Sonar For Practising Engineers PDF

Sonar for Practising Engineers
Third Edition
This page intentionally left blank
Sonar for Practising
bngineers
Third Edition
A. D. Waite
8
JOHN VVILEY & SONS, LTD
Copyright 02002 by Ashley Waite Published by John Wiley & Sons Ltd,
Baflhs Lane, Chichester,
West Sussex PO19 lUD, England
National 0 1243 779777
International (+44) 1243 779777
e-mail (for orders and customer service enquiries): [email protected]
Visit our Home Page on http://www.wiley.co.uk
This work is based on Sonar for Practising Engineers, Second Edition, published and distributed by Thales
Underwater Systems Limited (formerly named Thomson Marconi Sonar Limited), Ocean House,
Templecombe, Somerset, BA8 ODH (www.tms-sonar.com), 1998
All Rights Reserved No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, scanning or
otherwise, except under the terms of the Copyright, Designs and Patents Act 1988 or under the terms of a
licence issued by the Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London, UK W 1P OLE
without the permission in writing of the Publisher with the exception of any material supplied specifically
for the purpose of being entered and executed on a computer system for exclusive use by the purchaser of
the publication.
Neither the authors nor John Wiley & Sons, Ltd accept any responsibility or liability for loss or damage
occasioned to any person or property through using the material, instructions, methods or ideas contained
herein, or acting or refraining from acting as a result of such use. The authors and publisher expressly
disclaim all implied warranties, including merchantability or fitness for any particular purpose. There will
be no duty on the authors or publisher to correct any errors or defects in the software.
Designations used by companies to distinguish their products are often claimed as trademarks. In all
instances where John Wiley & Sons, Ltd is aware of a claim, the product names appear in capital or all
capital letters. Readers, however, should contact the appropriate companies for more complete information
regarding trademarks and registration
Ashley Waite has asserted his right under the Copyright, Designs and Patents Act 1988 to be identified as
the author of this work.
British Library Cataloguingin Publication Dpta
A catalogue record for this book is available from the Brisish Library
ISBN 0 471 49750 9
Typeset in 10/1 Ipt Times by Keytec Typesetting Ltd., Bridport, Dorset

Printed and bound in Great Britain by Antony Rowe Ltd., Chippenham, Wilts
This book is printed on acid-free paper responsibly manufactured from sustainable forestry,
for which at least two trees are planted for each one used for paper production.
To all my past and present colleagues within
the sonar community who have made
my career in sonar so rewarding
- and this book possible
Contents
Preface xiii
About the Author xv
Introduction xvii
1 Sound 1
1.1 Wave motion 1
1.2 Sound pressure 1
1.3 Reference intensity 3
1.4 Source level 4
1.5 Radiated power 5
1.6 Limitations to sonar power 5
1.7 Cavitation 6
1.8 Interaction 8
1.9 Changes to arrays 8
1.10 Projector sensitivity 9
I . 11 Hydrophone sensitivity 9
I . I2 Spectrum level 10
1.13 Sound in air and in sea water 11
1.14 Problems 12
2 Arrays 13
2.1 Need for projector arrays 13
2.2 Need for hydrophone arrays 13
2.3 Beam patterns 14
2.4 Directivity of a dipole 15
2.5 The general line array 18
2.6 Line array: beam pattern vs. steer angle 20
2.7 Broadside array: length and spacing 20
2.8 Beam pattern for a continuous line 21
2.9 Shading 24
2.10 Shaded arrays: transmit source levels 27
2.1 1 Directivity index 28
... Contents
Vlll
2.12 DI of a simple dipole 29

2.13 DI of a line array 31
2.14 DI of a planar array 32
2.15 DI of a cylindrical array 35
2.16 DI formulae based for simple arrays 36
2.17 Conformal arrays 36
2.18 Spherical arrays 36
2.19 Volumetric arrays 37
2.20 Beamformers 39
2.2 1 Domes and arrays 40
2.22 Problems 41
3 Propagation of Sound in the Sea 43

3.1 Propagation loss 43
3.2 Losses 43
3.3 Spreading losses 44
3.4 Absorption losses 46
3.5 Spherical spreading and absorption 48
3.6 Propagation in the real ocean 49
3.7 The speed of sound 50
3.8 Sound speed profiles 51
3.9 Deep sound channel 52
3.10 Reliable acoustic path 53
3.1 1 Surface duct propagation 54
3.12 Convergence zone propagation 56
3.13 Bottom bounce propagation 57
3.14 Propagation loss models 58
3.15 Ray theory and the Hodgson model 59
3.16 Hodgson example 61
3.17 Performance prediction 64
3.18 Multipath propagation 66
3.19 Problems 66
4 Target Strength 67
4.1 Definition 67
4.2 Formulae 67
4.3 Measurement 68
4.4 Dependence on pulse type and duration 69
4.5 TS of a sphere 69
4.6 TS of some simple shapes 70
4.7 TS of small targets 72
4.8 Mine target strength 72
4.9 Torpedo target strength 73
4.10 Submarine echoes 74
4.1 1 Beam aspect target strength 74
4.12 Bow aspect target strength 75
4.13 Submarine target strengths 75
Contents iX
4.14 Towed arrays 77

4.15 Target strength reduction 78
4.16 Practical values 80
4.17 Problems 81
5 Noise in Sonar Systems 83

5.1 Sources of noise 83
5.2 Thermal noise 84
5.3 Noise from the sea 86
5.4 Noise from a vessel 89
5.5 The sonar environment 90
5.6 Self-noise 90
5.7 Electrical noise 90
5.8 Machinery noise 91
5.9 Flow noise 91
5.10 Propeller noise 92
5.1 1 Variation with speed 92
5.12 Variation with frequency 92
5.13 Directivity 93
5.14 Self-noise and radiated noise 93
5.15 Addition of noise levels 93
5.16 Receiver noise factor 94
5.17 Noise factor of a sonar 95
5.18 Acceptable receiver noise level 97
5.19 Alternative calculation 99
5.2 1 Problems 101
6 Reverberation 103
6.1 Sources of reverberation 103
6.2 Scattering and reflection 103
6.3 Boundary roughness 104
6.4 Classes of reverberation 105
6.5 Backscattering strength 106
6.6 Reverberation target strength 106
6.7 Volume reverberation 107
6.8 Boundary reverberation 110
6.9 Scattering layers 111
6.10 Volume scattering strength 112
6.1 1 Sea surface scattering strength 113
6.12 Bottom scattering strength 114
6.13 Variation with fi-equency 116
6.14 Reverberation under ice 117
6.15 Problems 11:
7 The Sonar Equations 119

'7.1 What are they? 119
X Contents
7.2 What are their uses? 119

7.3 The basic sonar equation 120
7.4 The basic passive equation 120
7.5 The basic active equation 120
7.6 Detection threshold and detection index 121
7.7 Receiver operating characteristics 123
7.8 ROC curves 124
7.9 Problems 124
8 Passive Sonar 125

8.1 Radiated noise 125
8.2 Radiated noise: source level 125
8.3 Nature of radiated noise 126
8.5 Broadband and narrowband 129
8.6 Normalization 131
8.7 A Note on Swaths 131
8.8 Passive arrays 132
8.9 Passive aural 133
8.10 Passive displays 135
8.1 1 Formulae for detection threshold 136
8.12 Broadband square law detector 137
8.13 Broadband cross-correlator detector 138
8.14 Narrowband processor 139
8.15 Narrowband amplitude detector processor 140
8.16 Worked examples 141
8.17 Passive ranging 146
8.18 Triangulation 147
8.19 Vertical direct passive ranging 149
8.20 Horizontal direct passive ranging 153
8.2 1 Towed arrays 156
8.22 Bearing ambiguity 157
8.23 Self-noise 158
8.24 Problems 159
9 Activesonar 161
9.1 Range, pings and doppler shift 161
9.2 Pulse types 163
9.3 CW processing 163
9.4 FM processing 164
9.5 Active sonar equations 165
9.6 Reverberation index 168
9.7 FM pulses 169
9.8 CW pulses 171
9.9 Reverberation rejection by CW pulses 174
9.10 Reverberation and Target Echoes in the main lobe 175
9.1 1 Reverberation and Target Echoes in the sidelobes 178
Contents xi
9.12 Practical reverberation envelopes 179

9.13 Full- and half-beam processing 180
9.14 Beam forming 181
9.15 FM phase binning process 183
9.16 CW processing 184
9.17 Large aperture array 185
9.18 Detection performance 187
9.19 Noise and reverberation-limited detection ranges: 190
9.20 Ambiguity diagrams 192
9.2 1 Very long pulses 200
9.22 Operational degradation factor 20 1
9.23 Active displays 202
9.24 Unified detection and classification 206
9.25 Bandwidth 206
9.26 Beamwidth 208
9.27 CADAC 210
9.28 Levels of CADAC 21 1
9.29 CADAC and pulse features 212
9.30 Statistical analysis 213
9.3 1 Amplitude profiles 213
9.32 Multipath affects classification 214
9.33 Simple multipath example 215
9.34 Deep water: source and target in duct 216
9.35 Deep water: source and target below duct 217
9.36 Shallow water: source and target mid-water 218
9.37 Problems 219
10 Echo Sounding and Side Scan Sonars 221

10.1 Common features 22 1
10.2 Echo sounders 22 I
10.3 Echo sounder: design example 223
10.4 Side scan sonar 226
10.5 Side scan sonar: design example 229
10.6 Problem 23 1
11 Mine Hunting Sonars 233

11.1 Overview 233
11.2 Two broad classes of mine 234
11.3 Backgrounds to mine detection 234
11.4 Range and bearing resolution 235
11.5 Design of a mine hunting sonar 235
11.6 The threats 235
11.7 Design example 236
11.8 Performance 239
11.9 Classification 240
1 1.10 Mine avoidance 242
xii Contents
11.11 Mine avoidance sonars 243

11.12 Problems 248
12 Intercept and Communications Sonars 249

12.1 Intercept sonars 249
12.2 Communications sonars 250
12.3 Function of an intercept sonar 250
12.4 Intercept sonar equation 252
12.5 Worked examples 253
12.6 Reduction in the probability of intercept 255
12.7 Reduction in the probability of exploitation 256
12.8 Effectiveness of intercept sonars 257
12.9 Communications sonars 257
12.10 Communications sonar equation 258
12.11 Examples of communications sonars 259
12.12 Problems 260
13 Active Sonar Design 261

13.1 Introduction 26 1
13.2 Submarine detection 26 1
13.3 Hull-mounted surface ship sonar 26 1
13.4 Representativehull-mounted design 263
13.5 Longer ranges 267
13.6 Towed transmitters and towed array receivers 267
13.7 Representative design 268
13.8 Low frequency active: beware! 273
13.9 Torpedo detection 274
Conclusion 277
Solutions to Problems 281
Index 293
Preface
Most books on sonar - the use of underwater sound for the detection, classifica-
tion and location of underwater targets and for communications and telemetry -
have been written by physicists and mathematicians. They are not always easily
understood, nor are they immediately useful for solving the problems met by
engineers and technicians.
The aims of this book - written by a practising engineer for practising
engineers - are to provide an understanding of the basic principles of sonar and to
develop formulae and rules of thumb for sonar design and performance analysis.
No prior knowledge of sonar is assumed, and the physical principles and mathe-
matics will be readily understood by engineers and technicians.
The earlier editions were produced to be supplied as back-up material to a short
sonar course given by the author. This edition has been extensively rewritten to
facilitate its use by an individual reader. Several new topics have been included:
Echo sounding and side scan sonars for civil applications

Communications sonars
Low frequency active sonars
Many chapters contain worked examples and most chapters conclude with a few
problems for the reader to solve; solutions are given at the end of the book. I hope
these problems will be particularly useful to lecturers and students.
The book can be divided into three parts:
Equipment parameters: this part briefly describes the motion of sound in an

elastic medium, gives definitions of sound intensity and source level, explains
the use of projector arrays to increase the source level, and looks at the use of
hydrophone arrays to improve the signal-to-noise ratio of a wanted sound.
xiv Preface
Propagation in the sea and backgrounds to detection: this part considers

spreading and absorption losses, propagation modes and simple modelling,
noise and reverberation; it concludes with the sonar equations.
Practical sonar systems: this part develops passive and active sonar systems; it
looks at passive broadband, narrowband, intercept and communications sonars;
and it considers active sonars for the detection of submarines, mines and
torpedoes.
About the Author
Practising on HMS Jupiter!
Ashley Waite retired from the UK Ministry of Defence in 1990 after 40 years of
Underwater Warfare Research at Portland.
Primarily engaged in Sonar Research and Development, his experience spans
Surface Ship Sonars, Submarine Sonars, Surface Ship Torpedo Defence and a
lesser involvement with Helicopter and Minehunting Sonars. He has also worked
on Submarine Command Systems and the Underwater Aspects of Surface Ship
Command Systems.
xvi About the Author
He then joined Ferranti Thomson Sonar Systems (now Thales Underwater

Systems) as a Consultant on Active Sonar, working on improvements to the RN
Sonar 2050; the active concepts of a fully integrated submarine sonar system
(Sonar 2076); the Merlin helicopter sonar; and low frequency active sonar (Sonar
2087) for a further 10 years and still maintains an involvement with Sonar in a
freelance capacity.
Introduction
Many methods of detecting the presence of underwater targets in the sea have
been investigated. Here are some non-acoustic methods which have had varying
degrees of success:
Magnetic
Optical signatures
Electric field signatures
Thermal detection (infrared)
Hydrodynamic changes (pressure)
Magnetic methods include self-generated fields or perturbations of the earths

magnetic field, known as magnetic anomaly detection (MAD). Research continues
into these methods but underwater sound is still unsurpassed, in spite of formid-
able difficulties facing its propagation through a highly variable medium and from
the noise and reverberation backgrounds to detection. Sonar (sound navigation
and ranging) - by analogy with radar (radio detection and ranging) - uses
underwater sound for the detection, classification and location of underwater
targets.
Passive sonar listens to the sound radiated by a target using a hydrophone, an
underwater microphone, and detects signals against a background of the ambient
noise of the sea and the seZfnoise of the sonar platform (an omnibus term to
describe any vessel or site possessing a sonar system). Passive systems can be
made directional, therefore the azimuth (horizontal bearing) of a signal is known.
The nature of the signal - its frequency spectrum and how it varies with time -
will help to classifL the target.
xviii Introduction
Basic passive systems, however, give no information about the range of a target;
a signal may belong to a close, quiet target or a noisy, distant target. More complex
passive systems estimate range by the following methods:
Triangulation: measuring the bearings of a target from two well-separated

arrays.
Horizontal direct passive ranging (HDPR): based on the measurement of

wavefront curvature using three well-separated arrays.
Yertical directpassive ranging (VDPR): measuring the vertical arrival angles of

signals arriving at the same array via multiple paths as well as measuring the
time differences between them.
All of these methods are fundamentally dependent on the accuracy of the bearing
measurements and therefore demand large arrays and large separations to achieve
useful range estimates.
Active sonar uses a projector (an underwater loudspeaker) to generate a pulse
of sound which travels through the water to a target and is returned as an echo to a
hydrophone, often the same device as the projector and in this context more
commonly known as a transducer. The echo now has to be detected against a
background of noise and reverberation (unwanted echoes from the sea surface
and sea bed and from scatterers within the volume of the sea). Because the time
between transmission of a pulse and reception of an echo can be measured and the
speed of sound in the sea is known, the range of the echoing target is simply
calculated. Active sonars are sometimes known as echo ranging systems.
To survive in sonar, newcomers must become familiar with the decibel. They
will hear old hands discussing the design and performance of sonar systems in
exchanges where almost every other word seems to be deebee (dB or decibel).
What is this decibel?
First of all, the be1 is inconveniently large so it has been divided by 10 to
become the decibel. It simply compares the power or intensity of the sound at one
point in a system with that at another. The decibel defined:
Power gain = 10logl0 (2) I (dB)
The decibel is of course also used in electronics, communications, radar and

airborne sound, and will be familiar to most engineers. But it is in sonar where it
really thrives. Long expressions such as the sonar equations are assembled with
Introduction xix
many parameters all expressed in decibel form. And to avoid serious errors, the
decibel must be carefully defined and correctly applied to all of the terms in the
equations.
Power ratios in sonar systems are frequently very large numbers. Calculations
are greatly simplified when very large numbers are expressed in logarithmic form,
so that values can be added instead of multiplied. Interestingly, although radar
engineers may use decibels to describe the terms in a radar system (e.g. antenna
gain, receiver noise factor), the radar equation is more commonly written in linear
form rather than logarithmic form:
The individual terms may be given their linear values and multiplied, or expressed
in decibels and added.
Compare with an active sonar equation:
2PL = SL + TS N + DI + 10 log T - 5 log d
All terms are in dB form; here are some possible values:
PL, the propagation loss, might be 80 dB

SL, the projector source level, might be 210 dB
TS, target strength, might be 0 dB
All the examples, here and in the rest of the book, will use credible values for
these parameters, and the reader should thereby learn to recognize and question
suspiciously large or small values. (None of the terms in the sonar equations is
likely to have a value outside the range - 100 dB to +250 dB, at the very most.)
The speed of sound in water is about 1500 m/s (3355 mph), much faster than its
speed in air of 340 m/s (760 mph) at sea level, but very much slower than the
speed of light (electromagnetic radiation). Velocity, frequency and wavelength are
related by
xx Introduction
where
c = velocity in metres per second ( ds)

f = frequency in hertz (Hz) or cycles per second
R = wavelength in metres (m)
Wavelength is the distance travelled by the wavefront during one cycle. At

1000 Hz we have R = 1.5 m, and at 10 kHz we have R = 150 mm. The speed of
light (EM waves) is 3 X lo8 m/s. At a typical radar frequency of 2000 MHz the
wavelength is R = 150 mm. Sizes of transmitting and receiving arrays, which
typically have dimensions of several wavelengths, are therefore comparable for
many sonar and radar systems.
Introduction xxi
Why Is Radar Not Used to Detect Underwater Targets?
Figure 0.1 Why not radar?
The radar will detect the submarines periscope and, assuming that the electro-
magnetic energy spreads spherically, the inverse square law will apply and the EM
propagation loss (two-way) is 40 log 10 000 = 160 dB.
The sonar will detect the submarines hull and, again assuming spherical
spreading, the sound propagation loss is also 160 dB.
Therefore, given echoing surfaces above and below water, radar and sonar
systems have comparable performances. But if the submarine is completely
submerged, the EM losses over the final, say, 100 m are completely prohibitive at
any radar fiequency (Figure 0.1).
EM losses in the highly conductive sea water are given by 1400f1/* dB/km,
where f is in kilohertz and, for just a 100 m path in the sea, the losses at
2000 MHz are 200000 dB! Even at 30 kHz the losses equal 770 dB; the sea
effectively presents a short circuit to the EM energy.
A modern nuclear-powered submarine can remain completely submerged for
indefinite periods and during a mission may never expose any reflecting surface
above water, but will rely completely upon its sophisticated sonars to navigate, to
build up a complete acoustic scenario of its surroundings, to deploy counter-
measures, and finally to prosecute an attack. Radar systems, therefore, will never
be given opportunities to detect the submarine.
xxii Introduction
i_ I ) ) ) ;
Radiated noise at source = 120 dE3 c&
Radiated noise at hydrophone = 50 db
Background noise = 40 db
Figure 0.2 Simple passive sonar
The simplest passive sonar will have an omnidirectional wideband hydrophone

and its detection performance is given by the allowable propagation loss (PL)
between target and hydrophone (Figure 0.2). If we assume that, to make a
detection, the radiated noise at the hydrophone must be 10 dB greater than the
background noise, then
PL = 120 - 50 = 70 dB
and assuming spherical spreading and no losses,
20 log R = 70 dB
therefore R = 3000 m.
Practical passive sonars will have directional arrays and limited operating
bandwidths to reduce background noise, and they will integrate the signal (the
radiated noise from a target) over a period of time. These measures will all
improve the allowable PL significantly and greater ranges are possible. (But
120 dB represents quite a noisy submarine, so perhaps 3000 m is not too
unrealistic a range.)
Introduction xxiii
Transmittedpulse = 200 d
Received pulse = 200 - PL (dB)

Echo = 200 - PL + 10 (db)
\
target strength
Figure 0.3 Simple active sonar
The simplest active sonar will have an omnidirectional projector and hydro-
phone (Figure 0.3). Its detection performance is given by the allowable two-way
propagation loss (2PL), i.e. from projector to target and back to the hydrophone. If
now the echo at the projector is 10 dB greater than the background noise, which
for a surface ship sonar might be 60 dB, say, then
10 = (200 - 2PL + 10) - 60

PL = 70 dB
therefore R = 3000 m.
Practical active sonars will have directional transmit and receive arrays, and a
knowledge of the pulse will allow the processing to be matched to the returned
echo. These measures will all improve the allowable PL significantly, and much
greater ranges will be achievable.
1
Sound
1.1 Wave Motion

Sound is produced when an object vibrates and communicates its motion to the
surrounding medium (Figure 1.1). Consider a regularly vibrating sphere; as it
vibrates it alternately compresses and rarefies the surrounding medium, resulting
in a series of compressions and rarefactions travelling away from the sphere. These
are known as longitudinal waves since the particles move in the same dimension
as the travelling wave. (Transverse wave motion occurs, for example, in a vibrating
string, where the motion of the string is orthogonal to the direction of wave
travel.)
1.2 Sound Pressure

For a wave to be a plane wave, pressure changes only in the direction of
propagation of the sound; pressure is the same at all points in any plane normal to
this direction. Wavefonts are those normal planes - separated by one wavelength,
il - where p is at a maximum.
The speed of sound refers to the longitudinal motion of the wavefronts in the
medium and is related to wavelength and frequency by
The speed of sound is not to be confused with the particle velocity, u, which refers
to the movement of the molecules in the medium.
2 Sound
Elemental area of sound field
-
Particle motion
Longitudinal wave motion
Figure 1.1 Propagation of sound in an elastic medium
Given a plane wave IP = ( P C ) U ]

where
p = pressure (Pa or N/m2)

u = particle velocity (m/s)
p = fluid density = lo3kg/m3 for sea water
c = velocity of sound wave propagation
= I .S x lo3m/s in sea water
pc = specific acoustic impedance, 2
= 1.5 X lo6kg m-2 s-' for sea water

Reference Intensity 3
The sound wave carries mechanical energy with it in the form of the kinetic
energy of the particles and the potential energy of the stresses in the medium.
Because the wave is propagating, a certain amount of energy per second will flow
across unit area normal to the direction of propagation.
This energy per second (power) crossing unit area is known as the intensity of
the wave (power per unit area). For a plane wave, the intensity is related to the
pressure by
1.3 Reference Intensity

The reference intensity (Ir) in underwater sound is the intensity of a plane wave
having a root mean square (RMS) pressure equal to 1 pPa (one micropascal).
Inserting p = 10-6 and pc = 1.5 X lo6 in the above equation for I , we obtain
1 I, = 0.67 X
1
W/m2 J 1
Intensities are often loosely stated as re 1 p a . This is clearly incorrect since the
micropascal is a unit of pressure not intensity (power per unit area). Strictly,
intensities should be stated as re the intensity due to a pressure of 1 @a.
4 Sound
1.4 Source Level

The source level (SL) is defmed as
intensity of source at standard range

SL = lOlog
reference intensity
The SL of an omnidirectional projector is always referred to a standard range (1

metre or 1 yard) from its acoustic centre. At 1 metre the acoustic centre of an
omnidirectional source is surrounded by a sphere of surface area 4nr2 = 12.6 m2.
If the omnidirectional power output is P watts, then the source intensity at 1 metre
is P/12.6 W/m2 and SL becomes
SL = lOlog );(
= 1OlogP + lOlog(1.1846 X 10)

= 10logP+170.8dB
(If the standard range is 1 yard, then SL = 10 log P + 171.5 dB.)

If the projector is directional, then
DIt = 10 log (k)

lomni
where
DIt = transmit directivity index

Idir = intensity along the axis of the beam pattern
lomi = intensity of the equivalent non-directional projector
and then SL becomes
I SL = lolog P + 170.8 + DI,]

Limitations to Sonar Power 5
1.5 Radiated Power

P is the total acoustic power radiated by the projector, which is less than the
electrical power supplied to it, Pe, and the ratio of these is the projector efficiency,
E.
The efficiency depends on the bandwidth and may vary from as little as 0.2 to
as high as 0.7 for a tuned, narrow bandwidth projector. The radiated powers for
typical sonars may range fiom, say, 1 W to 40 kW and have DIt values of between
10 and 20 dB.
Extremes of SL are therefore
SL = lolog 1 + 170.8 + 10 = 181 dB

and
SL = lOlog40000 + 170.8 + 20 = 237 dB
1.6 Limitations to Sonar Power

To achieve maximum range with active sonars, we need to generate the maximum
amount of acoustic power - at least until the reverberation background limits the
detection range. Against this purely technical argument, however, it should be
remembered that when the power level is already high, say 20 kW, it may not be
cost-effective to double the power to the technical limit since the additional 3 dB
- which for active sonars means only 1.5 dB more in the allowable one-way
propagation loss - may only increase the range by a small percentage. Attempts to
increase the radiated power are eventually limited by two effects: cavitation and
interaction.
6 Sound
1.7 Cavitation
When the power applied to a projector or an array is increased, bubbles form on the
surface and there is a resultant loss in power by absorption and scattering within the
bubbles, degradation of the beam pattern and a reduction in the acoustic impedance
into which the projector generates (resulting in a mismatch with the transmitters
supplying the projector array). Cavitation is a function of depth (pressure) and can
be avoided by not exceeding the cavitation threshold (Figure 1.2).
Projector depth (m)
Figure 1.2 Cavitation threshold as a function of depth
Figure 1.2 gives a conservative estimate of the cavitation threshold for fiequen-
cies up to about 10 kHz and for pulses of at least 5 ms duration. It predicts the
start of cavitation at some point before it spreads to the total radiating surface. In
practice, higher power intensities are possible, perhaps 3 or 4 times higher, but
these should be confirmed by experiment before finalizing a design.
Because cavitation takes a finite time to build up, albeit very short, the threshold
also changes with pulse length and fiequency:
Pulse length: the threshold is increased for very short pulses, up to about 3
times for a 0.5 ms pulse compared with its 10 ms value, but is sensibly constant
for longer durations.
Frequency: the threshold is little changed up to about 10 kHz, but between

1OkHz and, say, 300 kHz the threshold increases roughly linearly with fie-
quency.
Cavitation 7
Example 1.1
What is the maximum radiated power for a 100 ms pulse of frequency 10 kHz to avoid
cavitation, for a cylindrical array of 1 m diameter and 1 m height, containing circular
elements closely packed and operating at a depth of 5 m?
Radiating surface area = (nX 1 X 1) X (n/4) = 2.5 m2

Maximum radiated power = 2.5 X 2 = 5 kW
And assuming DIt = 10 dB,the source level is
SL = lolog 5000 + 171 + 10 = 218 dB

Example 1.2
What is the maximum radiated power at 50 kHz to avoid cavitation at a depth of 5 m
for a planar array of radiating surface 0.1 m2 and pulse length 0.5 ms?
The radiated power intensity may be increased 3 times for the short pulse and 5 times
for the higher frequency. Therefore
Maximum radiated power = 0.1 X 2 X 3 X 5 = 3 kW
And assuming DI, = 20 dB, the source level is
SL = lOl0g3000 + 171 + 20 = 226 dB

Example 1.3
What is the maximum radiated power at 200 kHz to avoid cavitation at a depth of
50 m for a square array of radiating surface 0.01 m2 and pulse length 0.5 ms? At this
frequency, the array has a small size of 100 mm X 100 mm, which is 131 X 131.
The radiated power intensity may be increased 3 times for the short pulse and 20 times
for the higher frequency. Therefore
Maximum radiated power = 0.01 X 50 X 3 X 20 = 30 kW
And assuming DIt = 20 dB, the source level is
Note, however, that the power intensity is 3000 kW/m2 and for even moderate losses,
say 20 per cent, the array must dissipate some 600 kW of heat per square metre or, to
lapse into imperial, 6 kW over a 4 in square! Clearly the limiting factor at higher
frequencies andor greater depths is not cavitation but the power-handling capability
of the array.
8 Sound
1.8 Interaction
When a number of projectors are assembled together in an array and driven
electrically, the velocity of motion of the individual elements is not constant, but
varies from element to element in a complex manner due to the acoustic inter-
actions between them.
Unless this is compensated in the design, the interactive effects will reduce the
total power output; affect the transmitted beam pattern; and produce a mismatch
with the transmitters, which could damage transmitters, projectors or both.
Interactive effects may be reduced in three ways:
Separating the elements of the away produces a larger than optimum array and,
particularly if the spacing is much greater than A/2, deterioration in the transmit
beam pattern. This may be acceptable if the resultant transmitter design problem
is eased and total costs reduced.
Making the individual elements large so that their self-radiation impedances are
much greater than the mutual radiation impedances between elements.
Using individual amplifiers to drive each element at the correct amplitude and
phase to yield a uniform velocity of motion across the array.
1.9 Changes to Arrays

The problems of cavitation and interaction are far from trivial and have largely
ensured the survival and renewed application of designs which, to the reader, must
appear historical.
For example, the projector used in Sonar 2001 (an active and passive sonar
system fitted to the Royal Navys first nuclear-powered submarine, HMS Dread-
nought) and still being considered for new sonars, was originally designed, tested
and proven in the 1950s. The projector used in Royal Navy surface ship sonars
2016 and 2050 dates from 1972 and is highly likely to see service well into the
twenty-first century.
Modular changes, however, S U G ~ as halving the height or increasing the
diameter of existing arrays, which preserve the element spacings and power
intensity, or moderate frequency changes with appropriate scaling of dimensions,
are likely to be successful without demanding long and risky testing.
Hydrophone Sensitivity 9
1.10 Projector Sensitivity

If the voltage at the terminals of a projector is v, then the response SVin db/V is
given by
sv = lOlog ;( );
= SL - 20logv
In terms of power P, the response Sw in dB/W is given by
sw = S L - 10logP
Example 1.4
If SL = 200 dB re 1 pPa (strictly, re the intensity of a plane wave of pressure 1 @a)
and P = 30 W, then SW= 200 - 15 = 185 dB/W.
1.11 Hydrophone Sensitivity

If the sound pressure in micropascals at the hydrophone is p and the voltage at the
open circuit terminals of the device is v,the hydrophone sensitivity is given by
Sh =: 20 log(v/p) = 20 log LI - 20 log p (dB/V)
Example 1.5
If 20 log p = 80 dB re 1 Pa and v = 1 pV, then S h = - 120 - 80 = -200 dB/V. To
find the output voltage from such a hydrophone, we use
If 20 log p = 100 dB, then
20 log v = -200 + 100 = - 100 dB

v:= 10pv
10 Sound
1.12 Spectrum Level

The spectrum level (SpL) is the sound level in a 1 Hz band. It is expressed in dE3
relative to the intensity resulting from a sound pressure of 1 p a . The band level
(BL) refers to the total intensity of the sound in a band.
If the spectrum is flat (white noise) then
BL = SpL+ 10logAf
Example 1.6
If the spectrum level is 40 dE3 and Af = 3000 I-kq then BL = 40 + 35 = 75 dE3.
If the spectrum level is not flat, the band level can be obtained by integrating the
intensity over the complete band. For ambient sea noise, the intensity may fall by 6 dB
per octave. The total intensity is then given by
Example 1.7
What is the sound level (total intensity), BL, of sea noise in a band from 100 Hz to
10 kHz, given that the spectrum level at 100 Hz is 100 dB?
and in dB form, BL = 100 + 20 = 120 dB. This is 20 dB less than if the spectrum
were flat.
It is seldom necessary to calculate BL for a non-flat spectrum. A practical sonar

receiver will either equalize @re-whiten) its input, or process (simultaneously) bands
narrow enough for the spectra to be assumed flat.
Sound in Air and in Sea Water
1.13 Sound in Air and in Sea Water

Note that in Tables 1.1 and 1.2 the velocity of sound - the longitudinal wave
motion - is almost constant for a given medium, whereas the particle velocity is
directly proportional to the pressure.
Table 1.1 Sound parameters in air

SOUND IN AIR pc = 415 kgm-2 s-'
Sound level intensity Pressure Particle Particle
I (W/m2) p <N/m2> velocity displacement
24 (m/s) u / o (m at
800 Hz)
Threshold of hearing 10-12 2 x 10-5 5 x 10-8 lo-"
0 dB re 20 pPa
Conversation 10-6 2 x 10-2 5 x 10-5 10-8
60 dB re 20 pPa
Threshold of pain 1 20 5 x 10-2
120 dB re 20 pPa
Tame 1.2 Sound parameters in sea water

UNDERWATER SOUND p c = 1.5 X lo6 kg m-' s-'
~~~~ ~ ~-
Sound level intensity Pressure Particle velocity Particle

I (W/m2> p (N/m2) u (m/s) displacement
U/UJ (m at
10 Hz)
DSS2 at 10 kHz 0.67 x 10-14 10-4 6.7 x 10-11 1.1 x 10-15
40 dB re 1 pPa
Typical reverb at 2 km 0.67 x 1 6.7 x 10-7 1.1 x 10-11
range 120 dl3 re 1 pPa
Array: source level 0.67 x 10-4 10-5 6.7 X lo-* 1.1 x
220 dB re 1 Clpa (1 atm)
Note: it is interesting that 1.1 X 1Owl5 is only 3 millionths of the diameter of a hydrogen molecule (0.28 nm)
12 Sound
1.14 Problems
1.1 The pressure, p , of an underwater sound is 100 pPa. What is the intensity, I, of
the sound and its level expressed in decibels compared with the reference intensity, I,?
1.2 What is the source level, SL, of a projector radiating 40 kW of acoustic power and
with a directivity index, DIt, of 15 dB?
1.3 What is the safe maximum radiated power to avoid cavitation, given a 100 ms
pulse of frequency 20 kHz and a planar array of 2 m length by 1 m height, containing
circular elements closely packed and operating at a depth of 10 m?
1.4 The total intensity of the sound in an octave from 2000 to 4000 Hz is 80 dB
relative to the intensity resulting from a sound pressure of 1 pPa. Assuming the band
is flat, what is the spectrum level, SpL, of the sound?
2
Arrays
2.1 Need for Projector Arrays

Transducers operating in the transmit mode - projectors - are assembled as arrays
in order to increase the source level of the transmitted acoustic pulse. The
directivity of a projector array concentrates the transmitted sound in a given
direction.
In the horizontal plane the transmissions may be omnidirectional or, with a
corresponding increase in source level, they may take place over some smaller
sector. In the vertical plane there is clearly no point in transmitting omnidirection-
ally, so a typical array is designed to make the vertical beamwidth between 3" and
30" to the 3 dB points.
The required source level is seldom available from one projector, except
perhaps at very low frequencies where the surface area is likely to be quite large.
A single projector requires a large surface area to avoid cavitation, and at higher
frequencies this is not a simple engineering solution: directivity would be
sacrificed or at best compromised. Finally, the reliability of an array of projectors
is much better because several, perhaps even half, of the projectors can fail before
sonar performance is critically impaired.
2.2 Need for Hydrophone Arrays

Transducers operating in the receive mode - hydrophones - are assembled as
arrays to improve the response of the array in a desired direction, thereby increas-
ing the signal-to-noise ratio and indicating the direction of a signal source. The
14 Arrays
directivity of a hydrophone array is due to all signals impinging upon the array
from the same direction being in phase, and therefore reinforcing one another.
2.3 Beam Patterns

The beam pattern of an array plots the array response against angle, describing the
relative responses of a beam formed from an array to signals and noise from all
directions. Mathematically, the beam pattern of an array is identical for both
transmit and receive. The improvement in the signal-to-noise ratio due to the array
is known as the array gain. A special case of array gain, where the signal is
coherent and the noise is incoherent, is the directivity index. This parameter is
simpler to calculate and will normally be a satisfactory measure of the increase in
the signal-to-noise ratio due to the array.
Directivity of u Dipole 15
2.4 Directivity of a Dipole

S1 and S2 are two point sources (forming a dipole) vibrating in phase and at equal
amplitudes (Figure 2.1). A hydrophone placed in the far field in the direction 0
receives two sound waves of equal amplitude (since S1 and S2 are almost equidistant
from the hydrophone), but of different phase corresponding to the distance S2 H :
Figure 2.1 Simple dipole
S2H = dsin0
and the phase difference is
2nd
cp = -sin0
I.
The sound pressure at the hydrophone is the vector sum of the pressures p1 and
p2, of the same amplitude and with a phase difference cp (Figure 2.2):
For a single source S1, the pressure is constant whatever the direction. When a
second source Sz is present, the pressure varies between 0 and 2 p as a function of 0:
cp
= 2pocos- = 2pocos
2
16 Arrays
PI +P2
/
/
/
/
/
/
/
/
P2
Figure 2.2 Vector sum of pressures
2po is the pressure, pa, on the axis of symmetry of the dipole (where 8 = 0) and
we can write
- = cos
Pa
G sin e)
The intensity varies as the square of the pressure, therefore
and in dB form, we have
The 3 dB beamwidth (the width in degrees at the half-power points of the dipole
response) for different spacing of the dipole elements is obtained by putting
[
2010g cos -sin8
(7 11 = 3 dB
Cd ) 7
cos -sin8 =-
nd . n
--slnO =-
;1 4
Directivity ofa Dipole 17
Table 2.1 gives the 3 dB beamwidth at four element spacings and Figure 2.3 shows
the beam patterns. When the spacing is exactly L/2, nulls appear at 90" and 270".
As the spacing is increased, secondaries (sidelobes) appear which are soon
unacceptable, and when d = L they totally change the beam pattern so that the
major responses now occur at the previous null angles.
Table 2.1 3 dB beamwidths

Element spacing, d 6 (degree) 3 dB beamwidth
(wavelength units) (degree)
90 180
30 60
19.5 39
14.5
U-
270" 800
Figure 2.3 Dipole beam patterns as a function of spacing in wavelengths

18 Arrays
2.5 The General Line Array

When the simple dipole theory is extended to an array of n hydrophones of
spacing d, the beam pattern is given by
b(8) = 20 log
{sin(nnd/A)sin 0)
{ n sin(xd/A)sin e)1
The main lobes of an array can be steered by introducing phase or time delays in
series with the elements. When this is done the beamwidth and sidelobe structure
of the array is changed. The beam pattern of the line array is then modified to
b(0) = 20log
sin nn(d sin O / L - d sin OJL)
n sin x(d sin 8/A - d sin &/A) 1
where 6, is the angle of steer at broadside.
This function is used to plot Figure 2.4, the 3 dB beamwidth of a line array at
various steer angles. In endfire (OS = 90") the beam pattern is searchlight shaped
and there is no lewright ambiguity. In a transition region close to 90" the 3 dB
beamwidth is ambiguous. Although the beamwidth varies as the beam is steered,
the directivity index of the array stays approximately constant. An approximate
formula for the 3 dB beamwidth of a line array, useful for steers up to about 60", is
where
L = length (m)
f = frequency (kHz)
0, = steer angle (degree)
At broadside, OS = O", this formula simplifies to
I I:
2e3 = -
The General Line Array 19
3 dB beamwidth (degrees)
5 6 7 8 910 20 30 40 50
Array length (wavelengths)
Figure 2.4 Beamwidth of a line array of A/2 spaced elements at various steer angles
20 Arrays
2.6 Line Array: Beam Pattern vs. Steer Angle

The beam patterns in Figure 2.5 illustrate the transition from broadside to endfire.
Note that the 3 dEt ambiguity is already apparent at a steer of 60" for this short
(51) line array.
OdB
- 20
0" steer 30" steer

(broadside)
OdB OdB
60" steer 90" steer

(endfire)
Figure 2.5 Line array, 10 elements spaced A/2: beam patterns at various steers
2.7 Broadside Array: Length and Spacing

The beam patterns in Figure 2.6 further illustrate the effects of array length and
element spacing at broadside. Note again the total change to the beam pattern as
the spacing approaches 1.
Beam Puttern.for.a Continuous Line 21
Figure 2.6 Effects of length and element spacing on a broadside array: (a) element spacing
constant at A / 2 , array length increasing; (b) array length constant at 3L, element spacing
increasing
2.8 Beam Pattern for a Continuous Line

For a continuous line of length L, the pressure in the far field is found by
integration to be
pe
_.
-
-
sin{ (nL/il)sin S}
pa (nL/il)sin 8
This expression (Figure 2.7) is a sin x upon x fwnction, where x = (nL/A)sin 8; it

is somewhat easier to handle than the earlier expression derived from the
summation of n hydrophones of spacing d, an array whose length is given by
( n -- 1)d.
The intensity of the beam pattern in dB form is
b(0) = 20 log
sin{ (nL/iZ)sin 0)
(nL,/l)sin 8 1
22 Arrays
0
dB
- 10
-20
-30
(nLA) sin 0
Figure 2.7 Normalized beam pattern for a continuous line
There are nulls in the beam pattern where x = f n , f2n, f3n, . . ., and peaks in
the beam pattern where x = 0, fsn, f;n, . . ..
The sidelobe levels are given by
where rn is the sidelobe number. The results apply quite well to an array of spaced
elements provided their spacing, d, is not more than about A/2 and the array is at
least 3A in length.
Table 2.2
X x-' sin x 20 log(x-' sin x)
0 1 0
1.39 0.7 1 -3
2.27 0.32 - 10
2.85 0.1 -20
The numerical values in Table 2.2 can be used to derive some useful approxima-
tions for the shape of the main lobe. Beginning with the 3 dB beamwidth (Hz):
Beam Pattern-for a Continuous Line 23
nL
1.39 = -sni 03
A
Converting to radians, for small angles we obtain
1.39 = n-e3
L )'(
1
if the elements are spaced by 1/2,then
263 = 100/n
Similar expressions can be produced for 10 dB and 20 dB beamwidths and the

results are given in Table 2.3. Note that the 20 dB beamwidth is about double the
3 dB beamwidth.
Table 2.3 Beamwidth expressions
50A/ L 82A/ L 104A/ L

1OO/ n 164/ n 208/n
Note: the formula 100/n is an alternative to the approximate
formula of Section 2.5
Example 2.1
What are the 3 dB, 10 dB and 20 dB beamwidths for a line array of length 6;1? What
are the bearings of the first-, second- and third-order sidelobes relative to the direction
of maximum intensity?
First sidelobes occur where fin = 6 n sin 8, i.e., where 8 = f14.5"

Second sidelobes occur where ~ Z ~ J -=
C 6n sin 8, i.e., where 8 = f 2 5 "
Third sidelobes occur where f $ n = 6n sin 8, i.e., where 8 = f 3 6 "

24 Arrays
2.9 Shading
Except for the simple dipole with element spacings less than or equal to 1/2, all
the above beam patterns have significant sidelobe levels. The first-, second- and
third-order sidelobes are respectively only 13.5, 18 and 21 dB below the peak level
of the main lobe. Strong signals will be detected through the sidelobes of adjacent
beams as well as, correctly, in the main lobe of the beam at the bearing of the
signal. The resultant bearing ambiguity and additional, false, signals complicate
all further processes. The aim, therefore, should be to produce the narrowest
possible main lobe consistent with some reasonable level of sidelobes.
By using amplitude shading or weighting, i.e., varying the amplitudes of the
signals applied across the array (in transmit) or coming fiom the array (in receive),
it is possible to reduce the sidelobes, but always at the expense of widening the
main lobe and reducing the DI somewhat. In transmit, reducing the DI means that
the source level (SL) is reduced as a consequence. In an active system it is the
combined, transmit and receive, beam pattern which is important and it is hardly
ever desirable to hrther reduce the sidelobes at the expense of SL - far better to
adequately shade the receiver elements only.
All shading functions widen the main lobe by some factor; the factor is 1.36 for
the cosine shading hnction in Figure 2.8. Note also that the skirts of the main lobe
be (dB)
21 dB
.-
It 2n 3n
(nLA) sin 8
Figure 2.8 Normalized beam patterns: line array, elements spaced A / 2

Shading 25
extend into the first sidelobes of the unshaded pattern, and therefore simply
quoting the level of the first sidelobe is an incomplete description and may be
misleading (Section 12.6).
Mathematical procedures, based on work by Dolph and Chebyshev, determine
the amplitude shading coefficients which yield the narrowest main lobe for a
specified level ofsidelobes (Figure 2.9). For example, the coefficients for a six-
element line array to yield -30 dB sidelobes are 0.30, 0.69, 1, 1, 0.69, 0.30. And
the main lobe is broadened, at the 3 dB points by a factor of 1.3.
-10 -
-20 -
-30 -
( J c L A )sin 8
Figure 2.9 Unshaded and Dolph-Chebyshev shaded, normalized beam patterns
Some characteristics of commonly used weighting functions, normalized to the

unweighted array, are given in Table 2.4. All dB values in the table are rounded to
the nearest 0.5 dB. Note that all of the other weighting functions broaden the main
lobe more than the Dolph-Chebyshev -30 dB weighting function. The theoretical
sidelobe levels are generally lower but they are seldom fully achieved in practice.
Element failures and, in the case of a towed line, curvature of the array, ensure that
it is unsafe to expect sidelobes much better than, say, 25 dB down on the main
lobe.
Note the similarity between amplitude shading used to reduce bearing sidelobes
in a beam pattern and pulse shaping used to reduce frequency sidelobes in the
frequency spectrum of a pulse, as described in Chapter 9.
26 Arrays
Table 2.4 Commonly used weighting functions

Weighting function Reduction in Broadening 10 log BF3 Broadening
transmit power factor, 3 dB (a) factor, 20 dB
(a)
None 0 1 0 1
Dolph-Chebyshev 30 dE3 3 1.3 1 1.5
Cosine 3 1.4 1.5 1.6
Gaussian 3 1.4 1.5 1.6
Hamming 3.5 1.5 2 1.7
Squared cosine 3.5 1.6 2 I .9
Binomial 4 2.0 3 2.2
Example 2.2
If the array of the previous example is (a) Dolph-Chebyshev 30 dB shaded and (b)
Hamming shaded, what are the 3 dB and 20 dB beamwidths?
From the previous example, the unshaded beamwidths are 8.3" and 17.3". Therefore
the shaded beamwidths are as follows:
1.3 X 8.3" = 10.8"

(a) 1.5 X 17.3"= 26"
1.5 X 8.3" = 12.5"

(b) 1.7 X 17.3"= 29.4"
Shaded Arrays: Transmit Source Levels 27
2.10 Shaded Arrays: Transmit Source Levels

Shading a transmit array reduces the total output power (assuming that the power
outputs of unity-weighted elements cannot be increased). The total power is also
spread over a greater beamwidth and therefore the SL is significantly reduced
compared to an unshaded array. A reduction in SL will reduce the noise-limited
performance of an active sonar and should be avoided. However, if it is essential -
perhaps to broaden the vertical beamwidth of a sonar to achieve adequate depth cover
at close ranges - the reduced SL may be estimated quite simply using Table 2.4:
SL reduction = total power reduction + 10 log BF3

where BF3 is the broadening factor at 3 dB.
Example 2.3
A Dolph-Chebyshev array designed to achieve -30 dB sidelobes will have its SL
reduced due to the reduction in transmit power, 3 dB, and due to the broadening factor,
1 dB, Le., a total reduction of 4 dB.
Check
Given a line array of six elements, spaced A/2 and each capable of a maximum
output of 1000 W omnidirectionally, we obtain
SL = lOl0g6000 + 171 + 1010g6 = 217 dB

If the line array is now amplitude weighted,
0.3, 0.69, 1, 1, 0.69, 0.3
the total power output is
90 + 480 + 1000 + 1000 + 480 + 90 = 3140 W

and the power reduction is 10 log(6000/3 140) = 2.8 dB.
Also, because the 3 dB beamwidth is increased by a factor of 1.3, this power is
spread over a larger sector and the average power is reduced by 10 log 1.3 =
+
1.1 dB. The total reduction is therefore 2.8 1.1 = 3.9 dB, which shows close
agreement. The source level is now 2 17 - 4 = 2 13 dB.
28 Am y s
2.11 Directivity Index

When the hydrophones are assembled in arrays there is an improvement in the
signal-to-noise ratio (SIN). This improvement is known as the array gain and is
defined as
It is difficult to compute this quantity because it relies on knowing the coherences

of signal and noise across the dimensions of the array.
An important case is where the signal is a plane wave and coherent, and the
noise is isotropic and incoherent. Here the AG reduces to a more tractable and
easily visualized quantity known as the directivity index (DI):
I peak intensity of radiated pattern

D1 = l o log average intensity of radiated pattern
We will first derive the directivity index for a simple dipole and, avoiding complex
mathematical treatment, extend this result to cover the directivity indices for more
complex arrays.
DI of a Simple Dipole 29
2.12 DI of a Simple Dipole

The dipole, elements S1 and S2, is at the centre of a large sphere of radius R
(Figure 2.10). The sound flux passing through the shaded zone on the diagram is
then
Figure 2.10 Directivity index of a simple dipole
If I is the average intensity at the surface of the sphere, the total sound flux
passing through its complete surface is
and therefore
and the directivity factor, K (DI = 10 log K ) , is given by

30 Am y s
and substituting
(see Section 2.4), we obtain
Integrating by changing the variable, we obtain
and
When the elements are separated by d = J/2, this reduces to DI = 3 dB

When the elements are separated by d = 0, then DI = 0 dB
We would expect both of these results intuitively: two elements are twice (3 dB) as
good as one; and when the two become one (d = 0), the dipole is omnidirec-
tional and DI = 0 dB.
DI qf a Line Army 31
2.13 DI of a Line Array

For a line array of n elements spaced A/2,
IDI= 10lognl
Clearly the elements in the line are only spaced A/2 at one frequency. Therefore,
for a wideband system the DI will change with frequency. As the frequency is
reduced, the line approximates to a continuous line of length L = n&/2, where A0
is the wavelength at fo, the design frequency for the line array and
Note that the DI falls at a rate of 3 dB per octave.

An alternative expression for the DI of a line which will suffice for two or three
octaves below the design frequency fo (i.e., at a frequency where d = A / 2 ) is
Note that, from its definition, DI is always positive and cannot fall below 0 dB.
As the frequency is increased, the line can no longer be considered as
approximating to a continuous line, and excessive sidelobes appear as the spacing
approaches A. Therefore this equation should not be used for frequencies above,
say, 1.5f0.
Example 2.4
A line consists of 64 elements spaced A/2 at 1 kHz. What is its DI at 1 kHz, 100 Hz
and 1.5 kHz?
Length of line L = 32A = 32 X 1.5 = 48 m

At I kHz D I = 10logn= 1010g64= 18dB
At 100 Hz DI = lOlog(2L/d) = 1010g(2 X 48/15) = 8 dB
At 1.5 kHz DI = 18 + lolog 1.5 = 19 dB

32 Arrays
2.14 DI of a Planar Array

The DI of a planar array is simply calculated by considering it as the sum of a
number of linear arrays:
where m is the number of lines of n elements, all spaced by A / 2

The array is usually baffled, as when mounted on the flank of a submarine. For a
baffled array, add 3 dB to this result, so the wideband equation of the previous
section then becomes
Note again that DI is always positive and cannot fall below 0 dB. As the fiequency
is increased, the planar array can no longer be considered as approximating to a
continuous surface, and excessive sidelobes appear as the spacing approaches A.
Therefore this equation, in common with the equation for a line array, should not
be used for frequencies above, say, 1.5fo.
Example 2.5
A flank array has 8 rows of 32 elements spaced A/2 at 1 kHz. What are its DI at
800 Hz and its 3 dB beamwidths at 1 kHz and at steer angles of (a) broadside, (b) 30,
(c) 60?
DI = 3 + 1010g(8 X 32) + 2010g0.8 = 25 dB
The length of the array is 161 and its height is 4L The horizontal and vertical
beamwidths are given by the same formula:
DI of a Planar Array 33
Horizontal beamwidth = 76/(24 X 1) = 3.2"

(a) Vertical beamwidth = 76/(6 X 1) = 12.7"
+
8 h = (76/24)( 1 900/4000) = 3.9"
(b) 0, = 12.7" (steer is only horizontal)
4 = (76/24)(1 + 3600/4000) = 6.0"

(') 0, = 12.7" (steer is only horizontal)
34 Arrays
2.15 DI of a Cylindrical Array

A cylindrical array of elements comprising a number of vertical lines (staves) is a
good practical shape (Figure 2.11). The number of staves is simply related to the
beam spacing and is also, conveniently, a binary number. This makes the
beamforming simple because all beams are identical and provide 360" cover in
azimuth. The horizontal beamwidth is constant and the vertical beamwidth is a
function of the stave height.
Element
Stave
Ring
Figure 2.1 1 Cylindrical array
To form a beam, a certain number of staves - usually about one-third the

periphery of the cylinder - are phased to make them appear as a planar array.
Adjacent beams are then formed by stepping around the array at an indexing angle
equal to the required beam spacing.
A practical formula for the DI of a baffled cylinder is
DI of a Cylindrical Array 35
where
h = height (m)
d = diameter (m)
fo = design frequency (kHz)
This is for a 'full' beam, i.e., a beam formed by using all the usefully
contributing elements on the surface of the cylinder (an aperture of about 120" of
the cylinder). The design frequency, fo, is the fkequency where the elements are
spaced A/2.
Example 2.6
A cylindrical array has height 1 m, diameter 2 m and design frequency 5 H z , so
DI = 1010g(5 X 1 X 2 X 25) = 24 dB.
If 'half-beams' are formed for cross-correlation or phase comparisons (see later), the
DI of each half-beam is 3 dB less than the DI of the corresponding full beam. Here are
some practical formulae for the beamwidths of a cylinder.
Horizontal beamwidth
This formula uses one-third of the array's periphery, and we obtain
& = - 88 = 8.8"
Vertical beamwidth
This is the same as the formula for line arrays, and we obtain
76
0 -- = 15.2"
"-1X5
36 Arrays
2.16 DI Formulae for Simple Arrays

The above formulae for simple arrays are mostly in terms of numbers of elements,
spaced A/2. Because n = 2L/A, these formulae may also be expressed in terms of
the dimensions of the arrays. Table 2.5 lists the formulae for linear, planar, square,
circular and cylindrical arrays in both forms. Some of these formulae are
approximate but are perfectly adequate for practical sonar design.
Table 2.5 Formulae for DI

h Y Directivity index (a)
Line lolog n 10 log(2 L / A )
Planar 10 log rnn 10 10g(4fi/1~)
Square 20 log n 20 log(2Ll.A)
Circular 20 log n 20 log( 1.7742)
Cylindrical, baffled (see notes) 3 + 10 log rnn 10 log(5hdfi) or 10log( 1 1 h d / A 2 )
Notes: m is the number of elements in one-third of a ring, n is the number of elements in a stave; fo is in
m
2.17 Conformal Arrays

Conformal arrays follow the hull form of a vessel, usually a submarine. They have
been fitted to single-hulled submarines, and in order to provide a good surveillance
capability, they occupy both port and starboard flanks and extend around the bow.
Estimates of DI and beamwidths (which change significantly with azimuth) can be
made by approximating the array to an assembly of planar arrays (on the flanks
where the curvature is not great) and part cylinders (on the bow).
2.18 Spherical Arrays

Spherical arrays are used where greater vertical coverage is required (or to
compensate for ship movements). However, given the wastehl use of projectors or
hydrophones to provide coverage in this manner, and the reduced DI compared
with a cylindrical array of the same diameter, it is debatable whether adequate
vertical cover is not better achieved by vertical steering of a cylindncal array -
using complete staves rather than selecting elements as with a spherical array -
and accepting the small losses in DI which then exist.
Volumetric Arruys 37
2.19 Volumetric Arrays

When the elements occupy three dimensions, e.g., in two or more concentric
cylinders, the array is known as volumetric (Figure 2.12). It is then possible, by
suitable separation and phasing between the cylinders, to generate directivity
patterns and directivity indices equivalent to a baffled cylindrical array.
Consider an array of two concentric cylinders each made up of 16 X 6 element
staves (Figure 2.12). The elements are separated by A / 2 and the cylinders are
separated by A/4. The design frequency is 5 kHz. The overall size of the array is
therefore a cylinder of diameter 0.75 m and height 0.75 m. Staves A and B are
separated by d/4, and if the relative phase is 90,the horizontal directivity of the
pair of staves is then cardioid in shape, as shown. The resultant directivity index of
the complete array is therefore increased by 3 dB and is the equivalent of a baffled
cylindrical array given by
DI = 101og5hdf20
= lOlog(5 X 0.75 X 0.75 X 25)
= 18.5 dB
Helicopter dipping sonars
Helicopter dipping sonars often use volumetric arrays. The array is mounted on
ribs which fold (umbrella fashion) for stowage and lowering &om the helicopter
into the sea. When the array reaches the required depth, the rib fastenings are
released and the array configures as a number of concentric cylinders. A separate
transmitting array of omnidirectional projectors (omnidirectional in the horizontal
plane) may complete the assembly.
38 Arrays
I.75 m
t
I
Vectors add in quadrature
I
+----
(0)
-+-
A Y
\
A ------ 4
Vectors cancel
/ Vectors add
Figure 2.12 Volumetric, concentric cylinders, array

Beamformers 39
2.20 Beamformers
Analogue
Analogue beamformers are unlikely to be used except for very simple arrays in
modern systems. They are assembled using lumped constant (LC) networks which
must have a linear phase relationship with frequency. They are difficult to modify
and cannot, at least not simply, be made to respond to sound speed changes.
(When the speed of sound changes, because its frequency cannot change, its
wavelength must change. Therefore the parameters of the beam will also change
unless the speed of sound is used in the beamforming process.)
Digital
Digital beamformers use shift registers or random access memory store (RAM) to
hold sampled data. The beams are then formed by addressing the samples. Shading
functions and sound speed variations are easily accommodated.
Narrowband systems
If the beamformer is for use with a narrowband system then phase delays, which
are only exactly correct at one frequency, may be used to form the required beams
instead of the more usual time delays which are constant with frequency and hence
suitable for broadband systems.
40 Arrays
2.21 Domes and Arrays

Arrays seldom have an ideal hydrodynamic shape and they are not sufficiently
robust to survive without some protection. They are therefore normally housed in
domes made of acoustically transparent material. Modern domes are often built
up using glass-fibre epoxy resin mixes, which give a good compromise between
acoustic and mechanical properties. The acoustic windows in the dome may
simply be homogeneous, thinner areas or they may use a different material (e.g.,
rubber). No material transmits sound loss-free - some energy will be absorbed
and some reflected. The design of bafles, which are used to attenuate unwanted
noise fiom the propellers of a ship, for example, can also help to minimize the
unwanted effects of reflections (Figure 2.13).
Cylindrical array
\ \ \
Plane wave incident on dome
Figure 2.13 Baffled dome: plan view
The elongated dome meets the hydrodynamic requirements. The array, however,
can only be close to the dome over about half its circumference. Baffle 1 reduces
the stern noise. Performance astern (known in sonar as the non-operational
bearings) will inevitably be impaired and this determines the position of baffle 1.
It cannot be too close to the array without affecting too many non-astern beams.
Problems 41
The resultant gap between the array and baffle 1 introduces unwanted reflections.
Therefore baffle 2 is introduced to eliminate the more significant reflections.
Reflection A, in a region where the array is close to the dome, is harmless but
reflection B (without the baMle) would contribute energy to several incorrect
beams.
2.22 Problems
2.1 What are the beamwidths to the half-power points of a line array of length 10 m
and frequency 4 kHz (i) normal (broadside) to the array axis and (ii) at a steer of 60?
2.2 A line array has a length of 6L. What is the level of the fourth-order sidelobes (in
dB relative to the main lobe) and where do their peaks occur in the beam pattern?
2.3 What is the DI of a baffled planar array of length 5 m and height 2 m at 8 kHz?
The elements are spaced A / 2 at 10 kHz. First calculate the DI using the numbers of
elements and then compare with the result using the dimensions and wavelength.
2.4 What is the DI of a baffled cylindrical array of height 2 m and diameter 3 m at

5 kHz? How much would the source level of the array be reduced by halving its
height?
3
Propagation of Sound in the Sea
3.1 Propagation Loss

Propagation loss (PL) is a quantitative measure of the reduction in sound intensity
between the source and a distant receiver. If 10 is the intensity of the source
referred to a point one metre from its acoustic centre and I, is the intensity at the
receiver, then the propagation loss between source and receiver is
/ P L = lolog($) (dE3)l
3.2 Losses
A first approach to quantifling the likely PL is to consider it as the sum of a
spreading loss and a loss due to absorption. Other losses - not functions of range
- include scattering and refraction, and these will be considered later on.
44 Propagation of Sound in the Sea
3.3 Spreading Losses

The spherical spreading law
Refer to Figure 3.1(a). When the source is located in an unbounded and lossless
medium, the power is radiated equally in all directions and the total power, P,
crossing spheres of increasing radii surrounding the source does not change with
range. Therefore, since power = intensity X area,
and if rl = 1 m, the PL to range r is
PL = lolog(;) = 1010gr2
or in logarithmic form,
where Y is in metres.
Figure 3.1 Spreading losses

Spreading Losses 45
The cylindrical spreading law
Refer to Figure 3,l(b). When the source is bounded by parallel planes separated
by h metres, the power - again unchanging - crossing cylindrical surfaces of
increasing radii surrounding the source, is given by
and if' ~1 = 1 m, the PL to range Y is
PL = lolog(;) = 10logr
or in logarithmic form,
I P L = 10logrI
where Y is in metres.
3.4 Absorption Losses

When a sound wave travels through sea water, absorption losses occur through two
principal mechanisms:
Viscosity: losses due to viscosity are present in fresh water and salt water. This
contribution is proportional to the square of the frequency and accounts for the
straight line for fresh water when a is plotted on a log-log graph like Figure 3.2.
Figure 3.2 Attenuation coefficient (dB/km)

Absorption Losses 47
Molecular relaxation: losses due to molecular relaxation are only present in salt
water. The mechanism is a reduction of molecules to ions induced by the
pressure of the sound. At very high frequencies (greater than about 500 kHz)
pressure changes are too rapid for the relaxation to take place and therefore no
energy is absorbed. Magnesium sulphate relaxation is dominant over the
frequency range 2 to 500 kHz.Below 2 kHz boric acid relaxation contributes to
the losses.
Extensive measurements of these losses have been made and several empirical
formulae exist which relate them to frequency, depth (pressure) and salinity. The
total loss is given as an attenuation coefficient, a, in dB/ km.
The attenuation coe$cient, a, increases rapidly with frequency and changes
with temperature. It also varies with depth and salinity, but less strongly. Figure
3.2 plots the variation of a with frequency at a temperature of 10C and a salinity
of 35 parts per thousand (ppt), based on the formulae of Francois and Harrison.
An approximation for a, usehl between 0.5 kHz and 100 HZ,in 'standard' sea
water is given by
Table 3.1 is nut based on the approximation; it offers a selection of values for a
that will be useful in performance comparisons and in assessing the effects of
frequency changes on a design. Linear interpolation will be sufficiently accurate
to determine cr. at other frequencies.
Table 3.1 Values for a: not based on the approximation
0.5 MIZ 1 kHz 2 kHz 5 kHz 10 Mz 20 kHz 50kHz lOOkHz 200 kHz 500 kHz
5 0.02 0.06 0.14 0.33 1.00 3.80 15 30 55 120
10 0.02 0.06 0.14 0.29 0.82 3.30 16 35 60 125
15 0.02 0.06 0.14 0.26 0.68 2.80 17 40 65 130
3.5 Spherical Spreading and Absorption

The sum of spherical spreading and absorption losses is a useful working rule for
initial design and performance comparisons:
lPL=20logr+arx (dB)]
A remarkably good fit to measured propagation losses and to more complex

propagation models is frequently obtained. It is generally somewhat pessimistic,
as might be expected since there is usually some trapping which prevents spherical
spreading except at short ranges. Some workers use 15 log r as a compromise
between spherical and cylindrical spreading.
Figure 3.3 plots spherical spreading loss alone and with the addition of
PL = 20 log + ar x 10-3
a fWz)
a 0.06 1
b 0.2 3
c 1.0 10
d 7 30
e 33 100
f 64 300
Figure 3.3 Propagation loss curves: effect of absorption

Propagation in the Real Ocean 49
absorption at selected frequencies. Note that absorption is insignificant below

1 Mlz, even at 100 km,whereas at 100 kHz absorption is prohibitive even at 2 km.
It is interesting to compare sound and electromagnetic radiation losses in the sea:
Sound at 30 kHz, a = 5 dB/km

EM. wave at 30 kHz, aem= 7500 dB/km
(aeln
= 1.4 X 103f1/2dB/km)
This severe attenuation - even at low frequencies (low for electromagnetic

radiation) - makes an underwater radar no match for an acoustic (sonar) system
in the highly conductive oceans.
3.6 Propagation in the Real Ocean

Refiaction, scattering and the presence of the ocean boundaries - surface and
bottom - ensure that, except at very short ranges, free-field conditions never exist
in the real ocean and much effort is still being expended on developing reliable
models which take account of this.
The speed of sound in the sea - its absolute value and, more importantly, its
variation with depth - is fundamental to all models. A knowledge of the sound
speed profile (SSP) can help the sonar designer and operator to choose the
appropriate sonar propagation mode: a towed array or a variable depth sonar
(VDS) may be placed at an appropriate depth to avoid shadow zones, or the beams
of a hull-mounted sonar may be depressed to exploit bottom bounce or conver-
gence zone (CZ) paths.
3.7 The Speed of Sound

The speed of sound in the sea depends on the temperature, pressure (depth) and
salinity. A variety of empirical formulae exist for its calculation; here is one due to
Leroy:
c = 1492.9 + 3(t - 10) - 6 X 10-3(t - -4 X 10-2(t - 18)2
+ 1.2(s - 35) - 10-2(t - 18)(s - 35) + h/61

where
c = speed of sound (m/s)
t = temperature ("C)
s = salinity @pt)
h = depth (m)
The speed of sound at 1O"C, at zero depth, and for a salinity of 35 ppt is 149Om/s.
Here are some approximate coefficients for sound speed valid for use with this
'standard' sound speed:
Temperature Ac/At = +3.4 m/s per "C

Salinity Ac/As = +1.2 m/s per ppt
Pressure (depth) Ac/Ah = +17 m/sper 1000 m
Sound Speed Profiles 51
3.8 Sound Speed Profiles

Sound speed profiles (SSPs) are graphs of the speed of sound in the sea against
depth. The SSP depends on the location, the season, the time of day and the
weather. In most locations the salinity can be considered constant at 35 ppt, but
some environments have a different salinity. Near estuaries the salinity is highly
variable; in the Arctic melting ice means lower salinity near the surface; and in
parts of the Baltic very low salinities are present at all depths.
Low salinity also means very low attenuation coefficients and therefore reduced
propagation losses. A sonar designed for operation solely in the Baltic could, with
useful reductions in size and cost, use higher frequencies than required for normal
salinity.
It is the temperature of the sea as a function of depth that is the most variable
and the most difficult to determine. It is usually measured by a bathythermograph,
which may be deployed from vessels or aircraft, and typically has an accuracy or
resolution of about 0.25"C.
The speed of sound increases with temperature and with depth. When the water
near the surface is warmer than at greater depths, there are two opposing
tendencies as the depth increases:
The speed of sound decreases with decreasing temperature

The speed of sound increases with increasing pressure
The result of these opposing tendencies is to produce SSPs which vary widely
within the first few hundred metres of depth, and these SSPs are further
complicated by diurnal changes as well as mixing of the surface layer by wind and
waves.
Figure 3.4 shows a typical deep sea SSP divided into four principal layers:
Surface layer (duct): a layer of isothermal water mixed by the action of wind on
the surface of the sea. Sound tends to be trapped in this layer by surface
reflections and upward refractions.
Seasonal thermocline: temperature decreases with depth. During sumrner and

autumn the thermocline is strong and identifiable. During winter and spring it is
weak and merges with the surface layer.
Main thermocline: little affected by the seasons. This is where the main increase
in temperature over the cold depths of the sea occurs. Although the pressure
-
Speed of sound ( d s )
Surface layer (duct)
100 m
Seasonal thermocline
-------
Main thermocline
1OOOm
2000 m
Deep isothermal layer
5
e
P
1
Figure 3.4 Typical deep sea SSP
increases with depth, the net effect of temperature and pressure changes is to
reduce the sound speed through this layer.
Deep isothermal layer: constant temperature of about 4C right to the bottom.

The speed of sound increases with increasing pressure. At high latitudes the
layer extends closer to the sea surface, and in the Arctic it may completely
eliminate the other layers. The tendency is shown in the dashed curve of Figure
3.4 for a latitude of about 60".
3.9 Deep Sound Channel

Between the negative gradient of the main thermocline and the positive gradient
of the deep layer there is a sound speed minimum, where sound tends to be focused
by refraction. The depth at which this focusing occurs is known as the deep sound
channel (DSC). To exploit this channel, the source is placed close to the minimum
(which may be only a few hundred metres at high latitudes) and, because the
spreading is cylindrical, very long-range propagation is possible. Best results are
achieved when the receiver is close to the axis of the channel. Figure 3.5 shows a
simplified sound ray plot for a source placed at a depth close to the sound speed
minimum and with a vertical beamwidth of about 20".
Reliable Acoustic Path 53
Temp. Speed Range ___*
20" vertical beamwidth
Figure 3.5 Deep sound channel propagation
3.10 Reliable Acoustic Path

Placing the source deep in the sea (at least 1000 m) can improve detection of
shallow targets, i.e. targets at typical submarine diving depths (Figure 3.6). The
path is known as 'reliable' because it is insensitive to highly variable surface
effects and bottom losses. Conditions for reliable acoustic path (RAP) exist when
the source is placed at a depth, the critical depth, where the sound speed is equal
to the sound speed at the surface (Figure 3.7).
20 40 60 km
Depth
1500 m - 10" vertical beamwidth
Figure 3.6 Deep source: reliable acoustic path
Note that, particularly at high latitudes, the DSC may be close to the RAP
critical depth. A capability to steer the sonar beam vertically by, say, 5" (assuming
a 10" vertical beamwidth) as well as a capability to deploy the sonar at variable
depths will assist in reducing the shadow zones that would otherwise exist.
The great depths at which it is necessary to deploy both source and receiver - at
0 0 0 0 0 0
\ I
1000 1000 lo00
2000 2000 2000 -
3000 3000 3000 -
4000 4000 4Ooo-

Depth (m)
Figure 3.7 Critical depths for RAP: (a) tropical, latitude 20"; (b) temperate, latitude 50"; (c)
Mediterranean
least 500 m for DSC mode and 1000 m for RAP mode - impose severe engineer-
ing and mobility constraints on their use. The design of elements capable of
operating at depths in excess of, say, 1000 m is difficult if not impossible and
losses in the long cables will limit the achievable source levels.
3.11 Surface Duct Propagation

When surface winds and waves mix the shallow layers of the sea to produce a
nearly isothermal layer, the pressure effect dominates and the sound speed
increases down to a depth - the bottom of the layer - where the temperature starts
to fall and the sound speed begins to reduce until the minimum of the DSC is
reached. This isothermal layer, the su@ace duct, can be as small as 5 m and as
great as 200 m. Typically, ducts of 50-100 m are common in the colder waters of
the world.
Figure 3.8 shows the effect on sound transmitted fiom a source within the duct.
Rays that are projected close to horizontal are refiacted upwards and undergo
multiple surface reflections. On the other hand, rays which penetrate the layer are
refracted downwards at first, thus producing a zone - known as a shadow zone -
where hardly any sound energy penetrates. Targets within the shadow zone, i.e.,
Surface Duct Propagation 55
Figure 3.8 Surface duct propagation: shallow source
below the layer, are therefore difficult to detect. As with all propagation in the sea,
no mode is perfectly described by a simplified ray trace. The shadow zone is an
area where the sound intensity is greatly reduced and the transition from the
surface duct is not abrupt.
Increasing the depth of the source so that it is below the layer (Figure 3.9) has
the effect of increasing the range of the start o f the shadow zone, but it may then
extend into the duct. As with DSC and RAP modes, the ability to manoeuvre in
depth - a variable depth sonar (VDS) deployed from a moving surface ship or a
hovering helicopter - offers significant operational advantages.
Figure 3.9 Surface duct propagation: deeper source

3.12 Convergence Zone Propagation
Figure 3.10 Convergence zone propagation
Rays projected at steeper angles of depression, 5" or more, are bent downwards at
first - producing a shadow zone. At greater depths the pressure bends these rays
upwards to form annuli of high intensity (Figure 3.10); each annulus is known as a
convergence zone (CZ). The water must be deep enough for upwards refraction to
prevent the rays from hitting the bottom. Typically the water depth must be in
excess of 3000 metres. Depending on the bottom depth, the first convergence zone
will occur at around 30-50 km and it will be 3-5 km wide.
The propagation is sphericd spreading (there are no boundaries to confine the
sound) and the absorption losses are those appropriate to a temperature of about
4C (the temperature over most of the path) and the focusing effect gives a
convergence gain of typically 3-6 dB.
Successive convergence zones exist at multiples of the original range. The
propagation losses (two-way) will prohibit the use of any but the first convergence
zone for active systems, but passive systems can make detections at the range of
the second or even third convergence zone.
Bottom Bounce Propagation 57
3.13 Bottom Bounce Propagation

Propagation is possible by using bottom reflections (Figure 3.1 1). The sonar beam
is now deliberately trained downwards at relatively steep angles. The effectiveness
of the mode is determined by the nature of the bottom, whether it is absorptive or
reflective, and how the bottom loss varies with angle of incidence. As with
convergence zones, there exists a range annulus, which varies with depression
angle of the sonar beam, and at small angles of incidence this annulus can be very
wide. No focusing gain exists and the bottom reflection loss is typically between
10 and 20 dB. Therefore the mode is very demanding in projector power and array
size (because of the lower frequencies necessary to limit absorption over the long-
range paths).
Figure 3.11 Bottom bounce propagation
An active bottom bounce (BB) system will be large, very demanding in ships
power supplies and transmitters and requiring a large site for the installation of its
arrays. Successive range annuli exist at multiples of the first annulus. The
propagation losses (two-way) will prohibit the use of any but the first for active
systems, but passive systems can make detections at the ranges of the second
annulus.The range annuli for BB mode are significantly wider than for the focused
CZ mode. There must always be a shadow zone out to some minimum range, at
least 10 km, dependent on depth and the allowable angle of depression of the
sonar beam (allowable from the viewpoint of bottom loss).
3.14 Propagation Loss Models

The spreading, refraction and reflection of sound in the sea have so far been
discussed in a qualitative manner only, except with very simple - but often
adequate - quantitative expressions. The propagation of sound in an elastic
medium is described mathematically by solving the wave equation using the
appropriate boundary and medium conditions for a particular environment.
There are two approaches to solving this equation:
Wave theory: the propagation is described in terms of characteristic functions

called normal modes, each of which is a solution to the equation. The modes are
combined additively to satisfy the required boundary and medium conditions.
Ray theory: this postulates wavefronts and the existence of rays which indicate
where the sound from the source is going. It does not provide a good solution
when the radius of curvature of the wavefront or the pressure changes
significantly over the distance of a wavelength. It is therefore restricted to short
wavelengths.
The two approaches are compared in Table 3.2. Wave theory will not be consid-
ered further. The Sonar Modelling Handbook (published by DERA, UK Ministry
of Defence) is a useful source of information on the many wave theory and ray
theory models and their practical limitations.
Table 3.2 Wave theory versus ray theory

Wave theory Ray theory
Formal, complete solution Rays easily drawn
Difficult to interpret Easy to visualize sound distribution
Real boundary conditions difficult to handle Real boundary conditions easy to insert
Valid at all frequencies; in practice useful at Valid only at high frequencies (> 200 Hz?)
low fiequencies where ray theory fails
Computer program essential Computer program normally used, but rays
can be drawn manually using Snells law
Ray Theory and the Hodgson Model 59
3.15 Ray Theory and the Hodgson Model

Ray theory will be discussed in terms of the Hodgson model. This model is used
operationally by the Royal Navy and is applicable at frequencies above, say,
200 Hz. Sonar parameters that may be specified are frequency, depth, and beam
minimum and maximum (vertical) angles.
The sea surface is modelled as a reflector at which rays undergo an attenuation
per bounce dependent on frequency and surface roughness (sea state, wind speed
or wave height may be input). Bottom losses are calculated using a set of bottom
loss curves; the loss per bounce depends on fiequency and grazing angle.
Absorption losses are accurately modelled as a function of frequency, temperature,
depth and salinity and are continuously changed with range for each ray.
The sound speed field is calculated by interpolating the SSPs in depth and range
then the ray paths are calculated using Snells law, after splitting the environment
into a series of boxes within which the sound speed varies linearly with depth
(Figure 3.12). This is how a range-dependent solution is derived.
Figure 3.12 Refraction of sound
Snell j . law
Snells law is fundamental to ray theory models and describes the refraction of
sound rays in a medium where the speed of sound is changing. When the speed of
sound varies continuously with depth, the medium can be considered as a number
of thin layers of constant but different sound speeds within them. Snells law is
applied to the boundaries of the layers, and the sound ray is seen to be curved:
If speed continues increasing with depth, the ray will become horizontal and
c o s e = 1. The speed of sound where the ray is horizontal, cv, is known as the
vertex speed, and
This equation makes it possible to generate the path of a ray through the layers
into which the velocity profile has been divided.
Hodgson Example 61
3.16 Hodgson Example

Propagation losses in a central Mediterranean location are shown for February
(Figure 3.13) and August (Figure 3.14). The frequency is 5 kHz and the sonar
vertical beamwidth is 0-15". The source depth is 5 m.
February
The ray trace shows shadow zones between about 8 and 25 km and this is evident
in the PL curves for the two receiver depths of 50 and 200 m. Note that the simple
law of spherical spreading plus absorption follows the model predictions quite
accurately, particularly if the shadow zones are excluded. The reduced losses
centred on about 30 km are due to a CZ gain.
August
Severe downward refraction occurs near the surface, resulting in a shadow zone
from about 2 km until the CZ returns, centred on about 40 km.Note the first BB
returns and the second CZ returns (at greater than 110 dB each way, the losses will
be too great for an active system but passive will sometimes be possible).
Figures 3.13, 3.14 and 3.15 have been produced using the Wader-Hodgson
system. Intellectual property rights and copyright to the Hodgson Acoustic
Propagation Loss model are owned by the United Kingdom Secretary of State for
Defence. Intellectual property rights and copyright to the Wader Global Ocean
Information System are owned by Ocean Acoustic Developments.
Figure 3.13 Propagation losses in a central Mediterranean location: February

Hodgson Example 63
Figure 3.14 Propagation losses in the location of Figure 3.13: August

3.17 Performance Prediction

Propagation loss is one of the many parameters required to predict the detection
performance of a sonar system. Parameters such as source level (SL) and direc-
tivity index (DI) are fairly accurately known, to within say 2 dB, but propagation
loss (PL) is highly variable, and no matter how accurate the model and no matter
how well it represents the environment, any results can only be as good as the
input data.
Under operational conditions, it is only possible to measure the SSP at a limited
number of points, and usually only one (at the sonar platform itself). This SSP will
not apply precisely to all water columns between sonar and target.
PL graphs are highly dependent on the SSPs, which themselves rely primarily
on measurements of temperature against depth. The PL graphs in Figure 3.15
illustrate the differences for quite small inaccuracies in temperature measure-
ments, inaccuracies typical of bathythermographs
All the graphs are for an area in the North Arabian Sea during February. The
standard SSP for February yields Figure 3.15(a). When this SSP is modified to
represent possible bathythermograph errors, we obtain Figures 3.15(b) and (c).
Figure 3.15(b) is obtained by subtracting 1 m/s fiom all speed values down to
75 m depth and adding 1 m/s to all other values. Figure 3.15(c) is obtained by
adding 1 m/s to all speed values down to 75 m depth and subtracting 1 m/s fiom
all other values.
All three graphs are quite similar but there are significant differences in detail
which should illustrate the dangers of relying implicitly on the results fiom any
propagation model. Suppose a propagation loss of 100 dB is the allowable loss for
a given sonar system (the loss for which detection is just possible) then the
corresponding range limits differ quite markedly.
PeTformance Prediction 65
PL (dB)
60
1 00
140
20
Range (km)
PL (dB)
60
100
140
1 I
10 20
(b) Range (km)
PL (dB)
60
100
140
10 20
(C) Range (kin)
Figure 3.15 Propagation loss graphs for the North Arabian Sea: (a) standard SSP, (b) modified
SSP, (c) modified SSP
3.18 Multipath Propagation

The propagation of underwater sound always follows multiple paths in the vertical
plane. This is an inevitable consequence of the vertical beamwidth being finite,
and it causes differences between arrival times and propagation losses along the
possible paths. These differences increase as the distance from source to receiver
increases.
A quantitative knowledge of these differences is important. In active sonar
systems, large differences can produce multiple echoes separated in time (range);
small differences produce elongation of the target. In passive sonar systems, the
differences between times of arrival are used for estimating range. Multipath
propagation is considered further in Chapters 8 and 9.
3.19 Problem
3.1 What are the detection ranges achieved for a one-way propagation loss of 80 dB
assuming (i) spherical spreading plus absorption, (ii) cylindrical spreading plus
absorption? Give solutions for 5 kHz (a= 0.3 dB/km) and 20 kHz (a = 3 dB/km).
4
Target Strength
4.1 Definition
Target strength (TS) refers to the echo returned by an underwater target:
submarines, surface ships, torpedoes, mines, fish. It is defined as the log of the
ratio, in dB, of the reflected intensity referred to 1 m fiom the acoustic centre of
the target, to the incident intensity:
ITS = lolog($) 1
The use of an arbitrary reference distance of 1 m gives many underwater targets
positive values of TS. This does not imply that more sound is reflected from the
target than is incident on it. The acoustic energy appears to come from a
hypothetical point source within the scattering surface of the target, and for a large
target such as a submarine, this point may be some metres from the surface and
within the target.
4.2 Formulae
The correct value of TS to be used in the sonar equations should be chosen
carefully. In practice, TS is calculated using either the peak pressures of the
incident and reflected pulses or their total integrated energies. Here are the
resultant parameters:
68 Target Strength
PeakTS
Peak TS = 20 log( pr/ pi)
Integrated TS
where pi and pr are the peak pressures of the incident and reflected pulses; pi(t)
and pr(t) are the time hnctions of the pulses; Tp is the duration of the incident
pulse and T, is the time extent of the target.
4.3 Measurement
It is simplest to measure the peak pressures of the incident and reflected pulses.
Therefore peak TS is usually determined regardless of the method of measurement
and is the parameter normally used in the active sonar equations.
TS is defined at 1 m from the acoustic centre of a target. It is clearly impractical
and often impossible to make measurements at this distance. Therefore measure-
ments are made at greater distances and reduced to the definition range.
A method well suited to the measurement of the TS of small objects such as
mines is to compare the echo levels from the target with the level fiom a reference
target such as a sphere. For larger targets such as submarines and torpedoes, which
because of their size must be measured at longer range, say 1000 m, the nec-
essarily large size of the reference target (a sphere of TS = 0 dB has a diameter of
4m), the reference target may be replaced by a calibrated transponder or an
alternative method may be used.
Most TS determinations, particularly for these larger targets, have been made
by measuring the peak pressure of the reflected pulse at long range and then
reducing it to its 1 m value. The appropriate sonar equation is
EL = SL - 2PL + TS
The echo level (EL) and the source level (SL) are measured, the propagation loss
(PL) is either calculated, assuming spherical spreading plus adsorption, or meas-
ured using a calibrated hydrophone at the target, and the equation is solved for TS.
TS of'a Sphere 69
4.4 Dependence on Pulse Type and Duration

Integrated target strength (ITS) does not change with duration or type of pulse, but
fluctuations will occur from ping to ping, given the effect on TS of even small
aspect changes. These fluctuations will be averaged out so that data from the same
target should yield very similar ITS values irrespective of pulse type and duration.
Peak target strength (PTS), on the other hand, is very dependent on pulse type
and duration. It increases with pulse duration, reaching a constant value when the
pulse duration is comparable with the time extent of the target - in practice at
least 100 ms for a submarine and as short as 1 ms for a mine. For such long pulses
(long relative to the target dimensions) PTS is approximately equal to ITS.
Modern active sonars frequently use long CW (continuous wave) and long FM
(frequency modulated) pulses. It is important therefore to know which TS value to
use for these pulses:
Long CW: the small bandwidth and thus low resolution of this pulse type
ensures the PTS is approximately equal to the ITS.
Long FM: the large bandwidth and thus high resolution of this pulse type means
that the effective pulse length is short, of the order of a few milliseconds for an
antisubmarine pulse, and therefore the correct TS value will be given by the
PTS as measured by an equivalent short pulse.
4.5 TS of a Sphere
The meaning of target strength can be demonstrated by calculating the TS of a
sphere. Let a large sphere - large compared with a wavelength - be insonified by
a plane wave of intensity I,. If the sphere has radius a metres, the energy
intercepted by it from the incident sound is xa21j; where za2 is the cross-section
of the sphere, the scattering cross-section.
Assuming the sphere reflects this energy isotropically, the intensity of the
reflected wave at a distance Y metres from the centre of the sphere will be
I, = nu2II/4xr2.At the reference distance of 1 m this reduces to Ir/Ij = a2/4 and
the target strength of the sphere is given by
ITS = 1 0 1 0 g ( ~ )
Therefore a sphere of radius 2 m has TS = 0 dB.

70 Target Strength
4.6 TS of Some Simple Shapes

Formulae for target strengths of other simple forms have been derived (Table 4.1)
and these may be used to estimate target strengths of practical targets. The
minimum dimension of the shape should be large compared to a wavelength (say
at least 5A), although a useful indication of target strength may still be obtained
even when the minimum dimension is only 2R
Table 4.1 Target strengths of some simple shapes

Shape TS, dl3 Incidence Notes
Sphere h Y a is radius
Convex Normal to a1 and a2 are
surface surface principal radii
Plate of 10 log(A/A)' Normal A is area
any shape
Rectangular I0 log(ab/A)2 Normal aandbare
plate +
10 log(ab/A)2 20 log(x-' sin x) + 20 log(cos 6) 6 to normal sides a 3 b
where x = (2na/A)sin 6
Circular 10 log(na2/A)2 Normal a is radius
plate
Cylinder 10 log(aL2/22)2 Normal u is radius L
10 10g(aL~/22)~ + 20 log(x-' sin x) + 20 log(cos 6) 6 to normal is length
where x = (2nL/A)sin 6
Note that the scattering cross-section is known in radar as the radar cmss-
section (RCS or a). Expressions for the RCS of simple shapes are prevalent in the
radar literature. If helpful, these expressions can be translated into sonar TS
by dividing them by 4n.The target strengths of certain shapes at an angle 8 to
the normal may be determined with the aid of Figure 4.1, which plots
20 log(x-' sin x) dB against (L/A)sin 8.
Example 4.1
A cylinder has length L = 5 m, radius a = 1 m and A = 0.2 m. What is the TS at an
angle of 2" to the normal?
+
TS = 10 10g(aL~/2A)~ 20 log(x-' sin x ) + 20 log(cos 6 )
(L/A)sin 8 = 0.87
and from the plot we obtain
201og(x-' sinx) = -16 dB

TS of Some Simple Shapes 71
(Llh) sin 0
Figure 4.1 Plot of 20 log(x-' sin x) against (L/A)sin 8
+
Therefore TS = 18 - 16 0 = 2 dB. At the first null, (L//l)sin 8 = 0.5, therefore
8 = 1.15" and TS is theoretically -moo.
In practice, perfect nulls do not exist; the cylinder will not be perfectly uniform and
the incident pulse will have a finite bandwidth, i.e., the wavelength will vary during
the pulse. Broadband pulses (FM or short CW) will have significant percentage
bandwidths, perhaps 5 per cent and often greater. Therefore, particularly for broad-
band pulses, the nulls will be blurred and the peaks broadened. Narrowband pulses
(long CW) will have less effect. The dashed curve is a realistic plot avoiding the
extreme nulls and peaks. If it is used for this example, we obtain the following results:
F o r $ = 1.15" T S = 18- 1 8 + 0 = 0 d B
For 8 = 2" TS=18-24+0=-6dB
Example 4.2
A rectangle has sides a = 5 m, b = 3 m and I. = 1 m. What is the TS at an angle of 7"
to the normal?
TS = 10 log(ab/A)2 + 20 log(x-' sin x ) + 20 log(cos 6 )

72 Target Strength
Now the largest dimension is a, replacing L, so we use (a/ll)sin 8 = 0.61 and, fiom
the dashed plot, 20 log(x-' sin x ) = - 19 dB. Therefore
TS = 18 - 19 + 0 = -3 dB
It would be unsafe to attach too much importance to the accuracy of these theoretical
TS values; real targets are seldom regular, and their aspects are not precisely known.
f i e important point is how rapidly TS falls for very small departures from normal
incidence.
4.7 TS of Small Targets

A small target is simply a target whose minimum dimension is much less than a
wavelength; it is not necessarily small in absolute terms. Target strength is then
not only low but also strongly dependent on the wavelength; it is proportional to
A-4. Fortunately for sonar, as we shall see later on, other requirements ensure the
wavelength is usually less than the minimum dimension of the target so that this
region is avoided.
Resonance is possible when a target has one or more dimensions close to A; TS
can then vary wildly with A and may be greater or smaller than the values indicated
above. This resonance effect could be exploited by choice of frequency to improve
the detection of sonar targets at some possible cost to classification.
Suppose a submarine to have a minimum dimension of 8 m. To exploit a
resonance for this dimension, f = c/A = 188 Hz. To use other resonances, even
lower frequencies may be needed and sufficient bandwidth for classification will
not be possible. There are formidable obstacles to this approach: frequency
diversity would be needed to achieve resonances, and at the low frequencies
necessary for submarine resonances, both the transmit and receive arrays would
have to be very large.
4.8 Mine Target Strength

Mines may often be approximated by a sphere or a cylinder with hemispherical
ends. The TS of a finite cylinder is given by
71
Torpedo Target Strength 13
where
a = radius
L = length
2xL
x = -sin 8
/i
On the beam (normal to the axis) this reduces to
Suppose the mine is represented by a cylinder of length 2 m with hemispherical

ends of radius 0.15 m and with ;1= 0.15 (10 kHz).
TS on the beam = 10 log((O.15 X 4)/(2 X 0.15)) = 3 dB

0 TS on the ends = 1010g(0.15~/4)= -22.5 dB
As we have seen, the TS falls rapidly away from its value on the beam, but because
of the hemispherical ends it does not fall below -22.5 dB.
4.9 Torpedo Target Strength

A torpedo is basically cylindncal with a flat or rounded nose, and the same
formulae may be used to estimate its TS. Suppose the torpedo has length L = 5 m
and radius a = 0.5 m, and the frequency is 10 kHz (A = 0.15):
0 TS on the beam = 10 log((0.26 X 25)/(2 X 0.15)} = 13 dB

TS on the (rounded) nose = 10 10g(0.26~/4)= - 18 dB
Again, the TS falls rapidly away from normal aspect (on the beam), but because of
the nose, it will not fall below - 18 dB, except perhaps astern.
Some torpedo designs have flat noses to reduce flow noise and therefore
improve homing sonar performance. TS on the (flat, circular) nose at normal
incidence is
lOlog(x X 0.262/0.15) = 1.5 dB

74 Taeet Strength
Once again, this high TS value will fall off rapidly away from normal and a f i t
nose torpedo will be virtually undetectable by an active sonar at any aspects other
than close to beam or bow.
4.10 Submarine Echoes

Echoes from submarines include specular reflections from the outside profile of
the submarine; reflections or scattering from structures on or behind the casing or
pressure hull. The echoes are caused by a reflection from a surface normal to the
incident wave and are called normal incidence specular reflections. All other
reflections from the outside profile are at non-normal angles and directed away
from the receiver.
At sonar search frequencies the water-backed casings and all free-flood areas,
including the fin, are virtually transparent and therefore reflections from these
outer surfaces are at low intensities. The pressure hull, being air-backed, provides
good reflections and a significant echo.
At weapon frequencies (> 20 kHz) both casings and pressure hull are good
reflectors and echoes can be expected from the external casings, pressure hull, fin,
rudder, hydroplanes, stabilizers and propeller. Shadowing, due to target aspect,
will affect the number of highlights.
4.11 Beam Aspect Target Strength

For submarines (and mines and torpedoes) specular reflection appears to be the
dominant mechanism at beam aspect and the incident pulse is faithfully repro-
duced by the echo. If we take the dimensions of the uniform section of the
cylindrical pressure hull of a submarine to be 50 m and its diameter 8 m, then
using
TS = 10 log
(3
- (at normal incidence)
at 5 kHz we have
)::;r
TS = lolog =42dB
The 50 m aperture, equivalent to about 1501 at 5 kHz,ensures that this broadside

Sub mar ine Target Strengths 75
glint only occurs at this intensity over a fraction of a degree. But since the hull
tends to taper and is seldom exactly cylindncal, this large value is reduced and
exists over greater angles (compare with the narrow beamwidths achieved by long
towed arrays of large apertures). The fact that the TS varies rapidly with aspect
helps to explain the extreme variability of practical measurements.
4.12 Bow Aspect Target Strength

If the bow were hemispherical in form and 8 m in diameter, then the TS would be
6 dB over the entire bow sector. The submarine designer can easily reduce this
wide coverage in both azimuth and elevation by avoiding a hemispherical shape
and angling surfaces so that any glints occur at harmless elevations.
4.13 Submarine Target Strengths

Submarine target strengths are clearly directly related to the size and construction
of the class of submarine. Typical submarines are listed in Table 4.2 and have been
chosen based on their actual or potential exports or use by non-NATO navies. Six
R2 Mala, for example, have been delivered to Libya; they carry 250 kg of limpet
mines.
Table 4.2 Some typical submarines

Name Mass (tomes) Dimensions (m) Country
LARGE SUBMARINES
KILO 3000 74 X 10 X 6.6 Russia
209 1300 56 X 6.2 X 5.5 Germany
RUBIS (nuclear) 2700 74 X 7.6 X 6.4 France
SMALL SUBMARINES
Mini-sub 150 27 Italy, Maritalia
Mini-sub 90 20 X 2 X 1.6 North Korea
3GST9 30 10 Italy, Maritalia
R2 Mala 1.4 4.9 X 1.4 X 1.3 Former Yugoslavia
76 Target Strength
A submarine may be double-hulled, a complete external casing over an inner

pressure hull, or single-hulled with much of the pressure hull directly exposed to
the incident sound. The double-hulled construction may have main support
ribs that are internal or external to the pressure hull. If they are external, they will
add many potential highlights distributed along the length of the hull. For the
single-hulled vessel there may be an upper casing and support ribs with winches
housed beneath. The vessel may have a keel. All submarines have a bridge fin or
sail (except for the very smallest submarines), and because it is proportionately
larger, it is a more important echo source in smaller submarines. It is possible to
examine the details of each vessel and by theoretical analysis estimate the TS of a
particular design. The scattering mechanisms are diverse and often highly
dependent on frequency and aspect, therefore the sonar designer is unlikely to be
helped by such analysis.
Practical measurements of TS result in similar dependencies together with
measurement uncertainties. Many TS measurements are reported in the literature
and it is often claimed that they lend support to a standard butterfly pattern as
shown in Figure 4.2.
Figure 4.2 Aspect dependency of submarine target strength: butterflypattern

Towed Arrays 77
4.14 Towed Arrays

Towed arrays and cables are long, thin cylinders and echoes have been reported
fi-om them at sonar search frequencies. Maximum target strengths can theoretically
be as high as 10 dB, but again only very close to broadside and neglecting any
array curvature. Thin arrays will have lower maximum TS and this - combined
with submarine tactics which generally attempt to avoid presenting themselves
(and therefore their towed arrays) at beam aspect - implies that any contribution
fi-om a towed array either for detection or classification is unlikely to be usefirl
operationally.
78 Target Strength
4.15 Target Strength Reduction

Careful attention to the shape and orientation of the external surfaces of a
submarine will reduce the target strength and ensure that glints only occur at
harmless angles, particularly in elevation. But note that homing torpedoes may
approach at angles far from the horizontal, therefore harmless angles for search
sonars may not be harmless for weapon sonars.
For areas of a submarines external surface where only scattering returns are
likely, useful reductions in target strength are obtained by providing a coating with
a high transmission loss. Conventional decoupling coatings can be effective
down to about 3 kHz.
Sound absorption is needed to reduce specular reflections at normal incidence.
This can be provided by a material thin in terms of wavelength with limited
bandwidth, or a material thick in terms of wavelength but relatively broadband.
Figure 4.3 shows the theoretical performance of a thick-layer anechoic material.
Echo reduction at normal incidence (dB)
Thickness (U4)
Figure 4.3 Echo reduction: thick layer anechoic material
To obtain a useful reduction in echo strength, the material needs to be at least

1/2 thick. At weapon frequencies, say 30 kHz, ;1/2 = 50 mm, which is practical
for large submarines. At 3 kHz, however, 1/2 = 500 mm and this is clearly
impractical.
A submarine made less vulnerable to weapons by a 50 mm coating - perhaps
achieving a useful reduction of between 10 and 15 dB at many aspects - might
Target Strength Reduction 79
only achieve a reduction of about 3 dB at typical sonar search frequencies

(3- 10 kHz) and no reduction at all at frequencies less than 3 kHz.
Thin tuned coatings
A prototype thin tuned coating was used in the Alberich design employed by the
Germans during World War 11. It consists of an inner layer of rubber perforated
with a pattern of holes and covered by an outer layer of similar thickness (2 mm).
Resonances occur and the attenuation is high over limited bands. At search sonar
frequencies the resonant approach is essential, and in order to provide adequate
transmission loss, the sound absorber may be combined with a decoupling base
layer. The resonances depend on depth and temperature (Figure 4.4).
140 m, 13C
70 m. 21C
70 m, 13C
1 I I I I
20 40 60 80 kHz
Figure 4.4 Echo reduction: typical Alberich design

Target Strength
4.16 Practical Values

The following TS values are suggested for use in the design of sonar systems and
for performance calculations.
Submarines
Aspect Small Large: clad Large

Beam 5 10 25
Intermediate 3 8 15
Bow/stern 0 5 10
Cladding of large submarines reduces TS significantly, particularly for beam and

intermediate aspects. Small submarines are more difficult to clad successfully,
particularly against search sonar fiequencies; this is because the thicknesses of
coatings similar to those used on large submarines give weight and drag problems.
Others
Target Aspect TS (a)

Surface ship Beam +25
Off-beam +15
Mines Beam 0
Off-beam -10 to -25
Torpedoes Random - 15
Towed array Beam 0 (max)
Whale, 30 m Dorsal +5
Shark, 10 m Dorsal -4
Iceberg h Y + 10 (min)
Problems 81
4.17 Problems
4.1 What is the target strength of a sphere of radius 0.5 m? Above what frequency is
the result likely to be reliable?
4.2 A cylindrical mine has a length 2 m and radius 0.5 m. The ends are hemispheres.
What are the target strengths normal to the mine and end on? Calculate TS for
frequencies of 10 kHz and 100 kHz.
4.3 What is the target strength of an irregular plate of area 4 m2 and largest dimension
2 m at an angle of 5" to the normal and frequency 4 kHz? Use the practical dotted
curve of Figure 4.1 and take L to be 2 m, the largest dimension.
5
Noise in Sonar Systems
5.1 Sources of Noise

Noise is the background against which sonars, active and passive, must detect
signals fiom targets. For an active sonar, noise is augmented by reverberations
from unwanted sources (Chapter 6) and the signal is an echo from the target. For a
passive sonar, the signal is also noise - the radiated noise of the target.
This chapter will first discuss the sources of noise in sonar systems and then
consider the total noise as a background to detection - the self-noise of a sonar
system. There are three sources of noise to be considered:
Thermal noise
0 Noise from the sea (ambient noise)
0 Vessel noise
84 Noise in Sonar Systems
5.2 Thermal Noise

In common with any electrical receiving system, a sonar receiver adds its own
noise to the signals it receives. The designer must ensure that the noise introduced
from this source is negligible compared with the noise coming from the sea itself.
All the results established for the classical case of radio are valid for sonar when
aerial or antenna is replaced by hydrophone.
Any resistance, R, is the source of a thermal noise EMF (electromotive force)
resulting from the thermal agitation of its electrons. The value of this EMF, en
volts, is given by
where
k = Boltzmanns constant = 1.38 X J/K

T = absolute temperature (K) of the resistance
Sf= bandwidth (Hz)
A practical formula, valid at sea temperatures, is
where en is in microvolts, R is in kilohms and Sf is in kilohertz.

A passive circuit produces a noise proportional to the resistive component of its
equivalent impedance, even when this corresponds to a physical reality far
removed from a simple resistance. For an aerial, a component of its impedance
corresponds to the coupling with the surrounding space and to exchanges of
energy with this space. This component is known as the radiation resistance of the
aerial.
Similarly, for an underwater transducer, energy exchanges take place with the
sea through the motional resistance, R,, of the transducer, generating a noise
EMF given by the above formula where R is replaced by R,. Here the noise EMF
does not come from a thermal agitation of electrons, but from the thermal
agitation of the molecules of water producing pressure fluctuations at the face of
the hydrophone.
Thermal Noise 85
This thermal noise is an absolute minimum and would only be observed in the
absence of any other noise source, i.e.,
If the water had no other agitation except thermal agitation and if it were
completely isolated from any source of sound (a dead sea)
If the receivers were perfect and added no noise (the noise factor of the receiver
is NF, = 0 dB).
The thermal noise is given by
where Nthemalis in dB re 1 @a, and f i s in kilohertz.

5.3 Noise from the Sea

The thermal noise of the sea can only be the dominant background to a sonar at
high frequencies - at least 30 kHz and for all practical purposes at least 100 kHz,
where it equals the ambient noise expected from sea state 2 (SS2).
When the sea is not dead (Le., perfectly isolated from all sources of sound and
only subject to thermal agitation), even though it may appear perfectly calm, it is
subject to an agitation much greater than the thermal noise, particularly at the
lower frequencies (less than 30 kHz). Figure 5.1 plots the mean isotropic spectrum
levels as a function of frequency for various sea states. Departures from the
average of up to A10 dB are quite common, particularly in shallow water and
close to the surface. The ambient noise falls at about 5 dB per octave for
frequencies above 500 Hz.
6.2 0.5 I 2 5 10 20 50 100

Frequency (kHz)
Figure 5.1 Ambient noise of the sea
Note the thermal noise plotted in Figure 5.1 where it forms a lower bound to the
family of sea state noise curves. Sea state, wave height and wind speed are given
in Table 5.1.
Table 5.1 Sea state, wave and wind speed
Sea state 0 1 2 3 4 5 6
Wind speed (knots) Sl 5 13 16 19 22 28
Wave height (m),crest to 0 0.05 0.4 0.7 1.3 2 3
trough
Noise.from the Sea 87
80
- - - NoRainrain
-
4
i
I
Y
Rainfall rate (d)
k! 60
m
2
L_ 30
'u
'u
_I
E
2
4 4
0
3 40
\
.
0.1 I
'
1
DSSl
1 10
Frequency (kHz)
Figure 5.2 Rain noise
Rain can significantly increase the ambient noise. The spectra are quite flat with
frequency, and Figure 5.2 plots theoretical values which agree quite well with
measurements.
Shipping noise is important, particularly at frequencies below about 500 Hz.
In harbours and harbour approaches, the dominant noise is from shipping -
particularly small craft whose noise extends to frequencies of several kilohertz -
and from industrial activity along the shore.
Biological noise is produced by a variety of marine life. Besides the individual
sounds of sea mammals, snapping shrimps are a notable noise source. They
produce quite a flat spectrum of noise between 500 Hz and 20 kHz, and it can be
as high as 70 dB re 1 p a .
Figure 5.3 plots spectrum levels for these noise sources (except biological).
Some broad practical conclusions can be drawn:
0.1 1
Figure 5.3 Ambient sea noise
Heavy rain can dominate at all frequencies, but rainfall less than about 3 mm
per hour is not significant.
Noise resulting from the agitation of the sea due to wind and waves is dominant
above 500 Hz.It is difficult to relate the actual noise level to the prevailing wind
speed or sea state; the wind speed may be measured but the sea state is often
based on a subjective judgement of wave height, and although wind speed and
sea state are clearly related, rarely do their maximum or minimum values
coincide in time.
Shipping noise is unimportant above 1 kHz (radiated noise itself falls rapidly
with frequency and also falls with range due to absorption losses). The noise
below 1 kHz is likely to be dominated by discrete tones, but if there are many
ships then the noise may appear to have a continuous spectrum.
89
5.4 Noise from a Vessel

There are three principal sources of noise from a vessel (which is taken to mean a
surface ship, a submarine or a torpedo):
Propulsion machinery and auxiliary machinery

* The propellers
Flow noise
Figure 5.4 shows the relative contributions of these sources of self-noise at

operational bearings for a typical surface ship at 10 kHz, particularly one where
little has been done to reduce its radiated noise. Modern frigates can improve
significantly on these figures, perhaps achieving SS2 at speeds up to 10 knots. The
self-noise of a sonar is the sum of all noise sources, ambient and vessel, that are
present at the array/water interface.
10 20 30
Speed (knots)
Figure 5.4 Variation of self-noise with speed

5.5 The Sonar Environment

Sonar arrays - mounted within domes or behind acoustic windows - find
themselves in potentially quite noisy environments which can severely limit the
performance of the sonar. These noise sources have been described above and fall
within three classes:
The internal noise of the ensemble of hydrophone and receiver

The ambient noise of the sea
Own vessel noise
The designer will ensure that the internal noise of the sonar is negligible, nothing
can be done about the ambient noise of the sea, and therefore the design of the
total system - vessel, dome and array - must attempt to minimize the, usually
dominant, own vessel noise. A practical aim is to attempt to match the ambient
noise of the sea corresponding to, say, SS2 for surface ships up to operating speeds
of 15 knots and somewhat better for submarines.
5.6 Self-noise
The self-noise of a sonar is perceived as a noise voltage appearing at the output of
the receiver. It is more convenient, however, for both design and performance
analysis, to define it at the entry to the system, i.e., at the array/water interface.
The noise intensity is given in dB relative to the omnidirectional intensity of a
plane wave having a pressure equal to 1 @a in a 1 Hz band, the isotropic spectrum
level. If we assume that the vessel is in a calm sea and there are no parasitic noise
sources, the self-noise will be solely due to the combination of vessel and sonar.
We now discuss its components and indicate the principal methods for reducing
them.
5.7 Electrical Noise

The design of the sonar will normally ensure that the electrical noise introduced
by its receiver is negligible. Changes to the array andor the frequency band of the
sonar can cause problems - particularly if the sensitivity of the array is lower -
and the noise factor of the receiver must be checked and improved if necessary.
Flow Noise 91
Electrical interference on the cables between array and receivers is frequently a

problem. Effective electromagnetic screening of the cables at low frequencies is
very difficult. Fortunately, this problem can be eliminated or alleviated by modern
techniques where the analogue voltages from the hydrophones are converted into
digital form close to the array.
5.8 Machinery Noise

Machines closest to the array are the most troublesome; these may be propulsion
machines or auxiliary machines. There are several ways to reduce their contribu-
tion to self-noise:
0 Design the machines with quietness as a major parameter.
0 Isolate machines, shafts and piping from the hull using anti-vibration mounts.
Isolate the array and dome from the hull; provide damping between the array/
dome and the hull; dampen the hull itself.
Separate arrays and machines; site the arrays well forward of the main
propulsion machinery; avoid placing any auxiliary machinery close to arrays.
5.9 Flow Noise

The principal cause of flow noise is flow over the sonar dome and over the hull
close to the dome. Here are four ways to reduce it:
Good hydrodynamic design of the dome

Smooth interfaces between dome and hull
Reduced surface roughness of dome and adjacent hull
Coating the dome and the hull: the hull coating should continue significantly
beyond the dome
5.10 Propeller Noise

Propeller noise is produced by a purely hydrodynamic mechanism such as
cavitation at the tips of the blades or cavitation on the blades themselves, or by
mechanical vibration of the blades. Here are two ways to reduce its contribution to
self-noise and radiated noise:
Agouti: a device which emits air bubbles in the vicinity of the blades to replace
the water vapour bubbles created by cavitation. An improvement of the order of
10 dB is possible at high ship speeds (above 20 knots).
Bafles: these are mounted within the dome to shield the array from propeller
noise. They inevitably reduce the performance of the sonar over the stem arc
protected by the baffle; a useful rule of thumb is to assume the allowable
propagation loss is reduced by 10 dB over the stern arc.
5.11 Variation with Speed

Refer to Figure 5.4. At slow speeds (up to about 10 knots) machinery noise is
dominant unless the vessel is particularly quiet or in high sea states, when the
ambient noise of the sea will take precedence. Flow noise is dominant at medium
speeds (10-20 knots), above which the propeller noise begins to dominate the
self-noise even though flow noise continues to increase with speed. Cavitation at
the dome can also be important at high speeds, particularly if the dome is poorly
finished or has been damaged.
5.12 Variation with Frequency

The isotropic spectrum level of self-noise is observed to fall with frequency at
about 6 dB per octave (proportional to l/f2). At frequencies less than about
500 Hz this simple relationship should be used with care. The noise spectrum at
low frequencies is unlikely to be continuous and is influenced by marine traffic
and discrete fiequency components.
Addition of'hroise Levels 93
5.13 Directivity
Self-noise is markedly directive. There is clearly a maximum over the stem arc,
even with an intervening baffle. Measurements are frequently made in octave
bands and averaged over operational bearings (the forward 270), thus avoiding
any bias which would result from including measurements astern.
5.14 Self-noise and Radiated Noise

We need to be clear on the distinction between self-noise and radiated noise:
Self-noise: the total noise measured by a sonar system which forms the back-
ground to detection of a wanted signal.
0 Radiated noise: the noise radiated by a vessel and measured at some distance
from the vessel, typically between 100 and 1000 m. It is the source of signals
for passive sonars (Chapter 8).
Self-noise and radiated noise have many common sources and are frequently
simply two different aspects of the same phenomenon. For example, propeller
noise may dominate radiated noise at all speeds but only be an important
contributor to self-noise at higher speeds; flow noise may dominate the self-noise
when its source is close to the array but be an insignificant contributor to the
radiated noise.
5.15 Addition of Noise Levels

The total noise level from two sources of noise is obtained by their incoherent
addition on a power basis as given by the equation
where PT is the total noise power and PI and P2 are the individual noise sources,
all in dB re 1 pPa.
When P1 = Pz then PT = P2 + 3 dB
When P1 = P2 + 6 dB then PT = P1 + 1 dB
Therefore if one noise level exceeds the other by at least 6 dB, the other makes a
negligible contribution to PT. The equation is plotted in Figure 5.5.
Figure 5.5 Increase in total noise level due to a weaker noise
5.16 Receiver Noise Factor

The receiver noise factor, NF,, is defined as
where Noutis the noise power at the output of the receiver and N m i n is the un-
avoidable output noise, i.e., the noise at the input of the receiver due to the thermal
noise of the generator connected to the receiver input multiplied by the receiver
gain.
Noise Factor of a Sonar 95
5.17 Noise Factor of a Sonar

Consider first the hydrophone placed in dead water. The receiver is fed from the
equivalent resistance of the hydrophone (Figure 5.6). The equivalent resistance of
the hydrophone has two components, the motional resistance, R m , defined in
Section 5.2, and a resistance Rp, representing the losses in the hydrophone.
Receiver
Figure 5.6 Hydrophone and receiver
The efficiency, rj~,of the hydrophone is given by
+
where R = R m I$ is the total hydrophone resistance.
The noise factor of the receiver alone is NFr. The noise factor of the receiver
plus hydrophone, NFrh, is greater. This is because, in defining NF,, the thermal
noise due to R is unavoidable, whereas for the ensemble (receiver plus hydro-
phone) the only unavoidable noise is due to R , (which is smaller than R and
would apply to a perfect hydrophone where Rp = 0).
The noise power, Nmin, is therefore reduced by r j ~= R,/R, hence
Now consider the case where the sea state is not negligible. The noise factor
becomes NF, (at a given frequency), and
NF, =
isotropic spectral noise intensity - N,
-
+ Nmin
thermal noise (min) Nmin
and the resultant noise factor for the sonar, NF,,, for a given sea state, will be
and since
Nmin = unavoidable noise power

Nrh = noise power due to receiver plus hydrophone
N, = noise power at a given sea state
we obtain
NFres =
Nmin + Nrtr + Ns
Nmin
Using the expressions for NF, and NFrh obtained above, we can deduce the
following expressions for the overall noise factor for the sonar in a given sea state,
NFres
-4cceptable Receiver Noise Level 97
5.18 Acceptable Receiver Noise Level

The sonar designer is faced with the problem of determining an acceptable noise
factor for the sonar receiver. If the receiver plus hydrophone were perfect, we
would have NFrh = I(0 dB) and therefore
which is the minimum value of NF,,, for a given sea state. We lose, therefore, due
to the noise introduced by the sonar itself, in the ratio
Example 5.1
Consider a sonar operating at 5 kHz on a fi-igate whose self-noise is unlikely to fall
below the ambient noise, which is 50 dB re 1 @a for SS2.
We have
10 log NF, = 50 dB
since N,>> Nmin. Therefore
Suppose that we only lose 1 dB due to the internal noise of the sonar, then
NFres NFr = 0.26 X 100 000 = 44 dB

= 1.26 and -
NFS ?l
If the efficiency of the hydrophone is 40 per cent (-4 dB) then the noise factor of the
receiver itself will need to be less than 39 dB, which is easy to achieve.
The hydrophone sensitivity is given by Sh = 20logv - 201og p, and if S h =
-200 dB/V then
20 log v = -200 + 20 log p (dB)

= -200 + 40 = -160 dB/V
v = 10nV
i.e., the spectrum level of the receiver noise, referred to the input, must be less than
10 nV.
Example 5.2
Now consider a sonar, again operating at 5 kHz, installed on a submarine whose self-
noise, in its quietest state, can be less than the ambient noise of the sea, and take the
ambient noise in sea state 0 as 35 dB re 1 p a .
We have
lOlogNF, = 35 dB
since Ns >> Nmin. Therefore
Again suppose that we only lose 1 dB due to the internal noise of the sonar, then
NFr
5 = 1.26 and -= 0.26 X 3162 = 29 dB
N F S 7
and for the same hydrophone, the noise factor of the receiver has to be less than
25 dB:
20 log v = -200 + 20 log p (dB)

= -200 + 25 = -175 dB/V
v=2nV
2 nV.
Alternative Calculation 99
5.19 Alternative Calculation

Section 5.18 gave a somewhat complex procewre for calculating the required
spectrum noise level at the input to a sonar receiver. Here is a simpler approach
that produces the same results.
Example 5.1 by simpler approach

Consider a sonar operating at 5 kHz on a frigate whose self-noise is unlikely to fall
below the ambient noise, which is 50 dB re 1 pPa for SS2.
For 1 dB loss due to receiver noise, and again assuming 4 dB loss in the hydrophone,
the spectrum level of the receiver noise must be
50 - 6 - 4 = 40 dB
From Figure 5.5, 1 dB loss requires the receiver noise to be 6 dB down. Now we
substitute 40 dB into the sensitivity equation:
20 log v = -200 + 20 log p (dB)

= -200 + 40 = -260 dB/V
V = 10nV
10 nV:
Example 5.2 by simpler approach

Now consider a sonar, again operating at 5 kHz, installed on a submarine whose self-
noise, in its quietest state, can be less than the ambient noise of the sea, and take the
ambient noise in sea state 0 as 35 dB re I p a .
For 1 dB loss due to receiver noise, and again assuming 4 dB loss in the hydrophone,
the spectrum level of the receiver noise must be
35 - 6 - 4 = 25 dB
Therefore
20 log v = -200 + 25 = - 175 dB/V

v = 2nV
2 nV

Table 5.2 gives practical values for performance comparisons and preliminary
system design. All values are isotropic spectrum levels in dE3 re the intensity of
sound due to a pressure of 1 p a . The ambient noise figures apply to sonobuoys,
helicopter dipping sonars, towed arrays and submarines. The self-noise figures
apply to hull-mounted surface ship sonars; they are based on a ship moving at
speeds up to about 15 knots in sea states of about 1 or 2 (low) and 4 or 5 (high).
Table 5.2 Practical noise values for design and performance comparison
Isotropic spectrum level (dB)re the intensity of sound due to a
pressure of 1 Pa
0.5kHz llcHz 2lcHz 4kHz 8kHz 16kHz32kHz64kHz

Highambientnoise(SS4/5) 75 70 65 60 55 50 45 40
Low ambient noise (SS2) 65 60 55 50 45 40 35 30
High self-noise 84 78 72 66 60 54 48 42
Low self-noise 74 68 62 60 50 44 38 32
Atfiequencies below 500 Hz
Ambient noise tends to flatten but can be increased by shipping.
Self-noise is no longer a continuum, but is strongly influenced by discrete tones

from propellers and machinery.
Atji-equencies above 64 kHz
The thermal noise of the sea is increasingly the dominant source of noise and
reverses the steady falls in both ambient and self-noise with frequency.
Prt,hEerns 101
5.2i Problems
5.1 A hydrophone with a sensitivity of - 170 dB/V receives over a 1000 Hz band
centred on 100 kHz.What would its output be in SS2?
5.2 The self-noise of a sonar is 55 dB. If the ambient noise of the sea is also 55 dB,
what would be the level in a third-octave band centred on 6 kHz measured at the
output of a receiver beam of DI = 20 dB?
6
Reverberation
6.1 Sources of Reverberation

When sound is transmitted underwater it is scattered by marine life, inanimate
matter distributed in the sea and the inhomogeneous structure of the sea itself, as
well as by reflection from the surface and the sea bed. The component of the
incident sound energy reflected back to the source is known as bachcattering.
This backscattered energy is reverberation, which comprises both the background
to detection of a target and echoes from the target itself. Target echoes are simply
a special case of reverberation.
6.2 Scattering and Reflection

Scattering and reflection occur wherever there is a change in 2, the specific
acoustic impedance. The proportion of the energy reflected is given by
where Z, is the specific acoustic impedance of water, Zr is the specific acoustic

impedance of the reflector and Cr is the coefficient of reflection at the boundary.
When a reflection occurs at the sea surface, we have
Z,= 415 kgm--2s-l

Z, = 1.5 X 106kgm-* s-'
C, = -0.9995
104 Reverberation
Here Zr is the specific acoustic impedance of air and 2, is the specific acoustic
impedance of water. Almost all of the sound is reflected and only a tiny fiaction,
0.0005, escapes into the atmosphere. Note the phase change.
When a reflection occurs at the sea bed, the situation is much more complicated.
Because Z,, the specific acoustic impedance of the sea bed, is now much greater
than Zw, the specific acoustic impedance of water, the value of Cr is positive (but
still almost unity) and there is no phase change at the reflection. Once again, much
of the sound is reflected and only a fi-action penetrates the sea bed.
The sea bed, however, has highly variable acoustic properties due to its
composition and inhomogeneous nature, with the result that Z,, the specific
acoustic impedance of the sea bed, changes markedly with both depth of penetra-
tion and laterally. A sufficiently accurate and detailed knowledge of the bottom to
ensure useful predictions for scattering, and for propagation within the bottom
itself, is never realistically available.
6.3 Boundary Roughness

The nature of the reflection and scattering depends on the degree of roughness of
the boundary. If the incident sound wave strikes a perfectly smooth sea surface
or sea bed, nearly all the energy is reflected in the specular direction. As the
boundary roughens, more and more sound energy is scattered in non-specular
directions and the specular reflection is reduced in intensity until, for a very rough
surface, there is no discernible peak in the specular direction at all.
The roughness of a boundary is indicated by the Rayleigh parameter, 5, given
by
Rp = khsin8
where h is the wave height, measured from trough to crest, 8 is the grazing angle
and k = 2 n / A is the wave number. In practice when 4 << 1 the surface is taken to
be smooth and when l$ >> 1 the surface is rough.
Figure 6.1 illustrates the reflection and scattering of sound from (a) perfectly
smooth boundaries and (b) rough boundaries. The boundary shown is the sea
surface; inversion of the figure represents the similar situation when the boundary
is the sea bed.
C1assc.s of Reverberation
i 0.0005
Smooth
surface
/' i \
1.0 / \, 0.9995
/
1 \
+
Rough surface
f,
L
Specular direction
Figure 6.1 Scattering of sound from a boundary: (a) perfectly smooth, (b) rough
6.4 Classes of Reverberation

Along with noise, reverberation - the backscattered part of the total reflection -
forms the background to the detection of targets by an active sonar. In many
environments, reverberation can be the dominant component of this background.
It is therefore very important for the design of active sonars to have a good
measure of its magnitude and how it varies with roughness and grazing angle.
Reverberation may be classified according to its source:
106 Reverberation
0 Volume reverberation is where the reverberation originates from scatterers

within a volume of the sea.
0 Sea sugace reverberation is where the reverberation originates from scatterers

spread over an area of the surface of the sea.
Sea bottom reverberation is where the reverberation originates from scatterers

spread over an area of the sea bed.
For analytical purposes, sea surface and sea bottom reverberation may be lumped
together as boundary (area) reverberation.
6.5 Backscattering Strength

Backscattering strength is the fundamental parameter that decides the intensity of
the reverberation. It is defined by
1 S,," = 10 log -
I
;
;J
where Iscatis the intensity of the sound scattered (back to the source) by a unit area
or unit volume, referred to a distance of 1 m from its acoustic centre, and Ii is the
intensity of the incident plane wave.
6.6 Reverberation Target Strength

The target strength for reverberation, TSR, is a measure of the background to
detection of a target and is used later to derive the reverberation-limited active
sonar equation.
where A and V are the total reverberating area or volume, respectively, as defined
by the 'two-way' beamwidth of the sonar - the effective beamwidths of the
projector plus hydrophone combination in both azimuth and elevation. For
example, the horizontal two-way beamwidth of the combination of an omnidirec-
tional, in azimuth, projector and a 10" , in azimuth, receive array is 10" (the
smaller of the two).
Volume Reverberation 107
6.7 Volume Reverberation

To determine TSR for volume reverberation, it is necessary to determine 10log V.
The total reverberating volume (Figure 6.2) is given by
where Br and Bt are the receive and transmit beam patterns.
Figure 6.2 Reverberating volume
The length of the cylinder, dV (its extent in range), is such that the reflections
from all scatterers within it return to the source simultaneously. (The scattering of
the front end of the pulse by the rear scatterers in dV returns to the source at the
same time as the scattering of the rear end of the pulse by the front scatterers
in dV)
The extent in range is therefore cT/2, where T is the pulse length and c is the
speed of sound. (Note the division by 2, a constant source of confusion between
active, two-way, and passive, one-way, sonars.) The volume can now be written
V=-R2
I, B,Btd6
108 Reverberation
The integral is the equivalent beamwidth of the two-way (transmitlreceive)

combination. In this expression T is either
0 The actual pulse length, for a CW pulse

0 The reciprocal of the bandwidth, for a wideband pulse (see later)
and if R is large compared with the cross-section of the volume (say R > 4R@,
the volume is given by
cT nR2&Ov
V=--
2 4
where & and OV are the two-way horizontal and vertical beamwidths expressed in
radians. Figure 6.3 is a graphical representation of this simplified expression for
the reverberating volume.
Figure 6.3 Reverberating volume
Example 6.1
A sonar has two-way horizontal and vertical beamwidths of 12" (0.2 rad) and transmits
a pulse of length 1 ms. What is the TSR at 1000 m for S, = -80 dB?
We calculate
1500 X n X lo6 X 0.22

V= =2 x 104
2 4
W u m e Reverberation 109
and
TSR = -80 + 10 log2 X lo4 = -80 + 43 = -37 dB
When the beam patterns intercept a boundary, i.e., the surface or the bottom of the sea
- which, except in deep water, occurs at quite close ranges (perhaps 1000 m) - the
volume will be smaller than given by the formula. It is then possible, but not usually
necessary, to work out the cross-sectional area from a knowledge of the depth of water
and the depth of the source and to use this to determine the volume.
However, as soon as a boundary is intercepted by both transmit and receive beam

patterns, the boundary (area) reverberation is almost invariably dominant and volume
reverberation is no longer of interest.
110 Reverberation
6.8 Boundary Reverberation

To determine TSR for boundary reverberation we need to determine 10 log A,
where A is the area of the scattering boundary (Figure 6.4). Provided the axis of
the beam is only slightly inclined towards the scattering boundary, which is usually
the case, it is only necessary to consider the horizontal beam patterns of the arrays,
the solid angle for the volume case reduces to the horizontal beamwidth, and A
becomes
with @, in radians.
Figure 6.4 Boundary reverberation
Example 6.2
For the same sonar, what is the TSR at 1000 m for S, = -30 dB?
We calculate
Scattering Layers 111
and
TSR = -30 + lolog 150 = -30 + 22 = -8 dB
In deep water the dominant reverberation will be from the sea surface.
In shallow water (< 200 m) the dominant reverberation will be from the sea bed
unless the wind speed is high (see later).
0 Volume reverberation is seldom dominant; the exception is in calm, deep seas at

long range.
6.9 Scattering Layers

Some scatterers in the sea - such as the deep scattering layer, or a layer of bubbles
below the sea surface - lie in layers of limited thickness and are best considered
as a modified form of boundary reverberation. Quite wrong results would be
obtained if the scattering from these sources were considered to be uniformly
present throughout the entire volume, as previously calculated.
If the volume scattering strength of the layer is S, and the thickness of the layer
is N metres, then the scattering strength of the layer, Sp, is given by
I S[ = s, + 10 log H 1
and this value must be used in the expression for boundary reverberation.
112 Reverberation
6.10 Volume Scattering Strength
\
\
'\
/ Night
/
/
\
/
\+L
\
\ 100 m
/
/
Frequency 5 1<Hz
Figure 6.5 Migration of deep scattering layer
Volume scatterers are not uniformly distributed in the sea and tend to be
concentrated in the deep scattering layer (DSL). This source of reverberation is
overwhelmingly biological in nature and is a complex mix of different organisms
resulting in scattering strength versus depth profiles that change with frequency,
location and time. The representative S, profiles in Figure 6.5 show vertical
migration of the DSL with time of day. Estimates of thickness are shown and
St (day) = -76 + 10 log 80 = -57 dB

Se (night) = -74 + lolog 100 = -54 dB
Sea Surface Scattering Strength I13
6.1 1 Sea Surface Scattering Strength
Grazing an@e
ss (dB)
-40
-50
- 60
Wind speed 10 knots

2 4 6 x 10
(a) Frequency (kHz)
S, (dB) Grazing angle
-30
-40
-50
Sea state 6
Wind speed 30 knots
2 4 6 8 10
(b) Frequency (kHz)
Figure 6.6 Sea surface scattering strengths
The roughness of the sea surface, and the possibility of bubbles trapped just
beneath it, results in significant scattering of sound. The scattering strength varies
with the grazing angle of the incident wave and with frequency. Empirical
formulae by Chapman and Harris fit the many measurements quite well, particu-
larly at frequencies between 1 and 10 kHz.Figure 6.6 plots these formulae for two
representative sea states. Note that S, increases markedly with grazing angle and
also with frequency (around 3 dB per octave at the lower grazing angles which are
of particular interest to the sonar designer). The dashed lines show recommended
values for use in sonar design, representing high sea states (-40 dB) and low sea
states (-50 dB).
114 Reverberation
6.12 Bottom Scattering Strength

The sea bed is an effective reflector of sound. As with the sea surface, the
backscattering strength of the bottom (Sb) varies with the grazing angle and
frequency and also with the nature of the bottom. Early observations indicated that
the level of the reverberation was much greater over rocky bottoms than over sand
and mud, and it is now customary to relate s b , somewhat imprecisely, to bottom
type. The roughness of the bottom is at least as important as its constituent
materials and Figure 6.7 is a useful way of visualizing the redistribution of the
incident sound.
Figure 6.7 Sound scattering patterns
Figure 6.7 shows two directivity patterns for the sound incident upon the bottom
at point 0. The radius of the patterns in any direction is proportional to the
intensity of the sound reflected in that direction.
Pattern A is characteristic of a smooth bottom, where the specular reflection in

direction OR is large and the scattering in other directions including the
backscattering back to the source in the direction 01, at a grazing angle, 8,is
small.
Pattern B is characteristic of a rough bottom, where the specular reflection in

direction 01 is much greater. (The backscattering strength, Sb, is much greater.)
Sound is also absorbed where it penetrates the bottom in direction OK
Bottom Scattering Strength 115
Many, frequently discordant, measurements have been reported in the literature

and this, coupled with a lack of detailed knowledge of the sea bed at any particular
location, results in considerable uncertainty over the value of S b to use for a
particular problem.
Bottom scattering strengths may be considered roughly constant at frequencies
up to 10 kHz and grazing angles up to 10". They are, however, very dependent on
the material of the sea bed, varying from a low value of -45 dB for mud to
-25 dB for rock, as shown by the results fi-om a survey of UK shallow water areas
(Figure 6.8).
Figure 6.8 Distribution of S,, in UK shallow water areas
Bottom scattering strengths are generally greater than those from the surface
and therefore, when the geometry is such that the sonar vertical beam patterns
intercept the bottom (which is always the case for shallow water at practical
ranges), then bottom reverberation is often the dominant background to detection
and classification.
Figure 6.8 illustrates the wide range of S b which may be encountered in shallow
waters. We shall see later that s b is an extremely important factor in determining
the Performance and design of an active sonar and if, during its design, an
incorrect value is chosen for the maximum Sb likely to be encountered, the sonar's
performance will not be optimum.
Over a uniform bottom the basic detection performance will be determined by
the average value of Sb, which is little affected by comparatively rare, submarine-
116 Reverberation
like false alarms. The bottom may be mud, clay, sand, shingle, pebbles or a mix of
these and the calculated & (derived From measurements of reverberation levels)
falls within the range -35 to -50 dB at the low grazing angles encountered in
shallow water.
Over a non-uniform bottom (rocky outcrops, shorelines, undulations) the
average value of Sb is greatly influenced by such local features and it is suggested
that the actual values of &,, used to determine the initial detection ranges of a
sonar, should be some 5-10 dB less than the average measured values. Thus,
values for &, which are greater than -35 dB, i.e., falling within the range -35 to
-25 dB, and typical of non-uniform sea beds, should be reduced by 5 dB but
never to less than -35 dB. Table 6.1 should make this clear.
Table 6.1 Reduce St, values by 5 dB but never to less than -35 dE3
&, measured (dB) -25 -27 -29 -3 1 -33 -35
Use (dB) -30 -32 -34 -35 -35 -35
This argument is supported by initial detection trials where, over such difficult
bottoms, final detection and classification are delayed by a multitude of false
alarms but post-trials analysis Frequently reveals missed detection opportunities at
greater ranges (consistent with postulating lower values for Sb).
6.13 Variation with Frequency

At frequencies above 10 kHz, & appears to increase at about 3 dB per octave for
the smoother (mud, sand) bottoms, i.e., where the roughness is small compared to
a wavelength; but for rougher (shingle, pebbles, rock) bottoms, s b is independent
of frequency. This frequency dependency is an unfortunate result for mine hunting
sonars which, to achieve sufficient definition, must operate at high frequencies.
Problem 117
6.14 Reverberation under Ice
4- ~
5-10 kHz Summer
_cc____ 1-3 kHz polar ice 1
- 40 /
/
0
rr--
------ "
/ -
Ice-free
Wind speed = 30 knots
/
5 10 15 20
Grazing angle (degrees)
Figure 6.9 Under ice scattering strengths
The underside of the ice cover in polar regions has a very high backscattering
strength, S,, when it is rough and non-uniform (e.g., spring pack ice). When the
underside is smoother (e.g., summer polar ice), Si is significantly less. Most of the
Si measurements show an approximately 3 dB per octave frequency dependency
and are generally greater than S,measurements in ice-free water (Figure 6.9).
6.15 Problem
6.1 A sonar system has a horizontal beamwidth of 10" and transmits a pulse of
duration 100 ms. If the backscattering strength is -40 dB, what is the reverberation
target strength, TSR,at 4000 m? What is the TSR at twice the range?
7
The Sonar Equations
7.1 What Are They?

The sonar equations relate the parameters of the entire sonar system - the sonar
and its platform, the target and the environment - in order to determine its
performance. This chapter will develop the basic equations and explore the
concepts of detection threshold (DT), detection index, d, and their relationships to
the probability of detection, P d , and the probability of false alarms, Pfa.Later
chapters will develop the sonar equations to make them applicable to active,
passive, intercept and communications sonars. This is because it requires a
detailed knowledge of the signal processing to develop an expression for DT
appropriate to a particular class of sonar.
7.2 What Are Their Uses?

The sonar equations are used to predict the performance of a known design, and to
design a sonar for a given performance. For performance prediction, the para-
meters of the sonar are known and an estimate of its detection performance is
required. The sonar equations are solved for the allowable propagation losses at
specified probabilities of detection and false alarms, and the allowable propagation
losses are then converted to estimates of detection range in the environments of
interest.
When designing a sonar to detect at some predetermined range, the propagation
loss is first estimated for the environment of interest and is then used in the sonar
equations to solve for those parameters over which the sonar designer has some
120 The Sonar Equations
control. This process is often severely constrained by, for example, platform size
or equipment cost and is considerably aided by the intuition of an experienced
sonar designer.
7.3 The Basic Sonar Equation

The basic sonar equation simply expresses the difference between signal-to-noise
ratio at the output of the beamformer and the detection threshold (DT). This
difference is the signal excess (SE) and in dB form it is given by
SE = S - N - DT
Signal power is in the analysis bandwidth and noise power is in a 1 Hz band. This
threshold is defined such that a signal with S - N (at the output of the
beamformer) equal to DT, has a specified (often 50 per cent) probability of
detection ( p d ) for a required probability of false alarms (Pfa). The value of DT
depends on these probabilities and the sonar signal processing. Put another way, a
signal excess of zero corresponds to a pd of 50 per cent and a positive SE indicates
pd > 50 per cent.
7.4 The Basic Passive Equation

Passive sonars detect signals radiated by the target. The signal level is therefore
S = SL - PL, where SL is the source level of the target (the radiated noise in the
frequency band of interest). The basic passive sonar equation is therefore
ISE = (SL - PL) - N = DT(
7.5 The Basic Active Equation

Active sonars detect target echoes. The signal level is therefore S =
SL + TS - 2PL, where SL is now the source level of the projector array. Note the
active 2. The basic active sonar equation is therefore
I SE = (SL + TS - 2PL) - N - DTI

Detection Threshold and Detection Index 121
7.6 Detection Threshold and Detection Index

All sonar receiving systems conform to Figure 7.1. The minimum discernible
signal (MDS) is the S I N ratio at the array face which results in preset P d and Pfa
values:
MDS = DT - DI (if the noise is isotropic)
MDS = DT - AG (if the noise is directional)
Display/auto
Preset Pd for given Y,
system
T
MDS DT d
Figure 7.1 Basic sonar receiver
The detection threshoZd (DT) is the ratio of mean signal power to mean noise
power (expressed in dB form) measured after beamfoming, which results in the
preset P d and Pfa values. The signal power is in the analysis bandwidth and the
noise power is in a 1 Hz band. Detection index, d, is defined as
d=(
+
mean (S N) - mean (N)
std dev(N)
where the quantities are measured at the display or automatic system.

Figure 7.2 shows curves of probability density plotted for N alone and for
S + N.
+
The detection index, d, is equivalent to the ratio (S N ) / N of the envelope of
the signal processing output where the threshold, T, is established. The area under
+
the curve of S N above T is the probability that an amplitude in excess of T is
due to signal plus noise; this area equals P d . The area under the curve of N above
T is the probability that an amplitude in excess of T is due to noise only; this area
equals Pfa.These two probabilities therefore vary with the threshold setting and
depend on the value of d.
Figure 7.2 Pd, Pf,and threshold setting

Receiver Operating Characteristics 123
7.7 Receiver Operating Characteristics

Curves of receiver operating characteristics, ROC curves, express P d and Pfaas a
+
function of 5 log d for the appropriate distributions of N and S N. The standard
ROC curve (Figure 7.3) assumes Gaussian statistics for both distributions. Two
Pd 1 .o
0.5
0.2
Figure 7.3 ROC curves for Gaussian statistics: N and S +N
levels of signal variability are shown, defined in terms of the fluctuation index (FI)
given by
FI =
o(S + N)
uN
+
FI = 1 corresponds to a stable signal where the variability in S N is entirely due
to N.Increasing FI corresponds to increasing the signal variability.
7.8 ROCCurves
There is no one ROC curve applicable to all sonars. Figure 7.3, which assumes
Gaussian statistics, is applicable to sonars with large BT products (broadband
sonars). Table 7.1 gives values of 5logd taken from the Sonar ModeZZing
Handbook ROC curves together with their recommended usage. But here are some
useful rule-of-thumb values:
Broadband: 6 dB
Intercept and communications: 10 dB
Narrowband and active: 10 dB
Table 7.1 Values of 5 log d and their recommended usage

pfa= 10-5 pfa= USE FOR
Pd = 0.5 6 7 Broadband FI = 1
6 7 FI =4
12 13 Narrowband power IF =1
10 11 IF =2
9 10 IF =4
8 9 IF =8
8 9 Narrowband amplitude
10 11 Active, intercept and communications
Pd = 0.9 8 8 Broadband FI =1
>>8 >>8 FI =4
20 21 Narrowband power IF =1
15 16 IF =2
13 14 IF =4
9 11 IF =8
12 13 Narrowband amplitude
12 13 Active, intercept and communications
Integration Factor (IF) is defined in Section 8.5.
7.9 Problem
7.1 The Pd of a sonar is to be 0.5. What will be the change in Pf, which results from
reducing 5 log d from 6 dB to 5 dB? Use Figure 7.3.
8
Passive Sonar
8.1 Radiated Noise

Radiated noise is the noise emitted by a vessel and received by a hydrophone, or
an array of hydrophones, at some distance from the vessel. Radiated noise is the
source of signals for passive sonars which are designed to detect radiated noise
against a background of ambient and self-noise.
8.2 Source Level

The source level of radiated noise is analogous to the transmit source level for an
active sonar and is used in the passive sonar equations. Measurements are made at
some distance from the source (between 100 and 1000 m) and reduced to a
standard distance of 1 m by assuming spherical spreading and, where significant,
absorption. Measurements are normally made in third-octave bands over the
frequency range of interest - from 10 Hz to 100 kHz - and converted to spectrum
levels on the assumption that the intensity is constant over the band.
To investigate the Eine structure of the radiated noise spectrum, a much
narrower bandwidth is required. Discrete lines are important up to about 3000 Hz
and narrowband analysis is carried out over a selection of frequency bands of
interest such as those in Table 8.1.
Table 8.1 Analysis bandwidths are narrower

~~ ~
Frequency band (Hz) Analysis bandwidth

(Hz)
0-100 0.3
100- 1000 1
1000- 3000 4
126 Passive Sonar
8.3 Nature of Radiated Noise

The noise radiated by a vessel exists as a continuous spectrum, or continuum, on
which are superimposed narrowband discrete components, known simply as lines
or tonals. Both the continuum and discrete components reduce in intensity as the
frequency increases. Machinery noise and propeller noise dominate the spectra of
radiated noise in most conditions. The lower frequency end of the spectrum is
dominated by machinery lines and blade rate lines of the propeller. These lines die
away with increasing frequency and become submerged in the continuous
spectrum of propeller and machinery noise. As the speed of a vessel is increased,
the continuum increases in intensity and extends to lower frequencies.
Figure 8.1 illustrates the effect of increasing speed. For submarines, a decrease
in depth - because it increases any noise due to cavitation - has the same general
effect. For a given speed (and depth for submarines and torpedoes), a broad
crossoverfrequency exists, below which the spectrum is dominated by tonals and
above which the spectrum is the continuum of the cavitating propeller.
Figure 8.1 Radiated noise: continuum and tonals for (a) low speed and (b) high speed
For surface ships and submarines this frequency is between 100 and 500 Hz,
depending on speed and, for submarines, depth. For torpedoes the crossover
fiequency is higher, perhaps between 500 and 1500 Hz, and the tonals extend to
higher frequencies, say 3000 Hz, because of the higher speeds of machinery and
propellers.
Pructical Values 127

Information on the radiated noise of vessels, particularly submarines and torpe-
does, is highly classified and the sonar designer must resort to the classified
literature for details of levels and frequencies of tonals radiated from modem
vessels.
Figure 8.2 shows representative radiated noise spectra for sonar targets. The
radiated noise of a frigate is also shown for comparison. The spectra should be
considered as worst case, particularly in the regions dominated by discrete lines.
Submarine and torpedo quietening measures have succeeded in significantly
reducing or eliminating many discrete lines, particularly at the higher frequencies,
resulting in the need for passive sonars to operate at ever lower frequencies
I60
Frequency (Hz)
Figure 8.2 Representative radiated noise spectra for sonar targets

128 Passive Sonar
(therefore they need ever longer towed arrays) to detect the remaining lower
frequency lines, which are difficult to eliminate.
An older submarine, for example, may have a tonal at 200 Hz of intensity
125 dB, but tonals near this frequency for a modern submarine will have been
reduced to a value below the broadband continuum which tends to level off in
intensity at, say, 300 Hz (and perhaps 1000 Hz for torpedoes).
Broadband and Narrowband 129
8.5 Broadband and Narrowband

Passive sonar systems investigate the total spectrum of the noise radiated from
targets, using broadband and narrowband techniques. Broadband sonars examine
the total energy over a wide frequency band, which is generally divided into octaves.
The broadband sonar equations, derived later in this chapter, indicate that perform-
ance is improved by increasing the bandwidth. This is only true if the bandwidth
does not significantly exceed the spectrum of the noise radiated by the threat.
Narrowband sonars split the total energy into narrow frequency analysis cells in
order to search for discrete radiated lines. The narrowband passive sonar equa-
tions, also derived later in this chapter, indicate that performance is improved by
reducing the analysis bandwidth (until the signal is over resolved).
The broadband noise from a propeller may be amplitude modulated at the blade
rate fundamental and harmonic frequencies. A technique known as DEMON
(demodulation) exploits this by narrowband analysis over the band which covers
these modulation frequencies.
Because they provide a detailed knowledge of the threats radiated noise,
narrowband sonars give good detection and classification capabilities. Because
prominent tonals are radiated at comparatively low frequencies, particularly by
submarines, bearing accuracy is limited when compared with broadband sonars
operating at higher frequencies where beams are narrower and correlation
techniques may be employed to further improve bearing accuracy. (Very long
towed line arrays, however, achieve similar bearing accuracies at these low
narrowband frequencies.)
Figure 8.3 shows a complete passive sonar system. The array signals are
beamformed over the entire frequency range and output to audio, broadband and
narrowband sonars. The audio sonar will have facilities to select a beam and a
listening bandwidth.
The broadband sonar will detect signals from the beamformer (using energy or
cross-correlation techniques); integrate (incoherently sum the energy in the signal
band); normalize the signals and display them, typically in a bearing (beam)
versus time format. The integration gain for energy detection is 5 log BT,, and for
cross-correlation 5 log 2BTe, where B is the signal bandwidth (usually an octave)
and Te is the integration time (the time a signal is likely to remain within a beam).
The narrowband sonar performs a frequency spectrum calculation to convert the
time series beamformer output into power spectral data in two stages. The first stage
is applied to all data from all beams to produce a surveillance capability; the second
stage allows the operator to zoom in (increase the frequency resolution) to a part
of the total spectrum. This second stage is also known as vernier processing. The
gain fkom the total process is 10 log 3, where B is the final analysis bandwidth. If
130 Passive Sonar
2
a
M
iz
4 Note on Swaths 131
the signal bandwidth, B,, is greater than B, the signal is over resolved and there will
be a reduction in gain of 10 log B / B,. Electronic integration sums the energy from
each analysis cell over a number of time steps. The number of steps is known as the
integrationfactor (IF) and the gain is 5 log IF.
8.6 Normalization
There are significant differences in the mean noise levels with fiequency, bearing
(beam) and time, which must be reduced before display to the operator. Normal-
ization schemes all depend on using the information surrounding the cell to be
displayed to reduce these differences to the dynamic range of the display. Figure 8.4
shows an algorithm for normalizing a line of data in the frequency dimension before
display. The same process is repeated for all analysis cells.
Power
Cell i
Average these 8 cells to estimate local mean noise

Use to modify power of the ith cell
Figure 8.4 A typical normalization algorithm
8.7 A Note on Swaths

Swaths are signals with bandwidths significantly greater than the usual resolution
of a narrowband processor, but much smaller than the octave or more bandwidths
of broadband processors. Swath detection, therefore, requires wider analysis
bandwidths and modifications to the trend removal or normalization algorithms by
making similar increases in window widths and spacings.
132 Passive Sonar
8.8 Passive Arrays

Noise reduction programmes have succeeded in reducing the broadband noise
continuum and reducing or eliminating the higher frequency tonals radiated by
older designs of submarines and torpedoes. Modern passive sonars, therefore,
must employ large arrays to achieve sufficient gain and directivity at the low
frequencies necessary for the timely detection and classification of submarines
and torpedoes.
Surface ship hull-mounted arrays
Surface ship hull-mounted arrays will generally take the form of cylindrical arrays
mounted in bow or keel domes. Operating frequencies may extend down to about
100 Hz for narrowband tonals and up to about 10 kHz for broadband. The DI will
clearly be small at frequencies below about 1 kHz and the self-noise high.
Performance against submarines will be ineffective but torpedoes (apart from
quiet, electrically propelled, marks) will be detected at adequate ranges for the
deployment of countermeasures.
Submarine hull-mounted arrays
Submarine hull-mounted arrays again may be cylindrical but a submarine offers

much better opportunities to install large planar or conformal arrays. The DI will
still be small at the frequencies of submarine tonals, but the self-noise is generally
significantly less than for surface ships. Performance against submarines and
torpedoes is therefore effective against all but the quietest of submarines.
Line arrays
Line arrays towed from submarines or surface ships - by virtue of their length,
typically 32 or 64 wavelengths, and low operating frequencies, down to about
10 Hz - are very effective against both submarines and torpedoes. Separation
from the tow vessel assists in reducing the self-noise seen by the towed array,
particularly at bearings other than those of the ahead beams which receive the
stern radiated noise of the tow vessel in their main lobes.
133
8.9 Passive Aural

The principal characteristics of the human ear are as follows:
Dynamic range 120 dB (see Section 1.12)

Wide frequency range: 20 to 15 000 Hz
A powerful frequency analyser
As a frequency analyser, the human ear can resolve two sounds which differ by as
little as 50- 100 Hz over much of the audible band. Because of these character-
istics, the ear is still important in passive sonar, and perhaps unrivalled as the final
arbiter in the classification process.
The ear is not a simple broadband listening channel, but rather may be likened
to a contiguous comb of narrowband filters covering the entire audio spectrum.
The bandwidth of each of these hypothetical filters is known as the critical
bandwidth (Af,) of the ear and is between 50 and 100 Hz for frequencies between
300 and 2000 Hz and increasing somewhat outside this band.
The quantity AfC is the bandwidth such that an increase in spectrum level
(within reasonable limits), outside this band has no masking effect on a signal
Myithin the critical band (Figure 8.5). Therefore, the masking noise is given by
where N,(f) is the spectrum level ofthe noise at signal frequency f.

Experimental verification has shown that the recognition level of the ear is close
(b) f; (c) .fi

lncreasing signal level
Figure 8.5 Masking of a signal (wanted noise) by noise

134 Passive Sonar
to C(f>and, from the concept of critical bandwidth, it follows that as the level of
a signal (which is in fact the wideband radiated noise emitted by the threat) is
increased, as soon as this level reaches the noise level at any frequency within the
receiver bandwidth, it is perceived (at P d = 50 per cent) as a signal at that
frequency, known as the thresholdfrequency, f t .
Reducing the listening bandwidth risks eliminating the threshold frequency (or
moving it to another value where the noise is higher) and therefore the audio
bandwidth should cover the entire frequency spectrum of the expected signal.
The ear will have detected the signal at the level represented by Figure 8.5(b),
but a simple broadband receiver will not detect the signal until the level is
increased to about that represented by Figure 8.5(c), a positive signal-to-noise
ratio.
Passive Displays 135
8.10 Passive Displays

The ear has the advantage of being able to simultaneously use all the information
contained within the signal - harmonics, waveform, fluctuations in intensity and
frequency - which can often result in very low recognition differentials as well as
aiding in threat classification. Visual indicators, however, also benefit from the
quasi-permanent character of the signal by performing electronic integration after
detection and visual integration at the display.
The electronic integration time is limited by the time a threat may be expected
to stay within a beam - perhaps 2 s for a torpedo and 20 s for a submarine.
The visual integration time is not similarly limited because the integration can
now take place through many beams, although clearly it cannot be longer than a
few minutes for a torpedo and still offer a timely countermeasure response. The
visual integration time is still limited by practical considerations such as the
display capacity. There are also diminishing returns in the gain achieved; the
theoretical 1.5 dB for each doubling of displayed lines tends to level off at about
10 dB for 128 lines.
Any of the traces in Figure 8.6 could be a subsurface contact. A link with radar
(for surface ships) can eliminate surface contacts. Any remaining contact is highly
likely to be a torpedo or a submarine. Submarine detection is unlikely using a
surface ship hull-mounted array but possible using a submarine hull-mounted
array (because the self-noise is lower and much larger arrays are practical). Long
towed arrays deployed from either surface ships or submarines perform well
against both torpedoes and submarines.
Figure 8.6 Broadband (bearinghime) display

136 Passive Sonar
Narrowband signals are displayed in a format known, for historical reasons, as a

LOFARGRAM (low fiequency analysis recording gram) where the outputs from
all beams or a selected group of beams are displayed in a frequency versus time
format. LOFARGRAMS contain information which, to the skilled operator, can
assist in classification and analysis of contact motion. In Figure 8.7 the contact is
strongest in beam 2; there has been a change in doppler, which indicates either a
speed or a course change; and the frequencies of the tonals may classify the
contact.
.' '
; '..-.
; *:
**
Contact Own vessel
. 3
L
- \
-...
*
*.
*
*. *.
:
I
:
I
I
#
#
- - - - .
2
Time
*.
.
.
8.
9-
*
I
I
9
. } Beam 1
8.11 Formulae for Detection Threshold

The basic passive sonar equation (Chapter 7) may now be modified to include
expressions for the detection threshold (DT) which are specific to the type of
passive sonar. Formulae for DT are traditionally based on a square law detector
acting on a signal in Gaussian noise, whose gain is given by
+
mean(S N) - mean(N) signal power
( std dev( N)
Broadhand Square Law Detector 137
which, from the definitions of d and DT given in Chapter 7, becomes
d = BT(DT)2 or DT = 5 log d - 5 log BT (in dB form)
The product BT measures the number of independent noise samples combined in

the detector. Therefore the variance of the noise is reduced by BT relative to the
mean, resulting in the gain given by this expression.
8.12 Broadband Square Law Detector

This passive system simply beamforms using the full aperture of the array and
sums the energy in the bandwidth (usually split into separate octaves) of the
expected signal over a period Te, the electronic integration time. Therefore
and, substituting for DT, the sonar equation for this passive receiver, also called an
energy detector, becomes
SE = (SL - PL) - N + DI - (510gd - 510g BTe)
The expression for DT assumes a decision based on a single sample (or line for a
practical multibeam receiver), displayed to the operator or passed to an automatic
system. In practice, operators and automatic systems make decisions based on a
number of successive time samples. The equation must therefore be modified by
including a visual integration term, 5 log n, where n is the number of successive
time samples used by the operator or automatic system in making the decision.
Putting SE = 0 gives
IPL = SL - N + DI - 510gd + Slog BT, + 510g nl

Note that Te - the time that a signal may be expected to remain in a beam - is the
time between successive display samples (lines) and is greater than l/B.
138 Passive Sonar
8.13 Broadband Cross-Correlator Detector

Here the receive array is split into two equal halves to form pairs of codirectional
beams whose outputs are then complex cross-correlated. The phase outputs from
this process provide bearing measurements to a resolution of typically one-tenth of
the beamwidth (given adequate S/N ratio). Full beams can only improve on
bearing resolution by using amplitude comparisons between adjacent beams,
inherently inferior to phase comparisons.
The DT is now given by
and the sonar equation becomes
Note that the DI is now that of the half-beams, or 3 dB less than for the full
aperture. However, the factor of 2 in the 5 log 2 BT, term implies a net loss of
1.5 dB (3 - 1.5) compared with the energy detector, a small price to pay for the
improved bearing discrimination.
Narvo whand Processor 139
8.14 Narrowband Processor

For a narrowband signal processor, the square law detector must be modified to
allow for the different bandwidths of signal and noise. For the broadband case
Here signal and noise apply to identical bandwidths, but for narrowband the signal
is the intrinsic bandwidth of the line, whereas the noise is referred to a 1 Hz band.
Both levels must be referred to a common bandwidth, the analysis bandwidth, B ,
of each frequency cell. The input noise to the narrowband processor is therefore
narrowband, and for the narrowband case
and taking logarithms, DT = 5 log d - 5 log(Te/B).

The total processing time, Te, comprises two factors:
The analysis time, 1/ B
An integration factor, IF, which is the number of independent samples from the
signal processor summed before display.
So Te = IF/B; substituting for Te we obtain
DT=510gd-510g
(3
-
and simplifying gives
DT = 5 10gd + 10 log B - 5 log IF
If the signal bandwidth, B,, is greater than the analysis bandwidth, B, the signal
will be over resolved - it will appear in more than one cell - and there will be a
reduction in gain of 10 log( B/ 73,). Note that any mismatch cannot increase
processing gain and therefore any positive value for 10 log( B / B,) must be reduced
to zero.
I40 Passive Sonar
The expression for DT is then
DT = 5 log d + 10 log B - 10 log(B/B,) - 5 log IF
and the narrowband @ower) sonar equation is
IPL = SL - N + DI - 5 log d - 10 log B + 10log(B/B,) + 5 log IF + 5 log
8.15 Narrowband Amplitude Detector Processor

If the signal processor produces amplitude values (usually by simply taking the
square roots of the power values), the expression for DT becomes
DT = lolog [?+ l.05(&)1/2] + 10logB - lOlog(B/B,)
and the narrowband (amplitude) sonar equation is
PL=SL-N+DI-lOlog p';id+ ):(

- 1.05 1'2]
I - lOlogB+ lOlog(B/B,)+5logn
141
8.1 6 Worked Examples

Two passive sonar systems have been chosen to illustrate realistic performances
against torpedoes and submarines: (1) submarine-mounted flank arrays of length
8 m and height 4 m, and (2) a surface ship towed array, 322 at all frequencies by
selecting elements. The threat table (Table 8.2) will be used for both examples
and gives realistic radiated spectrum noise levels for a noisy, thermally propelled
torpedo (torpedo A); a quiet, electrically propelled torpedo (torpedo B); and a
typical submarine (submarine C).
Table 8.2 Threat table for Examples 8.1 and 8.2

Spectrum levels of radiated noise (a)
Torpedo A Torpedo B Submarine C

Broadband signal (Hz)
2000-4000 140 110 100
1000-2000 145 115 105
Narrowband signal (Hz)
400 155 125 .-
200 165 135 -

80 - -
130
40 - - 140
Example 8.1
The submarine passive sonar receives using flank arrays of length 8 m and height 4 m.
If the design frequency of the array is 2000 Hz then /2/2 = 375 mm and the array will
have 10 rows of 20 elements.
Assume the self-noise of the submarine to be equivalent to SS2. The background noise
will then be ambient sea noise or SS2, whichever is the greater. The DI will be given
by
At 3000 HZ +
DI = 3 10 log200 + 20 log 1.5 = 29 dB
At 400 HZ +
DI = 3 23 - 2010g5 = 12dB
At 200 HZ +
DI = 3 23 - 2O10g 10 = 6 dB
At 8 0 H z DI=3+23-2010g25 = 3 dB
At 40 HZ +
DI = 3 23 - 2010g50 = 3 dB
142 Passive Sonar
Because the flank arrays are baffled by the hull, DI cannot be less than 3 dB. The
performance of the system is given by the following tables. The ranges have been
calculated assuming spherical spreading plus absorption.
Narrowband
The narrowband equation is
(PL = SL - N + DI - Slogd - lolog B + lOlog(B/B,) + Slog IF + 5 log rzl

and the system performance is given in Table 8.3.
Table 8.3 Narrowband system performance for Example 8.1

Threat A Threat B Threat C
Tonal 200 400 200 400 40 80
SL 165 155 135 125 140 130
N 73 68 73 68 85 80
DI 6 12 6 12 3 3
5 log d 10 10 10 10 10 10
B 2 4 2 4 0.1 0.2
Bs 4 8 4 8 0.2 0.1
lolog B 3 6 3 6 -10 -7
10 log( B l B s ) -3 -3 -3 -3 -3 0
IF (= BT,) 4 8 4 8 1 2
5 log IF 3 5 3 5 0 2
n 30 30 30 30 60 60
5 log n 7 7 7 7 9 9
PL 92 92 62 62 64 61
R (km) 40 40 1.3 1.3 1.6 1.1
Broadband
The broadband sonar equation is
IPL = SL - N + DI - Slogd + Slog BT, + Slog nl

Worked Exumples 143
Table 8.4 Broadband system performance for

Example 8.1
SL 140 110 100
N 54 54 54
DI 29 29 29
5 log d 6 6 6
B 2000 2000 2000
re 2 2 10
5 log BT, 18 18 21
n 30 30 60
5 log n 7 7 9
PL ( a = 0.2) 134 104 99
x (km) > 100 50 40
Note that all three threats are detected by their broadband radiated noise at long
range but, except for the noisy torpedo (threat A), cZassiJication by narrowband
tonals is only possible at very close range. This is because, at the lower
frequencies of the tonals, the background noise is much higher and the DI of the
array is low (12 dB at best).
Ever quieter submarines and torpedoes are the stimulus behind the development
of towed arrays, which can achieve a good DI down to the very low fiequencies of
the residual tonals of modern threats. The second example will show how, even
when towed by a surface ship, towed arrays significantly improve classification
ranges.
Example 8.2
The towed array will be 321 at all frequencies by selecting elements. The background
noise at the array will be the ambient sea noise or the radiated noise of the tow ship at
the array, whichever is the greater. Note that a good towed array can be expected to
achieve self-noise levels equivalent to SS2. Take the spectrum level of the tow ship
radiated noise to be
140 dB from 40 to 400 Hz

130 dB from 1000 to 2000 Hz
The radiated noise at the array will be reduced by the separation from the ship (cable
length) and by the main lobe to sidelobe ratio of the array. Take the cable length as
600 rn and the ratio as -20 dB. Then, from 40 to 400 Hz, radiated noise at the array is
144 Passive Sonar
140 - 20 log 600 - 20 = 64 dB at operational array bearings

0 140 - 20 log 600 - 0 = 84 dB at ahead array bearings
And fiom 1000 to 2000 Hz, radiated noise at the array is
130 - 20 log 600 - 20 = 54 dB at operational array bearings

0 130 - 20 log 600 - 0 = 74 dB at ahead array bearings
Operational bearings for hull-mounted arrays are conventionally fiom R 135" around
to G135" (red for port, green for starboard), i.e., those bearings not directly affected
by stem noise. By the same token, operational bearings for towed arrays are from
R45"around to G45".
If the ambient sea noise is taken as SS2, the levels are as follows:
~~~ ~
Hertz 40 80 200 400 1500

Decibels 85 80 73 68 58
The directivity index will be constant with fiequency and given by
DI = 10 log 64 (64 elements spaced 1/2)
= 18 dB at all frequencies from 40 to 2000 Hz
For the array to be 321 at 40 Hz, its length must be
32 X1500
= 1200 m (using c = fA)
40
The performance of the system at operational bearings is given by the following
tables. The ranges have been calculated assuming spherical spreading plus
absorption. (Figure 3.3).
Narrowband
We have
[PL = SL - N + DI - 5 log d - 10 log B + 10log( B / B,) + 5 log IF + 5 log nl

Worked Examples 145
Table 8.5 Narrowband system performance for Example 8.2

Tonal 200 400 200 400 40 80
SI, 165 155 135 125 140 130
N 73 68 73 68 85 80
D1 18 18 18 18 18 18
5 log d 10 10 10 10 I0 10
H 2 4 2 4 0.1 0.2
23, 4 tl 4 8 0.2 0.1
10 log B 3 6 3 6 - 10 -7
10 log(B/ B s ) -3 -3 -3 --3 -3 0
IF (= BT,) 4 8 4 8 1 2
5 log IF 3 5 3 5 0 2
ri 30 30 30 30 60 60
5 log 12 7 7 7 7 9 9
PL 104 98 74 68 79 76
R (km) 160 80 5 2.5 9 6
Broadband
We have
IPL = SL - N + DI - 5 10gd + 5 log BT, + 5log T Z ~

Table 8.6 Broadband system performance for Example 8.2

~ ~~~~
Threat Threat A Threat R Threat C

SL 140 1 10 100
N 54 54 54
DI 18 18 18
5 log d 6 6 6
A 1000 I000 1000
re 2 2 10
5 log BT, 16 16 20
n 30 30 60
5 log n 7 7 9
PL 122 92 88
R (km) >lo0 30 20
146 Passive Sonar
Note that all three threats are now detected by their broadband radiated noise at
long range and also clussijied by narrowband tonals at useful ranges. This is
because, even at the lower frequencies of the tonals, the DI of the array is
maintained at 18 dB.
Over operational bearings, performance is limited by the ambient noise of the
sea and therefore the performance would be the same if the tow vessel were a
submarine.
8.17 Passive Ranging

When a passive sonar detects a signal from a radiating target, it gives a measure of
the bearing of the target and how it changes with time. The range of the target,
however, is not directly known; comparable signals at the array may result, for
example, fiom a distant noisy source or fiom a close quiet source.
Several techniques are exploited by passive sonars to estimate the range of a
target. They are particularly suitable for submarines whose bows and flanks
provide low noise sites for large arrays. A submarine is a very difficult target to
detect and therefore it will be reluctant to use active sonar, which immediately
advertises its presence, to determine the range of a contact that may have already
been detected and classified passively.
Triangulation 147
8.18 Triangulation
Given two well-separated arrays, the range can be estimated using simple
trigonometry. This technique, called triangulation, is not limited to sonar and is
best illustrated by a practical example. A submarine is in contact with a target at
range R from a flank array and from a towed array whose centre is S = 1000 m
behind the flank array. The maximum response angles (MRAs) are shown in
Figure 8.8.
Figure 8.8 Ranging by triangulation
From simple trigonometry, given one side and two angles, we have
1000 sin 145"

R= = 6581 m
sin 5"
This estimate is highly dependent on the accuracies of the bearing measurements.

If the beamwidths of both arrays are 4" then both MRAs could easily be in error
by f1" (one-quarter of the beamwidths), giving
1000 sin 146"

Rmax = = 10685 in
sin 3"
1000 sin 144"
k i n = = 4823 m
sin 7"
148 Passive Sonar
The accuracy of the range estimate is proportional to the angle subtended at the
target by the two arrays. Increasing the separation, S, of the arrays andor reducing
the range of the target increases this angle and improves the accuracy.
The arrays must be large enough for the beamwidths to be small at the
frequencies of interest. Furthermore, increasing beamwidths as steer angles move
towards endfire will limit the usefulness of the technique to angles not too far
removed from broadside.
Practical limits to the technique might be
Max R / S = 5
Max beamwidth = 5"
Max steer = f60" from broadside
The separation (or baseline) for measurements may be greatly increased by using
own vessel movement. This will greatly increase the baseline for the bearing
measurements. But the measurements are now also separated in time, and in
practice the target will also be moving. Initial assumptions are made for target
course and speed, e.g., constant values. Many mathematical procedures are avail-
able to refine these initial assumptions and the resulting range estimate will
improve with observation time. The technique is known as bearings only analysis
(BOA) and is clearly an implicit form of triangulation.
Vertical Direct Passive Ranging 149
8.19 Vertical Direct Passive Ranging

Measurements of the vertical angles of arrival at an array of a signal from a
radiating target and the time differences between signals reaching the array by
different paths can be used to estimate the range and depth of the target (Figure
8.9). The technique is known as vertical direct passive ranging (VDPR).
- - - - __ - --
R B
Figure 8.9 VDPR range estimation
Measuring the vertical angles of arrival to an adequate accuracy requires an

array with a good vertical aperture at all relevant frequencies. Flank and bow
arrays on submarines can therefore provide a complete VDPR facility but towed
arrays with no vertical directivity are restricted to measuring time differences by
autocorrelation only. The expressions used rely on the following assumptions:
Platform and target depths, p and t, are much less than the water depth, H
* The sea bed is flat at the bounce site
The range of the target in terms of arrival angle is given by the following
approximate expressions:
Bottom bounce only (B, solid path)
I R = (223 - p - t)/tan61
150 Passive Sonar
Bottom bounce and surface bounce (BS, dashed path)
SB (not shown)
SBS (not shown)
The vertical maximum response angle, 8, can be estimated by amplitude or by

phase:
Amplitude comparisons across a vertical fan of full beams (formed using

complete staves)
Phase comparisons between vertical half-beams (formed using top and bottom
halves of the vertical staves)
Example 8.3
Suppose the signal is contained within an octave from 1000 to 2000 Hz and the height
of the array is at least 5 wavelengths over this band, then from Figure 2.4, the vertical
beamwidth for steers up to 530"will be at most about 12". If we assume amplitude or
phase comparisons will improve on this by a factor of 4, then measurements of the
vertical arrival angle will be up to f 3 " in error.
For a simple bottom bounce only, H = 1000 m, p = 100 m, t = 0 (surface ship) and
8 = 30", the range is R = 3290 m. The limits (for 8 = 27" and 33") will be 3728 m
and 2925 m, or approximately f12%.
This might be seen as an acceptable result for a passive ranging system. But it does
depend fundamentally on the accuracy of 8, which in turn implies a narrow vertical
beamwidth. Unfortunately, detection - which must of course precede ranging - is
increasingly only likely at low frequencies where beamwidths will be large. Therefore
ranging using 8 only is likely to be very inaccurate.
The time diflerence of arrival, dT, between the two paths is given approximately by
Ertical Direct Passive Ranging 151
and for this example 6T = 67 ms. 6T is now used as an initial value for the off-
set between two 'windows' of time series data from the two paths shown.
Cross-correlation is performed between the two time series and the correlation peak is
used to improve on the initial value of 67 ms.
Suppose the improved value is 80 ms; using the above formula, 8 = 37" and the
range can now be determined from the earlier formulae:
For the downward-looking beam R = 1900/ tan 37" = 252 1 m

For the upward-looking beam R = 2100/ tan 37" = 2787 m
The average value is 2654 m.
There is an alternative method which can also yield an estimate of target depth.
Simple geometry in the vertical plane gives the following expressions for the sound
transit times for the four possible paths that include one bottom bounce:
1
Tss = ;[IP + (2H + p -- f)2] "2
and for this example, provided p and t are small compared with H , then 6T
= TBs - Ts is approximately
I I
If 6T = 80 ms then for this example R = 2667 m. This is very close indeed to the
average value obtained above (2654 m), by simply using the angle derived fi-om the
measured 6T in the coarse range formulae.
152 Passive Sonar
S S
Figure 8.10 VDPR depth estimation
Having obtained the range from one time difference, the depth of the target can be
estimated from the time difference between two other paths. If we take the difference
between the two paths in Figure 8.10,dT = TSBS- TB,then we obtain
Suppose 6T = 200 ms and the range, already obtained from different paths, is 2667 rn
then t = 150 rn.
Horizontal Direcf Passive Ranging 153
8.20 Horizontal Direct Passive Ranging

The curvature of the wavefront of a signal from a distant target determined by
measurements at three arrays can be used to estimate the range of the target
(Figure 8.11). The technique is known as horizontal direct passive ranging
(HDPR).
Target
,.+ f
/
+
Axis of arrays A
Figure 8.1 1 HDPR geometry
Consider three collinear equispaced arrays A , B and C. The target is at range R

and at an angle 8 with the axis of the arrays:
R2 = p2 + ( R - dr)
where d r is the additional path length due to curvature. We have p = dsine,

therefore
R2 = (dsin8)2 + ( R - (rr)2
and when R is large compared with d, we have

154 Passive Sonar
To find the range, we therefore need to determine dr and 8. For a distant target,
angle 8 is almost the same for all three arrays. In practice, determine 8 by
measuring and averaging the maximum response angles from the three arrays; dr
is given by
dr = BE - tCF
BE can be found by cross-correlating between arrays A and B, and CF can be

found by cross-correlating between arrays A and C.
The range estimate relies upon the accurate determination of dr, which is
dependent upon the precise measurement of the time differences between arrivals
at the three arrays. Therefore the positions of the arrays themselves must be very
accurately known in all three dimensions. The range estimate is less sensitive to
bearing errors, particularly at broadside where 8 = 90 degrees.
The delays measured are the sums of delays due to angle of arrival plus delays
due to wavefront curvature. On the beam, delays are largely due to curvature; close
to endfire, delays are almost entirely due to angle of arrival. The delays due to
curvature reduce as a hnction of sin2 8.
The method is optimized by making dr - the difference in path lengths due to
curvature only - as large as possible so that it can be accurately determined.This
occurs where A is large and is close to 90". In practice, d r needs to be some
significant proportion of a wavelength, say at least A/ 10, for an accurate measure-
ment, and for realistic values of A and array size, this limits the use of the
technique to quite high frequencies.
Target
fl
Figure 8.12 HDPR example

Horizonta/ Direct Passive Ranging IS5
Example 8.4
Three arrays, A , B and C, are mounted on the flank of a submarine (Figure 8.12).
Signals from a target at 30" to the axis of the arrays are detected in an octave centred
on 3000 Hz at all three arrays. The delays measured by cross-correlation are
Delay between arrays A and B = 0.0067 s

Delay between arrays A and C = 0.0133 s
Referring to Figure 8.1 1, we have
BE = 0.0067 c = 10.05 m
CF = 0.0133 c = 20 m
= 10.05 - 10 = 0.05 m
X = (20 sin 30q2/(2 X 0.05) = 1000 m
At 3000 Hz, the value of il is 0.5 m, so 6 r is a small but significant part of a

wavelength @/lo). If the target were actually f 3 " from its measured bearing, the true
ranges would be 1186 m or 824 m. (These are very similar errors as for VDPR.)
For a target on the beam (6 = 90") then, for the same 6v, we have I< =
(20 sin 90")'/(2 X 0.05) = 4000 m. Also, for the same bearing measurement errors,
the actual range would be 3988 m, i.e., a very small range error. Therefore, not only is
6 r largest for the same range when the target is on the beam, but also range accuracy
is least sensitive to bearing errors. The submarine can optimize the range estimate by
manoeuvring to put and keep the target on a bearing close to broadside.
In practice the arrays will not be equispaced and neither will they be aligned in the
remaining two axes. The full equations for range and bearing of a target will therefore
be three-dimensional and more complex, although based on the same simple
geometric relationships. The simple case for collinear, equispaced arrays has clearly
demonstrated the principle of HDPR and how a range estimate may be optimized.
I56 Passive Sonar
8.21 Towed Arrays

Towed arrays are essential for the detection of modem, quiet submarines and
torpedoes. This section discusses their construction, design and problems. A
simplified diagram of a towed array is shown in Figure 8.13.
Hose
Tail
VIM ~ o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o ~
Element 1 Element n
Figure 8.13 Simplified towed array
A typical towed array will comprise several hundred hydrophones, together with
electronic circuits to preampliQ, sample and digitize their outputs. The hydro-
phones are enclosed in a plastic hose, liquid filled to achieve neutral buoyancy in
the sea. The array is towed by a cable of a length decided by the speed of the tow
vessel and the desired depth of operation; it will be several hundreds of metres. A
vibration isolation module (VIM) reduces the tow vessel vibrations transmitted
down the cable, which would otherwise augment the noise of the array. The
assembly is completed by a tail whose purpose is to provide some drag to the
array so that it is maintained reasonably linear.
In practice the array will have significant curvature, which would affect the
beam shapes unless corrected in the beamformer. Heading sensors are placed at
intervals within the array and their readings used to correct for array curvature;
there will be at least three sensors (front, rear and middle) and possibly more in
very long arrays.
Towed arrays can have diameters as small as, say, 50 mm (thin arrays) and as
large as, say, 150 mm. The length of a towed array is determined by its fiequency
of operation and the desired DI. When used as a receive array in an active system,
the length might be from 10 to 50 m, whereas for a passive array the length might
be from 100 to 1000 m.
Passive towed arrays operate over a frequency range of several octaves. The
spacing between the elements is maintained at about A / 2 at the centre frequency
of each octave by a suitable choice among the available elements.
Example 8.5
Design a towed array covering 200 to 1600 Hz with a DI of about 16 dB at all
frequencies within this range.
DI = 10 log IZ = 16. Therefore n = 40 and each octave must use 40 elements spaced
ilJ2 at its centre frequency (Table 8.7). The complete array is shown in Figure 8.14.
The higher octaves are nested (a term possibly borrowed from computer program-
ming) within the lower octaves, and elements are chosen to meet the requirements of
each octave. To gain 3 dB in DI, or to operate at the next lower octave, the array must
be doubled in length.
Table 8.7 Design parameters for Example 8.5

Octave (Hz) fo (Hz) A/2 (m) y1 Length (m)
~~
200-400 300 2.5 40 100

400-800 600 1.25 40 50
800-1600 1200 0.625 40 25
Figure 8.14 Nested array

158 Passive Sonar
8.22 Bearing Ambiguity

A single line array is omnidirectional in the vertical plane and therefore when
horizontal beams are formed, they exhibit a Zefthight ambiguity.Figure 8.15 (top)
illustrates the source of the lewright ambiguity. There are several methods
available to resolve this ambiguity.
Course change
If the tow vessel, and therefore the towed array, changes its heading, it is possible
to resolve the ambiguity as shown in Figure 8.15. The array heading change is not
instantaneous and, particularly for a very long array, can take a considerable time.
Nor is the target stationary. Nevertheless, the true target bearing is often quickly
resolved, but may need confirming by making some assumptions about the targets
motion.
Array heading
Change to array heading
No
Figure 8.15 LeWright ambiguity

Twin urrays
Parallel twin towed arrays use the time delays between the signals arriving at the
two arrays to resolve the lefthight ambiguity. Maintaining the spacing between
two flexible arrays is a practical problem, particularly for very long arrays and
during a change of course. Precision, however, is not necessary provided some
horizontal spacing survives and the arrays do not cross over. If both arrays are also
used to form beams, an improvement in DI of up to 3 dB will result over a limited
frequency range (particularly usefid when the twin array is used as a receive array
in an active system which will have a comparatively small percentage bandwidth).
Triplets
In a triplet array, each element now comprises three hydrophones - a triplet - in

the vertical plane (Figure 8.16). Because time delays are measured between all
three pairs, the lefthght ambiguity can be resolved regardless of any rotation of
the array. The method needs a fairly large array diameter to house the triplets and
to produce measurable time delays.
Figure 8.16 Hydrophone triplet
8.23 Self-noise
Towed arrays are well separated from tow vessels and therefore the vessel radiated
noise is significantly reduced (given a cable length of 500 m, the value of 20 log R
is 54 dB) and, except for the ahead bearings of the towed array, this is further
reduced by the main lobe to sidelobe ratio of the beams (perhaps 20 dB). The
160 Passive Sonar
hydrodynamic noise of the towed array can be made negligible at normal tow
speeds of up to about 12 knots, and therefore the remaining and dominant noise is
the ambient noise of the sea.
8.24 Problems
8.1 The dimensions of the flank arrays of Example 8.1 are halved. How are the
narrowband and broadband performances affected?
8.2 Refer to Figure 8.9. The platform depth, p, is 300m, the target depth, t, is
unknown but small compared to the water depth, H, of 2000 m. If the time difference
between the paths shown is 100 ms, what is the range of the target?
8.3 Design a towed array to have a DI of 21 dB at the centre frequency of the octave
from 2000 to 4000 Hz. What would be the spectrum level of the radiated noise of a
target which is first detected at a range of 10 km against a background noise equivalent
to SS4? Assume broadband detection and 5 log d = 6 dB, T, = 10 s, n = 60. Propaga-
tion is spherical spreading plus absorption.
9
Active Sonar
9.1 Range, Pings and Doppler Shift

Active sonars use an array of projectors to transmit acoustic pulses into the water.
Underwater targets are detected, localized and classified by the echoes resulting
from their insonification by these pulses.
The time of echo arrival at the receiver array is used to determine the range of a
target (strictly, the slant range). The time, measured from transmission, is the time
taken for the sound to propagate to the target and back to the array. Hence the
range is given by
An active sonar transmission is known as a ping. A ping may be simply the

transmitted pulse or sequence of pulses, or it may be the total time between
transmissions, i.e., the sum of the duration of the pulses and the receive period.
The term ping is therefore ambiguous but the meaning is usually made clear by
the context in which it is used.
The ping interval is an alternative term for the time between transmissions and
is the reciprocal of the ping (or pulse) repetition rate (PRR). The frequency shift
of an echo is a measure of the relative velocity (doppler) of a target with respect to
the sonar platform. The total shift is the sum of four components (Figure 9.1):
162 Active Sonar
Figure 9.1 Components of doppler shift
A doppler shift to the transmitted pulse, equal to S, cos tIt

A doppler shift to the pulse received by the target, S cos cp
An equal doppler shift to the echo transmitted by the target, S cos cp
A doppler shift to the echo received by the platform, Sr cos 6 r
The first and fourth components are normally removed by own doppler nuZZiJica-
tion (ODN). The total frequency shift due to the target doppler, S cos cp, is given
by
lAf 2s cos cp
=flC
where f is the operating frequency. The factor of 2 results from the two equal
doppler shifts due to target motion (the second and third components).
Practical formulae for doppler shift are, for c = 1500 m/ s,
Af = f0.69 Hz per knot per kHz

Af = f1.33 Hz per m/s per kHz
Note that 1 m/s = 1.945 knots. And remember that doppler is a shift of relative
velocity not speed. For a passive sonar, only the third and fourth components are
relevant. The frequency shift is positive for an approaching target but may be
changed by heterodyning in the receiver.
C W Processing 163
9.2 Pulse Types

Active sonars know more about the signal to be detected, and therefore the
receiver is designed to match the signal, i.e., it uses matchedjilterprocessing. But
the background against which the signal has to be detected, contains reverberation
in addition to the ambient and self-noise of passive sonars. This additional
background is all important in active systems and the designer must be aware of
its magnitude and how to discriminate against it.
Active sonars employ two broad classes of pulse types:
* Continuous wave (CW): a pulse of constant frequency and duration T seconds.

The bandwidth of the pulse, and of the matched filter for optimum detection of
this pulse, is 1/ T Hz.
Frequency modulation (FM): the frequency of the pulse changes during the T
seconds duration of the pulse. The bandwidth, B , is not now the inverse of the
pulse length.
The replica correlation process used to detect an FM pulse is effectively a matched

filter. Independent outputs from the correlator occur every 1/ B seconds, i.e., at the
resolved pulse length.
9.3 CW Processing
CW processing is similar to narrowband passive processing where, after beam-
forming, the data is analysed using a fast Fourier transform (FFT) process and
then displayed to an operator andor input to an automatic detection and classifica-
tion system. Traditionally, active CW processing used a bank of analogue filters,
each matched to the bandwidth of the CW pulse. Sufficient filters were provided to
cover the expected frequency shifts due to doppler. In modern sonars the analogue
filters are replaced by an FFT processor.
164 Active Sonar
9.4 FM Processing
FM processing replaces the FFT by a process where the output of the beamformer
is correlated with a replica of the transmitted pulse. The signal processing for both
classes of pulse is essentially matched filter processing and the same expression
for signal processing gain therefore applies to both.
The signal-to-noise gain fiom replica correlation arises from a reduction in the
mean of the noise because, unlike the signal, it is uncorrelated with the replica
pulse shape. The DT equation assumes that the noise is completely uncorrelated
with the transmitted pulse and therefore the noise power is reduced by the number
of independent samples combined in the correlation process, which is BT Thus
power - 1 (mean noise power

after correlation BT before correlation
and the DT equation from this process is
IDT=510gd- 10logBTJ
Note that the gain is 10 log BT compared with 5 log BT for broadband passive
sonars. This is because replica correlation is a coherent process (complete know-
ledge of the signal) whereas broadband processes are incoherent (nothing known
of the signal).
A coherent gain is equal to the number of independent samples = BT
0 An incoherent gain is equal to the square root of the number of independent

samples = (BT>'/*
Active Sonar Equations 165
9.5 Active Sonar Equations

There are two active sonar equations; one is used to determine the performance
against a background of noise and the other against a background of reverbera-
tion. Although some models combine the two backgrounds to estimate perform-
ance, it is preferable to calculate noise and reverberation performance separately
for a better understanding of the effects of changing equipment parameters and
operating in different environments.
The noise-limited active sonar equation is
SE = SL + TS - 2PL - (N- DI + lolog B ) - (5logd - lolog BT - 510g n )
+
where ( N - DI 10 log B ) is the in-beam noise over the h l l bandwidth and
(5 log d - 10 log BT - 5 log n) is the detection threshold. Here y1 is the number of
pings used by the operator or automatic system to make the decision; in a passive
system, n is the number of lines used.
Putting SE = 0 and combining terms:
I2PL = SL + TS - N + DI + lolog T - 5 logd + 5 log n I

It is important to note that the noise-limited performance is independent of
bandwidth (given matched filter processing). Therefore, FM and CW transmis-
sions of equal duration have identical noise-limited performance.
The above equation was derived assuming perfect matched filter processing and
may require modification as follows:
CWpulse: If the analysis bandwidth is greater than the pulse bandwidth there will
be a loss (due to the increase in noise): If the analysis bandwidth is smaller than
the pulse bandwidth there will also be a loss (due to the loss of some of the signal).
The loss is given by a mismatch term:
either, - 10 log( B,/B,) when B,, > B, or, - 10 log( B,/B,) when B, < B,
LPFM Pulse: If the replica is extended (to allow for target Doppler) there will be
a constant loss (due to the increase in noise) for all dopplers of
166 Active Sonar
If the replica is not extended but there is a mismatch (an incomplete overlap) with
the signal return due to target doppler, there is a variable loss given by
where Bc is that part of the signal which overlaps the reference. Note that these
mismatch terms are always negative, so that - when added to the RHS of the
equation - they always reduce 2PL.
The reverberation-limitedactive sonar equation is
+
where (SL - 2PLR TSR) is the in-beam reverberation over the full bandwidth
and (5 log d - 10 log BT - 5 log n) is the detection threshold.
When the propagation paths for echo and reverberation are identical, 2PL
= 2PLR and these terms will cancel out.
TSR,the target strength for reverberation, is given by
The dominant reverberation source is usually at the sea surface, the sea bed or in a
horizontal layer within the water column (see Chapter 6) and therefore
and putting SE = 0 gives
10 log R = 1010g(l/T%) - Sb - 41 + TS - 5 log d + 10log BT + 5 log n

and combining terms:
I 10 log R = 10log(B/&) - St, - 41 + TS - 5 10gd + 5 log n I

Note that the 41 dB arises fiom
101og(cn/360) + 30 (to give R in km) = 11 + 30 = 41 dB

Active Sonar Equations 167
The term 10 log( B / O h ) is known as the reverberation index (RI) and is a useful
measure for comparing performance against a background of reverberation.
For a CWpulse, B = 1/T, where T is the pulse length.
For an FMpulse, B = 1/Tr, where T' is the resolved pulse length, or the time
between independent output samples from the correlator.
Example 9.1
CW pulse: T = 100 ms, 8, = 10"
RI = 1010g(10/10) = 0 dB
Example 9.2
FMpulse: B = 300 HZ 4 = 10"
The actual pulse length is irrelevant and RI = 10 log(300/10) = 15 dB.
The wider bandwidth of the FM pulse therefore results in a much improved perform-
ance against reverberation. Looking at it another way, the effective reverberating area
is determined by the bandwidth and not by the pulse length.
If the dominant factor is voEurne reverberation resulting fiom backscattering fiom that
part of the water column which is actually included in both transmit and receive beam
patterns, then the equation becomes
(2010gR = 10log(B/&.8,) - S, - 23 + TS - 5 logd + 510g nJ

Note that the 23 dB arises from
lOlog (i: I$) + 30 (to give R in km) = -7 + 30 = 23 dB

168 Active Sonar
9.6 Reverberation Index

The reverberation index is a measure of the effectiveness of the sonar system
against reverberation. The index may be increased for all pulse types by reducing
the effective horizontal beamwidth. The receive array must therefore be large
(many wavelengths) in the horizontal dimension.
Increasing the pulse bandwidth, B, will also increase RI but, for a simple CW
pulse, where increasing the bandwidth is only possible by reducing the pulse
length, T , the noise-limited performance is adversely affected. This, however, had
to be the approach in early sonars where the complexities of FM were beyond the
available technology. Long and short CW pulses would be used in the same ping.
Modern sonars employ two standard pulse design approaches to improve perform-
ance against reverberation:
A wideband FM pulse which spreads the reverberation power over the band-
width, B, of the pulse. When the signal is correlated with a replica of the
transmitted pulse - effectively coherent matched filter processing - the pulse is
compressed into an equivalent time - the resolved pulse length, Tr, which is
equal to the inverse of B. The area of the surface contributing reverberation is
therefore reduced and performance against reverberation is improved as B is
increased, increasing B in 10 log( B/&). Furthermore, the performance of the
wideband pulse is independent of target doppler - provided there is still an
adequate match with a replica.
A long, shaped CW pulse is used to produce a reverberation spectrum with a

narrow peak centred on zero doppler whose amplitude falls off rapidly with
increasing magnitude of doppler. Target doppler will ensure that target echoes
fall into regions of the reverberation spectrum where the reverberation power is
low and the performance eventually becomes noise limited only.
An alternative broadband pulse is the pseudorandom noise (PRN) pulse. Its

bandwidth will be close to or identical to that of an FM pulse designed for a
similar performance. The frequency-time structure of the pulse may be changed
in a random but known manner from ping to ping. Hence replica correlation will
still be possible.
9.7 FMpulses
In the strict radio communications sense, sonar FM pulses are not frequency
modulated at all. They are simply wideband pulses where the frequency changes
throughout the duration, T, of the pulse (Figure 9.2).
Frequency Reference Frequency

I
f
Time Time
Figure 9.2 Sonar pulses: (a) LPM and (b) LFM
Linear period FM (or h-yperbolicFM,
The periodicity of the pulse changes linearly with time (i.e., its frequency changes
hyperbolically with time). Therefore the echoes, shifted in frequency by target
doppler, will always correlate with a suitably extended replica. The extended
replica will result in a constant loss, perhaps as much as 2 or 3 dB, for the entire
range of target doppler and it is often preferable to use a replica equal to the pulse
bandwidth and to accept a loss increasing with target doppler. No doppler
information can be extracted from this pulse.
Linear FM
The frequency of the pulse changes linearly with time. The echoes shifted by
target doppler will not now fully correlate with an extended replica and a set of
replicas (analogous to the comb of filters or FFT cells used to process CW echoes)
are now required to cover the expected target dopplers. The best replica match
indicates the target doppler.
170 Active Sonar
Range error with target doppler
Target doppler moves the instant in time when h l l correlation of the echo and
replica takes place. There will only be an error when a single replica is used (the
LPFM case). To a first order, the time error is
At = (T/B)Af
where A j is the frequency shift due to target doppler, and
cT
Range error = - -Af (note the active 2)
2 B
Example 9.3
If T = 500 ms and B = 250 Hz,what is the error for 10 knots of doppler at 10 kHz?
Af = 0.69 X 10 X 10 = 69 HZ
1500 0.5
Range error = - -X 69 = 104 m
2 250
The target doppler may be determined by range rate calculations from two or more
pings or from a simultaneous CW transmission.
cw Pulses 171
9.8 CW Pulses
A rectangular shaped pulse is defined in the time domain as
A sin2n: ft for f f j d T / 2
otherwise
and the frequency spectrum of the pulse, normalized to 1 s and 1 kHz, is shown in
Figure 9.3. The 3 dB bandwidth is given by
0 4 8 12 16
Frequency (Hz)
Figure 9.3 Rectangular pulse spectrum

172 Active Sonar
A Hamming shaped pulse is defined in the time domain as
Ar0.54 + 0.46 cos(2nt/T)]sin 2nft for I tl d T / 2

otherwise
and again its frequency spectrum, normalized to 1 s and 1 kHz, is shown in Figure
9.4.
The 3 dB and 40 dB beamwidths are given by
-801 I I I I '
0 4 8 12 16
Frequency (Hz)
Figure 9.4 Hamming pulse spectrum
Note the significant reduction in sidelobe levels which is achieved, but at the
expense of a 50 per cent increase in the 3 dB bandwidth.
Figures 9.3 and 9.4 are obtained by analysis using a rectangular window. If
pulse and window have the same length, it is possible to exchange pulse and
window weighting functions without changing the final observed pulse spectrum.
In other words:
A shaped pulse analysed using a rectangular window has the same spectrum as
a rectangular pulse analysed using a shaped window.
CW Pzilses 173
The frequency spectrum of a Hamming shaped pulse analysed using a Hamming

shaded window, again normalized to 1 s and 1 kHz, is shown in Figure 9.5. The
3 dB and 40 dB bandwidths are given by
Af3 =-
dB Of I
-70,
-80- I I I 1
(J
0 4 8 12 16
Frequency ( Hz)
Figure 9.5 Hamming pulse: hamming window spectrum
A further reduction in sidelobe levels is achieved, but the 3 dB bandwidth is now

almost doubled compared to the equivalent rectangular pulse analysed using a
rectangular window.
All these spectra are theoretical. The need to tune projectors will 'knock the
corners' off a rectangular pulse, making the sidelobe levels lower than predicted;
and the limitations of electronics and digital processing will make it impossible to
achieve the very low sidelobe levels predicted for shaped pulses. It would be
unwise to expect sidelobes more than about 40 dB down on the main lobe.
174 Active Sonar
9.9 Reverberation Rejection by CW Pulses

As target doppler increases, detections take place against a reduced background
of reverberation, finally becoming noise limited. To reduce the possibility of
detection, a submarine will always attempt to minimize the magnitude of its
doppler as seen by a sonar receiver.
To achieve a good performance against reverberation, therefore, the CW pulse
must be designed to maximize reverberation rejection, Rj, at the target dopplers
of interest. Pulse bandwidth and sidelobe levels are the parameters which
determine the magnitude of Rj. Reverberation will be present in both the main
lobe and sidelobes of a beam, and this reverberation is modified by platform
movement. The importance and interplay of all these factors will now be consid-
ered.
Figure 9.6 shows the frequency spectrum of a 1 s Hamming pulse analysed
using a Hamming shaded window. It shows the reverberation rejection resulting
from target doppler for a carrier frequency of 10 kHz. The dotted envelope is that
of a rectangular pulse and demonstrates the importance of pulse and window
weighting to improve CW performance.
Rj (a)
- 20
I s, 10 kH2 pulses
-40
-60
I I I I I I I
-1.2 -0.8 -0.4 0 0.4 0.8 1.2
Target doppler (knots)
Figure 9.6 Reverberation rejection, R, versus target doppler
For example, -40 dB of R j is reached for 0.4 knots of target doppler using a
1 s Hamming pulse and window, but this is never reached using a rectangular
pulse for realistic submarine dopplers.
Figure 9.6 portrays a stable and stationary platform in a stable environment. If
this is not the case, the background reverberation in the main lobe and sidelobes of
a beam will be modified as follows.
Reverberation and Target Echoes in the Main Lobe 175
9.10 Reverberation and Target Echoes in the Main Lobe

Beamwidth
It is only possible to fully nullie own doppler at one bearing within a beam. The
finite beamwidth has two effects:
0 It increases the doppler spread of the reverberation within a beam. The

spreading is independent of pulse length and sea state. Echoes therefore need
more doppler to avoid the reverberation background.
There will be a related uncertainty in the measured target doppler.
If we assume a beamwidth of lo", the spreading and doppler errors in knots at

various relative bearings are as in Table 9.1
Table 9.1 Doppler spreadderrors due to beamwidth

Doppler spreadderrors (knots) Platform speed (knots)
Target on bow Target at 45" Target on beam

-
A0 f O Zt0 0
$10.04 h0.6 -1-0.9 10
ct0.06 i0.9 il.4 15
-1-0.08 f1.2 -1-1.8 20
Ship motion
Roll and pitch have little effect on reverberation spread but contributions due to
yaw are significant. NWS 1000 (UK Naval Weapons Specifications Publication)
gives a figure for yaw in large frigates and destroyers as 1.75" s - ~peak. (Yaw is
unlikely to be important in submarines or for towed arrays, except perhaps for
very long pulses.) If the array is 60 m from the centre of gravity then the linear
acceleration at a bow dome is 2 m/s2 and, by inference from other NWS 1000
figures, the doppler spreads for 1 s and 250 ms pulses are as in Table 9.2. These
spreads occur both at transmit and receive.
176 Active Sonar
Table 9.2 Doppler spreadderrors due to ship motion

Sea state Pulse length Doppler spreadderrors (knots)
(ms)
Target on bow Target at 45" Target on beam
1000 0 1.8 2.7
1000 0 0.9 1.3
250 0 0.4 0.6
250 0 0.2 0.3
Environment
In shallow water, bottom reverberation can produce spreads of between 0.2 and
0.5 knots. In deep water, surface reverberation is important and depends on wind
speed. A useful empirical formula is
IDopplerspread = 0 . 0 7 ~ 0.31 +
with doppler and wind speed, v, in knots. And, substituting for the appropriate
wind speeds, we obtain Table 9.3.
Table 9.3 Doppler spread due to sea state

~~~~ ~ ~ ~ ~
Sea state 2 4 6
Doppler spread (knots) 1.2 1.6 2.3
Total reverberation
The total reverberation spread in the main lubes results from the pulse spectrum
plus the three factors we have looked at:
Beamwidth
Ship motion
Environment
Their relative importance is highly dependent on the sonar platform. Beamwidth

Reverberation and Target Echoes in the Main Lobe 177
and ship motion are irrelevant to a stationary platform, such as a sonobuoy or a

helicopter dipping sonar, but the environment affects all platforms. Combining all
the factors can result in quite large, worst case, values for reverberation spread.
The likelihood of all factors coinciding, however, is not great and recommended
practical figures for a first pass at a problem are as follows:
Sonobuoy or helicopter f 0 . 5 knot
Towed arrays (ship or submarine) f 0 . 7 knot
* Hull-mounted sonar, submarine f 1.O knot
Hull mounted sonar, surface ship f1.5 knots

178 Active Sonar
9.11 Reverberation and Target Echoes in the Sidelobes

The reverberation (including target echoes) which enters a receiver through the
sidelobes of a beam is modified by an ODN frequency appropriate to the main
lobe axis, and for full-beam processing this produces bands of reverberation,
reduced only by the main lobe to sidelobe ratio (20-30 dB) appearing in the main
lobe. This is a performance degradation apparent in many beams, or given an
omnitransmission, in all beams; it can produce bands of reverberation which can
affect the detection of targets with dopplers up to twice platform speed.
Figure 9.7 shows the mechanism of spreading due to platform speed. For a
target in the ahead beam and a platform speed of v knots, both reverberation and
target echoes received in the ahead beam are shifted by a frequency equivalent to
v knots. An ODN frequency equivalent to v knots restores the reverberation in the
ahead beam entering through the main lobe to a spectrum centred on zero knots.
The reverberation entering through the sidelobes of the ahead beam, however, is
also shifted by v knots and therefore the ahead beam output has a series of
secondary reverberation spectra reduced by the main lobe to sidelobe ratios and
centred on various speeds around to -2v knots for reverberation entering through
the astern sidelobe.
v knots
t
Doppler Platform
Ahead 'y;
Beam 0 -+ -v
Astern - -v+ . -2v
Omni transmit beam pattern
Figure 9.7 Reverberation spreading due to platform speed

Pructical Reverberation Envelopes 179
9.12 Practical Reverberation Envelopes

By superimposing typical beam patterns onto the pulse spectrum, we can produce a set
of reverberation envelopes for different platform speeds and target bearings (beams).
Figure 9.8 shows an envelope for a broadside beam and a platform speed of 10
knots. Note that, far from being noise limited, detections can take place against
high levels of reverberation out to target dopplers of k10 knots. The pulse lengths
are necessarily long (for good noise-limited detection ranges and also for opera-
tion down to low magnitudes of doppler). The long pulse, however, results in a
very low RT. Therefore, except against very low reverberation, detection perform-
ance can be poor until the target doppler exceeds platform speed (at broadside, but
up to twice platform speed for the ahead and astern beams).
- 12 -- 8 -4 0 4 # 12
Target doppler (knots)
Figure 9.8 Typical reverberation envelope
Fortunately, this effect is only a problem when using full-beam processing.

When half-beam processing is used, the phase differences obtained by compari-
sons between pairs of codirectional half-beams can be used to gate out the
reverberation and target echoes in the sidelobes which, for a moving platform,
would otherwise form a significant part of the detection background (i.e., it would
be dominant over much of the likely target doppler spectrum).
Therefore, for half-beam processing, which will now be considered in detail,
we effectively revert to backgrounds represented by the pulse spectrum unmodi-
fied by beam patterns, e.g., the Hamming pulse in Figure 9.5, although the main
lobe spectrum will need to be broadened to allow for the effects of beamwidth,
platform motion and the environment.
180 Active Sonar
9.13 Full- and Half-Beam Processing

Full-beam and half-beam processing techniques are compared by postulating an
active sonar using a simple omnidirectional, in azimuth, projector for trans-
mission, and a 16-element (2 X 8 half-aperture) line array of hydrophones for
reception, and having the parameters listed in Table 9.4. The sonar uses both
LPFM and CW pulses. Identical pulse lengths and source levels are chosen so that,
given matched filter processing, both pulse types have identical noise-limited
detection performance.
Table 9.4 Sonar parameters

~~~ ~
Parameter Value Units

Frequency 10 kHz
Pulse length, T 1 S
Source level, SL 2 10 dB
FM bandwidth 400 HZ
CW bandwidth 2 Hz
Beamwidth, full aperture 8 degrees
Beamwidth, half-aperture 16 degrees
Phase bin width (2 bins) 2 degrees
Reverberation index, RI
FM, full beams 17 dB
FM, half-beams 23 dB
cw -6 dB
DI (receive) 12 dB
Integrated target strength, TS 10 dB
5 log d I0 dB
5 log n 3 dB
Background noise level 50 dB
The performance of the two pulse types is then assessed for both full-beam and
half-beam processing and the following benefits are shown to result from half-
beam processing. The last two items are applicable to both FM and CW pulses.
Enhanced FM detection performance for all doppler values
Noise-limited CW detection performance once the doppler spectrum of the

reverberation within the main lobe of a beam is exceeded
Beamforming 181
Improved bearing accuracy
Elimination of echo returns in adjacent beams: simpler autodetection and

tracking processes and reduced operator workload
Improved bearing accuracy
Amplitude comparisons between full beams give, at best, resolutions or accuracies

of around I /4 of the full aperture beamwidth, whereas phase binning can achieve
I / 16 of the half-aperture beamwidth, with a practical lower limit of 1". Autodetec-
tion and tracking processes clearly benefit from this improvement: fewer false
associations are made, resulting in fewer false alarms; weapon guidance predic-
tions are significantly enhanced; and weapon target acquisition rates are improved.
9.14 Beamforming
Segments of circular or cylindrical arrays are effectively reduced to equivalent
linear or planar arrays by applying time delays to the elements, hence a short line
of 16 elements spaced A / 2 is a useful and simple array with which to demonstrate
the principles of full-beam and half-beam techniques.
Full beams are fonned by the simple addition of all 16 elements, after shading
and time delays, to form the required set of beams. (For a circular array the time
delays are constant and different sets of elements are used to step around the array
in azimuth.) The resultant full beamwidths will be of the order of 8" and spacing
between beams is, in practice, somewhat less than this, to minimize scalloping
losses. For simplicity, we will ignore scalloping loss and make the spacing equal
to the full beamwidth (8").
Pairs o f codirectional half-beams are formed using the 2 X 8 elements of the
line in a similar manner to the full-beam case. Both left and right half-beams now
have widths of 16" and the spacing between each pair is also 16". The beam shapes
are shown in Figure 9.9.
182 Active Sonar
half-beams
\ \
/
/ / \
120"
60"
-0"
- 60"
- 120"
- 16" - 8" 0 8" 16"
Full beam
A \
f
Left half-beam Right half-beam
Receive line array 2 X 8 elements total

h/2 spacing. Hamming shading
Full beamwidth 8" and all 16 elements used
Half-beamwidth 16" and 8 elements used
Figure 9.9 Half-beam and full beam: amplitude and phase plots
FM Phase Binning Process I83
9.15 FM Phase Binning Process

Each pair of half-beams is processed as shown in Figure 9.10. The half-beam
outputs are correlated against a replica of the transmitted pulse and the correlator
outputs are bulk steered and added to form two full beams, together covering the
same azimuth as the original codirectional half-beams (Figure 9.9). Phase com-
parison between correlator outputs is used to select the appropriate (left or right)
full beam. The output of the selected full beam is then steered into one of 16 phase
bins which together cover 16" of azimuth. (The left full beam output, if selected,
will be steered into bins 1 to 8 and the right full beam into bins 9 to 16.)
Bulk steer and add to Select full beam and

form two full beams phase; bins
f - - - - A 7 r2
Left
half.
Phase bins
Select L/R full beam
Rig1
half.
16
Figure 9.10 LPFM signal processing for one pair of codirectional half-beams
The target appearing in one phase bin, or overlapping into two phase bins (and
here we will always assume two phase bins), is thus detected against a background
- noise or reverberation - reduced by the ratio of the half-beam width to the width
of two phase bins. The reverberation index is thus improved by 6 dB (8/2)
compared with the full-beam output that would otherwise be used.
Integration of the independent samples (every 1/B) from the phase binning
process will give a post-detection gain until the integration time exceeds the target
extent. If we assume the minimum target extent to be 15 m, we would sample and
reset the phase bins every 20 ms and
PD gain = 5 log(BTinty1) = 5 log(400 X 20 X = 4.5 dB
Be careful with this PD gain when calculating the performance of a broadband

system using the sonar equations. It is not appropriate if an integrated TS value is
used in the sonar equation (the integration is then 'in the water'). The peak TS
184 Active Sonar
value, which is perhaps about 6 dB less, should be used. In view of the uncertainty
surrounding TS measurements, although a PD gain certainly occurs, it is perhaps
better to err on the side of caution and ignore it when making performance
calculations.
9.16 CW Processing
Each pair of half-beams is processed as shown in Figure 9.1 1. The half-beams are
frequency analysed to form N doppler channels. Left and right full beams are then
formed for each doppler channel by a similar bulk steer and add process. Phase
Bulk steer and add to form two full beams

A
Left
channel
Phase Select L/R full beam

compare Doppler
channel 1
Doppler
channel 2
Left
half-beam Doppler
channel
(filter) N
Phase Select L/R full beam

compare 1 Doppler
ichannel .V
Right
half-beam '
--+ channel
I (filter) N 1 U U
Figure 9.1 1 Signal processing for one pair of codirectional half-beams
Lurge Aperture ilrruy 185
comparison is again used to select the correct full-beam outputs from each doppler
channel. These outputs are combined with the measured fine bearings (within a
beam) and passed to data processing and display at a rate depending on the CW
pulse length (e.g., every 250 ms for the 1 s pulse). There is no possibility of using
phase bins for post-detection integration because the length of the CW pulse is
already comparable with, or much greater than, the dimensions of the target. The
long CW pulse has effectively integrated the echo energy in the water and
therefore this time the integrated TS value should be used in the sonar equation.
However, there are still positive advantages in forming half-beams and deter-
mining target bearing by phase comparisons. The CW pulse will benefit from
improved bearing accuracy and the elimination of echoes from adjacent beams in
exactly the same manner as for the FM pulse, There is also an improvement in the
CW detection performance, i.e., it becomes noise limited, immediately the doppler
spectrum of the reverberation within the main lobe of a beam is exceeded, as
explained in Section 9.12
9.17 Large Aperture Array

Figure 9.12 shows the beam pattern of one of a pair of codirectional half-beams
formed using half the aperture of a 32A array of 64 elements, together with the
phase difference plot resulting from the complex cross-correlation of the pair of
half-beams. Hamming shading is used. The f 2 " beamwidth of each half-beam
corresponds to a phase difference of about .f120". By using the bulk steer and
phase binning process described earlier, phase differences greater than this can
eliminate all returns from not only the sidelobes but also the skirts of the main
lobe. Note that the phase difference remains greater than f120" for all values of
azimuth beyond the 3 dB points of the main lobe. In practice, imperfections in the
array, and unwanted curvature in the case of a towed array, may cause re-entrants
to occur in the phase difference plot beyond the i120" limits which will result in
returns through sidelobes. To minimize this, the sidelobe levels should be reduced
by beam shading - Hamming shading in this example.
186 Active Sonar
-33-
-36-
-39-
-42-
-45-
-48-
-80 -60 -40 -20 0 20 40 60 80

(a) Azimuth (degrees)
Phase difference (degrees)

180
150
120
90
60
30
0
-30
-60
-90
- 120
- 150
-180
-80 -60 -40 -20 0 20 40 60 80
(b) Azimuth (degrees)
Figure 9.12 (a) Half-beam response and (b) phase difference plot
Detection Performance
9.18 Detection Performance

Noise-limited performance
The noise-limited performance is identical for both pulse types and is given by
2PL == SL + TS - N + DI + lolog T - 5 10gd + 5 log TZ
and using the parameter values of Table 9.4:
2PL=210+10-50+12+0-10+3= 175dB
(The integrated TS value is used for both pulse types and, for the FM pulse, this is
justified by ignoring the PD gain due to the phase binning process.) If we assume
spherical spreading and absorption, the range is 9 km.
Reverberation-limited performance of the LPFM pulse
The reverberation-limited performance of the LPFM transmission (or any broad-

band pulse) does not change with target doppler, including zero, and is given by
Using the parameter values from Table 9.4, the LPFM reverberation-limited
ranges are given in Table 9.5 for three representative values of s b .
Table 9.5 LPFM reverberation-limited range

Reverberation-limitedrange (km)
Sb = -30 Sb = -37 Sb = -40

Full-beam processing 8* 40 80
Half-beam processing 32 160 320
"Note that, in high reverberation conditions, the noise-limited detection range, 9 km. is only achieved by
using half-beam processing
sb = -30 dB, a high bottom Sb
Sb = -37 dB, 75 per cent of UK inshore waters have sb less than this value
Sb = -40 dB, a low bottom, high surface value for S b
188 Active Sonar
At useful detection ranges (greater than say 5 km), submarines never subtend
angles greater than I", hence any processing and display scheme based on full
beams - which are inevitably at least a few degrees wide - will have no fine detail
in the bearing dimension to help with detection and classification.
For noise and uniform reverberation, where bearing extents are similar to the
submarine's, detection is not improved by using half-beam processing. But the
greater bearing accuracy and elimination of sidelobe returns will simplifL and
improve all subsequent processes.
For discrete reverberation, where bearing and bearing extents can differ
markedly from those of a submarine, detection is improved by using half-beam
processing.
ClassiJicationand display, against any background, are enhanced by the avail-

ability of fine bearing (phase bins) providing an extra dimension for echo shape
recognition and improved bearing accuracy for tracking.
Reverberation-limitedperformance of the C W pulse
The reverberation-limited performance of the CW transmission varies with

doppler and is given by
I10 log R = RI - S b - 41 + TS - 5 log d + 5 log t~ - Rjl

Note the additional term, Rj; this is the reverberation rejection and it determines
how the performance of the CW pulse varies with doppler. ( R j is zero for FM
pulses.) We are effectively calculating the detection range of a target with zero
doppler and then multiplying it (adding dB) by some factor determined by Rj.
For the CW transmissions, we are less interested in the reverberation-limited
range and pose the question, How much doppler before the detection is noise
limited? To answer this, assume the reverberation-limited range is 20 km (i.e.,
always longer than the noise-limited range) giving 10 log R = 13 dB. Then solve
the reverberation-limited equation for R j, again using the parameters of Table 9.4.
The target dopplers required to achieve the R j values (Table 9.6) are then obtained
from Figures 9.6 and 9.8, but adding another 0.7 knot as suggested in Section 9.5.
Note that the doppler magnitudes required using full-beams are for a moderate
platform speed, 10 knots and assume quite low sidelobes (20 to 30 dB down). In
practice, noise-limited detections will frequently only be achieved for target
Detection Pevfornzance
Table 9.6 Target doppler
Full-beam processing
-30 -27 8
-37 -20 3
- 40 -17 2
Half-beam processsing
--30 -27 1.1
-37 -20 1 .o
-40 - 17 1 .o
dopplers which exceed the platform speed for targets on the beam, and twice
platform speed for targets ahead or astern.
Half-beam processing is superior to hll-beam processing of CW pulses.
Detections remain noise limited down to target dopplers of about 1 knot for a 1 s
pulse. There is little to gain - in performance against low dopplers - from using
longer pulses. This is because the factors other than pulse length which cause
reverberation spreading - finite beamwidth, platform motion and the environment
- then become dominant.
Because of the target doppler requirement, it is operationally unsafe to use only

a CW pulse for any active sonar - unless there is a high probability of sufficient
target doppler - and the preferred mode of operation is to transmit both FM and
CW pulses every ping.
190 Active Sonar
9.19 Noise- and Reverberation-Limited Detection Ranges

Figure 9.13 is based on the parameter values of the hypothetical 10 kHz sonar.
Propagation is assumed to be limited by spherical spreading and absorption. Both
noise- and reverberation-limited detection ranges are plotted against TS for the
LPFM pulse. Note the significant difference in slopes. This is because the
reverberation-limited range is a function of 10 log R, whereas the noise-limited
range is a function of 40 log R (i.e., R4 compared with R).
ttxi'(s.S+ a )
- 10 0 + 10 +20
Target strength (dB)
Figure 9.13 Active sonar: detection ranges versus target strength
Noise-limited range is therefore comparatively insensitive to changes in TS or

N (a change of 3 dB results in about 20 per cent change in range), whereas when
reverberation limited a similar change in TS or S b will change the range by 100
Noise- and Reverheration-Limited Detection Ranges 191
per cent. The comparison indicates the importance of attempting to ensure that the
sonar is always noise limited. Clearly, there will be constraints which will often
make this impossible, but it must be the underlying aim in the design of any active
system.
192 Active Sonar
9.20 Ambiguity Diagrams

The concept of ambiguity diagrams provides further insight into the properties
of the pulses used in active sonar. Active sonars measure the range (time) and
doppler (frequency) components of a target by cross-correlating overlapping
segments of the received signal with a set of stored references. Each reference is a
replica of the transmitted pulse individually modified in time and/or frequency.
Sufficient references are used to cover the expected extent of target doppler.
Ambiguity functions provide a means of comparing the range and doppler
resolutions achievable from different pulse types and of determining the references
required for a given sonar task. Figure 9.14 shows the output power from a correlator
versus the time and frequency displacements of the received signal. The half-power
(3 dB down) extent of the correlator output is given approximately as follows:
In time (f) it is 1/B

In frequency (Sf,) it is I/T
Note that the output power is negligible at twice these displacements.
Figure 9.14 The ambiguity fimction
The 3D representation shows the correlator output above a 2D surface (f, Sfo).
The intersection at the detection threshold P (half-power) defines an ambiguity
contour within which a target cannot be located unambiguously, since all (f, bfo)
combinations result in detections.
A plot of this contour versus Sfo and f is known as the ambiguity diagram for
the pulse waveform. The ambiguity diagram, therefore, indicates the accuracies
and resolutions in range and doppler which are achievable from a given pulse.
The treatment which follows has been stripped of the somewhat difficult
mathematics necessary for even approximate derivations of the ambiguity dia-
grams for CW, FM and PRN pulses and concentrates on results and practical
examples.
A rnbigui? Diagrums 193
CW pulses
By equating the ambiguity function for a rectangular CW pulse to 0.5 (half-power)

and first putting t = 0 and then cffo = 0, we obtain dfo = 0.88/T and f = 0.6T,
and the resulting ambiguity diagrams for long and short CW pulses are sketched
in Figure 9.15. The area of the elliptical ambiguity diagram - for any length of
pulse - is approximately 0.6T(O.88/T)(n/4) = 0.4. The values for dfo and t can
be used to determine the resolution (extent of the ambiguity) of the pulse and the
number of references (in this case matched filters) required to cover the expected
target doppler.
Figure 9.15 Long and short CW pulses
Example 9.4
A sonar transmits a 4 kHz CW pulse of duration 0.5 s and is required to detect targets
with dopplers up to f 2 0 knots. What are the range and doppler resolutions of the
pulse and how many references (filters or FFT cells) are required?
We have
(sfo
2vfo
z --
0.88
z ___
C T
Therefore
0 . 8 8 ~ 0.88 X 3000
7Jz-=- = 0.7 knot
2Tfo 2 X 0.5 X 4000
194 Active Sonar
Note that here c must also be in knots, 3000. The doppler resolution (ambiguity) of
the pulse is 0.7 knot and therefore to cover f 2 0 knots requires 40/0.7 =
57 references.
We have
2R
f = 0.6T = -
C
Therefore
0.6 X 0.5 X 1500

R= = 226 m
2
The range resolution is 226 m and we need to sample at least twice during this range
to ensure we capture the peaks of the signal, i.e., at least every 0.6 X 0.5 = 300 ms
Note that, for the simple CW pulse, the concept of ambiguity functions is not
necessary to determine the range and doppler resolutions. For a rectangular CW pulse
the 3 dB width is 0.91/T (compare with 0.88/T above), which will determine doppler
resolution, and we could expect to sample at least twice every pulse length, i.e., every
250 ms, or every 188 m (compare with 226 m above).
FM pulses: linear period modulation
In linear period modulation (LPM) doppler changes the received signal in frequency
only, causing an overlap loss with the reference, and a slope mismatch does not
occur whatever the magnitude of BT. By equating the ambiguity fhction for the
LPM pulse to 0.5 (half-power) and putting first t = 0 and then dfo = 0, we obtain
dfo = 0.88/T and f = 0.88/B
The endpoints of the ambiguity diagram (Figure 9.16) are dfo = f 0 . 3 B and
t = f0.3T. And for any length of pulse, the area of the ellipse is approximately
0.6B(0.88/B)(n/4) = 0.4.
The doppler resolution (ambiguity) of the pulse is given by the maximum dfo
value of the diagram:
900 B
Sfo =-2vfo = 0.6B therefore v = - (knots)
C fo
Ambiguity Diagru m s
Figure 9.16 Ambiguity diagram: LPM pulse
The range resolution of the pulse is given by the intersections of the contour with
the time axis:
0.88 - 2R 660
t=-- - - therefore R = __ (metres)
B c B
Example 9.5
A sonar transmits a 10 kHz LPM pulse of 500 Hz bandwidth and 1 s duration. What
are the range and doppler resolutions of the pulse? What is the maximum target
doppler for 3 dB correlation loss?
660 900 B
R=-= 1.32m u=-------
= 45 knots
B fo
Because the doppler resolution is derived from the 3 dB contour, then the maximum
target doppler for 3 dB loss is f 22.5 knots (half of the resolution). The small
correlation losses resulting from large target dopplers explain why the LPM pulse is
known as a doppler-invariant pulse.
Once again a simpler derivation is possible. The range resolution is simply the inverse
of the bandwidth, i.e., 2 ms or 1.5 m. There will be a 3 dB correlation loss when only
0.707 of the signal overlaps the reference, which is equivalent to a doppler shift of
f 1 4 7 Hz or f 147/6.9 = f 21 knots (from Af = f0.69 Hz per knot per kHz),
which is in good agreement with the previous result of f 22.5 knots.
196 Active Sonar
FM pulses: linearfrequency modulation
In contrast to an LPM pulse, a linear frequency modulated (LFM) pulse which is

changed by target doppler will no longer perfectly correlate with any part of the
zero doppler reference and its ambiguity diagram computation assumes that the
slope mismatch is the limiting factor.
We have dfo = 0.88/T and f = 0.88/B, identical to the LPM case. The
endpoints of the contour are now dfo = 3.5fo/BT and f = 3.5fo/B2 and the area
of the ellipse is approximately (3.5fo/BT)(0.88/B)(;/4) = 2.4f0/TB2, which is
no longer a simple constant. The doppler resolution (ambiguity) of the pulse
is again given by the maximum dfo value of the diagram (Figure 9.17),
dfo = 2vfo/c = 3.5fo/BT, therefore v = 5200/BT knots. The range resolution is
R = 6 6 0 / B metres, as before.
Figure 9.17 Ambiguity diagram: LFM pulse
Note that, because of the assumption that slope mismatch is the limiting factor,
the equation for doppler resolution only gives valid results for large BT products.
For small BTproducts, a better result is obtained from the simpler derivation used
for the LPM pulse: calculating a doppler shift equivalent to 0.29 times the signal
bandwidth.
Example 9.6
A sonar transmits a 10 kHz LFM pulse of 500 Hz bandwidth and 1 s duration. What
are the range and doppler resolutions of the pulse? How many references are required
to limit correlation loss to 3 dB for target dopplers up to f 3 0 knots?
A mhiguity Diagrams 197
We have
R = 6 6 0 / B = " ~ 0 . 6m
6
ZI = 5200/BT = 5200/500 = 10.4 knots
Therefore we require 60/ 10.4 = 6 references
The area of the contour is 2.4fo/TB2 = 0.1, which is smaller than the corresponding
area of 0.4 for the LPM pulse. The smaller the area, the smaller the ambiguity of the
pulse (the better the resolution).
As we have seen, the smaller area of the contour for the LFM pulse indicates an
improved doppler resolution - about 10 knots instead of 45 knots for the LPM pulse.
The smaller ambiguity of the LFM pulse explains why it is sometimes known as a
doppler-intolerant pulse. Note, however, that while it may be intolerant from the
viewpoint of limiting correlation loss, the pulse only provides a poor measure of the
target doppler. (Compare this with a CW pulse of the same duration and frequency,
which would have a doppler resolution of 0.14 knot.)
Example 9.7
A sonar transmits 10 kHz LPM and LFM pulses of bandwidth 100 Hz and pulse
duration 1 s. What are the doppler resolutions of the pulses?
Suppose we begin like this:
0 LPM u = 900B/fo = 9 knots

LFM u = 5200/BT = 52 knots
This is clearly an invalid result; the LFM pulse is apparently more tolerant of doppler
than the LPM pulse of identical bandwidth and duration. A better result is now
obtained from the equation used for the LPM pulse. Both pulses, then, have the same
doppler resolution of 9 knots.
Example 9.7 highlights the problem of how to decide which equation to use. For
practical values of B and T, a valid result will be obtained using
;17 = 5200/BT knots for BT 3 500

0 u = 900B/f0 knots for BT S 100
For intermediate values of BT, or indeed for any BT value, a recommended, robust
approach is to use both equations and take the result of smaller magnitude.
198 Active Sonar
PRN p ulses
An alternative to a broadband FM pulse is the pseudorandom noise (PRN) pulse.

Its bandwidth will be close to or identical to the bandwidth of an FM pulse
designed for a similar task. The frequency-time structure of a PRN pulse will
change in a random but known manner from ping to ping. Hence replica cor-
relation will still be possible.
By equating the ambiguity hnction for the PRN pulse to 0.5 and putting
t = 0, Sfo = 0, we obtain the 3 dB (half-power) ambiguity diagram (Figure 9.18),
where Sfo = f0.44/T, which is identical to the expression for a CW pulse and
t = f0.44/ B, which is identical to the expression for FM pulses.
Figure 9.18 Ambiguity diagram: PRN pulse
The area of the circle is approximately (0.88/T)(0.88/B)(n/4) = 0.6/BT.

The doppler resolution of the pulse is given by Sfo = 2vfo/c = 0.88/T or
v = O.88c/2TfO(as for CW pulses). The range resolution is given by f = 0.88/B
or R = 660/B metres (as for FM pulses).
The PRN pulse therefore combines the desirable properties of both CW and FM
pulses, i.e., simultaneous good doppler and range resolutions. If BT is less than
2400 divided by the maximum target doppler, the references may simply be
frequency shifted, otherwise time-compressed or time-expanded references must
be used.
Compared with using a combination of CW and FM pulses, the drawbacks of
PRN are the need for a large number of references, each of which must be sampled
Ambiguity Diagrums f 99
at least twice in a period equal to the inverse of the bandwidth (at least every 1 ms
for a bandwidth of 500 Hz), and possible complications or compromises necessary
in the display of the data.
'The processing power requirement is unlikely to be a problem with current
technology but, as always, availability of an adequate bank of PRN trials data is a
strong argument for retaining combined CW and FM sonars.
Example 9.8
A sonar transmits a 5 kHz PRN pulse of 400 Hz bandwidth and 1 s duration. What are
the range and doppler resolutions of the pulse? How often must the signal be sampled
and how many references are needed for a maximum target doppler of &20 knots'?
We have
- 0.88 X 3000
v=- 0 . 8 8 ~- = 0.26 knot
2Tfo 2 X 5000
To cover f 2 0 knots, we need 40/0.26 = 154 references, each spaced 0.26 knot and
0.26 knot wide.
R = 660/B = 1.65 m
and BT = 400, which is greater than 2400/20, so time-compressed or time-expanded

references must be used.
Statistically independent signal samples are separated by the inverse of the bandwidth.
(Note that the approximate solution using the ambiguity fimction is 0.88/B, very close
to the inverse of the bandwidth.) Here the inverse of the bandwidth is 2.5 ms, and to
ensure that peaks of the signal are not missed, sampling should occur at least twice
during this period, say, every 1 ms.
Area of the contour = 0.6/BT = 0.6/400 = 0.0015
This area is very much smaller than for similar duration CW or FM pulses and
confirms the low ambiguity (good resolution) in both range and doppler of this class
of pulse.
200 Active Sonar
Table 9.7 Range and doppler resolution for different pulse types
Pulse type Range resolution (m) Doppler resolution (knots)
cw 450T
LPM 750/ B
LFM (BT < 100) 750/B
LFM (BT > 500) 750/ B
PRN 750/ B
Range and doppler resolutions of CK FM and PRlVpulses
Table 9.7 collates expressions for the range and doppler resolutions for the pulse
types considered. For intermediate BT values of LFM pulses, and preferably for
any values, calculate the doppler resolution using the two alternative formulae, as
discussed earlier, and use the smaller of the two results.
9.21 Very Long Pulses

Avery long pulse will, in practice, have a processing gain somewhat less than its
theoretical value. For pulses between 1 and 10 s duration, the degradation may be
of the order of 3-6 dB and is the result of instabilities in the environment, the
platform and the target. For example, a platform course change which resulted in
the target doppler changing during reception of an echo from a CW pulse, would
spread the echo energy into adjacent matched filters. Pulses up to, say, 2 s duration
may, at least under favourable conditions, have acceptable correlation losses but,
in order to maximize the transmitted energy and thus improve noise-limited
performance, the alternative of transmitting a train of pulses should be considered.
Example 9.9
To improve noise-limited performance, the term 10 log T in the active, noise-limited
sonar equation must be increased. Suppose a 10 s pulse is transmitted, then
10 log T = 10 dB. But if we assume the correlation loss to be 6 dB, the effective gain
over a 1 s pulse is only 4 dB. If the long pulse is replaced by 5 pulses, each of 2 s and
separately processed, then 10 log T = 3 dB but this can be followed by post-detection
integration, giving a further 5 log 5 = 3.5 dB gain. The effective gain over a 1 s pulse
is now 6.5 dB, an improvement of 2.5 dB over transmitting the single long pulse.
Operational Degradation Factor 20 I
9.22 Operational Degradation Factor

A figure of 4 dB is often suggested for the degradation in operational performance
of a modern computer-assisted sonar, compared with its theoretical performance.
However, 4 dB typically equates to a range reduced to about 0.7 of theoretical
range in noise-limited conditions, but only 0.4 of theoretical range in reverbera-
tion-limited conditions.
The factors contributing to the DF are unlikely to be identical for both
reverberation-limited and noise-limited conditions and, particularly in highly
reverberant shallow waters, it is the extra time required for the elimination of
false alarms and final classification which may decide the initial detection
range (the range may just as likely remain constant, or even increase,
depending on the relative velocity and bearing of platform and target during
this time).
Fluctuations in TS and PL from ping to ping are frequently very large, and
marginal initial detections seldom, if ever, occur. A contact is typically non-
existent during one ping and very strong on the next ping. Because of such
uncertainties, the whole concept of degradation is debatable for active sonars; but
if it must be applied, the recommended approach is to calculate both noise-limited
and reverberation-limited detection ranges and then to reduce the smaller by a
factor of 0.7 to allow for a final classification.
202 Active Sonar
9.23 Active Displays

Virtually all active sonar displays conform to the general format of Figure 9.19.
Range (time from transmission) is displayed on the vertical axis and bearing, or
+
bearing doppler, is displayed on the horizontal axis. Most active displays will
have raw sonar data, i.e., post-detection samples, forming a background to
synthetic sonar data, i.e., data on contacts generated by automatic detection and
classification processes.
In Figure 9.19 contact 3 is a possible submarine. The vertical stroke of the
marker indicates the expected position of the contact during the next ping, and the
magnitude of the horizontal stroke indicates the likelihood that the contact is
genuine. Computer-generated data on the most likely contacts (10, 20, 40?) are
displayed in the form of totes - alphanumerics of range, bearing, course, speed,
classification, confidence, etc., on all likely contacts, or on the history of a selected
contact. Typical active displays are best illustrated by postulating a hypothetical
but realistic sonar with the parameters in Table 9.8.
Bearing or bearing + doppler w
Figure 9.19 Active displays: general format

.4 ctive Displa~vs 203
Table 9.8 Parameters of a typical sonar

Frequency 10 kHz
Pulse length, T 300 ms
FM bandwidth 400 Hz
CW bandwidth -7
Hz
Beam spacing 11.25 degrees
No. of beams, 360" cover 32
Doppler channel width 1 knots
Doppler cover k30 knots
FM increments
classification (3 ms) 2 m
surveillance (20 ms) 15 tn
CW increment ( I 33 ms) 100 m
Phase bin width 1 degrees
Range scales 4, 8, 16 km
FA4 surveillance display
This typical sonar uses half-beam processing. The 1 ms correlator samples are
placed in phase bins and integrated to 3 ms for classification (3 ms is the inverse
of the bandwidth and the minimum time for independent samples). The 1 ms
samples are integrated to 20 ms for surveillance, i.e., roughly matching the
minimum target dimension. The FM surveillance display must therefore display
the integrated contents of the phase bins for each beam as a function of range
(time).
RANGE
5 12 pixels
BEARING
5 12 pixels
32 X 16 phase bins
Range scale 4 8 16
Range increments 267 533 1067
Pixels per increment 2 1 0.5
204 Active Sonar
For the 4 km range scale, display the same information in two consecutive lines.
For the 16 km range scale, use one line to display two consecutive range
increments. (The maximum ranges displayed will be slightly less than the nominal
range scales.)
C W surveillance display
The CW processing provides independent samples every 133 ms. This is equiva-
lent to a 100 m range increment, which is already large compared to the minimum
target dimension, hence it requires no integration before display.
5 12 pixels
1 I
BEARING, DOPPLER
5 12 pixels
Range scale 4 8 16
Range increments 40 80 160
Pixels per increment 12 6 3
Note that, because of the small bandwidths associated with the CW pulses used
for submarine detection, range resolution is poor and, for all range scales, the
same information is displayed in several lines if the display is to be the same size
as the FM display. Alternatively, these surplus lines could profitably be used to
display ping history. Whatever solution is adopted, equal height displays are
clearly desirable to allow the operator to readily associate FM and CW echoes
from the same target.
The maximum number of increments in the horizontal dimension (bearing +
doppler) is 32 X 60 = 1920. This number must be reduced by ORing beams and/
or doppler channels to match the available number of pixels, e.g., 16
+
beams X (30 doppler channels 2 spaces) = 512. The doppler channels may be
varied in width, increasing with the magnitude of the target doppler, to reflect both
the improved performance and the lower probability of high magnitude target
dopplers.
Active Displa-ys 205
The range, bearing and doppler samples, however, will be passed to the
automatic detection and classification processes at their full resolutions.
History displa-vs
Active history displays are analogous to the BT and LOFAR passive displays, where
successive time samples are replaced by successive ping samples. Figure 9.20
shows the simplest form of history display, used by the early electromechanical
range recorders. A slope equivalent to the platform relative velocity will indicate
reverberation or a zero doppler contact. Other slopes indicate non-zero doppler
contacts. Noise will seldom correlate from ping to ping. Derivatives of this display
are still used successfidly in modern sonars and are not lightly disregarded.
Noise Reverbs or zero doppler target Moving target
1 1 J \ Ping n
Pingn - 1
Ping n - 2
Pingn - 3
Range _______+
Figure 9.20 Range recorder trace
Geographically stabilized displa-ys
Geographically stabilized displays are a modern form of history display, made

possible by the processing power now available. If all the pixels of ping n - 1
are corrected for own ship movement in range and bearing and then summed,
at the display, with the corresponding pixels from ping n, the display will be
geographically stabilized. Both reverberation and stationary targets (truly
stationary, not simply having zero relative velocity) will be displayed at the
same coordinates for all pings, but moving targets will change position from
ping to ping.
Stationary targets will increase in brightness, but so will reverberation and
therefore, at least in reverberation-limited conditions, performance against station-
ary targets will be little enhanced. Moving targets will produce tracks on the
206 Active Sonar
display which will be easier to detect and class@. To avoid display whiteout after
several several pings, the individual pixel amplitude (brightness) values of the
samples from earlier pings must be decreased with time.
9.24 Unified Detection and Classification

The traditional approach to the design of an active sonar system treats the
detection and classification functions as separable and consecutive, but in
reality they are interdependent. Design decisions are then based on detection
alone, i.e., simply finding signals in noise or reverberation based on amplitude
only, and this may not be optimum for the composite detection and classifica-
tion task.
The traditional approach - optimizing detection before considering classifica-
tion - can be effective against noise and low reverberation but less so against high
discrete reverberation (clutter) such as is often present in coastal waters. This
discrete reverberation is the source of many false alarms, and a major problem not
properly addressed by the traditional approach where the sonar equations are used
to define the probability of detection ( P d ) at a specified probability of false alarms
(Pf,) based only on S I N ratios. Classification, which should be seen as the
reduction of false alarms, comes later.
The unified approach recognizes that detection and classification are not
separable. Design decisions must recognize this and the total task must be
optimized, not initial detection at the expense of classification. The parameters
fundamental to this unified approach are bandwidth and beamwidth.
9.25 Bandwidth
For detection we need to match the minimum target dimension (d). Suppose for a
submarine this is 10 m, then the resolvable pulse length is f = 13 ms from
-
d = c t / 2 ; as we have already seen, t can be much less than the actual pulse length.
Therefore bandwidth = l / t = 75 Hz. For classification we need to resolve the
structures of the target and any non-targets (false alarms). This will require a
higher resolution, to say 1 m,giving bandwidth = 750 Hz.
If the target is small, e.g., a mine, the required resolution will be much higher,
say 0.01 m, giving bandwidth = 75 kHz. We could use either a CW pulse of
duration 1/ B = 13p s or a longer, perhaps 10 ms, broadband (FM type) pulse.
Bandwidth 207
The point to emphasize here is that bandwidth is decided by the dimensions and
structure of the target and the false alarms. There is nothing to be gained by
increasing the bandwidth further. Note that if there are reflecting objects much
smaller than the target, they will be irrelevant because their target strengths will
be too small to compete with those of the target and Comparable false alarms.
208 Active Sonar
9.26 Beamwidth
Just as bandwidth determines the resolution in the range dimension, so does
beamwidth in the bearing dimension. Again, for detection we would like to match
the minimum target dimension, which is now the minimum angle subtended at the
array by the target. This is directly proportional to range, and for a submarine at a
range of at least 5 km this will always be less than about 1". Given practical arrays,
the smallest beamwidth will be at best about 4" and matching will only occur at
around 1-2 km, depending on submarine aspect. (Smaller beamwidths are possible
using towed arrays, but only at bearings within, say, normal f60".)
Figure 9.21(a) shows, in B scan format, submarine targets at two different
ranges and a patch of discrete reverberation which might, for example, represent a
rocky outcrop in shallow water. Only one beam of the display is shown: the
beamwidth is 4" and the range is from 0 to 5 km. Note that the width of the display
is proportional to range: at 5 km the width is about 270 m and therefore the
submarine at this range has a bearing spread of about lo, but at 1.7 km the width is
90 m and its bearing spread is much greater, about 3".
Figure 9.21 Comparison of bearings and bearing spreads for submarines and false alarms
Figure 9.21(b) shows one beam of a sonar display in the same B scan format
and produced using full-beam processing. Because fU-beam processing cannot
resolve bearing within a beam, the targets and the false alarm will all be given the
same bearing and no values at all for bearing spread. The submarine targets and
the reverberation patch, therefore, all appear at the same bearing and with the
same (unknown) spread in bearing. Note, however, that the differences in range
spread are displayed.
Beamwidth 209
Figure 9.21(c) shows one beam of a sonar display again in the same format but
produced using half-beam processing. We can now resolve bearing within a beam,
and the submarine targets and the reverberation patch are all displayed with their
correct bearings and bearing spreads. We now have, within the sonar's limits of
resolution, a fair approximation to the overall shape and position of contacts and
false alarms,
At 5 km range, the submarine is in one phase bin whereas the reverberation
occupies four phase bins. Since, at this range, a submarine cannot spread into
more than two bins, the bearing spread is a useful discriminant between targets
and f'alse alarms. At closer ranges, this discriminant may not be so useful. The
target spread, however, when considered together with other parameters such as
target motion, will still be a strong classification feature. For classification we
again need to resolve the structures of the target and any non-targets (false alarnis)
in the bearing dimension. Discrete reverberation (clutter) can have much greater
bearing spreads than submarines, and in order to exploit this difference for
classification (false alarm reduction), we require a higher bearing resolution up to.
say, 1/10 of the beamwidth. A practical minimum value for the composite
detection and classification task might be 0.5".
Full beams, using the greatest possible aperture of the array, cannot achieve
such small beamwidths, therefore detection and classification processing using
signals from full beams is not optimal against a background of high discrete
reverberation (clutter).
However, against a background of noise or uniform reverberation and only if
the other important advantages of the alternative, half-beam processing, are
ignored, then full-beam processing, which uses the highest possible array DI, is
marginally better: the 3 dB greater DI provides an extra 1.5 dB of allowable
propagation loss to use in the sonar equations.
In summary, half-beam processing can be used to effectively achieve small
beamwidths and therefore detection and classification processing using signals
from half-beams is optimal against a background of high discrete reverberation
(clutter). The effective beamwidth is the size of a phase or bearing bin.
The phase bins, as recipients of the output signals from the process, resolve the
structures of both targets and false alarms in both range and bearing dimensions,
to provide the best possible data for the composite detection and classification
task.
210 Active Sonar
9.27 CADAC
The output from the signal detection process is a time series of samples from
every beam of the sonar, and for modem high resolution sonars there are many
samples every ping. Suppose pd = 0.5 and Pfa = (this is for the simple
signal detection process only). For the typical sonar which was used to illustrate
active displays, the FM display has 512 X 500 = 2.5 X 105 pixels (samples).
Therefore there are 2.5 false alarms every ping, or 15 false alarms per minute!
What appears to be a very low Pfa has become an intolerably high rate of false
alarms. The post signal detection processes of CADAC (computer-aided detection
and classification) and display will help the operator to reduce this to an
acceptable false alarm rate to pass to the command system.
Levels qf'CADAC 21 1
9.28 Levels of CADAC

The post signal processing functions (CADAC and display) host many processes,
some of which have traditionally been thought of as detection and others as
classification. All of these processes contribute to a final classification and no
individual process is purely for detection or classification. (Compare with the
argument above that, for signal processing, detection and classification should not
be considered separately.) The combined CADAC and display task may be split
into several levels, each level building on the earlier ones. It would be wrong to
say that any level detects or any level classifies, rather that every level contributes
to the final classification.
Level 1
Level 1 is the lowest level; it detects signals within single cells; the signals may
originate from vessels, noise or reverberation. A fixed threshold is normally used,
at around 10- 13 dB above the average of the background (the threshold may be
lowered to dig deeper into the background in areas of interest, or increased to
inhibit inputs from improbable areas).
Level 2
If the sonar resolution is finer than the typical size of a contact (including false alarms
such as discrete reverberation) then returns will occur in several adjacent range and
bearing cells. Clustering these returns together forms the second level of classifica-
tion, since some estimate of the dimensions and shape of the contact can be formed
from the distribution of returned energy in the contiguous cells. This estimate is
commonly known as the individual weight of evidence of a contact. The parameters
used to make this estimate include cell amplitudes, total energy, range, range spread
and bearing spread. At level 2, additional processes can be applied to the cluster of
individual returns making up a contact. There are, broadly speaking, two ways of
doing this: statistical analysis and amplitude profiles (Sections 9.30 and 9.3 1).
Level 3
Level 3 examines how well contacts associate from one ping to another by seeking
consistent tracks. Goodness of track and individual weights of evidence of the
212 Active Sonar
contacts forming the track help to build up a parameter known as the weight of
evidence (WOE)of a contact. Courses and speeds of all contacts are also calculated
at this level (the speed of a contact is clearly an important classification clue).
Target doppler, if available, should be input to the WOEat this level.
Level 4
At level 4, information gathered from all other possible sources is used to assist in
the final classification of all contacts. Here are some possible sources:
Other modes of the sonar such as passive or intercept

Other sonars on the same platform
Sonars and other sensors on consort vessels or aircraft
Sonobuoys
Radar
Many of these sources will not help with the classification of the submarine itself.
They will, however, materially assist in the reduction of false alarms. The
reduction of false alarms can itself be viewed as classification. If all false alarms
are eliminated, surviving contacts must be submarines.
Level 5
Level 5, the final level, includes all the experience, knowledge of tactics, strategy,
and intelligence data possessed by the operator and the command, and it is only at
this, human, level that the classification can be positively confirmed. For this
reason, any automated classification system, no matter how clever its algorithms,
should be thought of as a classification aid not an automatic classifier.
9.29 CADAC and Pulse Features

For long CW pulses, the range resolution cell is very large - hundreds of metres -
and clearly provides no usefbl discrimination at CADAC level 2. A CW detection,
however, is highly unlikely without some target doppler to ensure that target
Amplitude Prqfiles 213
echoes fall into regions of the reverberation spectrum where the reverberation
energy is low. The important classification clue of target doppler is therefore
always available simultaneously with CW detections.
For the FM pulses, the range resolution cell is small compared with the typical
extent of a contact and therefore it is possible to make improved estimates of the
dimensions and echo profile of the contact. Furthermore, half-beam processing, by
improving bearing resolution, provides additional information (bearing spread)
towards this estimate, which can help to discriminate between targets and false
alarnis at CADAC levels 1 and 2.
For both pulse types, half-beam processing improves bearing accuracy and
eliminates echo returns in adjacent beams, from which ping-to-ping association
and tracking processes will benefit (CADAC level 3).
9.30 Statistical Analysis

Statistical analysis examines the joint distributions of amplitudes, ranges and
bearings of the resolved echoes of a large number of previously identified contacts.
Using specialized pattern analysis and recognition software, suitable discriminants
can be found to partition the returns into the desired classes. For example, it might
well be that the distribution of amplitudes with range for a submarine contact
follows a positively skewed probability hnction whereas a surface ship gives a
Gaussian distribution. The preliminary analysis phase examines all the possible
statistical measures and identifies the minimum subset which provides the required
discrimination among the known contacts. As long as the statistics of contact
echoes when the system enters service remain similar to those of the preliminary
training set, then it will be able to make successful attempts at classification into
the original categories.
9.31 Amplitude Profiles

The amplitude profiles approach imitates the way an experienced operator assesses
the contact echo. If the resolution is sufficiently fine (1 m or so) then the echo
from a contact will display some detail of the reflecting structure of the contact.
The amplitude of the returns along the length of the contact will be proportional to
the reflectivity of the contacts structure. A submarine may well return strong
echoes from the bow and the fin, and weaker echoes from the hull and the casing.
These reflections yield a set of strongly aspect-dependent amplitudehange profiles
2 14 Active Sonar
which, after normalizing and clean-up processes, are compared with memorized
profiles of known contacts to give an indication of the contacts type.
The method has the advantage that it can learn new profiles from just a single
exposure to a particular contact, and can dynamically adjust its repertoire of stored
profiles in accordance with the prevailing conditions. Furthermore, it has the
important attribute of being able to indicate to the operator why it has made its
hypothetical choice between alternative classifications in terms that the operator
can understand, rather than as obscure statistical technicalities.
There is a case for providing both statistical analysis and amplitude profile aids to
classification. The statistical analysis classifier is probably superior in classifying
those targets with which it has been trained. But it is not robust to changes in proc-
essing, it is sensitive to the environment (e.g., using different algorithms for deep and
shallow water), and from an operator viewpoint it makes its decisions in the dark.
The amplitude profile classifier is perhaps somewhat less than optimum when
operating against the specific targets with which the statistical analysis classifier has
been trained. On the other hand, it is robust to changes in signal processing, it is
insensitive to the environment, it can rapidly learn new profiles from just a single
exposure to a new target, and because the profiles are displayed to the operator, it
makes its classification in terms that the operator can see and understand.
9.32 Multipath Affects Classification

To interpret any classification, it is important to understand contact clusters and to
ask: What are the effects of multipath propagation on the individual echoes
returning from a target? Multipath returns are unlikely to significantly increase the
range or bearing extents of a cluster; path differences for similar strength returns
are unlikely to be greater than a few milliseconds. They can, however, confuse the
structure of the amplitude profile of a contact, by smearing and moving the
highlights, or even resulting in new ones, e.g., glints from features of the target
which are at normal incidence to the additional paths.
The operation of either type of classifier can therefore be impaired by multipath.
It has been suggested that the effects of multipath propagation could be deconvo-
luted ahead of the classification procedures which are affected by them. This relies
on a more precise knowledge of propagation than is likely in practice. Perhaps a
more promising approach would be to attempt to eliminate or reduce the confusing
returns. Narrow, steered, vertical beams could be employed, or if the vertical
beamwidth cannot be less than, say, 10 then phase binning in the vertical
dimension could be tried.
Simple Multipath Example 215
9.33 Simple Multipath Example

Classification displays often plot the amplitudes of individual returns (echoes)
from a target against range (target extent), to produce a range/amplitude profile of
the target. Multipath returns distort this profile in two ways (Figure 9.22):
0 Point a on the target submarine returns signals to the source via two paths, A
and AA, for this simplified case (Figure 9.22). The same point therefore puts
returns in two separated cells along the profile. Signals along paths AA and B
arrive simultaneously but from different points ( a and h) on the target, and are
detected against a background of reverberation returning along path C and also
arriving at the same time. Different point sources therefore put returns in the
same cell of the profile.
An echo from point b returned via the bottom will extend the length of the
profile. The magnitude of this extension will increase with range - as path
differences and the number of reflections from the boundaries increase.
Figure 9.22 Multipath example
Multipath propagation increases the total energy returning to the source from the
target without increasing the reverberation background, and therefore detection is
enhanced. On the other hand, classification or at least those features used by
classification which rely on the fidelity of the profile, will suffer to an extent that
depends on the energies of the misplaced returns.
The likely magnitudes of possible profile extensions and highlight displace-
ments will now be estimated for the most relevant propagation modes, i.e., surface
duct in deep water and in shallow water.
216 Active Sonar
9.34 Deep Water: Source and Target in Duct

The skip distance of the limiting ray (distance between surface reflections) is given
by
I [8(i) hd] 121

where
hd = duct depth
c = speed of sound
g = velocity gradient
In an isothermal duct then g (pressure change only) = 17 m/s per1000 m, so
112
Skip distance = [8 X (E)
0.017
X 1001 = 8400 m
From simple geometry, the extra path length over the direct path for the limiting
ray for each surface reflection is about 2.5 m (Figure 9.23). The limiting ray gives
the maximum difference as 0.3 m per km. So if the range of the target submarine
is 20 km, for example, its extent could be increased by up to 20 X 0.3 = 6 m (and
highlights misplaced by up to similar amounts along the profile). Given a large
submarine (at least 5 0 m long, say) the extension is unlikely to seriously affect
detection and echo shape recognition, but classification could suffer if it relies on
profile details.
Figure 9.23 Deep water: in duct example

Deep Water: Source and Target below Duct 217
9.35 Deep Water: Source and Target below Duct

The bottom is remote and therefore there will only be one reflection to consider, a
surface reflection (Figure 9.24). By simply approximating to a triangle, the length
of the indirect path exceeds the direct path by about 2 d m
- R (Table
9.9). Therefore, for this propagation mode, the target extension and profile
distortion is greatest for deep, close targets. An extra path length of 64 m is
comparable with the length of a submarine and will clearly strongly influence both
echo shape recognition and classification based on profile details.
t
Duct
Figure 9.24 Deep water: below duct example
Table 9.9 Path difference between direct and indirect

Depth (m) Direct path R (m) Indirect path (m) Difference (m)
200 10 000 10 008 8
200 5 000 5 Olh 16
400 10 000 10 032 32
400 5 000 5 064 64
218 Active Sonar
9.36 Shallow Water: Source and Target Mid-water

Figure 9.25 simplifies the geometry. The conclusions would not differ greatly if
the source were closer to the surface (e.g., in a surface ship dome). Two bottom
reflections are shown but in reality there will be many more; the longest indirect
path will have the most and result from the largest angle of depression, 8. Angle 8
is also the grazing angle for the reflections.
Figure 9.25 Shallow water example
Suppose 8 = 10" and h = 200 m. The distance between bottom reflections will
be 2268 m. For the two bottom reflections in Figure 9.25, R = 3403 m and the
indirect path is 3455 m. The path difference is 52 m. This is already quite large
and will increase linearly with the range, R. Bottom losses at each reflection,
however, reduce the amplitudes of the multipath arrivals. (Surface reflection losses
are much smaller and are neglected.)
Table 9.10 shows differences in path lengths for different angles at representa-
tive ranges of around 11 km and 23 km. Note that the large differences are
accompanied by large total reflection losses and can therefore be neglected.
Furthermore, bottom reflection losses increase rapidly at grazing angles greater
than about lo", so even if vertical half-beamwidths are greater than lo", there
should be no need to consider rays at angles greater than this.
Table 9.10 Path length differences at different angles

e Direct path, R Indirect path Difference Bottom Loss per Total loss
(m) (m) (m) reflections reflection (dB)
(dB)
1" 11 458 1 1 460 2 1 1 1
4" 11 449 11 464 15 2 2 4
9" 11 365 11 506 141 5 3 15
1" 22 916 22 920 4 1 1 1
4" 22 898 22 928 30 3 2 6
9" 22 730 23 012 282 9 3 27
Problems 219
9.37 Problems
9.1 A shipborne sonar transmits a pulse at 5 kHz and receives an echo at 5.025 kHz
from a target bearing G45". If the ship is moving at 8 knots, what is the doppler, or
relative velocity, of the target?
9.2 An LPFM pulse has a bandwidth of 400Hz centred on 4000Hz. Given own
doppler nullification, what are the target dopplers for processing losses of 1 and 3 dB?
9.3 An active sonar uses a Hamming shaped CW pulse to improve the reverberation
rejection, Rj, due to target doppler. How much target doppler would be needed to
achieve X, = -40 dB for a pulse of 1 s duration at 3000 Hz, 1000 Hz and 300 Hz?
10
Echo Sounding and Side Scan
Sonars
10.1 Common Features

Both echo sounding and side scan sonars use the motion of the sonar platform to
build up a representation of the sea bed and targets between the source and the sea
bed. The operating frequencies range between about 20 kHz and 500 kHz - the
frequency, as always, determined by the requirements of range (depth) and target
size.
10.2 Echo Sounders

The first echo sounders were used simply to measure the depth of water below the
sonar platform and replaced the traditional line and lead. Modern echo sounders,
with narrower beams and improved displays, are also used to detect schools of fish
below the sonar platform. For this application the echo sounder is often attached
to the trawl, enabling the trawlerman to control the depth of his trawl very
accurately; and because the distance to the sea bed is now the height of the trawl
above it, the trawler can avoid damage through contact with the bottom. When
attached to the trawl, the device is known as a netsonde
A typical echo sounder might operate at 100 kHz and, for a depth scale of
200 m, transmit a pulse four times per second. The standard display presentation
is known as an echogram. Depth (strictly, distance measured from the transducer)
222 Echo Sounding and Side Scan Sonars
is on the vertical axis, and time, or distance travelled over the sea bed, is on the
horizontal axis.
The echogram in Figure 10.1 displays some 30 ping returns and therefore, at a
ships speed of 10 knots (5m/s) the distance travelled to build up the echogram is
about 40 m. The school of fish can be seen to extend about 5 m in plan, and about
40 m in depth. The brightness or colour of the display will indicate the strength of
the echoes.
Figure 10.1 Echo sounder and display
There are no fundamental differences between the design of echo sounding and
side scan sonars and the design of a submarine detection sonar. The differences
are only of scale - frequency, range and target size. Because we are here
concerned only with short-range or vertical transmissions, refraction of sound,
which severely limits the performance of long-range antisubmarine sonars, is
unimportant and the propagation losses are well described by spherical spreading
plus absorption.
Echo Sounder: Design Example 223
10.3 Echo Sounder: Design Example

The correct choice of operating frequency leads easily into a straightforward
design procedure. A typical operating frequency for an echo sounder for use over
continental shelf areas, at depths up to 500 m, is 40 kHz. In the deep oceans, lower
frequencies will be needed to avoid excessive absorption losses which would
otherwise limit the maximum depth; 20 kHz is probably about the highest
practical frequency. In freshwater lakes where the absorption losses are very much
less, or to detect small targets like plankton close to the surface, higher frequencies
may be used, typically 100-500 kHz.
Let us begin with an outline specification (italic in Table lO.l), for an echo
sounder for continental shelf areas, and then build up a more complete specifica-
tion,
Table 10.1 Echo sounder: outline specification

Maximum depth 500 nt
Frequency 40 kHz
Beamwidth, solid angle 5"
PL 59 dB at 500 m
41 dB at 100 m
DI, transmit and receive 26 dB
Pulse length, T 133 ps
10 log T -39 dB
Ambient sea noise, N 35 dB
TS, sea bed +10 dB
TS, individual fish -50 dB
Source level, SL 187 dB
The propagation loss will be given by spherical spreading plus absorption (quite
accurately in this instance, due to the lack of boundaries, rather than the usual
working approximation).
PL = 20log Y + ar X lou3 (dB)

Y = 500 m, a =9 d B / h therefore PL = 59 dB
Y = 100 m, a = 9 dB/km therefore PL = 41 dB
A solid angle beamwidth of 5" will result from a square array of side L, or a
slightly larger diameter circular array. We have 83 = 76/Lf , therefore L =

76/& f = 0.38 m.
The DI, in both transmit and receive, is given by
DI = 2010g(F) = 2010g(~X030;:8) = 26 dB
The pulse length would have to be very short in order to resolve individual fish.
Suppose the individual fish to be separated by 0.1 m, the pulse length would need
to be
This may be too short to meet the propagation loss requirement but let us stay with
it for the present. Let the background noise, N, be the ambient noise at SS2, or
35dBat40kHz.Let5logd= 10dBand5logn=3dB.
Then to solve the noise-limited active sonar equation for the required source
level, SL, all we require is TS. The sea bed is a very strong target at normal
incidence and will be quite unmistakable. We will arbitrarily assign TS = 10 dB
to the sea bed. Fish, individually or in schools, will have very much smaller target
strengths, say -50 dB, but the depths will be less, say 100 m where PL = 4 1 dB.
12PL = SL + TS - N + DI + lolog T - 5 logd + 5 log n I

Therefore
SL = 2PL - TS + N - DI - 1Ol0gT + 5logd - 510g II
To detect the sea bed,
SL = 118 - 10 + 35 - 26 + 39 + 10 - 3 = 163 dB
and to detect fish down to 100 m depth,
SL = 82 + 50 + 35 - 26 + 39 + 10 - 3 = 187 dB
SL,,, is therefore 187 dB. Is this practical? We have
I SL = lolog P + 170.8 + DI, I

Echo Sounder: Design Example 22 s
Therefore
lOlOgP=187-171 - 2 6 z - l O d B
giving P = 100 mW.

The acoustic power intensity is 10-1/0.382 = 0.7W/m2 and, clearly, cavitation
will not be a problem. The array design is therefore practical and the remaining
parameters may be inserted in Table 10.1.
The echo sounder or fish finder would in practice have a selection of source
levels and pulse lengths under operator control. It would clearly be practical to
have a source level much greater than 187 dB - a value of 207 dB would still only
need I0 W of acoustic power.
The broadband, FM, techniques essential for the detection of many larger
targets at longer ranges, and offering excellent performance against a background
of reverberation, are not essential for these systems. The necessary range resolu-
tion has been shown to be obtainable using a very short CW pulse, and fish
detection is normally against a noise background only; the major source of
reverberation is the sea bed, which is beyond the range of most fish species of
interest. An exception is the volume reverberation due to dense concentrations
of plankton, which can limit the detection of individual fish.
10.4 Side Scan Sonar

A side scan sonar builds up a two-dimensional picture of the sea bed - together
with any-targets on the sea bed - using a combination of an asymmetrical
transducer and the motion of the sonar platform through the water. The transducer
may be mounted on the keel of a ship or be located in a towed body. A towed body
will have the advantage of depth capability (i.e., it may usefully be placed closer
to the sea bed). The installation, particularly if it is located on the keel of a ship,
will often include an echo sounder. Figure 10.2 shows the vertical beam coverage
provided by two transducers:
Echo sounder beam Side scan beam
Figure 10.2 'Vertical' beams: combined echo sounder and side scan sonar
An echo sounder transducer pointing directly downwards with a solid beam-

width of 5".
Sidc Scan Sonar 227
* A side scan transducer having a 'vertical' beamwidth of 45" and a 'horizontal'

(fore and aft) beamwidth of 2".
As the transmissions from the side scan transducer glance across the sea bed, each
ping builds up a single line on the display (Figure 10.3). As the ship moves,
therefore, successive pings build up a two-dimensional map of the sea bed. If the
bottom is smooth, the display will simply show a characterless, noise-like picture.
If, however, the sea bed has features, such as peaks and valleys, the picture will be
quite different; the peaks will backscatter strongly and the valleys, shielded by the
peaks, will display as shadows. Objects on the sea bed - wrecks, mines and
bottomed submarines - are frequently detected by their shadows. This mechan-
ism, the so-called shadowgraph eBect, is very important for the detection and
classification of bottom mines. Because of their very low target strengths, echoes
from mines are very difficult to detect against a background of bottom reverbera-
tion; the shadowgraph can, however, indicate the presence of a mine even when its
echo is hidden by the reverberation.
250 tK Ship's
track
I I
I I
I I
I I
I 1
I I
I 50m
4----'1
I I
I I
I I
2
Range (&)
Beamwidth 2" 'Beamwidth' 'Beamwidth'

4.5 m 9.0 m
Figure 10.3 Side scan display

A true geographic representation of the sea bed or a target requires that the
scales of both display axes shall be the same. This is achieved by choosing a
platform speed, V, which will match the track dimension to the maximum range
dimension of the selected range scale:
where
R, = selected range scale

n = number of pixcells [pixels?] or lines in the track dimension
PRI = 2&/c
But because the horizontal beamwidth is finite, albeit small, the beam patterns
from successive pings could overlap significantly (depending on PRI and ship's
speed) and therefore the data on successive lines may not be independent, except
at close range.
Example 10.1
Suppose the display has 250 lines, each one displaying a ping's worth of data, then
V = - l5Oo = 3 m/s (6 knots)

2 X 250
The pings or lines are therefore spaced by 1 m in the track dimension. Since the linear
beamwidth increases with range - from 0 at zero range to 9 m at 250 m range for the
2" wide beam in Figure 10.3 - the beam patterns from successive pings overlap
significantly; at 250 m, for example, the data from 8, i.e., 9 - 1, successive pings fully
overlaps, and completely independent data is only available every 9th ping.
A compromise, avoiding the storage and display of redundant data, might be to use the
data from every 3rd ping, thus ensuring independent data at about one-third of
maximum range and reasonable overlaps at greater ranges. The data from every 3rd
ping would then be displayed on three consecutive lines.
Side Scan Sonar: Design Example 229
10.5 Side Scan Sonar: Design Example

Let us again begin with an outline specification (italic in Table 10.2) for a side
scan sonar for continental shelf areas, and then build up a more complete
specification. The transducer is mounted in a towed body and therefore, because it
can be quite close to the sea bed, the maximum range requirement is 200 m. Let
the frequency be 100 kHz and the combined (transmit and receive) beamwidth 1"
fore and aft (horizontal) and 45" in the vertical plane normal to the track of the
towed body.
Table 10.2 Side scan sonar: design example

Maximum range 200 rn
Frequency 100 kHz
Combined beamwidth
fore and ajl 1"
normal to track 45"
PL 53 dB at 200 rn
DI, transmit and receive 21 dB
Pulse length, T 133 ps
10 log T -39 dB
TS, sea bed -40 dB
TS, individual fish -50 dB
The propagation loss will be approximated by spherical spreading plus absorp-

tion (not quite so accurately in this instance, because the sea bed presents one
boundary after the wavefront reaches it).
PL = 2010g Y + ur X lo-' (dB)

Y = 200 m, u = 35 dB/km therefore PL = 53 dB
Transmit and receive on separate arrays of identical size and tilted downwards by
45". To achieve the required combined beamwidths, each array must have double
the combined beamwidths and the dimensions are given by L = 76/f&. There-
fore, the width is 76/(100 X 2) = 0.38 m and the height is 76/(100 X 90)
= 0.0084 m. Hence each array will have two rows of 50 elements, all spaced by
112.
DI, in both transmit and receive, is given by
DI = 10 log(4Lh/A2)
= 10 log [(4 X 0.38 X 0.0084)/0.0152] = 18 dB
Baffling will add 3 dB to this, making DI = 21 dB. Once again, the pulse length
would have to be very short in order to resolve individual fish. Suppose the
individual fish to be separated by at least 0.1 m; the pulse length would need to be
2 x m - 2XO.l
T=- - -= 133
C 1500
This is probably shorter than necessary to map the sea bed or to display
submarines and wrecks but it may well meet the propagation loss requirement, so
let us stay with it for the present.
Let the background noise, N, be the ambient noise at SS4, equal to 30 dB at
100 kHz, and let 5 log d = 10 dB and 5 log n = 3 dB. Then to solve the noise-
limited active sonar equation for the required source level, SL, all we require is
TS. Backscattering from the sea bed at high grazing angles will be strong and its
target strength will be given by
TSR = S b + 10logA (dB)
where
cT
A = -R(3h ((3h in radians)
2
= (750 X 0.000 133)(200 X 0.017)
= 0.34 m2
Take the backscattering strength, &, to be -35 dB at the maximum range

(smallest grazing angle), then
TSR = -35 + 1010gO.34 = -40 dB
At shorter ranges, the reverberating area will be smaller but the backscattering
strength will be greater (higher grazing angle), therefore the TSR may not change
Problem 23 1
much. This is very much smaller than the TS of 10 dB used in the echo sounder
design, but remember that was for a normal incidence reflection rather than
backscattering. Again fish, individually or in schools, will have small target
strengths, say -50 dB.
I 2PL = SL + TS - N + DI + 1010g T - 5 logd + 5 log n I

Therefore
SL = 2PL - TS + N - DI - 1010gT + 5logd - 510g H

To detect the sea bed,
SL = 106 + 40 + 30 - 21 + 39 + 10 - 3 = 201 dB
and to detect fish out to 200 m range,
SL = 106 + 50 + 30 - 21 + 39 + 10 - 3 = 211 dB
SL,,, is therefore 21 1 dB. Is this practical? We have
I S L = lOlogP+ 170.8+DIt(
Therefore
10logP = 21 1 - 171 - 21 = 19 dB
giving P = 79 W.
The acoustic power intensity is 79/(0.38 X 0.084) = 2.5 kW/m2 and again
cavitation will not be a problem. The array design is therefore practical and the
remaining parameters may be inserted in Table 10.2.
10.6 Problem
10.1 An echo sounder is to be used to survey the deep ocean floor. The depth
requirement is 12 km to cope with the deepest ocean trenches. Given similar
parameters to Section 10.3 (but take SL = 2 10 dB, T = 10 ms and TS of the deepest
ocean floor as 40 dB), what is the highest possible operating frequency? What range is
achievable at 30 kHz?
11
Mine Hunting Sonars
1 1.1 Overview
Side scan sonars, operating at similar frequencies to the example in Chapter 10,
would appear to have the necessary performance to be used for mine hunting. The
short pulse lengths and narrow beams would provide sufficient discrimination in
the range and bearing dimensions to detect and classify mines. Operationally,
however, it is clearly preferable to detect a mine ahead of the ship - before it
becomes a danger - and the time taken to search an area using side scanning
techniques is a major liability.
A mine hunting sonar, then, is usually very similar to a hull-mounted anti-
submarine sonar. The array will be housed in a keel or bow dome and, in
surveillance, will have a wide arc of cover centred on the ahead bearing.
Classification of mine-like contacts is difficult, particularly bottom or close
tethered mines, where the background to detection and classification is frequently
high discrete reverberation (clutter) &om the sea bed. Target strengths are low and
therefore, to discriminate against this reverberation, narrow beamwidths and short
(actual or resolved) pulse lengths are required. The short pulse length may be
either a CW pulse, typically less than 1 ms, or a broadband FM pulse, in which
case the actual pulse length can be much longer. This will improve noise-limited
performance, but now the bandwidth can also be quite large; at least 10 kHz is
desirable, which gives a resolved pulse length of 0.1 ms.
Modern mines are very difficult acoustic targets; they are shaped and clad to
ensure very low target strengths. In extremis, detection and classification may only
be possible by closely approaching the mine using a remotely operated vehicle
(ROV) housing a very similar array.
234 Mine Hunting Sonars
11.2 Two Broad Classes of Mine
In-volume or near-surface: this class of mine must be detected against a

background of noise and sea surface reverberation. In shallow water the
background can also be bottom reverberation but it may be possible to avoid
this by the use of vertical directivity.
Bottom or close tethered this class of mine must be detected against a background
of noise and unavoidable bottom reverberation. It is by far the most difficult of
the two and definite classification may sometimes only be possible using an ROY
11.3 Backgrounds to Mine Detection

Figure 11.1 shows how the background to detection varies with depth and range.
In zone A the background is noise only. In zone B the background is noise and sea
surface reverberation. In-volume mines are detected and classified in this zone
relatively easily. Bottom mines will not be detected because the wavefront has yet
to reach the sea bed.
In zone C the background is noise, sea surface and bottom reverberation. Both
classes of mine may be detected and classified in this zone. Except on the rare
occasions when the surface reverberation is greater than the bottom reverberation,
the detection and classification of in-volume mines can be improved by steering
the beam up, as indicated by the dashed beam outline in the figure - at the range
of mine 1, the wavefront of the steered beam has not reached the sea bed and
therefore there will be no bottom reverberation (except through the beam sidelobes
at much lower levels).
Figure 11.1 Backgrounds to the detection of mines

The Threats 235
A mine hunting sonar, therefore, will benefit from having more than one fan of
beams available. A versatile design would include several fans, all or some of
which should be displayed simultaneously. Alternatively, split-beam processing in
the vertical plane may be used to achieve the same result.
11.4 Range and Bearing Resolution

A unified approach to detection and classification, as we saw in Chapter 9,
demands sufficient resolution in both range and bearing dimensions to resolve the
structure of both target and non-targets (false alarms) and, for mines, a bandwidth
of 75 kHz was suggested, giving a range resolution of 0.01 m.
A comparable resolution in the bearing dimension will be difficult - certainly
at 'long' range, which in this context might be 100 m or more. At 100 m a 1 wide
O
beam has a linear width of 1.7 m, acceptable for detection perhaps but too large to
resolve structure for classification. The use of a broadband FM pulse and phase
binning will improve the resolution; it is perhaps unrealistic to expect any better
than 0.25" from the phase binning, which would give about 0.5 m resolution at
100 m and 0.05 m at 10 m. Hence the suggestion that an ROY which can get close,
might be needed for the final classification of the most difficult targets.
11.5 Design of a Mine Hunting Sonar

The sonar must be capable of detecting and classifying both classes of mine at a
safe distance, at least 50 m? To ensure this, the complete system would ideally use
both hull-mounted and ROV arrays. The sonar may require a choice of pulse
length (certainly if only CW pulses are specified) and a choice of beamwidths,
comparatively wide for surveillance but much narrower for classification. The
operating fiequency follows from the bandwidth requirement. We can now
proceed to design the sonar.
11.6 The Threats

In-volume mine: spheroid, 1 m X 0.3 m, TS = -30 dB
0 Bottom or close tethered mine: irregular shape, minimum dimension 0.1 m,

maximum dimension 1 m, TS = -30 dB
11.7 Design Example

Let us begin with an outline specification (italic in Table 11.1) for a mine hunting
sonar to combat the above threats, and then build up a more complete specifica-
tion. The arrays are hull mounted and the maximum slant range requirement is
300 m. The operating frequency is 200 kHz.
Table 11.1 Mine hunting sonar: outline speci-

fication
~~
Maximum range 300 rn

Frequency 200 kHz
Transmit beamwidths
vertical 5"
horizontal 90"
Receive beamwidths
vertical 1"
horizontal 1"
PL 65 dB at 300 m
Directivity index, D1
transmit 16 dB
receive 43 dB
Pulse length, T FM 10 ms
cw 125 p
Bandwidth, B FM 80 kHz
cw 8kHz
10logT FM -20 dB
cw -29 dB
Reverb backscattering,%!, -20 dB
TS -20 dB
5 log d 10 dB
5 log n 3 dB
To achieve narrow beamwidths and wide cover:
Transmit:in azimuth over an arc of 90",vertical beamwidth 5"

Receive: horizontal and vertical beamwidths, minimum 1"
The propagation loss will be approximated by spherical spreading plus absorption

(not quite so accurately in this instance, because the sea bed presents one boundary
after the wavefront reaches it).
Design Example 237
PL =20logvfar X (dB)
r = 300 m, a = 50 dB/km therefore PL = 65 dB
Transmit array
To cover all depths and ranges, the transmit array must be capable of being steered
in the vertical plane. It must therefore have a number of separate elements. The
required number of elements will be given by 8 = 100/n. The transmit array is
therefore a single vertical stave of 20 elements at spacing L / 2 . To reduce the
maximum steer needed, both transmit and receive arrays will be mounted at an
angle of depression of perhaps 30". Wavelength L is only 7.5 mm and therefore the
radiating surface is about 75 X 4 = 300 X m2. To avoid cavitation the maxi-
mum power (from Figure 1.2) is 2 kW/m2.
Because of the high frequency and short pulses, this may safely be increased to,
say, 10 kW/m2 or P = 3 W. The array will be baffled and therefore DIt =
3 + lOlog20 = 16 dB and
SL = 10 log P + 171 + 16 = 192 dB
Receive array
Again, to cover all depths, azimuths and ranges, the receive array must be capable
of being steered in both the vertical and horizontal planes. The beamformer will
produce fans of beams to meet the requirements of surveillance (the initial sweep)
followed by a more detailed look with narrower beams for classification. The array
must be large enough to form the narrowest beams; wider beams will be formed
by subsets of elements or by combining beam outputs. The required number of
elements will be given by 8 = 100/n, and since the narrowest beams are 1" in
both dimensions, n = 100. The receive array is therefore a square of side 0.75 m.
Again, the array will be baffled and therefore
DI, =3 + lOlog(100 X 100) = 43 dB

Pulses
The major problem for a mine hunting sonar is cZassiJication, particularly for
mines on the sea bed, which have to be detected and classified against a back-
ground of bottom reverberation. Therefore the pulses must have sufficient
bandwidth to be able to discriminate against discrete reverberation (clutter). We
will propose two pulse types and compare their performances:
FMpuZse: bandwidth 80 kHz, duration 10 ms. The pulse needs to be quite short
to avoid excessive dead range; a 10 ms pulse will give a dead range of 7.5 m.
The target will not have any doppler, hence a linear FM pulse will suffice. (Note
that the dead range is not absolute; as soon as the system goes into receive
mode, processing of the pulse begins but the full gain is not achieved until a
complete pulse is processed, i.e., 10 ms after the start of the receive period.)
CW pulse: a comparable bandwidth would imply a very short pulse

(1 /80 000 = 12.5 p).Such a short pulse would have negligible noise-limited
performance and a compromise is necessary. Let the CW pulse have a duration
of 125 ps. The bandwidth is then 8 kHz.
239
11.8 Performance
I2PL = SL + TS - N + DI + 1010gT - 5logd + 5 10gn I
0 + +
FMpulse: 2PL = 192 - 30 - 30 43 - 20 - 10 3 = 148 dB therefore PL =
74 dB. Using this pulse, the requirement for PL = 65 dB is easily met.
0 + +
CWpulse: 2PL = 192 - 30 - 30 43 - 39 - 10 3 = 129 dB therefore PL =
65 dB. Using this pulse, the requirement for PL = 65 dB is now onlyjust met.
The reverberation-limited active sonar equation is
1 lolog R = lOlog(B/&) - S b -- 11 + TS - 5 10gd + 5 log TZ I

When R is in metres, the constant is 11 dB.
* F-M pulse: 1 0 l o g R = 4 9 + 3 0 - 1 1 - 3 0 - 1 0 + 3 = 3 1 d B
therefore R =
1300 ni. The reverberation-limited range requirement is easily met with this
pulse.
CW pulse: 1OlogX = 3 9 + 30 - 11 - 30 - 1 0 + 3 = 21 dB therefore R =

139 m. The reverberation-limited range requirement is not met with this pulse.
Note that a shorter CW pulse, about 50 ps, would meet the reverberation-limited
range requirement, but not the noise-limited range requirement. More importantly,
however, even a 50 ps duration CW pulse has poorer range resolution than the FM
pulse. We therefore conclude:
The best solution for the composite detection and classification task is to use the
FM pulse.
If this conclusion appears to conflict with some in-service mine hunting sonars, it
is because of technological limitations when they were designed.
11.9 Classification
Classification demands the best possible discrimination in both range and azimuth.
In the range dimension, the resolved pulse length for the FM pulse is 12.5 p,
giving a runge resolution of about 9 mm. This is, of course, very small compared
to the dimensions of a mine and there will be many returns in the range dimension
to help with classification.
In azimuth, the linear resolution is proportional to range - at 100 m the
resolution is 1.7 m for the 1" wide receive beams. Only at very close ranges is the
resolution in azimuth comparable with the range resolution. Phase binning can
improve this, but realistically not more than a factor of 4 (0.25"). So the best we
can expect at 100 m is an azimuth resohtion of about 0.5 m. This resolution,
although quite large compared with the dimensions of most mines, means that,
dependent upon aspect, several returns from different bearings are still possible
from larger mines every ping.
An important aid to the classification of mines is the acoustic shadow projected
onto the sea bed by bottom objects. The effect is illustrated in Figure 11.2. At the
beginning of the receive period, the background is just noise. When the lower edge
of the vertical beam intercepts the sea bed, reverberation adds to the noise as the
pulse sweeps along the sea bed. Once the pulse has reached the mine then echoes
- which may not exceed the background - will be added to the background. The
Figure 11.2 Shadowgraph effect

Classlfication 24 1
presence of the mine interrupts the bottom reverberation, resulting in a shadow

where the background is noise only. At the end of the shadow range, bottom
reverberation returns until possibly interrupted again by another mine or other
sources of discrete reverberation. Note that the shadow can still be produced by a
mine or other source which is not strong enough to be detected from its echoes.
Shadows can therefore betray the presence of mines possessing insufficient TS for
normal detection from echoes. A shadow will only occur if the target extends (in
azimuth) across all or most of the beam or phase bin width.
11.10 Mine Avoidance

Mine hunters are specialized vessels dedicated to the location and clearance of
mines. Other vessels will clearly benefit fiom the provision of a sonar that can
detect at least in-volume or near-surface mines so that avoiding action may be
taken. Bottom mines are far more difficult to detect and classiQ, and a mine
avoidance sonar will almost certainly not detect the worst of them; if bottom
mines are suspected, a mine hunting sonar will normally have to be used.
A mine avoidance sonar will differ from a mine hunting sonar primarily in its
range and azimuth cover requirements: the fitted vessel must detect mines
sufficiently early to take avoiding action, and the maximum range of a mine
hunting sonar is more akin to the minimum range of a mine avoidance sonar.
It is possible to adapt a sonar whose primary function is to detect submarines
for the role of mine avoidance. It would employ an FM pulse using all the
available transducer bandwidth and having a shorter than usual duration to avoid
too long a dead range. A useful performance against in-volume and near-surface
mines is possible, but because of the limited bandwidth available, bottom and
close tethered mines will be very difficult to detect, let alone classify, at avoidable
ranges. Furthermore, the TS of a mine falls with reducing frequency (the mine is
smaller in terms of the wavelength), although this is partially compensated by the
backscattering strength of bottom reverberation also falling with reduced fie-
quency. Steering the beams in the vertical dimension will again be useful in an
attempt to ensure that the background to detection is noise only. The resulting
design will not be optimum and the preferred approach is to use a dedicated sonar
designed for this task.
The following two sonars - one an adaptation and the other a dedicated design
- will indicate the performances achievable from the two approaches.
Mine Avoidance Sonars 243
11.l 1 Mine Avoidance Sonars

1 1.11.1 Adaptation of an antisubmarine sonar
The hypothetical sonar used for comparisons between full- and half-beam proces-
sing will be used again here. Because we need the smallest possible beamwidths,
phase binning is used and a CW pulse will not be considered. The bandwidth
available from the omnidirectional projector is assumed to be half an octave, or
5 kHz. Table 11.2 shows the parameters, modified where appropriate (modified
parameters are in italics).
Table 11.2 Adapting an antisubmarine sonar

Frequency 10 kHz
FM bandwidth 5000 Hz
Beamwidth, full aperture 8 degrees
Phase bin width 1 degrees
RI for FM with halfbeams 37 dB
DI (receive) 12 dB
Target strength, TS -30 dB
5 log d 10 dB
5 log n 3 dB
Background noise level 50 dB
sb -30 dB
Modified parameters are in italics
Detection performance
The noise-limited performance is given by
and using the parameter values of Table 1 1.2, we have
2PL =I 210 - 30 - 5 0 + 12 - 10 - 1 0 + 3 = 125 dB

If we assume spherical spreading and absorption, the range (from Figure 3.3) is
1000 m.
The reverberation-limited performance is given by
I 10 log R = RT - sb - 1 1 + TS - 5 log d + 5 log n I

1010gR=37+3O- 11 - 3 0 - 1 0 + 3 = 19dB
The reverberation-limited range is only 80 m. So a mine with TS = -30 dB

against a reverberation background of s b = -30 dB will not be detected at a
realistic avoidance range. This is quite a high reverberation background but it does
show the advantage of being able to steer the beams upwards to avoid bottom
reverberation. In-volume and near-surface mines can then be detected against a
background of sea surface reverberation that is unlikely to have a backscattering
strength of more than -40 dB. The reverberation-limited range would then be
800 m, comparable with the noise-limited range. The most difficult bottom mines
may well have target strengths even less than -30 dB and therefore, even given a
benign reverberation background ( s b = -40 dB, say), medium Frequency sonars
are unlikely to be able to detect all mines on or close to the sea bed. Furthermore,
the poorer resolution, implicit in the smaller bandwidth, will make the classifica-
tion of all mines more difficult.
11.11.2 Dedicated design
The range necessary for reliable mine avoidance is unlikely to be achieved at the
very high frequencies used for mine hunting. The frequency chosen must, how-
ever, be high enough to provide sufficient bandwidth for a good reverberation-
limited performance. We have already seen that an FM pulse gives superior
performance in all environments. (A CW pulse can only be better given target
doppler, which is not normally the case with mines.)
Let us begin with an outline specification (italic in Table 11.3), for a mine
avoidance sonar to combat the threat posed by a mine of TS = -30 dB, and then
build up a more complete specification. The arrays are hull mounted and the
maximum range requirement is 1000 m. The operating fiequency is 80 kHz and
the bandwidth 40 kHz (half an octave).
Table 11.3 Mine avoidance sonar: outline

specification
Maximum range 100 m
Frequency HO kHz
Bandwidth 40 kHz
Transmit beam widths
vertical 20"
horizontal 90"
Receive heamwidths
vertical 30
horizontal 2"
PL 83 dB at 1000 m
R1 (= 10 log(40 000/2) 43 dB
Directivity index, Dl
transmit 10 dB
receive 45 dB
Source leve, SL 193 dB
10 log T -10 dB
TS -30 dB
5 log d 10 dB
5 log n 3 dB
To achieve narrow beamwidths and wide cover:
Transmit: in azimuth over a forward arc of 90, vertical beamwidth 20"

* Receive: horizontal and vertical beamwidths, minimum 2"
The propagation loss is calculated using spherical spreading plus absorption (not
very accurately because the sea surface and the sea bed will present boundaries for
much of the ping).
PL = 201og Y + aY X (dB)
Y = 1000 m, a = 23 dB/km therefore PL = 83 dB
Transmit array
The height of the transmit array is given by 0 = 20" = 1 OO/n. The transmit array
is therefore a single vertical stave of 5 elements at spacing d/2 and mounted at an
angle of depression of 10". Wavelength ;1 is only 18.75 mm and therefore the

radiating surface is about 300 X 5 = 1500 X m2 (i.e., 5 circular elements of
diameter 18.75 mm). To avoid cavitation the maximum power (from Figure 1.2) is
2 kW/m2. Because of the high frequency and the short pulses, this may safely be
increased to, say, 10 kW/m2 or P = 15 W. The array will be baffled and therefore
+
DIt = 3 lOlog5 = 10 dB and
SL = 1010gP+ 171 + 10 = 193 dB
Receive array
To cover all depths and azimuths, the receive array is an incomplete cylinder made
up of a number of staves spaced at 1/2 (Figure 11.3). Each stave will have
100/2 = 50 elements. The diameter of the array is given by &, = 88/df. There-
fore diameter d = 88/(2 X 80) = 550 mm. Each full beam, or codirectional pair
of half-beams, is produced using 120" of the array. To provide 90" of azimuth
cover, the staves must occupy 240" around the periphery of the cylinder. There will
therefore be a total of (nX 550 X 2)/(3 X 9.375) = 120 staves. The ahead beam,
for example, will use the 60 staves in segments B and C. Adjacent beams are
formed by stepping around the array by one stave (the staves have an angular
spacing of 2"). Several fans of beams in the vertical dimension will be formed to
cover the required depths and range. The near-horizontal fans will avoid bottom
reverberation until quite long ranges, well beyond avoidance range.
The directivity index of the array is given by the formula for a cylindrical array
used previously:
DI, = 10 log(5hdf2) = 10 log(5 X 0.9375 X 0.55 X 802) = 42 dB
3
Ahead beam uses the 60 staves in segments BandC
Figure 11.3 Receive array: plan view

FM pulse
The FM pulse has bandwidth 40 kHz and duration 100 ms. The pulse needs to be
quite short to avoid excessive dead range; a 100 ms pulse will give a dead range
of 75 m.The target will not have any doppler, hence a linear FM pulse will do.
Detection performance
I2PL = SL + TS - N + DI + lolog T - 5 logd + 5 log TZ I
and using the parameter values of Table 1I .3, we have
2PL 193 - 30 - 35 + 42 - 10 - 10 + 3 = 153 dB

giving PL = 77 dB and the range is about 800 m. To achieve 1000 m, PL must be
83 dB, i.e.,we need another 6 dB. Several possibilities are open to us:
The noise level is SS4; if we had taken SS2, the noise level would have been
30 dB.
The pulse length could be increased (but watch the dead range); increasing T to
300 ms would increase PL by 2.5 dB.
Increasing SL by 10 dB should be possible; the receive DI is already high (large

aperture array). A selection of these changes should provide the extra decibels.
Note that because of the 2PL in the equation, an extra 12 dB is needed to gain
6 dB in PL.
The reverberation-limited performance is given by
1010gR=43+30-11 -30- 10+3=25dB
The reverberation-limited range is therefore 3 I 6 m. So against a reverberation

background of Sb = -30 dB, a mine with TS = -30 dB is detected at a range

where avoidance is still possible, but less than the noise-limited range of
800-1000 m. This is quite a high reverberation background but again it shows the
advantage of being able to steer the beams upwards to avoid bottom reverberation.
In-volume and near-surface mines can then be detected against a background of
sea surface reverberation, which is unlikely to have a backscattering strength of
more than -40 dB. The reverberation-limited range would then be 3 160 m, much
greater than the noise-limited range. The most difficult bottom mines may well
have target strengths even less than -30 dB, so even a dedicated mine avoidance
sonar will find it difficult to detect all mines on or close to the sea bed.
Classification
Classification demands the best possible discrimination in both range and azimuth.
In the range dimension, the resolved pulse length for the FM pulse is 25 p,giving
a range resolution of about 18 mrn. This is very small compared to the dimensions
of a mine, and there will be many returns in the range dimension to help with
classification.
In azimuth the linear resolution is proportional to range. Classification can be at a
closer range than initial detection; if we assume 300 m, the resolution is 10 m for the
2" wide receive beams. The resolution in azimuth is never remotely comparable with
the range resolution. Phase binning can improve this, but realistically not more than
a factor of 8 (0.25"). So the best we can expect at 300 m is an azimuth resolution of
about 1.3 m. This resolution is comparable with the dimensions of most mines,
therefore the majority of returns from a mine will be at the same bearing and a
complete picture of a mine will not be available. However, even this resolution is
useful for classification because it will discriminate against large discrete patches of
bottom reverberation, which can extend over much greater dimensions.
11.12 Problems
1 1.1 How would the mine hunting sonar of Section 1 1.7 perform if TS were -25 dB?
11.2 The mine avoidance sonar design of Section 11.1 1.2 took the PL at the centre
frequency of the pulse (80 kHz). At the highest frequency of the pulse, what SL is
necessary to achieve a range of 800m? And assuming SL does not change with
frequency over the bandwidth of the pulse, what equalization is needed to maintain a
flat response at the receiver?
12
Intercept and Communications
Sonars
12.1 Intercept Sonars

An intercept sonar detects active transmissions from other sonar platforms -- a
submarine intercept sonar, for example, will have to be capable of detecting and
classifying active transmissions from surface ships, other submarines and helicop-
ter dipping sonars, and the active transmissions from torpedoes in their homing
stages.
An intercept sonar, unlike an active sonar, will have no a priori knowledge or
incomplete a priori knowledge of the signals it has to detect. Hence matched filter
processing cannot - at least initially - be used and the basic intercept receiver
simply employs energy detection.
The arrays and processor will have to accept a wide frequency spectrum - from
about 1 to 20 kHz for search sonar transmissions, and from about 20 to 100 kHz
for weapon homing transmissions. The arrays will not be optimum throughout this
band and typically the DI might be between 3 and 10 dB.
Nevertheless, because the intercepted pulse has only undergone a one-wail
propagation loss - from source to receiver -- compared with the two-wuy
propagation loss of the intercepted active sonar, intercept ranges will be at least
equal to, and typically several times greater than, the ranges achieved by the active
sonar.
250 Intercept and CommunicationsSonars
12.2 Communications Sonars

A communications sonar receives signals (voice or encoded messages) from
another platform. But now, unlike an intercept sonar, the receiver will have an a
priori knowledge of the signals it has to detect. Hence matched filter processing
can now be used.
The bandwidth of the receiver will therefore only need to be wide enough to
accept the frequency spectrum of the messages. Intelligible speech has a band-
width of about 3 kHz (100-3000 Hz) but encoded messages may have much
larger bandwidths, particularly if the communications channel must be covert - to
be secure from intercept, the message must be fast and of short duration, which
implies a large bandwidth.
As with intercept, the signals only undergo a one-way propagation loss, and
long ranges are possible using only modest source levels.
12.3 Function of an Intercept Sonar

An intercept sonar examines detected pulses to determine the following:
Carrier frequency
Bandwidth
Pulse type (FM, CW, PRN, etc.)
Pulse duration
Pulse shape
Pulse interval
Multipath structure
With the exception of multipath structure, all serve to help identify the transmitter
and hence the likely platform.
The carrier frequency may be between, say, 1-100 Wz. A typical receiver to
cover this spectrum might do its surveillance processing in octave bands
(1 -2 kHz, 2-4 kHz, etc.,), but should also have the capability to vary the exact
bands given some a priori knowledge of likely threats.
Bandwidths may be between, say, 1 Hz (a long CW pulse) and 1000 Hz (a short
CW pulse or broadband FM or PRN pulses). The intercept sonar should at least
Function of an Intercept Sonar 2s 1
measure start and finish frequencies and ideally it would completely analyse the
amplitude and frequency structure of the pulse.
Pulse interval, and any change in the interval, are important clues to the likely
range of the intercepted sonar. Suppose pulse interval = 15 s (corresponding to a
maximum range of about 10 km) the transmitter will be within this range,
assuming it is in contact. If the pulse interval changes to say 8 s, this is a strong
indication that the transmitting platform is closing range and is within 5 lun of the
intercept sonar.
Time delays between direct and bottom bounce arrivals in deep water can also
indicate range, but again only very broadly given that the simple broadband
intercept receiver is unlikely to be able to make a good measure of the vertical
angle of arrival. The large planar or conformal arrays fitted to submarines,
however, can measure vertical angles of arrival to sufficient accuracy to give a
good estimate of range, either from time delays between direct and indirect
arrivals of the pulse, or from the vertical angle of maximum response (compare
with VDPR). Given a knowledge of the original pulse length of a transmission, it
may be possible to infer the range of its source from any stretching of the pulse
resulting from multipath. The range estimate may be no more than that if the pulse
is stretched then the source is distant, and if the pulse is not stretched then the
source is close.
252 Intercept and Communications Sonars
12.4 Intercept Sonar Equation

An intercept sonar is a passive sonar optimized for the detection of active
transmissions from other platforms. The basic intercept sonar equation is there-
fore
SE = (SL - PL) - N + DI - DT
This is the basic passive sonar equation: SL is no longer the radiated noise of a
target but is the source level of the active transmission.
Because nothing can be assumed about the signal to be detected, the detection
process for the simplest intercept sonar is based on the received signal-to-noise
ratio without any processing gain, therefore
Both signal and noise are referred to the same bandwidth, hence the term 10 log Br
where B, is the bandwidth of the receiver.
Combining these equations and putting SE = 0, we obtain the intercept sonar
equation:
Worked Examples 253
12.5 Worked Examples

Example 12.1
An intercept receiver mounted on a submarine processing an octave band from 6 to
12 kHz detects a signal of bandwidth 400 Hz centred on 9 Wz. Given spherical
spreading and absorption, what is the maximum range of the source of the signal,
assuming it is from a sonar with an SL of 220 dB? (The submarine is aware that the
sonar is likely to be Type XXX on a surface ship with a standard SL of 220 dB.)
Assume that DI =6 dB at 9 kHz and background noise N = 40 dB:
PL = 220 + 6 - 40 - 10 - lolog60 = 138 dB
PL = 2010g Y + 0.9~
X = 138 dB
Intercept range = 50 km
How does this compare with the detection range from the ship?
2PL = SL + TS - N + DI + 1OlogT - 5logd + 51og n

Take N = 60 dB, TS = 10 dB, DI = 20 dB, T = 1 s, 5 logd = 10 dB, n = 5:
2PL=220+10-60+20+0- 10+3=183dB
PL = 20 log Y + 0 . 9 ~
X = 92 dB
Detection range = 12 km
Therefore the presence of the ship can be known to the submarine long before it
becomes an actual threat.
Example 12.2
60 W z detects a signal of bandwidth 1000 Hz centred on 50 kHz with a ping interval
of 0.5 s. What are the probable source and range of the signal?
Assume DI = 3 dB and N = 30 dB. The frequency and ping interval indicate that the
source is probably a torpedo in its active homing phase. The SL may therefore be
about 200 dB.
PL=200+3-30- 10- lOl0g30000= 118dB
PL = 20 log r + 16r X = I18 dB

Intercept range = 3 km
What is the probable detection range of the submarine by the torpedo?
2PL = SL + TS - N + DI + lolog T - 5 logd + 5 log n

Take N = 4 0 d B , T S = 10dB,DI = 20dB, T = 20ms, 5logd = IOdB, n = 5 :
2PL=200+ 10-40+20- 17- 10- 166dB
PL = 20logr + 16r X = 83 dB
Detection range = 1.3 km
But because the ping interval is only 0.5 s (corresponding to a range scale of about
400 m), the submarine may well infer that the torpedo is well within this maximum
range.
So the submarine intercept receiver is capable of detecting the torpedo active

transmissions well beyond the torpedos active acquisition range. Therefore the
torpedo launch platform (submarine, surface ship, aircraft) should attempt to closely
localize the target and delay homing transmissions until the torpedo is within a few
hundred metres of the target.
Reduction in the Probability of Intercept 25s
12.6 Reduction in the Probability of Intercept

Submarines are becoming increasingly difficult to detect passively, even from
other submarines, and the active capability of submarine sonars may perforce have
to be used for detection purposes (rather than simply to localize a target
immediately before an attack).
The active transmissions should therefore be designed for a low probability uf
intercept (LPI) and also, recognizing that this is very difficult and may not be
achieved, for a low probability oj' exploitation (LPE), i.e., having detected the
transmission it should be difficult for the intercept receiver to infer anything about
the identity of the intercepted platform and its movements.
Unfortunately, designing for LPI is in direct conflict with designing for
detection. A low probability of intercept may be achieved in several ways:
e Reducing SL
Increasing frequency to limit range by absorption losses
e Increasing the bandwidth
Using sector transmissions to confine the energy within azimuths of interest
If increased bandwidth is achieved by reducing the pulse length, as with CW, then
detection performance will also suffer. If it is achieved without reducing the pulse
length, detection performance will not suffer, but since the intercept receiver
bandwidth is likely to be larger still, the performance of the intercept sonar will
not suffer either.
Clearly, with the possible exception of sector transmissions, all these measures
severely prejudice the detection performance of an active sonar. But a simple
example will suffice to show that LPI may not be usefully improved even by the
use of sector transmissions.
Example 12.3
8 kHz detects a signal centred on 5 kHz from a sonar with a known sector SL of
230 dB. Given spherical spreading and absorption, what is the maximum intercept
range of the sector transmission and the reduced intercept range for bearings outside
the sector, for transmission sidelobes of -30 dB?
Assume DI =6 dB at 5 kHz and assume N = 40 dB:

Within the sector, SL is 230 dB and PL = 230 + 6 - 40 - 10 - 10 log4000

= 150 dB. PL = 20 log r + 0.3r X low3= 150 dB therefore r = 150 km.
Outside the sector, SL is 200 dB and PL = 200 + 6 - 40 - 10 - 10 log4000 =

120 dB therefore r = 70 km.
So the intercept range outside the sector (although only about half the range within the
sector) is still very long and much greater than the probable maximum range of the
intercepted sonar, perhaps 20 km.A useful goal would be to reduce the out-of-sector
intercept range to this value, i.e., 20 km. What level of sidelobes would achieve this?
PL = 20 log 20 000 + (0.3 X 20) = 92 dB
Therefore the sidelobes would need to be 150 - 92 = 58 dB down. This sidelobe level
is quite impractical and, even theoretically, it could only be achieved by severely
compromising the sector SL and beamwidth.
Since all measures to reduce LPI compromise the performance of an active sonar, it is
reasonable to conclude:
Virtually any sonar transmission will be intercepted at a range far in excess of its
own detection range.
The emphasis must therefore fall upon LPE. Note, however, that mine hunting sonars,
which operate against close targets using high frequency transmissions, have a very
low probability of intercept. This is because absorption severely reduces the range
where the sound intensity becomes negligible.
12.7 Reduction in the Probability of Exploitation

Useful LPE measures will include those designed to hinder platform identification
and platform movement and location. The features of an intercepted transmission
examined by an intercept receiver (listed above) are exploited to identzh the sonar
and hence the likely platform types. The identification may, at best, be made non-
specific by ensuring that transmissions from the different platforms within a force
are as similar as possible. Once a sonar has been designed, transmitted and
intercepted by an adversary, there is little that can be done to hinder any future
identification of the sonar.
It is possible to infer range rate (relative velocity) from the doppler shifts to
intercepted transmissions of known frequencies. The doppler shift is 0.35 Hz per
knot per kHz given ODN. (Compare this with 0.69 Hz per knot per kHz for an
Communications Sonars 257
active sonar.) A useful LPE measure, therefore, is to randomize the frequency of a

pulse so that the intercept receiver infers an incorrect range rate. If, for example,
the frequency of a transmission from a platform closing at a relative velocity of 10
knots is reduced from its nominal value of 5 kHz by 35 Hz, the intercept receiver
could infer that the platform is opening at a relative velocity of 10 knots.
The use of a range scale greater than necessary will reduce the intercept
opportunities and can confuse the intercept receiver about the range of the
intercepted platform, and whether or not the active sonar is in contact (particularly
if the range is known).
12.8 Effectiveness of Intercept Sonars

Intercept sonars can be exceedingly effective; they dissuade the use of active
sonars - particularly by submarines with a strong desire to remain covert -
because it is difficult if not impossible to avoid being detected and identified
(usually long before the active sonar is in contact itself) by even a mediocre
intercept receiver. Probably the best that an active sonar can do is to attempt
to mislead the intercept receiver by making small random changes in transmit
frequencies and making range scale changes which are at odds with the real
tactical scenario.
12.9 Communications Sonars

Communications sonars are normally used for communicating between surface
ships and submarines, or between submarines. They may simply use voice
channels (underwater telephones) or, particularly when security and integrity are
important, encoded messages. The techniques used for encoding and modulation
will only be considered here insofar as they impinge on the design of the
underwater communications channel itself, i.e., the bandwidths needed for voice
and message channels.
Voice channels suffer significant distortion during underwater propagation,
particularly at low frequencies (say 1- 12 kHz) where the attenuation in the water
can differ markedly across the 3000 Hz bandwidth required for the channel.
Reverberation and multipath also add to the distortion. This distortion may be
removed by digital encoding before transmission but, inevitably, at low frequen-
cies the communications link will be slowed down. A low frequency channel has
the incidental advantage that it can often use an existing search sonar array,
transmitters and receivers.
Secure encoded message channels must be of short duration to reduce intercept
opportunities. Therefore they need more bandwidth than a simple voice channel.
A typical bandwidth is 10-50 W .The actual operating fiequencies are corre-
spondingly high and the range much reduced.
12.10 Communications Sonar Equation

A communications sonar may be considered as a special case of an intercept sonar.
Both sonars receive active transmissions, but the communications sonar knows the
nature of the signal to be detected whereas the intercept sonar does not. The basic
communications sonar equation is therefore
SE = (SL - PL) - N + DI - DT
This is the basic passive sonar equation: SL is no longer the radiated noise of a
target but is the known source level of the active transmission.
The detection process for the simplest communications sonar is based on the
received signal-to-noise ratio, in the bandwidth of the communication channel,
and therefore
D T = 5 1 0 g d + 10logBc
Both signal and noise are referred to the same bandwidth, hence the term
10 log B, where B, is the bandwidth of the communication channel.
Combining these equations and putting SE = 0, we obtain the communications
sonar equation:
I PL = SL + DI - N - 5 10gd - lolog Bc I
Examples of Communications Sonars 259
12.1 1 Examples of Communications Sonars

Example 12.4
An underwater telephone is required to use an existing surface ship search sonar array for
communication with submarines. The array is tuned over a band from 8 to 10 kHz where
its SL = 215 dB, and its response falls by 12 dB per octave away from resonance. The
telephone is to use a band firom 12 to 15 kHz. Assume that a dedicated system on the
submarine will deliver the same performance. What is the SL needed for a range of4 km?
The communications sonar equation is
I PL = SL + DI - N - 5 log d - 10 log B, I
Let DI = 20 dB, 5 log d = 10 dB, N = 45 dB (SS4 at 12 kHz, the vessels will
be stationary or slow moving and the background will be ambient sea noise),
B, = 3000 Hz. Assume spherical spreading and absorption:
PL = 2010g4000+ (1.6 X 4) = 78 dB
SL = 78 - 20 + 45 + 10 + 35 = 148 dB
This very modest SL results from propagation being one-way only, and the transmitter
power will need to be significantly reduced for this application of the array in order to
minimize the probability of intercept.
The dedicated system on the submarine will also have modest requirements. It may
only be necessary to provide a suitable projector and use an existing passive array
(possibly an intercept array). The SL of the submarine array must be adjustable to
minimize the probability of intercept, and in practice both vessels would reduce their
transmitted powers to levels consistent with intelligible communication.
Example S2.5
An encoded message channel uses a band fi-om 50 to 1OOkHz. What SL is now
required for a range of 4 km, the same range as in Example 12.4?
Let DI = 30 dB, 5 logd = 10 dB, N = 35 dB (SS4 at 50 kHz), B, = 50 kHz and

assume spherical spreading and absorption:
At50kHz
PL = 2Olog4000 + (16 X 4) = 136 dB

S L z 136-3O-t 3 5 + 1 0 + 4 7 = 1 9 8 d B
260 Intercept and CommunicationsSonars
At 1OOkHz
PL = 2010g4000+ (35 X 4) = 212 dB

SL = 212 - 30+35 + 10 + 4 7 = 274 dB
The large difference in SL between the ends of the band is range dependent, even if a
source level of 274 dB were possible, it would be very difficult to equalize success-
fully. Therefore a channel of this bandwidth is only capable of much shorter ranges. A
more realistic design would aim for a range of 1 km,say, then:
At 50 kHz
PL = 2010g 1000 + (16 X 1) = 76 dB

SL = 76 - 30 + 35 + 10 + 47 = 138 dB
At 1OOkHz
PL = 2010g 1000 + (35 X 1) = 95 dB

SL = 95 - 30 + 35 + 10 + 47 = 157 dB
Now the difference in SL between the ends of the band is only 19 dB and easy to
equalize.
Fortunately, a modest range is often acceptable: it is easy for a quiet submarine to

detect and locate a relatively noisy surface vessel and then approach to, say, 500 m
before establishing communication. Both vessels should reduce their source levels to
the minima required for reliable communication (perhaps 10 dB greater than the usual
values determined on detection criteria alone).
12.12 Problems
12.1 An intercept receiver mounted on a submarine processing an octave band fiom
10 to 20 IcHz detects a CW pulse of duration 1 s at a frequency of 15 kHz. Given
spherical spreading and absorption, what are the range limits of the source of the
signal, assuming it is fiom a sonar whose SL can be varied between 180 and 220 dB?
Assume that DI = 10 dB at 15 kHz and background noise N = 40 dB.
12.2 If the submarine has a TS = 15 dB and the DI of the intercepted sonar is 20 dB,
what are the intercepted sonars likely noise-limited ranges at these limits of SL?
13
Active Sonar Design
13.1 Introduction
This chapter will use the concepts of the previous chapters to design representative
active sonar systems for the detection of submarines and torpedoes. The inter-
actions between the initial requirements and the limitations imposed by platform
size and the environment will be demonstrated by considering practical systems.
13.2 Submarine Detection

The requirement for the detection of submarines may be stated simply as: to detect
and classify at beyond (the submarines) weapon range, at an acceptable (low)
false alarm rate. This requirement will, in practice, not be met for all submarines,
in all environments, at all times. The designer will have to consider the constraints,
primarily the platform size, and design for the best performance possible.
13.3 Hull-Mounted Surface Ship Sonar

Hull-mounted antisubmarine sonars are widely used in surface ships (and sub-
marines). When fitted in ships, this class of sonar generally has the following
characteristics:
Cylindrical array; 1-3 m in diameter; 0.5 to 2 m in height; the size is

constrained by the platform.
262 Active Sonar Design
Mounted in a keel dome or, more commonly for modern sonars, in a bow dome;
the bow site offers the lowest self-noise.
Operating frequencies between 3 and 15 kHz; the frequency is chosen (it almost
chooses itself) to give an acceptable balance between performance against noise
and reverberation backgrounds for the array size.
Detection ranges from 5 km( 15 kHz) to 20 km(3 kHz) but highly dependent on
the environment.
Representative Hull-Mounted Design 263
13.4 Representative Hull-Mounted Design

The largest possible array for the ship has been determined to be 2 m in diameter.
For the greatest possible noise-limited range, the operating frequency should be
low. Too low, however, and the horizontal beamwidth will be too small for
adequate performance against reverberation. The receive DI will also fall with
reducing frequency, which will reduce the noise-limited range, thus creating an
opposing requirement to increase the frequency. By iteration through a design, an
optimum frequency will finally emerge. An experienced designer might start by
specifying the beamwidths and proceed as follows:
Horizontal beamwidth 8h = 8"
Vertical beamwidth 8, = 12"
We have 6 h = 88/dfo therefore fo = 5.5 kHz, and we have OV = 76/hfo therefore

h = 1.15 m.
The receive DI is 10 log 5 h d f i = 25 dB. For all-round cover use an omnidirec-
tional transmission. The transmit DI will then be that of a line of length I . 15 m
(the height of the array):
Transmit DI = 10log(2h/l) = 9 dB
The source level is SL = 10 log P + 171 + DIt
Ifwe specify SL = 224 dB then the total acoustic power is P = 25 kW
The array elements should be spaced no more than l / 2 = 136 mm. Make the
actual spacing 120 mm. A stave can therefore have 1 150/120 = 10 elements.
(Note all dimensions are approximate and would be modified by practical
considerations.)
At the crossover points of the beams there will be a 'scalloping loss' of 3 dB. To
limit this loss to about 1 dB, the staves must be spaced less than a beamwidth, say
6", so there will be 360/6 = 60 staves. There will therefore be a total of 600
elements in the array and, in transmit, each element must be capable of transmit-
ting 25 000/600 = 40 W of acoustic power. Suppose the radiating surface of each
element to be ~ ( 0 . 0 5 then
) ~ , the total radiating surface is 600r~(0.05)~
= 4.7 m2.
The radiated power intensity is therefore 25/4.7 = 5.3 kW/ m2 and Figure 1.2
indicates that cavitation could be a problem unless the array is at a depth of at
least 10 m. Practical arrays, however, exist and operate at about this intensity at
lesser depths, so it is safe to continue with this design.
Based on the above analysis and the procedures and recommendations from
previous chapters, we are now in a position to attempt to complete Table 13.1, a
table of parameters for the sonar.
Table 13.1 Parameters of hull-mounted surface

ship sonar
Frequency 5.5 kHz
Pulse length, T 1 S
FM bandwidth 750 Hz
CW bandwidth 2 Hz
Horizontal beamwidth, 8 degrees
Reverberation index, Ri
FM 20 dB
cw -6 dB
DI (receive) 25 dB
Target strength, TS 10 dB
5 logd 12 dB
5 log n 3 dB
Background noise 55 dB
-33 dB
Pulse length
The pulse length, T, has been chosen to be 1 s for both FM and CW pulses.
Increasing T beyond this might increase the noise-limited range slightly but it
does have operational implications - the sonar would be blind out to 1500T/2
metres, or even longer for the first pulse if both pulses are transmitted sequentially.
The other argument for increasing T - improved doppler resolution and hence
possibly improved performance against very low target doppler - is particularly
important for a stationary platform (e.g., a helicopter dipping sonar). Chapter 9
discusses the CW target doppler problem in detail but note here that a 1 s shaped
CW pulse will have a bandwidth of about 2 Hz, which corresponds to a doppler
width of about 0.5 knots at an operating frequency of 5.5 kHz. This is already
comparable with spreads in the reverberation background, so increasing the pulse
length will not be useful.
Representcitive Hull-Mounted Design 265
FM bandwidth
The FM bandwidth is chosen to resolve the structure of targets and non-targets, as

discussed in Chapter 9.
Reverberation indices
For FM, 10 log(750/8) = 20 dB

0 For CW, 1010g(2/8) = -6 dB
Detection index
The detection index is 5 log d = 12 dB. This value, for an active sonar, has
P d = 0.9 and Pf, = A decision after n pings, made by the operator or an
automatic detector, results in an incoherent gain of 5 log n(dB). For n = 5 we have
3 dB.
Background noise
The background noise, 55 dB, is the spectrum level of sea state 4 (SS4) ambient
noise at the operating frequency of 5.5 kHz. It is likely to be the dominant source
of self-noise for a typical fkigate at speeds of up to about 15 knots.
Target strength
The target strength (TS) is an average value for a typical submarine. Should the
same value be used for both pulses? It is usual for the TS to have been measured
using a CW pulse of a duration which roughly matches the target extent, say
100 ms for a submarine, and it is the peak value which is quoted. For long CW
pulses, the peak TS equals the integrated TS and the suggested value of 10 dB is
therefore appropriate.
The large bandwidth of the FM pulse means that its effective pulse length is
short, 1.3 ms, and the peak TS does not equal the integrated TS. The correct value
to use for TS is now somewhat less (the peak TS as measured by an equivalent
short pulse, which is perhaps about 5 dB less). However, we can use the same,
10 dB, value if we integrate the post-detection samples, but do not include a term
in the sonar equation for this. (In other words, the signal processing does the
integration and not the target.)
Performance
The noise-limited performance of the sonar is given by the following equation:
I2PL = SL + TS - N + DI + lolog T - 5 logd + 5 log n I

Substituting the values fiom Table 13.1, we obtain
2PL = 224 + 10 - 55 + 25 + 0 - 12 + 3 = 195 dB

The allowable PL is 98 dB, and assuming spherical spreading plus absorption
(a = O S ) , the noise-limited detection range is about 22 km. This range is achieved
for both the FM and CW pulses because they have the same duration.
The reverberation-limited, performance of the sonar is given by the following
equation:
I 10 log R = 10 l~g(B/@,) - &, - 41 + TS - 5 log d + 5 log n I

Substituting the values from Table 13.1, we obtain
1010gR=20+33-41+10- 12+3=13dB
and the reverberation-limited range of the FM pulse is 20 km.

But consider the effect of reducing the TS to 6 dB. The allowable PL is now
96 dB and the noise-limited detection range is about 20 km (reduced but not a
dramatic change). The reverberation-limited range, however, is now only 8 km,
and the sonar is reverberation limited. The design, therefore, is only well balanced
given a 10 dB target and Sb d -33 dB.
The sonar may still, however, have a perfectly adequate performance. A value
of & = -33 dB is only exceeded in about 20 per cent of UK shallow water
areas and in deep water a value of &, = -40 dB is high. Note that the
reverberation-limited performance can also be enhanced by using half-beam
processing, which effectively reduces the value of & to be used in the equation
for RI.
The small, 2 Hz, bandwidth of the CW pulse results in a very low RI, and its
Towed Transmitters and Towed Array Receivers 267
reverberation-limited range is negligible. The CW pulse is only useful given

sznfJicient target doppler to ensure that detections are noise limited.
13.5 Longer Ranges

To detect submarines at ranges significantly higher than 20 km, it is necessary to
use a much lower frequency. Simply to use a lower frequency and the same array
size, however, would not be sufficient. In noise-limited conditions the reduced
absorption losses would be more than offset by a lower receive DI and increased
self-noise. For many ship designs, a 2 m diameter array is the largest practical and
a different approach is required.
13.6 Towed Transmitters and Towed Array Receivers

The requirement for increased range (to counter increased weapon ranges) is often
met using a combination of a towed body transmitter and a towed line array
receiver (Figure 13.1). This combination will have a, perhaps limited, variable
depth capability and operate at frequencies between, say, 300 and 2000 Hz. The
receive array will be quite short compared with a passive towed array and it may
be attached to the transmitting body.
Figure 13.1 Towed active sonar system

13.7 Representative Design

Consider a low frequency active sonar operating at 1000 Hz. The transmit array
comprises a vertical stack of elements of height 3 m, or 21, and transmitting
omnidirectionally in azimuth. The source level is again 224 dB. The vertical
beamwidth is 0, = 501/L = 25". The receive line array is 321 in length, which is
48 m at 1000 Hz. At broadside the horizontal beamwidth is @, = 501/L = 1.6";
@, increases at angles away from broadside, as shown in Figure 2.2. If we define
operational bearings as f60", then the maximum value for &, is approximately 3".
The receive DI is 10 log n = 18 dB. Once again we can complete a table of
parameters for the sonar (Table 13.2), based on the above and the procedures and
recommendations from previous chapters.
Table 13.2 Parameters of low frequency active sonar

Frequency 1 kHz
Pulse length, T 2 S
FM bandwidth 300 Hz
CW bandwidth 1 Hz
Horizontal beamwidth, 3* degrees
FM 20 dB
cw -2 dB
DI (receive) 18 dB
Target strength, TS 10 dB
5 log d 12 dB
5 log n 3 dB
sb -33 dB
*Maximum
Pulse length
The pulse length, T, has been chosen to be 2 s for both FM and CW pulses. A 2 s
shaped CW pulse will have a bandwidth of about 1 Hz, which corresponds to a
doppler width of about 1.5 knots at an operating frequency of 1 kHz. This is larger
than possible spreads in the reverberation background, so increasing the pulse
length of the CW pulse may be useful. Note well that as frequency is reduced,
Representative Design 269
there needs to be a greater target doppler (relative velocity in knots) to achieve

noise-limited performance with a CW pulse.
FM bandwidth
The FM bandwidth is now limited to 300 Hz by the available bandwidth for an

acceptable projector transmit efficiency. This is less than optimum to resolve the
structure of both targets and non-targets, but it is acceptable.
Reverberation indices
* For FM, lOlog(300/3) = 20 dB
For CW, 1010g(1/3) = -5 dB
Target strength
The target strength (TS) is an average value for a typical submarine.
Detection index
The detection index is 5 log d = 12 dB. This value, for an active sonar, has
P d = 0.9 and Pfa= A decision after n pings, made by the operator or an
automatic detector, results in an incoherent gain of 5 log n(dB). For n = 5 we have
3 dB.
Background noise
The background noise, 70 dB, is the spectrum level of sea state 4 (SS4) ambient
noise at the operating frequency of 1 kHz. The self-noise of a towed array is that
of the ambient sea noise at towing speeds up to perhaps 10 knots, except at ahead
bearings where the radiated noise of the tow ship may be dominant (see Example
8.2 1.
Perjormance
I 2PL = SL + TS - N + DI + 10 log T - 5 10gd + 5 log n I
2PL = 224 + 10 - 70 + 18 + 3 - 12 + 3 = 176 dB
The allowable PL is 88 dB, and assuming spherical spreading plus absorption

(a= 0.06), the noise-limited detection range is about 26 km. This range is
achieved for both the FM and CW pulses because they have the same duration.
This is significantly greater than the 22 km range of the hull-mounted sonar of the
previous example, operating at a fiequency of 5.5 Hz. Note that it has been
achieved at a much lower PL (88 dB compared with 98 dB). This is due to the
much smaller absorption losses at 1 Wz.
The reverberation-limited performance of the sonar is given by the following
equation:
I 10 log R = lOl~g(B/&)- &, - 41 + TS - 5 10gd + 5 log n I
10 log R = 20 + 33 - 41 + 10 - 12 + 3 = 13 dB
and the reverberation-limited range, using the FM pulse, is 20 km. It is the same
as the hull-mounted sonar because the RI is unchanged. (The smaller bandwidth is
compensated by the narrower horizontal beamwidth.)
The small, 1 Hz,bandwidth of the CW pulse results in a very low RI. and its
reverberation-limited range is negligible. As always the CW pulse is only useful
given sufficient target doppler to ensure that detections are noise limited. The
remarks on reverberation-limited performance made above for the hull-mounted
sonar apply equally here.
Representative Design 27 1
Here are some ways we could use to improve the noise-limited performance of
this sonar:
0 Double the length of the towed array = +3 dB
0 Increase SL = +3 dB
0 Transmit a train of five 2 s pulses

and incoherently add (5 log 5) = $3 dB
This is a total increase in 2PL of 9 dB, and PL becomes 93 dB. The noise-limited
range is now 35 km. Note that a similar increase in 2PL for the hull-mounted
sonar (which is probably only possible by resorting to sector transmissions) would
have increased its range to about 28 km.
A useful insight into the performance of any active system is obtained by
plotting the levels of echo, noise and reverberation against range:
Echo level = EL = + TS - 2PL

SL
Reverb level = RL = SL + (Sb + 10 log A) - 2PL
Noise level = NL = N - lolog T - DI
Using the parameters fiom Table 13.2 and assuming spherical spreading and
negligible absorption, here is what we obtain:
EL = 234 - 40 logR
RLCW 224 - 33 + 10 log(40R) - 40 log R = 207 - 30 log R
R L ~ ~ = 2 2 4 - 3 3 +lOlog(O.13R)-40l0gR= 182-3OlogR
The echo, noise and reverberation levels are plotted against range in Figure 13.2.
The performance results using the plots are for a single ping and are therefore not
directly comparable with the results fiom the fbll sonar equations derived above.
Note in particular how excluding the 3 dB gain from multiple pings halves the
reverberation-limited range. The effects of modifying parameters (TS, Sb,
SL, B, N ) can be easily and quickly seen by drawing new lines parallel to the
appropriate existing lines. For example, increasing EL by 5 dB - by changing SL
and/or TS - results in equal noise- and reverberation- limited ranges of 30 km.
120
12 dB echo excess at
reverb-limited range
100
12 dB echo excess at
noise-limited range
80
60 -
----- Noise level --

\
..
I 1 I I I I I
1 2 5 10 20 30 50 100
Range (km)
Figure 13.2 Echo, noise and reverberation levels plotted against range
Low Frequency .4ctive: Beware! 273
13.8 Low Frequency Active: Beware!

As the frequency of operation of an active sonar is reduced, the noise-limited
range will tend to increase because the reduction in absorption losses can outweigh
the increase in the noise background (as demonstrated by the above examples).
There are, however, two problems directly related to reducing the frequency and
these can adversely affect performance:
It is only possible to maintain a high SL over a limited bandwidth; it would be

unwise to expect the usable bandwidth to be more than, say, one-third of an
octave. The available bandwidth is therefore proportional to frequency. At
frequencies of at least 2000 Hz, a bandwidth of 670 Hz (one-third of an octave)
is realistic. This will meet the requirements for both detection and classification.
At 300 Hz, however, the bandwidth will only be 100 Hz, sufficient for detection
but not for classification.
The frequency shifts due to target doppler (target motion relative to the sonar
platform) are directly proportional to frequency. Target doppler is therefore
inversely proportional to frequency. This is an unfortunate and important result
for active sonars because it means that as the frequency is reduced to improve
noise-limited detection range, there is an increase in the magnitude of target
doppler needed to achieve noise-limited performance using a CW pulse.
(Remember that the performance of a long CW pulse against a background of
even low reverberation is always negligible.) Suppose a target echo needs a
frequency shift of 2 Hz to avoid the main lobe of the reverberation spectrum
(making any detection noise limited), this will be achieved by a target doppler
of about 1 knot at a frequency of 3000 Hz but not until about 10 knots at
300 Hz, a very unlikely target doppler indeed!
13.9 Torpedo Detection

Modem torpedoes are frequently very quiet and, as we saw in the passive sonar
examples, difficult to detect and classifL in time to take evasive action or launch
countermeasures. Active sonar can, with advantage, be used against such targets.
Reducing target strength produces a rapid fall in performance in even moder-
ately high reverberation areas. If the background is predominantly noise, the fall
will be much less. Fortunately, torpedoes are high doppler targets for most of the
time and a well-chosen CW transmission will ensure that detection is against a
noise background. Furthermore, the target doppler (without which a detection will
not be made) gives an immediate classification, and because it is an active system,
the location of the torpedo is known immediately.
Almost any active sonar able to detect submarines can be modified to detect
torpedoes. As an example we will take the representative hull-mounted sonar and
modify some of its parameters to get Table 13.3; the modified parameters are in
italics.
Table 13.3 Modified parameters of hull-mounted sur-

face ship sonar
Frequency 5.5 kHz
FM bandwidth 750 Hz
CW bandwidth 20 Hz
Horizontal beamwidth, @, 8 degrees
FM 20 dB
cw -6 dB
DI (receive) 25 dB
Target strength, TS -15 dB
5 log d 12 dB
5 log n 3 dB
sb -33 dB
Modified parameters are in italics
Pulse length
The pulse length has been reduced to 100 ms. A 1 s pulse would have a dead range
of 750 m, which is obviously too long for this application, and the range accuracy
Torpedo Detection 275
would also suffer. The 100 ms CW pulse will have an acceptable dead range of
75 m and the range resolution of the pulse will be somewhat better than this, say
20 m.
Clcy bandwidth
The CW bandwidth is therefore about 20 Hz ( 2 / T ) ,giving a doppler cell width of

about 3 knots. Therefore quite small dopplers (for a torpedo) will ensure noise-
limited detections. For half-beam processing, about 5 knots will avoid the pulse
main lobe; but for full-beam processing, target dopplers up to at least platform
speed will be necessary to avoid reverberation in the beam sidelobes. Neverthe-
less, even this is unlikely to be a problem as torpedo dopplers will usually be
greater than platform speed.
Target strength
The target strength is the suggested value for a random aspect torpedo, - 15 dB.
Performance
I2PL = SL + TS - N + DI + lolog T - 510gd + 5 log TZ I

Substituting the values from Table 13.3, we obtain
2PL = 224 - 15 - 55 + 25 - 10 - 12 + 3 = 160 dB

The allowable PL = 80 dB, and assuming spherical spreading plus absorption
(a = 0.5), the noise-limited detection range is about 7 km. This range is achieved
for both the FM and CW pulses because they have the same duration.
The reverberation-limited performance of the sonar is given by the following
equation:
1 10log R = 10 log(B/&) - S b - 41 + TS - 5 log d + 5 log nl

Substituting the values from Table 13.3:
0 +
For the FM pulse 10 log R = 20 33 - 41 - 15 - 12 + 3 = - 12 dB and the
reverberation-limited range is only 60 m
For the CW pulse lolog R = -6 + 33 - 41 - 15 - 12 + 3 = -38 dB and the

reverberation limited range is only 0.2 m
Neither of these results should be surprising. For the FM transmissions the short
range when compared with the submarine target is entirely due to the much lower
target strength of the torpedo (25 dB less than the submarine value), and for the
CW transmissions the negligible reverberation-limited ranges for either target are
entirely due to the much lower reverberation indices of the CW pulses.
The optimum transmission for torpedo detection and classification, then, is a
fairly long CW pulse (as long as possible consistent with dead range and
resolution requirements). But there must be some target doppler. Because of its
wide bandwidth, an FM pulse cannot use target doppler to avoid reverberation, so
unless the reverberation background is very low, an FM pulse will not detect a
torpedo at any target doppler.
Conclusion
The science of Underwater Sound had its beginnings in the 19thcentury - if we

ignore the oft quoted use of listening tubes by Leonard0 da Vinci in the 15fh.
However, the use of Underwater Sound to locate targets (Sonar) is comparatively
recent and the lineage of present day systems can be traced back to the Second
World War.
By the end of W 2 , active sonars operating at around 20 kHz and using
transducer arrays of only a few elements, were able to detect submarines at ranges
of 1 or 2 km at best. The elements of the system were little different from today.
After WW2, improvements to submarines and torpedoes forced the sonar
designer to use lower frequencies to increase detection ranges and, in the fifties,
sonars like the RN Type 177 operating at between 6 and 12 kHz were entering
service in NATO and Warsaw Pact Navies. These sonars, when used in controlled
trials with alerted operators, were shown to be capable of ranges of up to about
20 km, but, operationally, this performance was seldom achieved. This Opera-
tional Degradation, which could result in detection ranges of only about 30% of
theoretical values, has been largely overcome by the use of computers first
introduced during the seventies to help the naval operator perform his demanding
surveillance and classification tasks.
During the final decades of the 20th century, ever quieter submarines and longer
range weapons have driven sonar designers to hrther reduce operating frequencies
- for both active and passive systems - resulting in the present generation of long
towed arrays and low frequency towed projectors.
What of the future? More decibels could be gained by, for example, increasing
SL and using long multiple arrays. Such steps would be very demanding in cost,
power and difficulties of deployment: current active sonars already use about
50 kW of electrical power, and more complex towed arrays for passive sonars
would present severe handling problems.
Sonar is a mature discipline that would appear to have reached fundamental -
or at least practical - limits: Active sonars will rarely detect submarines beyond
278 Conclusion
about 40 to 5Okm range. Effort should not be squandered in near impossible

attempts to extend this limit. Rather, the immediate future tasks should be:
(1) To ensure that this limit is more often reached

(2) To reduce false alarms in all environments
Neither task is easy. The first would be helped by exploiting favourable propaga-
tion modes - for example, VDS and Bottom Bounce (not only in an attempt to
increase maximum range but also to fill in any intermediate shadow zones). The
second would be helped by a unified approach to detection and classification; the
careful choice of pulses and processing to best achieve this unified approach; and
thorough integration of data on contacts from all available sources. Submarine
detection needs to be optimised against both noise and reverberation, but without
compromising classiJication(which needs the best possible resolutions in range,
bearing and doppler).
It is the author5 firm belief that these requirements are best met by a
combination of an FM pulse (for best range resolution) and a CW pulse
(for best doppler resolution), and half beam processing (for best bearing
resolution).
Although much effort in recent years has gone into the design of alternative
pulses in attempts to meet the above requirements with one pulse, it has not
completely succeeded because, inevitably, any alternative pulse will compromise
the best performance obtained from combined FM and CW pulses. Furthermore,
careful study into the performances of alternative pulses against all possible
dopplers will frequently reveal significant ambiguities.
Despite the clear advantages of half beam processing, there are two arguments
frequently used against it:
(1) The half beam DI is 3 dB less than the full beam DI. Note, however, that the
bulk steer and add process after detection restores most or all of this reduction
(compare with passive broadband energy and cross correlator detection,
Sections 8.12 and 8.13). In any event the gains in the unified detection and
classification task from having additional information on the bearing spreads
of targets and reverberation will more than compensate for this loss.
(2) The success of the phase comparison process relies upon a good signal-to-
noise ratio. But this is inevitably the case for an active sonar system where
marginal (low SNR) detections are not a consideration: typically there will be
no discernible echo on one ping and a strong echo on the next, due mainly to
Conclusion 279
fluctuations in target strength caused by perhaps quite small changes in target

aspect from ping to ping. (Contrast with passive systems where the signal
energy is - in the short term at least - constantly available and a track will
gradually, at low SNR, become apparent against the noise background.)
Finally, dear reader, I hope that you have enjoyed reading my book as much as I
have enjoyed writing it. I remain dedicated to the advancement of sonar - which,
in these cost conscious days, may simply be to arrest any backward move! - and
would welcome dialogue with interested readers, particularly those who point out
the inevitable errors for correction in future impressions.
Solutions to Problems
Problem 1.1
I = p 2 / p c = (10-4)2/(1.5 x lo6) = 0.67 x W/m2
1, = 0.67 X lo-'* W/m2
]pi= 1010g 104 = 40 d~
Problem 1.2
SL = 101og P + 170.8 + DIt
= IOl0g40000 + 171 + 15
= 232 dB
Problem 1.3
From Figure 1.2 the safe power density is 5 kW/m2. This may be doubled because of
the operating frequency. The radiating surface is 2 X 1 X (n/4) = 1.6 m2. Therefore
the total safe radiated power is 2 X 5 X 1.6 = 16 kW. The factor 3t/4 is needed to
calculate the actual radiating surface resulting from the use of circular elements.
Problem 1.4
BL = SpL + 10logAf
Af = 2000 Hz and BL = 80 dB
Therefore SpL = 80 - 10 log 2000 = 47 dB
282 Solutions to Pmblems
Problem 2.1
The 3 dB beamwidth of a line array is given by
where L is length in metres, f is frequency in kHz and 8, is the steer angle.
(i) 28, = 76/(10 X 4) = 1.9"

(ii) 28, = [76/(10 X 4)]( 1 + 0.9) = 3.6"
Note that a steer of 60" has almost doubled the beamwidth.
Problem 2.2
The sidelobe levels are given by 2010g[2/n(2m + l)]. For m = 4 the sidelobe levels
are 23 dl3 below the main lobe. Their angular positions are given by
(nL/il)sin8 = f 9 n / 2 , and since L = 6il we have 8 = 3149". Note that the sidelobe
levels do not depend on the dimensions of the array. The higher-order sidelobes only
exist, however, if the array is large enough (in terms of wavelengths) to generate them
Problem 2.3
At lOkHz, il/2 = 75 mm. The array therefore has 2000/75 = 27 rows of
5000/75 = 67 elements. At 8 kHz we have
DI = 3 + lOIog(27 X 67) - 2010g(10/8) = 33.7 dB

Alternatively
DI = 3 + 10 log(4Lh/A2) = 10 log(4 X 5 X 2/0. 18752) = 3 + 30.6 = 33.6 dB

The trivial difference is due to the element numbers being the nearest integer values.
Problem 2.4
The DI of a baffled cylinder is given by
DI = 10 log 5hdfi = 10 log(2 X 3 X 5 2 ) = 22 dB
Halving the height of the array reduces the transmit DI by 3 dB. The number of
elements in the array is halved and therefore the transmitted power is also reduced by
3 dB. The source level is therefore reduced by a total of 6 dB.
283
Problem 3.1
(i) The equation for spherical spreading plus absorption is
At 5 kHz 80 = 20 log r + ( 0 . 3 X
~ therefore r = 7 km
At 20 kHz 80 = 20log r + ( 3 r X therefore Y = 3.3 km
(ii) The equation for cylindrical spreading plus absorption is
PL = lolog Y 4- (ar x lo--3)
At 5 kHz 80 = lolog + (0.3r X

Y therefore r = 100 km
At 20 kHz 80 = 10 log r + (3r X therefore r = 13 km
A value of PL = 80 dB is realistic for many sonar systems, and so are the ranges
obtained by assuming spherical spreading plus absorption. The ranges obtained by
assuming cylindrical spreading plus absorption, however, are usually far too large in
practice. Such ranges would only be achieved using the deep sound channel (DSC)
mode where the sound is confined to cylindrical spreading by refraction alone. For all
other modes there will be losses at the boundaries, and these losses will significantly
reduce the range.
Problem 4.1
The target strength of a sphere is given by TS = 1010g(a2/4). Therefore TS =
10 log(OS2/4) = -12 dB. The minimum dimension of the sphere is 1 m, therefore the
maximum wavelength for a reliable value of TS = 0.2 m. The fi-equency must
therefore be at least c/A = 1500/0.2 = 7500 Hz.
Problem 4.2
The target strength of a cylinder is given by TS = 10 log(aL2/2il):
At 10 kHz TS = IOlog[(0.5 X 22)/(2 X 0.15)] = -2 dB
At 100 kHz TS = 1010g[(0.5 X 22)/(2 X 0.015)] = 8 dB
End on, the target strength is that of a sphere, TS = 1010g(a2/4). The TS does not
change with frequency (provided the diameter is at least 31) and therefore, for both
frequencies, TS = 10 log(OS2/4) = - 12 dB.
284 Solutions to Problems
Problem 4.3
The target strength of any plate at an angle 8 to normal is given by
TS = 10 log(A/A)2 + 20 log(& sin x ) + 20 log(cos 8)
The largest dimension is 2 m (replacing L), so we use x = (2/L)sin8 =

(2/0.375)sin5" = 0.46 and, from the dashed plot of Figure 4.1, we have
20 log(x-l sin x ) = - 16 dB.Also 20 log(cos 5") = 0 dB, therefore
TS = 1010g(5/0.375)~- 16 + 0 = 22 - 16 + 0 = 6 dB
Problem 5.1
At 100 kHz there will be two noise sources: thermal noise and ambient sea noise. The
thermal noise is given by - 15 + 20 logf (f in W ) ,therefore = 25 dB.
Ambient noise, from Figure 5.1, is also 25 dB. So the total noise is 25 + 3 = 28 dB.
The output voltage from the hydrophone is given by
20 log v = sh+ 20 log p + 120 (adding 120 dB puts v in pV)

= -170 + 28 + 120 = -22 dB
To find the output in a 1000 Hz band, add 10 log 1000 = 30 dB, so 20 log v = 8 dB
and v = 2.5 pV.
Problem 5.2
The isotropic spectrum level of the noise at the array/water interface is 55 +
3 = 58 dB because self-noise = ambient noise = 55 dB.One third-octave is 2000 Hz,
therefore the isotropic band level is 58 + 10 log 2000 = 91 dB.The noise measured at
the output of the receiver beam is 91 - 20 = 71 dB.Because self-noise measurements
are traditionally made at the beam outputs of a sonar receiver, they are sometimes
misleadingly quoted without adding the DI of the beam. So beware of any suspiciously
low values for self-noise.
Problem 6.1
Start with
where R is in metres and @, is in radians. At 4000 m the reverberating area is

285
A = (1500 X 10-'/2) X 4000 X 0.17 = 51 000 m2

TSR = -40 + 47 = 7 dB
Doubling the range doubles the reverberating area, therefore
The TS of a vessel is constant with range. Suppose the echo needs to be 10 dB greater
than the background reverberation to be detected, then for a TS of 20 dB, the system
would be reverberation limited at 8000m. At greater ranges, because the TSR
increases, reverberation would mask the vessel and it would not be detected.
Problem 7.1
Using P d = 0.5, we can read off Pf,:
For 5logd = 6 dB, Pfa= 2 X
For 5 log d r= 5 dB, I>f, 1
Therefore Pfa is increased 50 times. Note how a small reduction in the detection
threshold (of which 5 log d is one of the terms) results in a large increase in false
alarms. This will have important implications for an automatic detection system.
Problem 8.1
The only parameter to change is DI:
At 3000 HZ DI = 3 + lolog50 + 2010g 1.5 = 23 dB

At 400 HZ DI = 3 + 17 - 2010gS = 6 dB
At 200 HZ DI = 3 + 17 2010g 10 = 3 dB
-
At 80 HZ DI = 3 + 17 2010g25 = 3 dB
-
At 40 HZ DI = 3 + 23 - 2010gSO = 3 dB
Because the flank arrays are baffled by the hull, DI cannot be less than 3 dB. The PL
and R values are then as follows.
Broadband
PL, a = 0.2 (dB) 128 98 93

R (km) >loo 35 25
Narrowband
PL(dB) 89 86 59 56 64 61
R (km) 30 20 0.9 0.6 1.6 1.1
Broadband detection performance is still very good, even with the much smaller
flank arrays. Narrowband performance is still good for the noisy torpedo, poor but
unchanged for the submarine, and still further reduced for the quiet torpedo. Had the
size of the array been doubled, which would hardly be practical, performance would
still be unsatisfactory against the quiet targets, thus demonstrating the need for long
towed arrays.
Problem 8.2
The time difference between the two paths is given by
This solves to give R = 15490 m. Note that because neither path includes a
component due to a surface reflection from the target, t does not appear in the
equation. To find t one of the paths must include a surface reflection fiom the target.
Problem 8.3
DI = 10 log n = 21 dE3, therefore the array must have 126 elements. They will be
spaced i2/2 at 3000 Hz. A/2 = c/2f = 1500/6000 = 0.25 m, and the array length
is 126 X 0.25 = 31.5 m. At 3000 Hz, PL = 201og 10000 + (0.2 X 10000 X
= 82 dB. The sonar equation to use is
PL = SL - N + DI - 5logd + Slog BT, + 51og n

Therefore
SL = 82 + 6 0 - 21 + 6 - 22 - 9 = 96 dB
This is the spectrum level of the radiated noise of a fairly quiet submarine, implying
that a quite short towed array may be adequate to detect submarines. The problem,
however, is classification and the need to discriminate against other, probably noisier,
targets such as surface ships, for which a capability at much lower frequencies is
essential.
287
Problem 9.1
The frequency shift due to ship (platform) motion is Af = 2Scos6 X (f/c). Be
carefid with the units: S must be in m/s if c is in m/s.
Af = 2 X 4 X 0.707 X (5000/1500) = 18.8 HZ
The total shift is 25 Hz, therefore the shift due to the target is 6.2 Hz:
6.2 = 2 X (S cos cp) X (5000/1500)
Therefore S cos cp = 0.93 m/s or 1.8 knots. This is the relative velocity of the target
(its doppler); we cannot separate the target speed from its bearing without more
information.
Problem 9.2
There will be a 3 dB processing loss when the replica only matches the echo for 0.7 of
its bandwidth. This will result from a 120 Hz frequency shift:
Af = 2 X ~ ~()f / ~ )
(SCOSX
120 = 2 X (S cos cp) X (4000/1500)
S cos cp = 22.5 m/s or 45 knots
There will be a 1 dB processing loss when the replica matches the echo for 0.9 of its
bandwidth. This will result from a 40 Hz frequency shift:
40 = 2 X (S cos cp) X (4000/1500)

S cos cp = 7.5 m/s or 15 knots
Even 15 knots is a lot of doppler for a submarine (remember that doppler is relative
velocity, not target speed). Therefore there is seldom any point in using an extended
reference.
Problem 9.3
For a Hamming shaped pulse, Af40 = 3.5/T. Therefore the doppler shift must be
i l . 9 Hz to achieve Rj = -40 dB.
At 3000 Hz
1.9 = 2 X target doppler X 3000/1500

target doppler = 0.475 m/s or about 1 knot
At 1000 Hz
1.9 = 2 X target doppler X lOOO/l500

target doppler = 1.425 m/s or about 3 knots
At 300 Hz
1.9 = 2 X target doppler X 300/1500

target doppler = 4.75 m/s or about 10 knots
Doppler shift is proportional to frequency. Target doppler is therefore inversely

proportional to frequency, an unfortunate and important result for active sonars
because it means that as the frequency is reduced to improve detection range, it
increases the magnitude of target doppler needed to achieve noise-limited perfor-
mance when using a CW pulse.
Problem 10.1
Make a guess and then iterate. Suppose f = 20 kHz. Propagation loss (dB) will be
given by
PL = 201og r + a r x 1 0 - ~
=20log12000+12a=82+12a
For noise-limited detection,
2PL = SL + TS - N + DI + 1010gT - 5 10gd + 5 log n

= 210 + 40 - N + 26 - 20 10 + 3 = 249 - N (dB)
-
At 20 kHz, N = 40 dB therefore PL = 105 dB. Equating to (i), we have 105 =

82 + 12a, so a = 2. Clearly, given T = 10 ms then 20 kHz is too high (a = 3.8 at
20 kHz). We would need to reduce the frequency to about 15 kHz. Such a compara-
tively low frequency implies a large, expensive array and a better solution would be to
use a long broadband pulse.
Try a 1 s, 20 kHz, 100 Hz bandwidth, FM pulse. The resolution would be the same
but now
2PL = 210 + 40 - 40 + 26 + 0 - 10 + 3 = 229 dB

289
+
Again, equating to (i), 115 = 82 12a therefore u = 2.8 and the frequency is now
about 18 kHz. Note the significance of absorption at these higher frequencies: it is
not practical to achieve, say, 10 km range at frequencies in excess of about 15 M z ,
even given a very strong target like an ocean floor at normal incidence. At 30 kHz
we have
PL = 115 = 20 log r +7r X (dB)
+
therefore r = 5700 m. The TS of the bottom is given by TSB = SB 10 log A. At
normal incidence SB will be at least -20 dB, and at a depth of 12 000 m and for a
beam of solid angle 5" then 10 log A = 60 dB. Therefore TSB = -20 +
60 = 40 dB.
But this target strength does assume the bottom to be flat (within 7.5 m for the
effective pulse duration of 10 ms) throughout the area A.
Problem 11.1
Noise limited
2PL = SL + TS - N + DI + 10 log T - 5 log d + 5 log ~1

For the FMpulse
2PL = 192 - 25 - 30 + 43 - 20 - 10 + 3 == 153 dB

therefore PL = 77 dB:
PL = 77 = 2010g r + ur X ( a = 50 dB/km)
giving r = 470 m.
For the CWpulse
2PL = 192 - 25 - 30 + 43 - 39 - 10 + 3 = 134 dB
therefore PL = 67 dB:
PL=67=201ogr+arX (a=50dB/km)
giving Y = 330 m.
Reverberation limited
The reverberation-limited active sonar equation is
10 log R = 10 log(B/&) - - 11 + TS - 5 log d + 5 log TZ

where R is the range in metres.
For the FMpulse
1 0 1 0 g R = 4 9 + 3 0 - 11 - 2 5 - 10+3=36dB
Therefore R = 4000 m. The reverberation-limited range requirement (to be greater

than the noise-limited range) is easily met with this pulse.
For the CWpulse
lOlogR=39+30- 11 - 2 5 - 10+3=26dB
Therefore R = 400 m. The reverberation-limited range requirement (to be greater

than the noise-limited range) is met with this pulse.
Increasing the TS by 5 dB meets the original detection range requirement using
either pulse. The higher resolution of the FM pulse, however, means that it is still the
best pulse for the total task of detection and classification.
Problem 11.2
Use the approximate formula for absorption attenuation, a and
= 0.05f'.4, assume
spherical spreading.
Frequency (kHz) 60 80 100

a (dB/km) 15 23 32
PL (dB) 70 76 84
SL (dB) 205 205 205
2PL = SL + TS - N + DI + 1010gT - 510gd + 510g TZ

At 100 kHz we obtain
2PL = 168 = SL - 30 - 35 + 45 - 10 - 10 + 3
29 1
and SL = 205 dB. To flatten the response at the receiver, the attenuation at 60 kHz
must be 2 X (84 - 70) = 28 dB.
This equalization is only correct at 800 m. In practice it would probably be better
to equalize at, say, half maximum range (400m) and accept the smaller changes in
response at other ranges. SL will change with frequency and the equalization should
also take this into account.
Problem 12.1
The intercept sonar equation is
PL = SL + DI - N - 5 10gd - lolog Br
For SL = 180 dB we have
PL = 180 + 10 - 40 - 10 - lolog 10000 100 dB
At 15 kHz, a = 1.6and
PL = 2 0 1 0 g ~ +1 . 6 X
~ lo-' = 100 dB
Y = 12000 m
PL = 220 + 10 - 40 - 10 - lolog 10000 = 140 dB
PL = 2010g Y + 1 . 6 X
~ lop3 = 140 dB
Y = 31 000 m
Note that the pulse type and duration is irrelevant to intercept (provided its bandwidth
is less than B r .
Problem 12.2
The active sonar noise-limited equation is
2PL = SL + TS - N + DI + lO10gT - 510gd + 5 log IT

2PL= 1 8 0 + 1 5 - 4 0 + 2 0 + 0 - 1 0 + 3 = 1 6 8 d B
PL=201ogv+1.6rX 1 0 - 3 = 8 4 d B
Y = 5000 m
2PL = 220 + 15 - 40 + 20 + 0 - 10 + 3 = 208 dB
PL = 20 log r + 1.6r X = 104 dB

r = 13000m
As always, this (noise-limited) performance will only be achieved if the target

submarine shows sufficient doppler. Note that the 40 dB (100 times) increase in
transmitted power only increases the range by a factor of 2.6.
Index
Absorption of sound range error, 170

mechanisms, 46 replica correlation, 163-4
Active detection and classification reverberation index, 168
computer assisted detection and Active sonar design
classification hull mounted, ship, 26 1-7
amplitude profiles, 2 13- 14 low frequency towed system, 267-72
levels of, 2 11- 12 torpedo detection, 274-6
pulse features relevant to, 2 12- 13 Acoustic shadow, 240- 1
statistical analysis, 2 13 Addition of noise levels, 93-4
effects of multipath on, 2 14- 18 Agouti, 92
unified approach to, 206-9 Ambient noise, xvii, 86-8
Active displays Ambiguity diagrams, 192
CW surveillance, 204-5 CW long and short pulses, 193-4
FM surveillance, 203-4 LPM pulses, 194-5
general format, 202-3 LFM pulses, 196-7
history, range recorder, 205 PRN pulses, 198-9
Active sonar, xxiii table of range and doppler resolutions, 200
CWpulse, 163 Anechoic cladding, 78-9
Hamming pulse, 172-3 Arrays
rectangular pulse, 171 comparison with radar, xxi
reverb. rejection by, 174 conformal, 36
doppler, 161-2 cylindrical, 34-5
equations, dipole, 15- 17
detection threshold, 165 directivity, receive, 14
noise limited, 165 directivity, transmit, 13- 14
reverb. limited, 166 line, 18-21
FM pulse, 163 modular changes to, 8
linear period, 169 need for, 13- 14
linear frequency, 169 planar, 32-3,
full beam processing, 180-9 spherical, 36
half beam processing, 180-9 volumetric, 37-8
matched filter processing, 163-4 Array directivity, 13
ping, 161 Array gain, 14,28
PRN pulse, 168 Attenuation coefficient, 47
pulse types, 163 approx. formula, 47
Index
Attenuation coefficient (continued) Deep scattering layer, 112

table of values, 47 Deep sound channel, 52-3
variation with: DEMON, 129
frequency, 47 Detection index, 121
salinity, 47 Detection threshold, 121
temperature, 47 Detection threshold formulae
Azimuth, xvii active, 164
communications, 258
Baffles, 40- 1,92 intercept, 25 1
Band level, 10 passive broadband, 137 - 8
Bathythermograph, 5 1 passive narrowband, 139-40
errors, 64 Directivity
Beamformers, 39 in receive, 13- 14
Beam pattern, 14 in transmit, 13
continuous line, 2 1-3 Directivity index, 28-37
sidelobe levels and positions, 2 1-2 of cylindrical array, 34-5
dipole, 17 of dipole, 29-30
effects of steering, 18- 19 of line array, 31
effects of shading, 24-7 of planar array, 32-3
line array, 20- 1 of volumetric array, 37
Beamwidth table for common array shapes, 36
continuous line, 23 Domes
cylinder, 35 baffles, 40- 1
dipole, 17 reflections, 40- 1
line array, 19 Doppler shift, 161 -2
Bearings only analysis, 148
Bottom bounce, 57,278 Echo ranging, xviii
Bottom scattering strength, 1 14- 16 Echo sounders, 22 1-5
Boundary reverberation, 110- 1 1 Echogram, 221
Boundary roughness, 104-5 Electrical noise, 90
Electronic integration, 8.5
Cavitation, 6-7
threshold, 6 False alarm reduction, 206-9,278
variation with depth, 6 Fish Finder, 22 1-2
variation with frequency, 6 Flow noise, 89,9 1,93
variation with pulse length, 64 Fluctuation index, 123
Cladding, 78-9 Fluid density, 2
Communications sonar Frequency, xix, xx
bandwidth requirements, 257-8 Full beam processing
examples of, 259-60 beamforming, 181 -2
sonar equation, 258 comparison with HB processing, 180- 1,
Convergence gain, 56 209
Convergence zone, 56 detection performance, 187-9
Critical bandwidth of ear, 133-4
Grazing angle, 104
Decibel
defined, xviii Half beam processing
use in sonar equations, xix beamforming, 181 -2
Index 295
benefits, 180-1,208-9,278 dedicated design, 244-8

comparison with FB processing, 180- 1. overview, 242
209 Mine hunting sonar
detection performance, 187-9 acoustic shadows, 240- 1
signal processing arrays for, 236-7
FM, 183-4 backgrounds to detection, 234-5
CW, 184-5 classes of mines, 234
refutation of criticism, 278-9 classification, 240- 1
Hearing overview, 233-4
threshold frequency, I34 performance, 239
threshold of, 1 1 pulses for, 238
Helicopter sonar, 37, 177 range and bearing resolutions, 235
Hodgeson model, 59-60 Minimum discernable signal, 12I
example, 6 1-3 Motional resistance, 84
Horizontal direct passive ranging, 153-5 Multipath propagation, 66,2 14-2 18
Hull mounted ship sonars, 261 -7 in deep water, 2 16- 17
performance, 266-7 in shallow water, 218
representative design, 263 -6
Hydrophone, xvii Netsonde, 22 1
sensitivity, 9 Noise
addition of levels, 93-4
lntegration factor, 124, 131, 139 ambient, xvii, 86-8
Intensity of sound, 3 biological, 87
Interaction effects, 8 directivity of, 93
Intercept sonar electrical, 90- 1
effectiveness, 257 factor, 94-8
function, 250- 1 flow, 89, 91,93
overview, 249 isotropic spectrum level, 90
reducing probability of exploitation, 256-7 machinery, 89, 91
reducing probability of intercept, 255-6 propeller, 89, 92
by sector transmissions, 255-6 radiated, 93, 125-8
sonar equation, 252 rain, 87--8
Isotropic spectrum level, 90 self, xvii, 83, 89-93
shipping, 87-8
Large aperture array, half beam plots, 185-6 thermal. 84-5
Layer scattering strength, 1 11-2 vessel, 89-93
LOFARGRAM, 136 Noise factor
Longitudinal wave, I calculations, 97 - 100
Low frequency active sonar, of receiver, 94
bandwidth limitations, 273 of receiver & hydrophone, 95
doppler requirements, 273 Noise levels, practical values, 100
performance, 270-2 Normalisation, 131
typical design, 267-9
Operational degradation, 20 1,277
Machinery noise, 9 1
Maximum response angle, 147, 154 Particle velocity, 1, 1 1
Mine avoidance sonar, Passive arrays, 132
adaption of A/'S sonar, 243-4 surface ship hull mounted, I32
296 Index
Passive arrays (continued) Propeller noise, 92

submarine hull mounted, 132 Plane wave, I
towed arrays, 132, 156-60, 143-6
Passive displays, 135-6
broadband, 135 Radar
narrowband / LOFAR, 136 cross section, 70
visual integration on, 135 equation, xix
Passive ranging, xviii, 146- 155 detection of submarines, xxi
BOA, 148 Radiated noise, 93, 125-8
HDPR, 153-5 continuum, 126
triangulation, 147-8 line structure, 125
VDPR, 149-52 nature of, 126
Passive sonar, xvii, xxii, 125-60 representativevalues, 127-8
aural, 133-4 source level, 125
broadband, 129-30,135, 137-8, 141-6 tonals, 126
complete system, 130 Radiated power, 5
DEMON, 129 Radiation resistance, 84
electronic integration, 135, 137 Rain noise, 87-8
equations Range recorder, 205
broadband cross-correlator, 138 Ranging
broadband energy, 137 Active, xviii
narrowband amplitude, 140 Passive, xviii, 146-55
narrowband power, 139-40, Ray theory, 58
narrowband, 129-31, 136, 139-40, 141-6 Hodgeson model, 59-60
vernier processing, 129 examples, 6 1-63
Performance prediction, 1 19 Receiver operating characteristics, 122-3
accuracy, 64 recommended values for 5 log d, 124
Ping, 161 Reference intensity of sound, 3
Ping interval, 161 Reliable acoustic path, 53-54
Ping/pulse repetition rate, 161 Remotely operated vehicle, 1 1.1
Plankton, 10.1, 10.3 Reverberation, xviii
Pressure of sound, 1 backscattering strength, 103, 106, 1 1 1- 17
Probability of detection, 120-2 boundary, 1 10- 1 1
Probability of false alarms, 120-2 in shallow water, 1 14- 16
Projector, xviii index, 166, 168
interaction effects, 8 reflection, 103-4
sensitivity, 9 sea-surface, 113
Propagation loss, 43-47 sea-bed, 114-16
allowable, xxii target strength of, 106, 166
definition, 43 under ice, 1 17
models, 58-60 variation with frequency, 116
two way, xxiii volume, 107-9, 167
Propagation modes Reverberating area, 1 10
bottom bounce, 57 Reverberating volume, 107-8
convergence zone, 56 Reverberation spreading
deep sound channel, 52-53 beamwidth, 175
reliable acoustic path, 53-54 environment, 176
surface duct, 54-55 practical envelopes, 179
297
ship motion, 175-6 variation with temp. and depth, 50- 1

sidelobes, 178 Speed of EM waves, xx
Spreading losses
cylindrical, 45
Scattering and reflection, 103-4 spherical, 44
Scattering cross section, 69 Submarine target strength, 74-6
Scattering of sound from boundary, 104-5 Surface duct, 54-5
Scattering layer, 11 1 Swaths, 131
Sea state, 86
Sea surface scattering strength, 113
Self noise, 83, 89-93 Target strength
Shading of arrays, 24-7 definition, 67
broadening factor, 24-6 dependence on pulse type and length, 69
reduction in SL, 27 integrated TS, 68
weighting functions, 24-6 measurement, 68
Shadowgraph, 227,240- 1 near resonance, 72
Sidescan sonar, 226-3 1 of mines, 72-3
Sonar equations, xix, 1 19-20 of reverberation, 106, 166
active, noise limited, 165 of simple shapes, 70-1
active, reverb. limited, 165-7, 188 of small targets, 72
basic active, 120-1 of sphere, 69
basic passive, 120 of submarines, 74-6
broadband energy, 137 of torpedoes, 73 -4
broadband cross-correlator, 138 of towed arrays, 77
communications, 258 peak TS, 68
intercept, 252 practical values, 80
narrowband amplitude, 140 reduction, 78-9
narrowband power, I39 ThreshoId frequency of hearing, 134
Sonar, origin of name, xvii Threshold of hearing, 1 1
Sonar platform, xvii Thermal noise, 84- 5
Sonar power, limitations to, 5-8 Torpedo active detection
Sonobuoy, 9.1 1 optimum pulse, 276
Sound using modified A/S sonar, 274-5
band level, 10 Torpedo passive detection, 141-5
intensity, 3 Towed arrays, 132,143-6,156-60
parameters in air, 11 bearing ambiguity, 158-9
parameters in sea water, 1 1 cable, 156
pressure, 1 curvature of, 156
reference intensity, 3 heading sensors, I56
scattering patterns, 1 14 hose, 156
spectrum level, 10 hydrophone triplets, 159
speed profile, 5 1 -2 nested, 157
Source level, 4-5 resolving bearing ambiguity, 158-9
of radiated noise, I25 self noise of, 159-60
Sources of reverberation, 103 tail, 156
Specific acoustic impedance, 2, 103 target strength, 77
Spectrum level, 10 twin arrays, 159
Specular reflection, 104-5 VIM, 156
Speed of sound, xix, 50 Transducer, xviii
298 Index
Transverse wave, 1 Volume reverberation, 107-9, 167

Triangulation, 147-8 Volume scattering strength, 1 12
Under ice scattering strength, 1 17

Underwater telephone, 12.8 Wavefiont, f
Wave height, 87
Vertical direct passive ranging, 149-52 Wavelength, xx
Very long pulses, 200 Wave theory, 58
Vessel noise, 89-93 Wind speed, 87
Vibration isolation module, 156
Visual integration, 135, 137 Yaw, 175

Sonar For Practising Engineers PDF

Uploaded by

Copyright:

Available Formats

Sonar For Practising Engineers PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Sonar For Practising Engineers PDF

Uploaded by

Copyright:

Available Formats

What is the book about?

What is the book about?

What topics are covered in the book?

What topics are covered in the book?

Sonar for Practising Engineers

British Library Cataloguingin Publication Dpta

ISBN 0 471 49750 9

Typeset in 10/1 Ipt Times by Keytec Typesetting Ltd., Bridport, Dorset

2.12 DI of a simple dipole 29

3 Propagation of Sound in the Sea 43

4.14 Towed arrays 77

5 Noise in Sonar Systems 83

7 The Sonar Equations 119

7.2 What are their uses? 119

8 Passive Sonar 125

9.12 Practical reverberation envelopes 179

10 Echo Sounding and Side Scan Sonars 221

11 Mine Hunting Sonars 233

11.11 Mine avoidance sonars 243

12 Intercept and Communications Sonars 249

13 Active Sonar Design 261

Echo sounding and side scan sonars for civil applications

Equipment parameters: this part briefly describes the motion of sound in an

Propagation in the sea and backgrounds to detection: this part considers

Practising on HMS Jupiter!

He then joined Ferranti Thomson Sonar Systems (now Thales Underwater

Magnetic methods include self-generated fields or perturbations of the earths

Triangulation: measuring the bearings of a target from two well-separated

Horizontal direct passive ranging (HDPR): based on the measurement of

Yertical directpassive ranging (VDPR): measuring the vertical arrival angles of

Power gain = 10logl0 (2) I (dB)

The decibel is of course also used in electronics, communications, radar and

2PL = SL + TS N + DI + 10 log T - 5 log d

All terms are in dB form; here are some possible values:

PL, the propagation loss, might be 80 dB

c = velocity in metres per second ( ds)

Wavelength is the distance travelled by the wavefront during one cycle. At

Why Is Radar Not Used to Detect Underwater Targets?

Figure 0.1 Why not radar?

Figure 0.2 Simple passive sonar

The simplest passive sonar will have an omnidirectional wideband hydrophone

and assuming spherical spreading and no losses,

Received pulse = 200 - PL (dB)

Figure 0.3 Simple active sonar

10 = (200 - 2PL + 10) - 60

1.1 Wave Motion

1.2 Sound Pressure

Elemental area of sound field

Longitudinal wave motion

Figure 1.1 Propagation of sound in an elastic medium

Given a plane wave IP = ( P C ) U ]

p = pressure (Pa or N/m2)

p = fluid density = lo3kg/m3 for sea water

c = velocity of sound wave propagation

= I .S x lo3m/s in sea water

pc = specific acoustic impedance, 2

= 1.5 X lo6kg m-2 s-' for sea water

1.3 Reference Intensity

1.4 Source Level