characteristics of different fish sizes, species or sea-floor types
at different frequencies, might in some sense provide improved diagnostic
information. Ideally, one might wish for an echo sounder capable of emitting
an ultra broad-band sounding pulse, extending from some tens of kHz to
the high hundreds of kHz. The acoustic returns from such a pulse could
then, effectively, be spectrum-analysed within the epoch containing energy
reflected from a particular feature of interest, such as a fish-shoal. The major
problem in devising such an equipment lies in the design of adequately
broadband, efficient and cheap acoustic transducers with a suitably uniform
frequency response. Currently, multifrequency sonars use but a restricted
set of narrowband transmissions and are capable of only modest manipulation
of the data so acquired.
 Sonar Engineering and Applications 161

Most fishing echo sounders operate in a vertical transmission mode. Some

are, however, adaptable to provide a rotating sweep at some reasonably
acute angle to the horizontal. This makes possible a plan-position indication
of sub-sea features, which could be either on the sea-floor or in mid-water.
The method is thus particularly attractive for the remote location of shoals
of fish or other commercially valuable shoaling marine life such as shrimp
or squid. Whilst it is clearly simplest to rotate the scanning soundhead
mechanically, there has been much interest in either totally electronic
rotational scanning, or partially electronic and partially mechanical scanning.
The technique used is referred to as sector scanning sonar and involves the
design of arrays and beam-forming electronics such as was described in
section 8.5. A major obstacle to the wider adoption of sector scanning sonar
is the high cost of the multi-element transducer arrays and beam-forming
electronics, and the rapid escalation of this cost with demand for increased
cell resolution.

In the context of horizontally-directed plan position indicating sonars for

fisheries research, interesting work has been conducted on the development
of miniature transponding tags which can be sewn without harm to the skin
of a fish to enable fish-tracking experiments to be conducted. The tags, an
example of which is shown in figure 9.3, are activated by the transmitted

Figure 9.3 Miniature transponding fish tags (courtesy of Fisheries

Laboratory, Ministry of Agriculture, Fisheries and Food, Lowestofl)
162 Underwater Acoustic Systems

sonar beam, and return a signal at the same frequency, but with a much
greater signal strength than would the fish being tracked. So sophisticated
has become the development of these transponding tags that measurement
and re-transmission of compass direction data, to a resolution equal to "eight
points of the compass rose", has been achieved [9.8] .

Yet another aspect of fishing sonar development has to do with quantifying

the size of a marine population, either by counting or by some attempt at
biomass estimation. The former approach involves the use of narrow-beam,
high-frequency echo sounders of high spatial resolution. The latter involves
monitoring the acoustic backscatter from a population, when insonified by
a directional high-frequency echo sounder, and using this signal to infer
the biomass density [9.9].

9.5 Side-scan Terrain-mapping Sonars

The side-scan sonar utilises much the same electronic system architecture
as the basic echo sounder. It provides an alternative solution to the plan-
position indication problem referred to in the previous section. Side-scan

Towina and 1n01111ission·line cable

Figure 9.4 The side-scan sonar towfish, typically about a metre long
and with a narrowband pulsed transmission at a centre frequency in
the band 100-500 kHz
 Sonar Engineering and Applications 163

(a)

(b)

Figure 9.5 Side-scan sonar images: (a) a wreck in the North Sea,
showing deck detail and, thrown into the sand, the pale acoustic
shadow. showing details of superstructure (courtesy of Gardline
Surveys, Great Yarmouth); (b) sand water ripples on the sea-floor
(courtesy of Fisheries Laboratory, Ministry of Agriculture, Fisheries
and Food, Lowestoft)
164 Underwater Acoustic Systems

sonar is widely used in marine site- and pipeline-surveying, and is important

in mine countermeasures, in locating and identifying enemy mines. The
principal features of the side-scan system are illustrated in figure 9.4. The
most important feature of the equipment is the soundhead itself which, being
many (typically about 50) wavelengths long, is of unusually narrow beamwidth
in the horizontal plane, and thus provides good azimuth resolution.

Whilst it might be possible to mount such a soundhead on a surface vessel,

much as the ordinary echo sounder soundhead is mounted, but set in a
sideways looking attitude, several factors militate against this. First, servo
stabilisation of the soundhead would be needed, to counter the effects of
pitch and roll while the vessel was under way. This is entirely feasible, and
is done for some hydrographic survey echo sounders of a specialised nature.
However, it is a technique which adds substantially to cost. Utilising the
towfish principle effectively eliminates the stabilisation problem.

It also removes the soundhead from the self-noise of the towing vessel. Since
echo returns will be at oblique incidence and therefore rather weak, this
is at least beneficial. Furthermore, the soundhead can be towed nearer the
sea-floor than if ship-mounted, and this improves the short-range swath
coverage. Finally, the entire equipment can be transported easily and can
be rapidly set up on even quite small vessels of convenience.

The side-scan sonar will typically operate to generate about 100 W of pulsed
acoustic power at a frequency in the range 100-500 kHz. The swath width
will typically extend to about 100 m on either side of the towfish, which
will carry identical soundheads on each side. Each outgoing pulse will
illuminate a lateral patch on the sea-floor and the return signal from this
patch will establish one scan-line of a "raster scan" picture of the sea-floor,
which builds up as an image on, usually, a dry-paper facsimile-type recorder,
as the vessel proceeds along its tow. As the image shown in figure 9.5 shows,
considerable detail of the sea-floor may be obtained. Sandwave ripples,
rock-scaurs, exposed pipeline, cable and anchor chain, as well as wreck and
other debris, may be clearly identified by a well-executed survey. The higher
frequency sidescans will offer the best resolution and in the band 100-500
kHz there will be relatively little degradation of range. The author is aware
of sidescan designs in the low MHz range, with a swath width of the order
of 25 m, which have been investigated for mine identification purposes.

At the other extreme of operational scale, "GLORIA", the Geological Long

Range Inclined Asdic, developed at the United Kingdom Institute of
Oceanographic Sciences, and currently operated under contract by Marconi
Underwater Systems Ltd, utilises a sidescan array which is about 5 m long
 Sonar Engineering and Applications 165

by 1.5 m high, operating at a transmission frequency of 6.5 kHz with a pulse

length of 30 ms and a pulse transmit power of 50 kW. It can cover a swath
of width up to 27 km to one side of the towfish and, as its name suggests,
was originally designed for rapid sea-floor mapping activities for geological
investigations [9 .10].

9.6 Seismic Survey [9.11]

In this sphere of activity we move farthest from the structure of the

conventional echo sounder, although the operating principle remains broadly
the same. We have seen how, by using transmission frequencies of a few
kHz, penetration of the sea-bed may take place and sediment layering and
other features may be revealed. For geological studies and, particularly, for
offshore gas and oil prospecting, dramatically deeper penetration of the sea-
bed is needed, often to depths in excess of a kilometre. In order to achieve
such penetration, two major changes in system design are necessary. First,
it becomes mandatory to use low-frequency (1 kHz and below) "seismic"
shock waves, rather than piezo-electrically generated acoustic pulses. These

Figure 9.6. The survey ship detonates a shock wave (short tow cable)
which passes on many paths to (three of many) hydrophone elements
on the long towed array. Sediment acoustic properties (sound speed,
density) vary within the layers. A salt-dome enclosing a gas pocket is
also shown
166 Underwater Acoustic Systems

shock waves are typically generated either by an air-gun, which releases

compressed air rapidly into the water, producing a bubble field very similar
to that released by a dynamite detonation, or by an oxygen-propane gas-
detonation in a flexible-walled chamber within a shallow-towed submerged
raft. Second, it is necessary to use a long towed array as the receiver. The
array will typically contain some two to four dozen hydrophone elements
within a flexible, oil-filled plastic tube of diameter (about) 5 em and length
up to 3 km.

The shock wave generated on each firing by the survey ship passes downwards,
reflecting off the sea-floor to produce, at the various hydrophones in the
array, a first reflected shock wave. Further reflected shock waves will result
from interaction with the sediment interfaces. Particularly strong reflections
will be identified from gas pockets beneath impermeable salt-domes, because
of the marked change in acoustic impedance at such an interface.

The data processing activity which takes place in analysing the time sequences
resulting from successions of shock-wave detonations as the ship executes
a survey is one of the classical "inverse problems" of mathematics. Time
of flight on many reflective and refractive passes through many sediment
layers will be known, but layer thickness, sound speed and density will not.
The task is to establish, and refine the parameters of, a model predicting
these unknown quantities. This is done by numerical calculation, with
human interaction.

9.7 Acoustic Positioning and Navigation [9.12]

In discussing the use of fish-tags for tracking purposes, in section 9.4 above,
the concept ofthe transponder was introduced. Transponders are equipments
which, deployed in pairs or in larger networks, can interact with each other,
to allow determination of their separation by acoustic pulse time-of-flight
measurement. The simplest transponders are acoustic beacons of the type
used in conjunction with aircraft flight recorders. Such beacons remain in
a passive, listen-only mode, until awoken by an interrogation pulse, usually
transmitted by a searching recovery vessel. They then respond to further
interrogation pulses allowing, to some degree, the surface vessel to position
itself vertically above the beacon, at which time the transponding round
trip delay will be a minimum.

The method is extended in underwater navigation systems, by emplacing

a net of at least three slave transponder units at particular geographic
locations. The master transponder is able to activate coded responses from
each of the beacons so that, as figure 9.7 illustrates, slant ranges between
 Sonar Engineering and Applications 167

[Q Mobile Master Transponder ''

''
''
'''

\ Slave 2
''
''

Figure 9.7. The circular acoustic navigation principle

the master and each of the slaves may be measured and, by a process of
triangulation, location within the survey area determined. The literature of
the subject distinguishes between long-baseline and short-baseline systems.
This distinction is less than completely fundamental. It does not (generally)
relate the terms "long" and "short" to, for example, transmission wavelength.
Rather, it tends to reserve the former terminology for sea-floor slaves and
the latter for ship- or platform-mounted equipments.

The method illustrated in figure 9.7 is unambiguous spherical navigation.

The mobile transponder may be thought to be at the unique point of
intersection of three hemispheres centred on each of the three slaves and
of radii equal to each of the respective slant ranges. The intersection of
only two such hemispheres would produce a vertical semicircular locus of
possible position, rather than a uniquely determined point of fix.

Circular, or transponding navigation carries with it the penalty that the

mobile must carry an active transmitter. In radionavigation applications,
this would prove to be a severe disadvantage, with many users competing
to communicate with the slave stations. The problem can be avoided, and
the active transponder replaced by a passive receiver, if hyperbolic rather
than spherical navigation is used. Consider for example, the geometry
illustrated in figure 9.8. Here, two beacons and the mobile are shown located
within a single inclined plane surface. On this surface are shown the
hyperbolic loci which mark out lines of constant time difference in reception
of signals which had been emitted simultaneously from the two beacons.
If the mobile measures what it "hears" as the time difference, it can
168 Underwater Acoustic Systems

determine on which hyperbola it lies. Naturally, with only two beacons the
system is ambiguous. The ambiguity may be entirely or contextually removed
by overlaying one or more further patterns of hyperbolae and/or utilising
additional information, such as depth data. Hyperbolic navigation in
underwater survey is less frequently employed than spherical navigation.

Figure 9.8. Hyperbolic acoustic navigation

9.8 Doppler Measurements

The doppler effect is the frequency shift which occurs in perceived sound
when either an observer moves with respect to the transport medium, or
the medium itself moves. The effect is employed to advantage in several
sonar measurement equipments among which is the doppler current meter.
This device, illustrated in figure 9.9(a), is used to measure water velocity
at a point, or movement of an object through the water. The operating
principle is that of a high-frequency continuous wave sonar with spatially
separated transmit and receive transducers. The horizontal flow component
along the sound axis between heads A and B and heads C and D of the meter
generates doppler shifts of magnitude

From these equations, speed v and the direction angle e are easily obtained.
 Sonar Engineering and Applications 169

A loosely similar principle is used in doppler logs. The Janus configuration,

illustrated in figure 9.9(b), is frequently used in this application. It allows
the forward speed and sideways drift of a submersible to be determined.
The calculations involved are similar, except that an additional angular
dependence is introduced by the downward depression angle of the beams.
The doppler shift information is contained in backscattered, rather than
reflected, sound.

(a) (b)

Figure 9.9. Applications of doppler measurement in underwater

acoustics

A range-gated doppler log has recently been successfully operated for ocean
remote current sensing. Here the concept is to inspect the doppler shift on
signals backscattered in the consecutive range cells of a pulsed high-
frequency sonar. The instrument takes advantage of the signal processing
power available using modern microelectronics.

Finally, the doppler principle has been employed with considerable success
in monitoring sea-floor geotechnical properties [9 .13]. Here, a constant
frequency 12 kHz transmitter is housed in the tail of a free-fall, torpedo-
shaped projectile. As the projectile, which is known as a penetrator, descends
it accelerates, with corresponding doppler shift, until it reaches a terminal
velocity which, for a two-tonne penetrator can exceed 100 miles per hour
(50 m s-1). At this speed the received signal at the surface is approximately
11.6 kHz. On impact, the penetrator decelerates and the doppler shift
decreases to zero. This allows the deceleration profile to be measured and
the depth of penetration to be calculated. This in turn allows the sediment
strength to be estimated remotely, without coring.
170 Underwater Acoustic Systems

References

[9.1] G. Haines, Sound Underwater, David & Charles, Newton Abbott, 1974

[9.2] R.J. Urick, Underwater Sound for Engineers, McGraw-Hill, New York, 1975

[9.3] W.S. Burdic, Underwater Acoustic System Analysis, Prentice-Hall, New Jersey, 1984

[9.4] V.M. Albers, Underwater Sound, Dowden Hutchinson and Ross, Stroudsberg, Penn.,
1972

[9.5] E. MacCurdy, The Noteboo/cs of Leonardo da Vinci, Garden City Publishing Co., New
York, 1942

[9.6] R.A. Fessenden, US Patent Application 744,793 (1913)

[9.7] R.B. Mitson, Fisheries Sonar, Fishing News Books, Farnham, 1983

[9.8] N.D. Pearson and T.J. Storeton-West, The Design of an Acoustic Transponding Compass
Tag for Free-Swimming Fish, Proc. 5th !ERE Inti. Conf. on Electronics for Ocean Technology,
Edinburgh, September 1987, pp. 83-92

[9.9] R. Coates and C. Orgill, Population Density Measurement by Acoustic Backscatter,

IEEE Oceans '87 Conference, Halifax, Nova Scotia, September 1987

[9.10] R.B. Whitmarsh and A.S. Laughton, A Long-range Sonar Study of the Mid-Atlantic
Ridge Crest near 37"N (FAMOUS Area) and its Tectonic Implications, Deep Sea Research,
Vol. 23, 1976, pp. 1005-1023

[9.11] E.A. Robinson and T.S. Durrani, Geophysical Signal Processing, Prentice-Hall,
London, 1985

[9.12] P.H. Milne, Underwater Acoustic Positioning Systems, Spon, London, 1983

[9.13] R. Coates, A Deep-ocean Penetrator Telemetry System, IEEE J. Oceanic Engineering,

Vol. 13, No. 2, April 1988, pp. 55-63
10 Acoustic Communications

10.1 Introduction

The literature surrounding the subject of underwater acoustic communications

is, in some respects, surprisingly scant. This is particularly the case if one
concentrates only upon that material directly concerned with actual underwater
communication systems as opposed to more general aspects of propagation
and channel modelling. The major difficulties with which the communication
engineer is concerned, when attempting to design underwater acoustic
communication systems, revolve around the problems of reverberation and
multipath transmission and high attenuation at high acoustic frequencies.
An extensive bibliography on the subject has been published by the author
[ 10.1]. A subset ofthat bibliography, dealing with selected specific underwater
communication systems is presented here [10.2-10.13]. However, the reader
is referred to the original source for more information on specific
communication systems, and papers on general aspects of channel modelling
and more detailed mathematical treatments than can be handled in a text
of this nature.

Acoustics is not, of course, the only method of obtaining underwater

communication: cable systems (in many different contexts) are widely used
-but carry the obvious and frequently unacceptable disadvantage oftethering
the remote end of a link. Fibre-optic methods have also been used, to a lesser
extent, as yet - though doubtless that situation will change in the future.
The problem of tethering remains, as does the need for copper, in providing
a supply of electrical power to the sub-sea installation.

Electromagnetic propagation, using both radio and laser transmission, has

been considered. However, except in particular and unusual circumstances,
neither approach is of great value. Because salt water is conductive, only
the lowest- barely usable- radio wavelengths will propagate any distance.
These are emanations in the ELF (Extra Low Frequency - 30 Hz to 300
Hz) band. Signals propagated in the ELF band require large transmitter
powers and large antennae. ELF band propagation has been studied intensively
during the past two decades with a view to establishing "bellringer"
171
172 Underwater Acoustic Systems

communication with the nuclear submarine fleet. That is, given a channel
of such restricted information bandwidth, simply use it to request, from a
normally covert submarine fleet, use of surface-deployed high-frequency
radio antennae, to download a rapid stream of short-term, tactically valuable
information.

Curiously, visual light frequencies are the least attenuated of all

electromagnetic emanations, by salt water. The idea has even been proposed
- the reader must judge the significance of the proposition - that the eye
evolved as it did because it evolved in primitive sea-creatures, for which
the only significant stimuli for nerve-cell evolution, when the animals
themselves were submerged in the sea, were in the optical waveband.
Unfortunately, from the communication engineer's viewpoint, the difficulty
with optical communications lies less with attenuation than with scattering.
It is the presence of the numerous scattering particles in the sea which
militates against long-range optical communication. The use of laser light,
because of its narrow, pencil beam, certainly serves to minimise the scattering
volume between transmitter and receiver, but the technique remains fraught
with difficulties and, at this time, is largely impracticable except in very
specialised application areas.

10.2 The Gross Attributes of the Received Signal

If, as figure 10.1 suggests, the received waveform consists of the sum of
a main path, plus reverberation paths, then certainly the received signal will
exhibit some degree of fading behaviour. This we may classify in terms
of fading statistics which will describe the probability density distribution
of envelope and phase respectively.

Figure 10.1. Multipath structure in a highly reverberant environment

 Acoustic Communications 173

If the main path is not grossly the dominant path, then the classical model
for such phenomena, based upon the application of the central limit theorem,
will presume Rayleigh amplitude and uniform phase distributions. That is

p(A) = (N<A">) exp(-A7'/2 <A2·>); 0 ~ A~ oo

= 0; --oo ~ A ~ 0

where <N·> is the variance, or mean square value, of the envelope A(t).

p(cl>) = l/27t;
If the main path is in fact dominant, then the amplitude distribution may
tend towards being of Rician [10.14] form.

Another attribute which may be of value in gauging the effectiveness of

possible solutions to the multipath problem will be the spectral bandwidth
of the fading envelope. Here, we attack the problem by considering an
heuristic view of the processes giving rise to the envelope fluctuations
themselves. Consider the situation in which a main-path and a single,
dominant, grazing incidence surface bounce occurs. Suppose surface wave
action is slight: just a slow, small amplitude shift of the reflective surface
of the sea. Then we may anticipate long, slow and deep fades whenever
the multipath and main received signals move into phase opposition. However,
the frequency with which these fades will occur will be greater than the
surface wave frequency, if the length difference between the direct path
and the reflected path spans many cycles at the carrier frequency. Thus given
range, R, depth of receiver and transmitter, h, and assuming an average
surface wave height cr, the differential length will be

Sl "'4h aiR

If both transmitter and receiver are firmly located in the water, then it is
interesting to consider the frequency of envelope fading caused by a single
surface reflection interference path acted upon by surface waves. The
geometry of the problem is illustrated in figure 10.2.
R

Figure 10.2. Transmission dominated by a single surface

bounce and the main path
174 Underwater Acoustic Systems

The number of cycles of fade per surface wave cycle will be ~1/A where
Ais the transmission wavelength. Both the average height a and the average
frequency f•vs of surface wave cycles are in tum dependent upon wind speed,
w [m s-1 ], being given approximately by the relations

CJ =5 X 10-3 X W2. 5

and it follows that the number of cycles of fade per second or, loosely, the
fading frequency of the envelope, will be of the order of

Thus if, for example, we transmit at 15 kHz, for which A= 0.1 m, over
a range of 500 m, at a depth of 10m and with a modest wind speed of 20
m s-1 , we find that the fading frequency will be about 0. 7 Hz.

Another interesting condition relates to the vertical spatial frequency of

interference maxima, since this quantity gives at least some indication of
the vertical interspacing of receiver array elements, if some form of space
diversity is contemplated as a method of combating the fading problem.
Here, we assume the transmitter to be at depth h1 and the receiver to be
at depth h2• If the differential length is equal to some integer number of
wavelengths ~I = nA/2; n = 0,1, ... then a condition for constructive
interference, incorporating the 180" phase shift at the sea-surface, will be
met, as figure 10.3 shows, at receiver depths of the order of h2 = nAR/4h 1 •

t
R

_r2.5m
~

~ ~'
''
-~---------------------------

Figure 10.3. The fan lines are lines of destructive interference in the
R-z plane. Shaded blobs correspond to vertically defined intensity nulls
at rangeR. Unshaded blobs correspond to horizontally defined
intensity nulls at depth h2 (the moving ROV situation)
 Acoustic Communications 175

Let us again consider the example of a 15kHz source at 10m depth and
500 m range. The vertical interspacing of intensity nulls will then be 2.5
m, suggesting array element interspacing of this order, if space diversity
reception is contemplated.

Yet a third condition relates to the frequency of envelope fading if either

the transmitter or receiver is moving; mounted, for example, on an autonomous
ROV. Then, we find that the fading frequency for a receiver at depth h2
moving towards or away from a transmitter at depth h1 and at range R will
be approximately

which for our 15 kHz source, with both transmitter and receiver at 10m
depth, and with their horizontal separation being 500 m, results in a fading
frequency of 0.03 Hz if the ROV speed is 4 knots, or 2 m s-1 •

In conclusion, these simplified examples of operation may lead us to

suppose that the required response bandwidth of electronic equipments
designed to combat fading need only be modest. For our example, the fastest
fading frequency noted above, to surface wave induced fading, was only
0.7 Hz. For safety, and bearing in mind the simplicity of the model, we
might imagine, given other conditions, that assuming a fastest fading
frequency in the regime 10-100 Hz should allow the system designer some
latitude in formulating schemes to combat multipath.

10.3 The Channel Transfer Function

Let us take, again by way of example, the situation depicted in figure 10.3.
The differential path length Bl = 2h 1h:fR imparts a differential delay Bt =
Bl/c. Note also that the two paths will be subject to attenuation caused by
spreading and loss but that, for our present example both loss factors will
be comparable. There will be a surface reflection loss, k, but at grazing
incidence and modest wind speed, this will also be modest. The channel
impulse response will thus be of the form

h(t) + k h(t + Bt)

Here h(t) is the main path transfer function, dictated primarily by the
transmission and reception equipments. The loss factor is given by
k = -1011120 where J.1 is the intensity reflection coefficient and where the
negative sign models the 180• phase shift on reflection from the sea-surface.
Fourier transforming, we find that the channel frequency response is
176 Underwater Acoustic Systems

H(jw) + k exp(jro~t) H(jro) = A(ro) exp(jcj>(ro))

where

A(ro) = IH(jro)l {2(1 + k cos(ro~t)) p12

Spectral modulation

l - ===j=-----r-------------------------

1i :' f \

1"11----1/0 t ___.: Frequency

Figure 10.4. Spectral modulation resulting from the interference

between a surface bounce and a main path

For our example (R = 500, h1 = 10, h2 = 10, A. = 0.1) we find that ~t = 267
J.LS. It follows, as figure 10.4 shows, that the transfer function IH(ro)l has
imposed upon it, in forming A( ro), a frequency ripple. For the stated channel
properties, this ripple has a frequency period of some 3.745 kHz.

If the (15 kHz) transmit transducers exhibit a nominal (-3 dB) bandwidth
of 1.5 kHz then, if k is unity, intersymbol interference of uncertain but
possibly considerable severity may be anticipated. Thus constructive
interference (a maximum of the spectral modulation) at the transmit centre
frequency will result in the relatively unaffected transmission characteristic
illustrated in figure 10.5, whereas destructive interference (a minimum of
the spectral modulation) at that frequency will engender severe spectral
distortion and serious intersymbol interference.

For our example, a higher transmission frequency and thus wider bandwidth
will produce a spectral modulation which may introduce several cycles of
frequency variability in the transmission characteristic. If the chosen values
of R, h1 and h2 were altered, so that ~t increased, then again we should expect
faster amplitude fluctuations across the transmission characteristic A(ro).
Finally, with yet further multipaths entered into our model, higher order
anharmonic ripples in the transmission characteristic will result, leading
 Acoustic Communications 177

to a complex modification to the mainpath transfer function, itself fluctuating

at a rate not dissimilar, we might presume, to the fading frequency discussed
above, and presenting some significant problems in adaptive equalisation.

Mainpath transfer
function IHI

Spectral modulation :

--------- -,-------------- I
-~--------.

Frequency

Multipath transfer Multipath transfer

function IAI with function IAI with
destructive interference constructive interference

Figure 10.5. The effect of a single surface bounce plus the main path,
in bringing about spectral distortion on the channel transmission
characteristic
178 Underwater Acoustic Systems

10.4 Combating Multipath

The simplest approach to overcoming multipath, in whatever context the

transmission occurs, is just to transmit the data sufficiently slowly and using
a sufficiently robust modulation technique that the channel has adequate
time to settle and allow detection decisions to be made at the receiver. For
example, if we employ pulse interval modulation, then we seek to detect
the presence of packets of energy consequent upon each transmitted pulse,
distinct from each other but smeared to some extent by the channel
reverberation time. It is thus the channel impulse response, or effective
bandwidth, which limits the transmission rate, rather than the bandwidth
of the transmit and receive equipments themselves.

Clearly, both diversity reception, with its potential for selecting situations
of constructive interference, and equalisation, with its ability to minimise
intersymbol interference and maximise the use of available mainpath
bandwidth, are attractive propositions whether invoked independently or
simultaneously.

10.5 Diversity Reception

In the context of underwater acoustics we may visualise three broad classes

of diversity: spatial, spectral and temporal diversity. Spatial diversity
systems take cognisance of the fact that the multipath structure of the
channel can, indeed, produce regions in the water where acoustic intensity
is high, albeit only for some limited period of time. A simple diversity
system, such as is shown in figure 10.6, may thus be employed to allow
the receiver to select the hydrophone element producing the highest output.
Such a system implies no beamforming, is simple to implement and thus
offers the merit of some degree of element redundancy, in the event of
hydrophone failure, and thus enhanced reliability.

Naturally, such a system may be modified, as figure 10.7 suggests, to allow

far more sophisticated use of the available hydrophone elements. The
function of the weighting networks is to provide gain and phase adjustment
on the signals received from each hydrophone. Here a preferred technique
is maximal ratio combining [10.15] which attempts to maximise received
signal-to-noise ratio. Maximal ratio combining carries with it a substantial
computational penalty which may be outweighed for underwater applications
by the simplicity of either the switched diversity receiver or an equal gain
combining scheme. It should be noted that, in a sense, the law of diminishing
returns applies. If switched diversity picks out the strongest signal, relatively
small benefit accrues from adding in several other signals of increasingly
dubious quality.
 Acoustic Communications 179
Surface buoy equipped
for data storage or
re-transmission ----...._

To receiver
...___ Hydrophone selection
mechanism

Hydrophone array - . . . .

0
Transmit transducer

Figure 10.6. Simple switched (spatial) diversity reception

Receiver circuitry Data output

''
''
''
''
''
'
'''
'' ''
''
''
'

L
-------------------------------------~
Weight modification

Hydrophone array

Figure 10.7. Maximal ratio combining in linear diversity detection

180 Underwater Acoustic Systems
It is possible, of course, to imagine the weighting as a beamforming
operation. It would appear that maximal ratio combining has the effect of
steering nulls of the array polar response into the direction of the multi path
"image" sources. One would imagine, in any case, that the beamforming
weightings would have to account for the somewhat random amplitude and
phase factors experienced in the sound-field as a consequence of the
multipath environment.

Spectral diversity takes account of the fact that, because of the varying
multipath structure ofthe channel, so also will the channel transfer function
vary as a function of time. If transmission may be made, either switchable
or simultaneously, on more than one frequency, then thus may we anticipate
or utilise a momentarily preferable frequency slot for signalling. Just as
beamforming "looks a bit like" some forms of spatial diversity, so frequency-
hopped spread spectrum communications resemble, in some respects, spectral
diversity signalling. Some care should be exercised in making such intuitive
jumps, if for no other reason than that transmitter and receiver transducers
may be of distinctly limited bandwidth. We return to this topic, in the context
of parametric transmission, in the last section.
Data power

l__
spectrum
Spread power
spectrum

~
Data Input Modulator
I
I
I
I
I
Pseudo-random I
sequence
generator
/ ,....----,
I
~Demodulato

Synchronisation
Control

Figure 10.8. Direct sequence spread spectrum

 Acoustic Communications 181
Temporal diversity acknowledges the transport-delay model of the channel.
If coding can be applied such that information is spread over a time-frame
significantly longer than the channel reverberation time, then it is possible
that error corrective properties may be utilised in overcoming the multi path
problem. Again, an analogy can be drawn with direct sequence spread
spectrum communications used, in this instance, specifically to reject the
multipath interfering signals. Figure 10.8 illustrates the basic system
configuration.

A data source with bit-rate R 1 is added, modulo-2, to the output of a pseudo-

random digit generator, with a digit or "chip" rate, R 2 • Normally, R 2 »R 1 ,
so that the spectral bandwidth of the PRDG output would be much greater
than that of the data sequence. The digital "modulation" thus impressed on
the data sequence can be removed at the receiver by adding exactly the same
PRDG sequence. Synchronism is thus a vital presumption. The nature of
the synchronism must be such as to account for the main-path transport
delay. If the multipaths add in yet further delay, the signal returns arriving
by those routes will not synchronise and will be suppressed in the ratio
(R 1:R2), in terms of their relative power, by the receiver modulo-2 addition
operation. Of course, for this suppression to be unambiguous, there is an
implication that the sequence length T = L/R2 must be at least greater than
the longest significant additional multipath delay. If it is not, then there
is the possibility that a multipath return may occur at a time equal to some
multiple of the sequence length and be decoded as though it were a valid
main-path return.

10.6 Equalisation

Equalisation is the process of compensating for channel-imposed amplitude

and/or phase distortion of the received signal spectrum. The simplest form
of equalisation, passive equalisation, may be applied as a compensation
filtering, wherever stable propagation frequency distortion is encountered.
The one circumstance where passive equalisation would appear to be of use
is in tailoring an otherwise inadequate transmit transducer response to
improve amplitude uniformity and/or phase linearity. Otherwise, because
in the sea, stable propagation would be the exception rather than the rule,
some form of active, adaptive equalisation would appear to be inevitable.

In radio receiver design, an early equaliser strategy consisted of combining

an amplitude slope equaliser, which compensated for in-band gain slope,
and a space diversity receiver which compensates for notches or dips in
the received frequency response [10.16]. Such a technique is adequate for
use with relatively simple modulation techniques and might thus be
182 Underwater Acoustic Systems

commended for underwater applications. More recently, more sophisticated

equalisers have been developed. These include adaptive transversal equalisers
[10.17, 10.18] and decision feedback equalisers [10.19, 10.20] used either
alone or in combination with amplitude slope equalisers and, usually, space
diversity receivers.

The equalisation problem is most frequently presented in the context of a

main path plus a single interfering path, with the result that the effective
channel transfer function H(f) becomes modified as

H'(ro) = (1 + k(t) exp(-jrot)) H(ro)

where k(t) is a time-varying gain-factor for the multipath channel and tis
the multipath excess delay. The equalising filter is required to establish a
transfer function

G(ro) = (1 + k(t) exp(-jro·t))-1

which may be expanded, according to the Binomial Theorem to yield

k2 (t) exp (-j2rot) k3 (t) exp (-j3<.o't)
G (ro) = 1 - k (t) exp (-jrot) + - - - - - - - - - - - + ......
2! 3!
This expansion provides us with the basic structure of a finite impulse
response filter capable of acting upon a sequence of values sampled at
intervals t from the received waveform. Of course, we may be largely
ignorant of the value of t or, indeed, we may well expect to encounter a
more richly multipath environment than that suggested above.

Another way of approaching the equalisation problem is to recognise that

the end-result of channel distortion- and the only significant consideration
with digital signalling - is intersymbol interference. First, we design the
transmitted digit shape (by means of some appropriate pre-transmission
filtering) so that, as figure 10.9(a) shows, it exhibits no intersymbol
interference. That is, the amplitude of the filtered pulse is zero at time
instants corresponding to the adjacent pulse midpoints. This is usually
achieved by defining the pre-transmission filter to impose a transmitted
pulse spectrum with "cosine rolloff''. During transmission, the channel will
induce distortion of the pulse spectrum and bring about intersymbol
interference, figure 10.9(b). In a digital system, we seek to sample the
received bit-stream at the point of some received digit "maximum eye
aperture". All adjacent bit-centre values should thus be zero. This can be
achieved by means of the transversal equaliser circuit illustrated in figure
 Acoustic Communications 183

Transmitted Received
digit digit

(a) (b)

Demodulator

: Corrected
! output
''

~ :'
'
~
....... ! ........... ! ____•_____•_..
' ..~
I'
:
:'
!------------!
____ -~~.::::~~ ____ _r-----~...... !,,___
I #,'"•
: Adjustment • •• :
~ Target
., pattern

(c)

Figure 10.9. The transversal equaliser

10.9(c). For pre-set equalisation, the target pattern is a once-off transmitted

training sequence. For adaptive equalisation the target pattern may be
obtained directly from the output of the equaliser by passing the output
through a slicing or hard-limiting circuit.

10.7 Communication using Parametric Transmission

Quazi and Conrad [10.21] make the suggestion that parametric transmission,
because of its ability to establish pencil-beam transmission at relatively low
frequencies, with physically small transducers, might have particular
184 Underwater Acoustic Systems

advantage in avoiding surface and sea-floor reflections and thus minimise

or eliminate the corruptive effects of multipath transmission.

Parametric sonar might be less attractive than Quazi and Conrad suggest,
since high directivity at a frequency approaching one of the parametric
primaries is, in any case, readily achieved using conventional transmission.
One of the remaining advantages of the parametric method is that transmissions
using the lower, secondary frequency are less strongly attenuated, in water,
than conventional transmissions at the primary frequency. For many
applications this advantage would be offset by the poor power efficiency
of parametric conversion. Another potential advantage is the possibility of
making use of the extreme frequency agility of the secondary frequency.
At least in so far as bandwidth is concerned, the absolute width of sweep
of the secondary cannot in any case exceed the primary bandwidth. Finally,
the added complexity of a parametric projector would increase cost and
could adversely affect robustness.

The one potentially significant consequence of parametric transmtsston

could be attached to the aforementioned frequency agility. Thus although
the absolute sweep capability is probably very similar, whether a parametric
secondary transmission or ordinary transmission at the primary frequency,
the proportional sweep capability of the parametric secondary is much larger
than that of the innately narrowband primary. Two possibilities then ensue:
the application of parametric transmission in a true spread spectrum context
(since any other really broadband transmit transducer is difficult to envisage)
and also use in a frequency hopping diversity receiver structure, where a
proportionately large frequency hop is more likely to result in disassociation
from a previously deleterious multipath combination.

References

[10.1) R. Coates and P. Willison, Underwater Acoustic Communications: A Bibliography

and Review, Proc. bJSt. Acoustics, Vol. 9, Pt 4, December 1987, pp. 54-62

[10.2) R. Coates, Acoustic Data Telemetry from Beneath the Ocean Floor, Proc.IEEE Oceans
'87 Conf., Nova Scotia

[10.3) S.D. Morgera, Digital Filtering and Prediction for Communication Systems Time
Synchronisation, IEEE J. Oceanic Eng., Vol. OE-7, No. 3, July 1982, pp. 110-19

[10.4) D.C. Brock, S.C. Bateman and B. Woodward, Underwater Acoustic Transmission of
Low-Rate Digital Data, Ultrasonics, Vol. 24, No. 4, July 1986, pp. 183-8

[10.5) J. Capotvic, A.B. Baggeroer, K. Von der Heydt and D. Koelsch, Design and Performance
of a Digital Acoustic Telemetry System for the Short Range Underwater Channel, IEEE J.
Oceanic Eng., Vol. OE-9, No. 4, Oct. 1984, pp. 242-52
 Acoustic Communications 185

[1 0.6] J. V. Chase, A Tracking and Telemetry System for Severe Multipath Acoustic Channels,
Proc. IEEE Oceans '81 Conf., Boston, Mass., pp. 35-39

[10.7] D. Garrood, Applications ofthe MFSK Acoustic Communications System, Proc.IEEE

Oceans '81 Conf., Boston, Mass., pp. 67-71

[10.8] P.O. Kearney and C.A. Laufer, Sonarlink - A Deep Ocean, High Rate, Adaptive
Telemetry System, Proc. IEEE Oceans '84 Conf., Washington, D.C., Sept. 1984, pp. 49-53

[10.9] B. Leduc and A. Glavieux, Long Range Underwater Acoustic Image Transmitting
System, Institut Fran~ais de Recherche pour !'Exploitation de laMer, BP 337, 29273 Brest
Cedex.

[10.10] G.R. Mackelburg, S.J. Watson and A. Gordon, Benthic 4800 Bits/s Acoustic Telemetry,
Proc. Oceans '81 Conf., Boston, Mass., p. 72

[10.11] C.S. Miller and C.E. Bohman, An Experiment in High-Rate Underwater Telemetry,
!ERE Conf. on Eng. in the Ocean Environment, 1972

[10.12] R.B. Mitson, T.J. Storeton-West and M.G. Walker, Fish Hean-rate Telemetry in the
Open Sea Using Sector Scanning Sonar, Biotelem. Patient Monitoring, Vol. 5, No. 3, 1978,
pp. 149-53

[10.13] R.M. Dunbar, S.J. Robens and S.C. Wells, Communications, Bandwidth Reduction
and System Studies for a Tetherless Unmanned Submersible, Proc.IEEE Oceans '81 Conf.,
Boston, Mass., pp. 127-131

[10.14] S.O. Rice, Mathematical Analysis of Random Noise, Bell Sys. Tech. J., Vol. 24, No.
46, 1945, An. 3.10

[10.15] W.C. Jakes, Jr., Microwave Mobile Communications, Wiley, New York, 1974

[10.16] Y.Y. Wang, Simulation and Measured Performance of a Space Diversity Combiner
for 6 GHz Digital Radio, IEEE Trans. Commun., Vol. COM-27, Dec. 1979, pp. 1896-1907

[10.17] M. Shafi and D. Moore, Funher Results on Adaptive Equaliser Improvements for
16-QAM Digital Radio, IEEE Trans. Commun., Vol. COM-34, Jan. 1986, pp. 59-66

[10.18] F. de Jager and M. Christiaens, A Fast Automatic Equaliser for Data Links, Philips
Tech. Rev., Vol. 37, No. 1, 1977, pp.10-24

[10.19] P.P. Taylor and M. Shafi, Decision Feedback Equalisation for Multipath Induced
Interference in Digital Microwave LOS Links, IEEE Trans. Commun., Vol. COM-32, March
1984, pp. 267-279

[10.20] C.A. Belfiore and J.H. Park, Decision Feedback Equalisation, Proc. IEEE, Vol. 67,
No. 8, August 1979, pp. 1143-1156

[10.21] A.H. Quazi and W.L. Conrad, Underwater Acoustic Communications,/EEE Commun.
Magazine, Vol. 20, No. 2, March 1982, pp. 24-30
Index continental shelf 52
continental slope 52
absorption 20 correlation function 33, 42, 45
abyssal plain 52 cross-correlation 33,47
acoustic centre 16 CfD measurement 5
acoustic communication 171-85 Curie temperature 116
acoustic impedance 9 cutoff frequency, mode 78
acoustic intensity 9, 16 cylindical spreading 18
acoustic navigation 166
acoustic positioning 166 decibel 8
acoustic telemetry 162, 171-85 deep-ocean transmission 58-64
admittance, transducer 130-2 depth dependence of noise 94
ambient noise 28, 90-103 dipole receiver 140
angular distribution function 34 dipole source 96
angular distribution of noise 95-100 directivity
ARMA analysis 42 receiver 113, 147
array, transducer 121, 136-52, 178-80 source 17
array polar response 140-5 dispersion, waveguide 82, 87
array shading 144 diversity
array steering 146, 180 frequency 180
attenuation 18, 19 space 178
anomalous 21 doppler measurements 168
sea-water 19
sediment 21 echo sounder 154
auto-correlation 33, 45 eigenfunction 86
eigenvalue 86
backscatter elasticity 3
surface 109 endfire array 141, 146, 149
volume 107-9 energy spectral density 33, 45
bandwidth energy time-frequency plot 33, 45
signal 7 equalisation 181
transducer 125
bathythermograph 5 fading behaviour 174
beamformer 146 "fast" bottom 14
beamsteering 146, 180 , Fast Fourier Transform 36
beamwidth 113, 136, 144, 147 ferro-electricity 116
Beaufort Scale 92 FFT 36
Beckmann-Spizzichino loss 27 filter-bank spectrum analysis 36
Bessel equation 86 finite energy process 33
bistatic sonar 30 finite power process 33
blade-rate tonal 101 fish tags 162
broadside array 141, 146 fishing sonars 160
bulk modulus 3 flextensional transducer 115
free-field spreading 18
caustic 62 frequency
cavitation 100, 148 spatial 7
cavitation noise 100 temporal 7
cepstrum 34,47 frequency hopping 180
circular navigation 167
Collias equation 4 Gloria 165
communication, acoustic 171-85 Goll-type transducers 127-9
complex cepstrum 50 group velocity 78
conductivity 3

186
 Index 187

half-beamwith 113, 136, 144, 147 thermal 92

Helmholtz equation 85 noise level 29, 93
heterodyne spectrum analysis 34 noise spectrum level 92
homomorphic deconvolution 34, 49 noise variability, angular distribution
hydrophone design 132-4 95-100
hydrophone line array 137-43 noise variability with depth 94
hyperbolic navigation 168 noise variability with time 93
non-stationary process 33
image interference 70-2, 173-7 normal modes 73--89
impedance,acoustic 9 NUC loss model 25
intensity 9
interference, image 70-2, 173-7 parametric communication 183
inverse-square loss 18 parametric source 147
isothermal propagation 59 particle velocity 1
isovelocity propagation 55 pattern synthesis, array 143-5
period
Janus velocity measurement 169 spatial 7
temporal 7
Langevin projector 117-22 piezo-electricity 114, 116
lead zirconate titanate 116 Poisson's ratio 119
line array 137-43 polarreponse 113, 129, 140-5
Lloyd mirror 70 polar response measurement 129
loss poly-vinylidene fluoride 116
Beckmann-Spizzichino 27 porosity 21
NUC 25 positioning, acoustic 166
Rayleigh 25 power spectral density 33, 45
sea-floor reflection 25 propagation equation 6, 83
sea-surface reflection 26 propeller noise 100-1
spreading 18 PVF 116
transmission 18 PZT 116
magnetostriction 114 Q-factor, transducer 125
mainpath multipath 105 quartz 116
matching, transducer 127
meta-cepstrum 49 ray coefficient 54
mode cutoff frequency 78 ray-tracing 53-72
mode theory 53 Rayleigh distribution 173
moisture content 21 Rayleigh parameter 27
molecular resonance effects 20 Rayleigh reflection loss 25
monopole receiver 139 receiver directivity 113, 147
monostatic sonar 30 reflection coefficient
multipath, mainpath 105 intensity 12, 24
multipath propagation 105, 172-84 pressure 11
multiple matching layers 127-29 refraction 54
multiple source images 66 resonant transducer 123-7
reverlxrration 29, 104-10
navigation, acoustic 166 surface 30, 104
new reference unit 8 volume 30, 108
noise 28,90-103 Rician distribution 173
ambient 28,90-103 rigid bottom 77
self 28
shipping 92, 100-3 salinity 3
surface agitation 92 scattering
188 Index

surface 109 thin-disc transducer 127-9

volume 107-9 time variable gain 158
sector-scanning sonar 157 tonpiltz transducer 122
seismic surveying 165 transducer 112-52
self-noise 28 flextensional 115
shading 144 free-flooding ring 122
shadow zone 62 Goll-type 127-9
shelf-sea transmission 65-70 Langevin-type 117-22
ship noise 100-3 multiple layer 127-9
side-scan sonar 163 polarresponse 129
signal excess 30 thin disc type 127-9
sing-around velocimeter 6 tonpilz 122
singing 101 transducer arrays 121,136-52,178-80
"slow" bottom 15 transducer matching 127
Snell's law 10, 53 transducer terminal admittance 130-2
sonar transducer testing 130-2
bistatic 30 transducer tuning 123-7
mono-static 30 transducers
sonar equations 16 air-gun 115, 159
sound speed 1 atresonance 123-7
sound velocimeter 3, 6 explosive 115, 159
source directivity 17 sparker 115, 159
source intensity 16 transformer, design criteria 156
spatial correlation 50 transmission coefficient
spatial period 7 intensity 13
specific heat 3 pressure 11
spectral density 33 transmission loss 18
spectral estimation 41 transmission modelling
spectrum, wavenumber 50 shallow-sea 65-70, 73-89
spectrum analysis wedge-sea 70
FFf 36 transmitting response 123-7
filter-bank 36 transponders 162, 166
heterodyne 34 travelling wave 6, 83
Prony 40 tuning, transducer 123-7
spectrum level, noise 92 TVG circuit 158
specular reflection 10
spherical spreading 18 velocimeter 3, 6
spread spectrum communication 180 vertical directivity of noise 95-100
spreading loss 18 viscous friction 19
stationary process 33 volume reverberation 30, 108
Sturm-Liouville problem 86 volume scattering 107-9
sub-bottom profiling 159
surface reverberation 30, 104 wake noise 100
surface scattering 109 wave equation, 2-D 84
surveying 166-8 waveguide dispersion 82, 87
synthetic aperture 149 waveguide propagation 62, 73-89
wavelength 7
targetstrength 22 wavenumber 7
telemetry, acoustic 162, 171-85 wavenumber spectrum 50
temporal period 7 window functions 39
terminal response measurement 130-2
terrain mapping sonar 157, 163 XBT 5
thermocline 60