.AN IUVESTIGATIOli OF UNDERWATER ACOUSTICAL

P liEU 0 r:TEli A.

.-

A thesis by

BRIAn RAY.

Submitted for the deg~ee of Doctor of Philosophy_ -/

in the University of London.

Physics Department,

Imperial College,

London S. W. 7.

August 1970.

1

2

ABSTRACT

The aspect of underwater acoustics which the writer has chosen

to study concerns the problems of human communication underwater, and

in particular, communication between divers. The complete communicatioi

chain between the mind of the speaker and that of the listener is

discussed. As a first step an acoustic telephone was constructed

and tested. This revealed a lack of fundamental understanding

concerning the formation of words inside a closed cavity (such as a

diving facemask) and the manner in which the human ears function

underwater. A tentative model for the formation of words inside a

mask is proposed.

Following from this general appraisal, a series of experiments

were designed to test binaural hearing underwater and the results

of two independent series of subject tests showed, for the first time,

that a diver could located a sound source within a target error of

20°. This demonstration of binaural ability is considered important

in understanding the mechanism by which a diver interprets a voice

signal under noisy reverberant conditions.

During these directional hearing tests some subjects reported

that a sound source in front of the body appeared louder than one

behind. This was confirmed with the aid of a suLmersible audiometer.

A second series of audiometer tests using a ubmerged 'dry' laboratory

as a calibration facility, provided additional evidence of a sensitivity

differential and demonstrated the usefulness of this type of

-structure in underwater acoustical observations.

CONTENTS

1. Introduction

.1.1 Introduction

1.2 Resume of diving physiology and technology.

2. A review — and a plan for research.

2.1 Early experiments.

2.2 Development of ultrasonic communication.

2.3 Direct audio communication.

2.4 Electromagnetic communication.

2.5 The effect of helium on speech production.

2.6 The formulation of a programme of research.

3. A complete underwater communication system.

3.1 Introduction.

3.2 The microphone.

3.3 The amplifier

3.4 The transducer.

3.5 A receiving hydrophone.

4. Initial experiments with communication equipment.

4.1 Introduction.

4.2 The formation of words underwater

4.3 Noise and range of communication.

4.4 Hunan hearing underwater.

4.5 Discussion.

4.6 The formation of words.

4.7 The range of communication.

4.8 Directional hearing.

3

5. Directional hearing experiments (1)

5.1 Introduction.

5.2 The sound source.

5.3 Free choice experiments.

5.4 Two choice experiments.

5.5 Results of free choice experiments.

5.6 The results of two choice experiments.

5.7 Conclusions.

6. Directional hearing (11)

6.1 Introduction.

6.2 Different source angles around the subject's head.

6.3 The effect of a rubber hood.

6.4 The detection of an obstacle.

6.5 The effect of the subject hearing the source switched on

6.6 The effect of a reference on the free choice tests.

6.7 Subject inverted.

6.8 Subject signalling the confidence of his judgement.

G.9 The effect of different sound sources.

6.10 Conclusions.

7. Audiometry and observations from an underwater laboratory.

7.1 Initial experiments.

7.2 The construction of•an improved audiometer.

7.3 The second series of audiometer tests.

7.4 Results.

7.5 Other observations from the underwater laboratory.

4

8. Discussion.

8.1 Articulation underwater.

8.2 Propagation.

8.3 Human hearing underwater — the threshold.

8.4 Human hearing underwater — directional hearing.

8.5 Applications to communication equipment design.

8.6 Suggestions for further research.

Acknowledgments.

Appendix 1. The signal to noise ratio from an underwater microphone.

Appendix 2. Significance of the mean vector.

Appendix '3. The design and operation of an underwater laboratory.

References.

List of symbols.

Glossary of diving terms.

5

LIST OF PHOTOGRAPHS

Firr. Photon;raph.

3.9 Four different electrodynwnic transducers.

3.13 Two hydrophones used for recording underwater signals

3.14 Direct audio system in use. Malta 1966.

4.1 Three different types of diving facemask.

5.1 Buzzer sound source.

6.2 Photograph taken fran'free choice' film record, showing subject inverted.

7.1 Crude audiometer in use.

7.3 The improved audiometer.

7.4 Inside the underwater laboratory.

8.2 Steriophonic hearing aid.

A.1 The underwater laboratory.

6

7

Chapter 1

INTRODUCTION

"There is a whole world of sound beneath the waves

waiting to be explored" (Alexander Graham Bell)

1.1 Introduction

Once embarking on any study concerning the underwater

environment, one becomes aware that there are an enormous number

of unknown or unexplored problems. Unlike some other branches of

science these problems are often not obscure, they can be as

obvious to a person who puts his head underwater today as they were

to Bell when he first tried it in the middle of the last century.

Within these few pages it is only possible to recount a study of

one small aspect of underwater acoustics. The aspect that was

chosen is possibly one of the most significant if man is to

continue to explore the oceans. This concerns the problems that

men experience in attempting to speak to one another while working

beneath the waves.

Up to the last decade diving was the province of the sportsman,

the salvage contractor and the naval frogman. The first could not

afford the expense of communication facilities, the second managed

without, and the third did not want his presence revealed at all!

In receht years the scientist has ventured underwater both to gain

first hand experience of marine problems and to study man in an alien

environment. The discovery of petroleum resources under the

sea has brought in the engineer and technician. All these people

need to .communicate to perform their allotted tasks. After all the

8

human society has largely evolved through the use of the spoken

word.

This study will examine the physics and technology of

voice communication between men in the underwater environment. In

this context, it will he necessary to examine the mechanjsm of

word formation under the restrictions that underwater brenthing

. apparatus may impose, methods of transmitting the speech signal

through water and the mechanism of human hearing- underwater. Certnin

items of equipment that were not previously available were developed

to aid this research. The results of exp~riments in the open sea

shed new light on the workings of the human ear underwater. A

submerged laboratory was developed to aid this study of the human

aspects of underwater acoustics.

1.2 Resnmp. ()f otvinp," nhvsinlogv and technolog;v

Before discuss ing the prohlems of human cf)mrnunication underwD.ter

it is appropriate to consider briefly the physiology and technl')logy

of diving. In this introduction emphasis. will be placed ori those

aspects which have a direct. bearing on communicRtil')n. For a more

p-eneral introduction the reader is referred to the BoS.AoC. Diving

!.Ianual (1), or for a co::nprehensi"e study to "The physiology ann

medicine f)f diving- and compressed air wnrk" edited by Bennett

and Elliott (2).

9

Physiolorly

The human body is not sensitive, to an absolute pressure

as such; although it will not tolerate a pressure differential.

Consequently, to survive underwater, man in the first instance

need only provide himself with a supply of breathing gas at

the ambient pressure. The fundamental difference between the

submariner and the diver (or someone in an underwater habitat)

is that the former is working in an environment at the same

pressure as the surface, whereas the latter is exposed to the

pressure of the water, roughly one atmosphere for every ten

metres of depth.

One apparent exception to this rule concerns the middle car

cavities. In the normal way these connect with the throat by two

tubes — the eustachian tubes. These are normally closed and only

open occasionally during such acts as swallowing. It is for

this reason that a change in barometric height, such as descend—

ing a steep hill, is accompanied by a sensation in the ears

which is relieved by swallowing. The trainee diver must be

taught to recognise this sensation and to deliberately "Clear

His Lars" by swallowing or more often 'by pinching his nose and

blowing into it while he is descending. (If his diving equipment

renders nose pinching impractical, he will probably wear a nose

clip.)

Failure to clear the ears properly can result in a

temporary loss of hearing while gross failure can rupture the

ear drum (Miles 3)

10

If these precautions are taken then the average diver,

breathing compressed air, would be comfortable to a depth of

50 metres while the experienced professional might dive to 90 metres.

Three factors limit the use of compressed air to these depths.

Nitrogen has a narcotic effect when breathed under pressures of more

than about six atmospheres, the human system is intolerant to a

partial pressure of more than two atmospheres of oxygen and the densit

of air under pressure may be sufficient to impair breathing.

-To overcome these problems, an atmosphere containing a

reduced percentage of oxygen diluted with helium instead of nitrogen

is normally used when diving below 70 metres. The resulting mixture,

known as "heliox”, is often described by the percentage of the main

constituents. For example a 7/93 mixture would 'contain 7% oxygen

and 93% helium and would be suitable for use at depths between

20 metres and 300 metres. Such a mixture could not, of course, be

breathed at surface pressure as the partial pressure of oxygen would

be insufficient to support life. This brings about logistic

difficulties in recording the.human voice in mixtures such as these.

When the body is subjected to pressure, the breathing gases

will tend to dissolve in the tissues. The longer the time spent

under pressure or the higher the pressure, then the greater the

amount of dissolved gas. When the diver surfaces it is necessary to-

prevent a rapid release of this gas which may otherwise give rise

to the symptoms of decompression sickness (the diver's bends). To

11

prevent this possibility the diver must reduce the pressure on

his body slowly. The time spent decompressing may vary from a few

minutes after spending half an hour diving to 30 metres, to a week

after spending a day at 300 metres. Clearly the longer decompression

times cannot be spent in the water and in these cases it is usual

to bring the diver to the surface in a pressure vessel and to

transfer him, under pressure, to a surface decompression chamber.

Before leaving this brief summary of diving physiology,

mention must be made of cold. Present day protective clothing is

only partially effective in preventing hypothermia and the writer

has been forced to abandon many experiments due to the subject or

experimenter becoming excessively chilled. This phenomena is

not confined to cold water and one series of hearing tests had to

, be called off after only 35 minutes in water at 2oo C.

Diving Equipment

Broadly speaking, present day diving can be put into three

classes depending on the equipment used:

a) Standard diving dress

b) Self contained compressed air open circuit breathing

apparatus generally known as aqualung or scuba diving.

.0 Closed or semi-closed circuit breathing apparatus using

a breathing gas of either oxy-nitrogen (in different

proportions to those in air) or heliox.

The standard diving dress has changed little in the last

century.and is still the most popular for dock and harbour diving.

To high pressure

air supply

Non—return exhaust valve

Rubber diaphragm Tilt valve

!:outh bit

The equipment consists of a loose fitting rubber suit sealed

to. the famous large copper helmet. Heavy lead boots allow the diver

to stand and walk on the sea-bed. Compressed air is fed from

the surface through an airline and the pressure inside the diving

dress is automatically maintained at that of the surrounding water -

by allowing excess gas to escape into the sea through a non-return

valve.

The popularity of scuba equipment, particularly among amateur

divers, since the second world war is due to its intrinsic

simplicity and relative safety. The air, which the diver carries in

cylinders on his back, is not supplied continuously, but is available

on demand. The key to the scuba equipment is the demand valve and

a simple one is shown in fig. 1.

Fig 1.1 A simple demand valve.

12

13

On inhalation, the reduction of pressure in the chamber

will cause the flexible rubber diaphragm to deflect inwards and

open the tilt valve allowing a supply of air all the time the

diver is inhaling. The exhaled air is expelled to the sea water

through a non-return valve. Needless to say, such a simple

device is inherently very noisy. On inhalation there is the noise

of compressed air being released and on exhalation the generation

and release of bubbles.

The scuba diver normally wears a close fitting, free flooding,

foam rubber suit. He does.not generally use a helmet but instead •

a mask covering his eyes and nose. He will usually hold a mouthbit

between the teeth which is coupled to the demand valve. If the

demand valve is attached directly to the mouthbit with just a single

thin. high pressure air hose, connected to the air supply then it

. is known as a "single-hose demand valve", whereas if it is mounted

behind the diver and connected to the mouthbit by large diameter,

low pressure, inhalation and exhalation hoses, it is known as a

"twin-hose demand valve". An alternative to the mouthbit is a

mouthcup which is designed to help the diver speak underwater. This

takes the form of a rubber cup that fits over the mouth instead of

between the diver's teeth. One further alternative is a large

mask known as a full-face-mask which as the name implies, covers

the whole face.

When diving deep, it is uneconomic to expell the exhaled gas

into the water. One breath of 2 litres at.200 metres depth is the

14

equivalent of 40 litres at surface pressure. One method of

overcoming this is to feed the exhaled gas through a chemical bed

to remove carbon dioxide and into a flexible bag or "counterlung".

After a small amount of oxygen is added to make up for that used

by the human body the gas in the counterlung can be rebreathed.

In principle, no gas need be released into the water in such a

closed circuit system. However, there are practical problems and

such closed circuit equipment is rarely seen outside military circles

where there is a requirement for equipment which does not produce

bubbles and is sufficiently quiet to use in the presence of hostile

acoustic detection equipment.

Some of the problems of a closed circuit system can be

overcome if, instead of supplying pure oxygen to the counterlung,

a mixture of oxygen and inert gas is used. As the inert gas (either

nitrogen of helium are described as "inert" in this context) is not

consumed by the body, there must be an escape of gas to the sea in

order to establish an equilibrium in the counterlung. The relative

economy of gas of this semi-closed system falls between that of a

scuba and a closed circuit set. The design of a semi-closed set

is discussed by Williams (4).

The closed, or semi-closed circuit diver normally wears a

close fitting rubber suit either sealed, or like the scuba diverts

free flooding. Although he could use any of the masks of the scuba

diver, he will normally use a full-face-mask or possibly a lightweight

fibreglass helmet.

15

1.3 The problems of human measurements underwater

Under normal surface conditions measurements on speech

and hearing are often made in some form of anechoic chamber.

The lowest speech frequencies have wavelengths in air of about 2 metres

and the design of a chamber with, say 1 metre of sound absorbing

material, can be effective at removing unwanted reflections. In water

the wavelength of this same frequency would be nearer 10 metres and

the equivalent chamber would require 5 metres of absorber on all

sides including the water surface. Even if the idea is not

mechanically impractical then the thought of an operation with

human subjects without a free air/water surface would cause concern

on safety grounds.

Having abandoned attempts to find an anechoic tank, the

writer considered the use of swimming pools, but these exhibit a strong

standing wave pattern at audio frequencies apart from being rather

noisy. However for some recordings, particularly those involving'

a microphone inside the subject's facemask, a pool is suitable and

has been used for this type of work. Although a fresh-water lake

on the outskirts of London was used for some tests, the water

temperature in this country is normally too low for subjects to

remain stationary for long periods without wearing protective clothing

on the head. As the design of such clothing varies considerably,

with some almost completely impairing hearing both underwater and

on the surface, it was decided that for most of these investigations

the subject would not wear any form of hood.

16

The remaining sites that the writer used were in the

Mediterranean, either off the south - coast of France or off the

coast of Malta. In the latter area the water temperature in

Summer is between 22 and 26oC and the weather relatively reliable.

Furthermore, and most important, there were willing divers

available to act as subjects.

The selection of subjects is difficult. For safety reasons

they must be reasonable divers but furthermore they must be capable

of applying themselves to the task of answering simple questions

underwater. This may seem straightforward until it is realised

that the average amateur diver who is in the seas off Britain

would be, and should be, spending 90% of his time making sure

he stays alive. This same diver may well not be aware that his

ears are capable of functioning underwater and if not warned

beforehand may not even recall the sound of a "thunderflash" detonated

in the vicinity.

Previous workers have found a serious fall-off in the

performance of divers performing simple tests underwater when compared

with the surface. Although this was first assumed to be entirely

inert gas narcosis, it has now been shown to be partly just the

effect of putting a man underwater in the open sea. (Baddeley 5).

However this degradation can be small under almost "ideal" diving

conditions. "Ideal" appears to mean warm, calm, clear water with

a safe landing place.

By combining experimental work with a scientific

expedition, the writer was able to obtain the willing help of

experienced divers. Divers on these expeditions would invariably

be keen undergraduate science students in good health, and the

nature of the expeditions are such that the personnel are diving

on most days and are acclimatised to working underwater.

Occasional use has been made of personnel on similar expeditions

from other establishments.

17

18

Chapter 2.

A REVIEW - AND A PLAN FOR RESEARCH

2.1 Early Experiments

The claim to the first scientific experiments in underwater

communications probably belongs to Alexander Graham Bell (Born 1847)

He is reported (6) to have said that as a boy he had clicked stones

underwater and was startled by the loud report. He sent another

boy up to half a mile across the bay and distinctly heard the

stones with his ear submerged. Although the French physicist,

Paul Langevin, with his colleague Chilovsky and the Canadian

Fessenden, had demonstrated active sonar systems before the First

World War, these were not put into effective anti-submarine use until

the post war era. During the Great War the principal submarine

detection systems made use of the human ear. One such system

involved two spaced hydrophones coupled to a pair of earpieces mounted

on separate trombone slides. The operator placed the trombone

mouthpieces against his ears and localised the sound source by adjusting

the two slides until the sound was apparently coming from a central

position. A graduated scale on the slides enabled the bearing of

the submarine to be ascertained (7).

19

By the outbreak of the second world war a considerable

wealth of information was available on the propagation of sound

and ultrasound through the sea. For long range submarine

communication it was possible to pulse a sonar transmitter with a

morse key. This became known as SST or supersonic telegraphy (8).

Over short ranges the human voice could be transmitted into the

water in the same manner as a public address system. By the end of

the war acoustic torpedos, directional hydrophone arrays and sona-

bouys had all proved themselves under operational conditions (9).

In the field of diving, the end of the second world war saw

most of the diving equipment that is used to date in service. Heliox

mixtures had been used on the salvage of the U.S. submarine Squalis,

the Cousteau-Gagan aqualung had been used to over 60 metres and a

Swede, Arne Zetterstrom had dived to over 150 metres on a hydrogen-

oxygen mixture before losing his life on his way to the surface

(through a human error).

2.2 The Develonment of Ultrasonic Communication

The aftermath of the second world war left most of the navies

in the western world with a standardised ship to submarine ultrasonic

voice communicator (13). A carrier of about 8 KHz is amplitude

modulated and the upper sideband is transmitted. At the receiver

a locally generated 8 KHz tone is used to demodulate the incoming

signal. This is completely analogous to a single-sideband radio

telephony channel.

20

Examination of the attenuation of ultrasound in seawater

(fig 2.1) demonstrates the reason for the choice of such a relatively

low carrier frequency. The right hand side of fig 2.1 is a scale

showing the propagation distance corresponding to an attenuation

of 120 db. This may be regarded as being typical of the gain available

from a good receiver and is therefore an indication of the range

available. This does not take into account the "spreading" loss

which would follow an inverse square law in deep water and an

inverse law in shallow conditions where the spreading is in two

dimensions only. No mention has been made of directional transmitting

or receiving arrays as these are seldom applicable to diving operations.

Finally this simplified picture ignores the bandwidth of the signal.

(To be strict, the choice of the figure 120 db implies that the

bandwidth is of the order of that required for voice communication).

There are two further sources of difficulty in propagation which

will limit the use of an ultrasonic communication link. These are

forward scattering and multiple path propagation (fig 2.2). The sea

is not a uniform media and forward scattering is caused by signals

arriving at the receiver after being refracted by layers of,water of

slightly differing temperature. These signals will add to give a

resulting signal that will fluctuate at random in phase and amplitude.

Berktay and Gazey (11) have shown that this should be negligable for

short range communication (<100 metres). In comparison, the parts

of a signal subject to multiple path propagation will arrive at the

receiver very much later than the direct path signal. Multiple paths

are often far more serious in shallow water and around wrecks and

1. air

0.1 - Attenuation

dbimetre

0.01

Multiple paths

sea surface

Receiv- / er

........., '...„.--...1- Receiv-

, ...• -------.4 er ..r ..-.. ~'-•.i .... / or" -. , ‘.,4 _,.- /

..- ///

,..., ....- -7' .04 tb /

....... .0. ... .' .../ i -,r . Tra,nsm- o' -- itter 0- ----P .....

, Transm-6,-- itter

Forward scattering

21

Fig 2.1 The attenaution of sound in seawater (after Tucker Et Gazey,10)

0.12

1.2 Range

Itm

- 12

- 120

• 10 30 100 300 1000

frequency

Fig 2.2 Forward scattering and multiple path distortion.

0.001 ,

harbour installations and these are just the areas where divers are

likely to be working most.

One of the earlier successful communications sets to be designed

specifically for divers was that of Gazey and Morris (12). This

employed a 120 KHz carrier which was frequency modulated. The range

was up to about one killometre. As a precaution against multiple

path signals reflected from the surface or the sea bed, cylindrical

transducers were employed. These were arranged to float vertically

in the water and had an omnidirectional response in the horizontal

plane but were limited to 300 about the horizontal in the vertical

plane.

The debate between frequency modulation (F.M.) and amplitude

modulation (A.M.) is still unresolved. An A.M. system will resolve

multiple path signals and pass them to the listener who will perceive

them as echos or reverberation similar to that experienced in an

empty concert hall. In contrast an F.M. system, although better at

discriminating against noise, will resolve multiple path distortion

in a way with which the human ear is not familiar.

The majority of ultrasonic communication equipment is not capable

of transmitting and receiving a signal simultaneously. This will mean

that the divers must use some form of "Procedure" for operating their

send-receive switch. During a conversation it is not possible for

one party to interupt the other. In radio telephony this is known

as "simplex" operation. It has been proposed to replace the send-

receive switch with an electronic circuit which will perform this

22

23

function automatically (12). This is known as a voice operated

switch and has been used for a number of years on loudspeaking

telephones to prevent a "howl" when the microphone and loudspeaker

are operating simultaneously. The problem with such a circuit is

that it will respond to other loud noises and operate falsely.

Underwater, where the sound of the divers breathing apparatus may

be considerably louder than his voice, voice operated switches

are unlikely to prove successful.

There is one ultrasonic communicator which uses a "duplex" system

where a separate transmitter and receiver can operate simultaneously

(14).

2.3.Direct Audio Communicators

These systems transmit amplified human speech through the sea.

There is no carrier or modulation process involved and the unaided

human ear can be used for reception. It follows that "duplex"

operation is possible and both parties can interrupt one another

while speaking. The main limitations lie with the limited

sensitivity of the human ear and with the difficulty of producing a

transmitting transducer with a sufficient bandwidth to cover the

whole of the audio range.

The threshold of underwater hearing has been measured by many

workers, notably Hamilton (15), Wainwright (16), Montague & Strickland

(17) and Brandt & Hollien (13). Fig 2.3 is taken from this last

author.

db ref

0.0002 dyngs cm

Hearing aid

30.

70 Unaided ears

50 .

To diver's air supply

Flexible membrane

Diaphragm & coil

Fig 2.3. Human hearing underwater (From 18 & 16)

24

p 1 ■ . 125 500 2K 8K

frequency Hz

Fig 2.4 Two types of electrodynamic transducer with pressure

compensation.

Iron

Electromagnet diaphragm

25

There have been two attempts to produce an underwater

"hearing aid". The threshold for a diver wearing that designed

by Wainwright (16) is shown on Fig 2.3. The other, designed by

Bauer and Torick (19) is a binaural device and was designed to

produce the interaural delays that a subject would normally experience

in air.

In order to cover the relatively large bandwidth, from 100 Hz

to 5 KHz, most designers have used electrodynamic transducers. The

more common piezoelectric and magnetostrictive devices that are used

for ultrasonic generation, are rejected because it is difficult to

arrange for the moving element to have the required displacement to

radiate sufficient energy at low audio frequencies. Although

variable reluctance electrodynamic transducers have been used for

underwater communication (20), the moving coil arrangement is more

popular (21)(22). (Fig 2.4).

If the internal mechanism of the electrodynamic transducer is

allowed to flood with water, or even if it is filled with oil or a

liquid with similar acoustic impedance, then the transducer can he

considered as a vibrating plate. It can be shown (23) that where

the dimensions of the transducer are small compared with the acoustic

wavelength, then the intensity radiation from this dipole source

will be given by:-

I 4 4 G 2 Tr f a U cos20 3 2 c r

2.1

Where the velocity of the vibrating plate is:-

U U e 2 rr ift 0

On the other hand if the medium has only access to one side

of the plate which is the situation if the transducer mechanism

is air filled, then the transducer will act as a simple source, and

the intensity will be:-

I= /, f2 02

8 c r2

Where 0 is the net flow of fluid from the source.

Now it can be seen from 2.1 and 2.2 that the radiation from

a simple source will fall off at low frequencies with the square of the

frequency, whereas that of the dipole will fall off as the fourth

power of the frequency. It is for this reason that practical designs

of transducers for audio frequencies are invariably air filled and

operate as simple sources.

As the diaphragm of the transducer must be free to move, some

means of pressure compensation is needed to prevent the hydrostatic

pressure of the sea from forcing the diaphragm into the mechanism

of the transducer. The two transducers shown in Fig. 2.4 illustrate

two different methods of pressure compensation. In the first a

rubber membrane at the rear of the transducer can flex and provide

the compensation; in the second the transducer is provided with a

supply of gas at the ambient sea pressure from the diverts breathing

apparatus.

26

2.2

27

2.4 Electromagnetic Communication

Although conventional radio waves in the 15 to 30 KHz region

are used for submarine communication, the aerial requirements

(up to 10 Km in some cases) render this impractical for diver

communication. There are however, two methods which can be used

over short ranges between underwater swimmers. These use either

the magnetic field produced by a small coil or the electric current

field produced by two spaced electrodes. In the magnetic case

the receiver is a similar coil and in the electric case a second pair

of electrodes are used (28).

The range over which an electric or magnetic field decays in a

conducting medium is described by the classical "skin-depth" relation.

In M.K.S. units this can be shown to be (24):-

2 2.3 (.4.) T>AA

Taking the conductivity of sea water to be 0.5 mho/m, then the

distance at which the field has dropped to 1/0 can be found at any

frequency. Now 1/e corresponds to 8.7 db. By using the same criteria

as was used in Fig 2.1, that the 120 db loss between transmitter

and receiver, a table of the range at various frequencies can be

drawn up. As in the acoustic case this ignores spreading loss

Fig 2.5 Range of magnetic & Electric field communication equipment

Frequency Theoretical Range

1 KHz 280 metres

10 KHz 100 metres

100 Miz 28 metres.

28

Although not liable to the acoustical problems of an

audio or ultrasonic system, these electrical communication links

are very prone to interference from electric power installations

and nearby radio transmitters.

2.5 The effect of helium on speech production

If one breaths a gas with a velocity of sound higher than

air, there is a marked change in voice quality. This distortion is

due primarily to the increased velocity of sound in the exhaled gas.

It must, of course, be borne in mind that the composition of the

exhaled gases will differ from those inhaled due to the metabolic

processes of the human body. The human voice can be considered

as the larynx, a low frequency generator very rich in harmonics,

followed by the vocal tract, a series of cavity resonators (25).

The spectrum of a vowel or voiced consonant will show the line

spectra of the larynx within an envelope function produced by the

relatively low Q resonances of the vocal tract. The most obvious

effect of helium is to shift the envelope up in frequency. The

larynx pitch is not appreciably altered. (Fig 2.6)

One instrument that can be used to analyse helium speech is

the "sonagram" or "voice-print". This produces a trace of frequency

against time, the amount of darkening of the paper being related to

the amplitude at that particular frequency. Fig 2.7 is a trace

of the author's voice recording the words "GOAT BAIT" while breathing

air and a 20/80 oxy-helium mixture at atmospheric pressure. It can

be seen that the instrument has not revealed the fine structure of

the larynx overtones but it has given some indication of the position

"AIR" SPEECH

SHIFT

"HELIUM-MIXTURE" SPEECH

SHIFT

SHIFT

VOICE HARMONICS

VOCAL-CORD PITCH: 100 Hz 100-Hz SPACING----1 (FUNDAMENTAL)

Goat bait Goat bait

3.0K 4 ..i.....;;...i.;

2 .51( -I 'i ,. '' - - 1 1, i„,, • , z 1.1,,iifor L!'ll

•!, 1.5I: 1, -

1.0K

50 0

oxyLhelium

2.0K

air

29

Fig 2.6 The effect of helium on the voice.

Pig 2.7 A recording of some helium speech (retouched for reproduction)

30

of the envelope. The horizontal lines that have been drawn on

the trace to show an estimate of the first three peaks of the

envelope function. These are normally called the voice formants.

At present there are three methods used for attempting to

improve or as it is sometimes described - unscramble - helium speech:

a) Frequency shifting

It is relatively easy to shift the spectrum of a signal by

a fixed amount. This is accomplished by modulating the original

signal with a carrier whose frequency lies well outside the audio

band; for example 50 KHz. If we now select the lower sideband and

perform a second modulation process with a different carrier, in

our example this could be 49,900 Hz, we will recover the original

signal but shifted down in frequency by 100 Hz. This technique in

itself would not be a satisfactory solution as 100 Hz shift would

reduce the larynx pitch to almost zero hertz while hardly improving

frequencies around 3 KHz which may require a reduction of over 1000 Hz.

What is required is proportional frequency reduction. The

"Band-shifting" unscrambler attempts to simulate this by breaking

the incoming signal into three or four frequency bands and by shifting

each of these bands by a different amount. The lowest frequency band

containing the larynx pitch is not normally shifted.

b) Time domain shifting

A tape recorder running at slow speed is one method of obtaining

a proportional frequency shift. It is often observed that the

naturalness of helium speech is improved by replaying it on a tape

recorder at a slower speed than that which was used for recording.

31

This method, although simple, would be impossible for any

real time processor. However, it is possible to break the incoming

speech into short sequences, slow these down, throw away the overlaps

that will result, and finally add these sequences to produce an

unscrambled signal in real time. If the length of sequence is

chosen to correspond to one period of the larynx fundamental, then

the reconstituted signal will retain the correct pitch. This

technique is after W.R. Stover (26).

c) Vocoder techniques

The vocoder is an instrument for analysing the synthesising

speech. It breaks the incoming signal into a large number of

channels (typically between 15 and 30) and can distinguish between

the envelope and the fine structure of the pitch overtones shown

in Fig 2.6. The vocoder does not perform a frequency shift in the

normal sense of the word. When used as an unscrambler it is arranged

to detect the envelope of the speech spectrum, to shift this

envelope down the frequency scale proportionally, and to adjust the

levels of the fine structure to correspond to this new envelope.

This concept was originated by R.M. Golden (27).

2.6 The formulation of a programme of research

What is the chain of communication between two underwater

swimmers who are speaking to one another? Fig 2.8 is an attempt to

break this chain that exists between the mind of the first diver

and that of the second, into as large a number of individual links

as it is possible. Those parts that have a closer connection to

physiology or psychology than to physics or engineering have received

1

Ears A

Psychological > Thought "Noise" 1 1

1

Articulation

Speech in Mask 1

Acoustic Noise Demand valve etc.

Microphone

Transmitter

40

Fig 2.8. The communication chain.

Propagation

Noise in Sea

Receiver

If Necessary.

Earphone 1

Acoustic Noise Bubbles, Breathing etc.

Ear

Psychological

• "Noise"

Perception

32

DIVER ONE

DIVER TWO

33

brief mention in chapter 1. All these links must be considered

when designing a communication system. After all, it is of little

use if the world's finest electronics are operated by a switch,

too small to be grasped by the hands of the diver who is trembling

with cold or anxiety:

The writer has set out to examine this chain, to determine

its weakest links and to concentrate his attention on these. With

this in mind the first stage was to design and construct, from

first principles, a communication system and to compare this with

'those designed by other workers. This initial step was to produce

a better understanding of the whole field and some knowledge as

to the areas where present hypotheses are failing.

The outcome of the early pilot experiments was that the writer

considered that there were two links in this chain that were

particularly weak and furthermore were applicable to study with the

resourses available in a physics laboratory. These aspects were

the formation of speech in a confined cavity and the mechanism of

human hearing underwater. The writer chose to examine the

second of these in considerable detail.

After the initial experiments with complete communication

systems, the difficulties of making measurements in open water were

appreciated. The main problems concern the sea surface. When taking

low level acoustic measurements from a boat, wave motion and the

noises generated on the surface would inevitably be fed down the

hydrophone cable. Further, ironical as it may seem, the problems

of communicating with human subjects underwater proved formidable.

34

It was for these reasons that attempts were made to

construct an underwater laboratory to use for acoustic observations.

This is not the place to describe the designs which were not built

through lack of finance. However, attention will he turned to

one successful design which was used to measure the threshold of

human hearing underwater.

33

Chapter 3

A COMPLETE UNDERWATER COMMUNICATIONS

SYSTEM

3.1 Introduction

The simplest method of voice communication would be to

shout loudly into the water and to use the unaided ears for

reception. However, two serious problems arise. Due to the impedance

mismatch between air and water, only a very small fraction of the

sound energy will be radiated into the water. Furthermore, the

restrictions on the face imposed by most types of diving masks would

make it difficult to shout anyway.

The simplest useful diver communication system is that which

has been referred to in section 2.3 as "direct audio". The diver

who is speaking is given a microphone, amplifier and transducer to

transmit his voice in a similar manner to a conventional public

address system. No receiver is required. As a first stage to

examining the communication chain, a direct audio system of this type

was constructed. The design will be discussed in the next few pages.

3.2 The Microphone

It might be suggested that almost any conventional microphone

mounted in the air space of the face mask, would prove suitable.

However, a diving mask is liable to flood at any time and the microphone

must be capable of withstanding not only immersion but also the

hydrostatic pressures involved. Three possible solutions are ,

illustrated in figs 3.1, 3.2, and 3.3. The first is to use a

36

+ 10 volts.

Fig 3.1

Electrodynamic insert

Resin

Air space Flexible membrane

Fig 3.2

Telephone earpiece mechanism

Fig 3.3

—100db -

—110db

Fig 3.5

input

Diaphragm open to sea on both sides

Hollow piezoelectric cylinder

Fig 3.4

non

ring

'Crystal' microphone insert

Response (reL 1 volt / microbar)

500 1K 2K 4K 8K Hz.

111 220

2N3702

2.2K

1K

2N3810

100p±'

37

conventional microphone with a relatively small internal air

volume. This is mounted inside an enclosure, one wall of which

is made from a flexible membrane. The membrane both transmits

the sound and can flex to accommodate the hydrostatic pressure.

The microphone in fig 3.2 is constructed like a telephone earpiece

and the moving part, the diaphragm, need not be protected from the en—

vironment as it is free flooding on both sides. Finally it is

possible to produce a microphone sufficiently rigid to withstand

the hydrostatic pressure. The small sealed piezoelectric ceramic

cylinder shown in fig 3.3 is possibly the simplest example of this

technique. Incidentally, one advantage of the "stiffn microphone

is that the system self resonances are generally well above the

audio range. This will tend to produce a relatively level frequency

response.

The writer constructed several microphones of this latter

type. The construction technique and a typical response curve is

shown in Fig. 3.4. One may recognise the active part as being the

insert from a domestic crystal microphone. These were chosen because

they were readily available, an important point when one considers

the high mortality rate of any equipment used in the sea.

Unfortunately, although the microphones that were constructed

had an excellent frequency response and were mechanically robust,

they were very insensitive. In Appendix 1 it is shown that even with

a well designed amplifier, such as the design in Fig. 3.5, the

electrical noise will correspond to a sound pressure field of around

2 ±64 db with reference to the usual origin of .0002 dynes/cm. The

writer was of the opinion that, under the conditions of use

where the noise level produced by the diverts breathing equipment

is high, this disadvantage is outweighed by the advantages.

Up to now it has been assumed that the microphone would be

placed inside the facemask to nick up airborn vibrations in the

conventional manner. On the other handl there are two alternative

microphone positions that have been used. The microphone can be

pressed against the boney structure of the head (29) or against

the side of the throat (31). Although the former, known as a bone

conduction microphone, is capable of better reproduction than the

throat microphone, both of these are positioned outside the diver's

mask and consequently have a tendancy to pick up water borne sound.

When attempts were made to use these forms of microphone with a

direct audio communication system there was a tendancy to "howl""

because they were receiving the transmitted signal through the

sea water.

3.3 The Amplifier

The amplifier circuit is conventional It employed a

quasi-complementary, output stage capable of a nominal 10 watts

R.M.S. sine wave drive into a 3 ohm load. This form of amplifier

was first described for audio work by Toby and Dinsdale (32).

In detail the design differs from the original in the use of an

output transformer to provide flexibility in the choice of load

impedance, and in the addition of a low-noise preamplifier.

38

39

Power for this amplifier was provided by a 22 volt nickle-cadmium

secondary battery with a capacity of 500 mAH. It is not

proposed to discuss the circuit further as almost any similar design

would have been suitable.

The mechanical design of the amplifier is rather unusual.

The battery and the amplifier are housed in a perspex tube (fig 3.8).

Perspex was chosen, not for its strength, although this design is

adequate for use to 60 metres but because it is transparent. This

enables any leak or ingress of moisture to be detected at an early

stage.

The method of sealing the microphone and transducer cables

and the method of sealing the control spindles should be clear from

the diagram. These are not the only possible methods but are,

in the writer's opinion, the most reliable. However, there is one

fault which this type of seal will not give protection against.

This is known as "hosepiping" and occurs when water leaks through

a break in a cable sheath and travels down the inside of thn cable,

through the sealing glands and into the instrument. (Probably the

best known example of hosepiping was the accident with the habitat

Sealab 111 where helium travelling up the inside of the electrical

power cables caused the project to be abandoned (33). As the cable

lengths are short in this equipment it was decided to accept this

risk to obtain the ease of changing transducers and microphones that

these seals offered.

Battery Amplifier Control knobs

Brass insert

ring

'C' rings

Brass/ spindle

Fig 3.8 The construction of the underwater amplifier

Cables to transducer & microphone

'Perspex' case

40

Detail of cable seal Detail of control seal

Fig 3.10 Response of 'Subaqua 10' transducer (taken from manufacturers data)

40 '

30 ' db.

20 •

10 -

200 500 1K 2K 5K 10K

frequency Hz.

3.4 The Transducer

A pressure equalised moving coil transducer of similar

design to that discussed in 2.3 was chosen. The construction of such

a device is similar to that of a conventional loudspeaker and best

performed with the facilities normally found in a specialist firm.

Consequently all the devices that were used in the course of this

research were either converted from loudspeakers or constructed by

a loudspeaker manufacturer. The photograph (fig 3.9) shows four

different moving coil transducers; one was a converted loudspeaker;

the others, specialist products. One of these latter, the Goodmans

"Subaqua 10" proved the most effective and reliable and unless

otherwise mentioned it can be assumed that this unit was employed

on all the tests recorded here. The response of the "Subaqua 10" is

shown in fig 3.10.

Before passing on it might not be out of place to qualify

the word "reliable" as it has been used in the preceding paragraph.

All the transducers illustrated in fig 3.9 leaked in service, the

worst leaked more often than not. The "Subaqua 10" leaked once and

with help from the manufacturers was rebuilt with improved seals.

It has since proved satisfactory over a number of years.

3.5 A Receiving Hydrophone

In the foregoing paragraphs the components of a simple direct

audio communication system have been discussed. For the monitoring

of subjects using this equipment and for the purpose of recording,

some form of hydrophone is required. Initially a hollow ceramic sphere

41

Fig 3.9.

Four electrodynamic transducers.

Goodman's "Sub—Aqua 10"

Modified loudspeaker

Raytheon "Yack—Yack" University sound underwater

loudspeaker

42

43

with a nominal resonance of 80 KI-Iz and a sensitivity of around

-103 db ref. 1 volt/pbar, below resonance, was sealed onto a length

of coaxial cable. However, although this proved useful in the

laboratory it was found to have a disappointing performance in the

field. The main problem was noise generated by the long length of

cable. This not only propagated as an acoustic signal down the cable

to the hydrophone but also appeared to give rise to an electrical

sig-nal directly. This latter problem was found somewhat surprising.

One might have expected an electrical signal to be generated in a

similar manner to that generated by a condenser microphone if there

had been a potential across the cable. However care had been taken

to see that this situation had not arisen. When a selection of

coaxial cables were tested by connecting them to an amplifier with

an input impedance of 1 megohm, the least satisfactory (unfortunately

a popular brand) generated several millivolts when handled and was

capable of generating a signal of over one volt when cracked like

a whip.

To overcome these problems a preamplifier was constructed

and attached to the lower end of the main cable. This was designed

to amplify the hydrophone signals and to apply them to the cable

from a low impedance source. The piezoelectric sphere was coupled

to the preamplifier by about 20 ems of lightweight cable and

provided with a small float so that it would hang just above the

preamplifier, the latter acting as an anchor.

When used in relatively shallow water (less than 10 metres)

44

with the preamplifier laying on the sea bed, this system proved

very satisfactory in use. Cable noise, both electrical and

mechanical, was completely eliminated. The circuit is reproduced

in fig 3.11 along with the electrical and acoustic performance.

The latter was measured by comparison with a known condenser

microphone in an airborne sound pressure field. When examining

this design it should be recalled that this preamplifier was constructed

during trials in Malta in 1966 and the availability of components

was limited. This was one of the reasons that the method of

waterproofing the preamplifier was to encapsulate the complete

circuit inside a film cassette can, using epoxy resin as shown in

fig 3.13.

On return to the U.K., it was decided to construct a hydrophone

along similar lines to the above but using a design capable of a

better noise figure, utilising low noise cable, and providing some

useful additional features such as headphone- monitoring. One

addition was the provision of a high pass filter. This can be

switched to remove frequencies below 200 Hz and is useful when

listening with headphones to discriminate against sea—state noise.

In the illustration, fig 3.13, this hydrophone and preamplifier

can be seen alongside its predecessor. The somewhat unusual

circuit is reproduced in fig 3.12.

Unfortunately all was not to prove well for this improved

design. When it had its first serious use in connection with

hearing trials in 1968, the piezoelectric sphere leaked sea water.

45

T2.2 1<

-r

-Ft v.

11 /c

Submerged unit

V/ 3707 /N3707

90 IC 0,F 3.9 k

Fig 3.11 Hydrophone pre-amplifier

r 5 K'

2 1-1

99,-r 4-70 f~ f K 910

f Fl

33

Submerged unit

n

try

2 N 317 /

33k

J

I Re?,

4.4 •

.o33,.F

TO 0

Eltic

AN 3704- //I/

2143707

3•7h

77i /

Electrical perfor:iance with 12 m of cable between.pre-amp and surface.

-72db ref 1 volt/microbar. Gain 33db, Input impedance 800X GOpf. Acoustic sensitivity

Fig 3.12 Hydrophone pre-amplifier

/0.2

J 10k•

1/7 3702.

Gain 40db, Optional filter (18db/octave below 20011z) shown switched out.

46

Fig 3.13.

Hydrophones used for recording underwater signals

Clevite CH-13

Used in underwater laboratory.

Spherical hydrophone with

Spherical hydrophone

pre—amplifier built in with pre—amplifier using

film cassette tin. FETE and high—pass filter.

Although this hydrophone has not played a significant part

in the experiments discussed in this thesis, the writer feels that

the design is sufficiently unusual to warrant inclusion.

47

Fig 3.14.

Direct audio system in use. Malta 1966.

48

49

Chapter 4

INITIAL L^_i_PartElE7:TS WITH CO,a1UNICATION ONIPMENT

4.1 Introduction

In the summer of 1966 various types of communication equip—

ment designed for use by divers were transported to the island of

Malta. These field trials were organised as one of the main projects

of an expedition sponsored by the iloyal Geographical Society and

Imperial College Exploration Board. The team consisted of two

postgraduate and five undergraduate students under the leadership

of the author.

The apparatus used in Malta included communication equipment

built to the design described in the previous chapter and two

commercial communication sets. A series of tests were conducted to

probe the workings and shortcomings of these types of devices.

4.2 The formation of words underwater •

The subjects were seated in about 6 metres of water rearing

conventional open circuit aqualung equipment. A cable from a micro—

phone of the type shown in fig 3.4 was led to an amplifier',

loudspeaker and tape recorder on the surface. The subjects' were

asked to- read a test paragraph, the editorial fro:: the Daily

Telegraph was popular, as distinctly as possible. For comparison,

recordings were made with subjects wearing diving equipment on the

shore. The tests were repeated with different types of facemask

and demand valve. These are illustrated in fig 4.1. There was no

disagreement by the subjects as to the merits of the different

50

Fig 4.1.

Three different types of diving 'facemask'

Full—face mask with

Mouth—cup (or mouth—mask)

demand valve attached attached to twin hose

demand valve

Mouth 'bit' (shown on single hose demand valve)

( The mouth bit and mouth—cup would be worn with a conventional

mask covering the eyes and nose.)

51

types of faceaask. All these tests were performed with a single

hose demand valve.

Condition Comment

Intelligible All masks in air

Full face mask underwater Words missed but meaning generally clear

Many sentences had to be repeated to understand them

Some divers could convey simple words but most could achieve nothing.

Mouth mask underwater

A conventional mouth—bit

Although in not such a drastic manner as the face masks, a

change in demand valve did effect the ability of the diver to

articulate. The actual noise produced by the valve did not seem

too important as most subjects chose a very simple, very noisy,

single—stage valve as being the most convenient for voice commun—

ication *. In general, single hose valves were preferred from twin

hose models and the least liked was a twin hose valve that was

notable for a high exhalation resistance.

4.3 Noise and ran!,:e of communication

ever a range of '25 metres the direct audio set was compared

with an G 17.1z upper—side—band A.M. ultrasonic communicator. The

latter contained a send—receive switch and was operated in a similar

way to an ordinary "walkie—talkie". The table compares the effect

of noise on these two systems.

* The valve referred to is the "Normalair" single hose model. It is made almost entirely of plastic and the pressure reduction from cylinder pressure to ambient pressure is accomplished within 3 ma of the diver's mouth:

52

Origin of Noise Description ._ „ 17e,',1,- icr n effect

on Iiirect .udio ?,:asking.., -1inr, effect :

on Carrier System

Sea state, and waves on shore

Continuous, Low frequency Very little None

Diver's demand valve

Wide band (hiss)

Complete masking

Complete masking

Biological (snapping shrimps)

Impulsive '

(crackling) Small Considerable

human movement :bubble noise

Intermittent Easks weak signals

Small

Reverberation Related to the signal

Not obvious Very noticeable

As can be seen, the serious noise sources over which the

designer has some control are the denand valve, movement of the

body in the water and e:thalation bubbles. It would seem unlikely

that there is any way of silencing the conventional demand valve

sufficiently to avoid masking incoming signals. It follows that

for reliable communication both parties must synchronise their

breathing. Ironically this is aided by the bursts of noise received

from the breathing equipment of the other party. It was for this

reason that steps were not taken to prevent breathing noises from

being transmitted. The use of closed, or semi—closed circuit

breathing apparatus (see 1.2) should considerably reduce the self

generated noise. Unfortunately this type of diving equipment was

not available for comparative testing.

The maximum range of he ultrasonic set was limited to

about 25 metres by the signal becoming garbled with multiple

path distortion. Although this is considerably loss than the

manufacturers specification of up to 5 kilometres, it is felt

that this latter figure probably refers to the use of this

equipment in the open ocean, not in shallow rocky conditions.

nen the moving coil transducer that was used in the direct audio

communicator was provided with a good electrical signal from a

surface microphone and amplifier it could be clearly heard at

ranges of over 100 metres. nen used by a diver the range of

reliable communication was reduced to about 30 metres. This

reduction was probably mainly due to the added distortions from

the diver's racamask. Finally, the rane of a commercially produced

direct audio set (Raytheon Yak—Yak) was found to be between

5 and 10 metres. The particular model available gave a relatively

low output and produced a considerable amount of (dstortion.

the device was completely encapsulated, there was no way of

investigating these shortcomings).

4.4 Human hearin-r underwater

One property of multiple path Listortion is that in general

the false signals will arrive at the receiver from directions

that will differ from the direct signal. If the human hearing

faculty is able to perceive the direction of a sound source

underwater then it is possible that this may be of some help in

reducing the effects of multiple paths.

53

34

To test directional hearing wide—band sound, tape

recorded music was the best available signal, was radiated from

a transducer positioned about 5 metres below. the surface. It

was necessary to use a signal with a wide bandwidth as a narrow—

band one, such as a pure tone, would produce a standing wave

pattern. Under these conditions it may be impossible, in

principle, to define the origin of the sound field. The subject

divers were positioned about 10 metres from the Source and at

the same depth. With their eyes closed, the subjects were first

spun round and then asked to point to the direction of the source.

ro protective clothing was worn on the head. The divers were

requested to try and remain motionless while making their

judgement; fig 4.2 shows the result of 40 judgements. The angles

were measured to the nearest 45 degrees, the figure on the left

is a polar plot of the subjects judgements with respect to the

true direction of the source, and that on the right, a plot of

the direction that the subjects were facing. These results were

tested for significance usinz a method that will be described in

detail later. suffice it to say at this stage that these' results

fail to meet a criteria of significance (i.e. there is more

than one chance in 20 that the subjects indicated random directions).

4.5 Discussion

These trials were in tie nature of pilot experiments and were

in consequence, relatively primitive. After allowing for this there

are three main questions that arise. Is there a physical explanation

Direction of source in both cases

A

CD

0

0

1" •

C4. O

p

CD p w.

CD cn cf-

0)

Direction that the subjects indicated

Direction that the subjects were facing

SCATW, *" represents one indication

56

to the problems of forming words underwater? Nhat limits the

range of a direct audio communicator, and finally, what part,

if any, does directional hearing play. in communication between

divers? These rill be considered in order:

4.6 The formation of words

It was seen in section 4.2 that it was easier for a diver

to form words into a large facemask than into a small one, and

most difficult of all when there was no mask over the mouth.

Both Webb (20) and Hainan (30) have also reported that some types

of mask have a distinct advantage for voice communication. Apart

from accounting for this observation it is also necessary to

explain why there is a difference between the same equipment worn

in air and underwater. An interesting demonstration of this latter

point was performed by taking a fully kitted diver and monitoring

his voice as he slowly lowered himself into the water. The most

noticeable drop in intelligibility came, not when the exhaust

ports of his breathing apparatus submerged as might be expected,

but when his head and mask went underwater. As far as is known

there is not a published mod.el to account for the difference

in facemasks or this last observation.

The writer proposes to adopt a model of the vocal tract

proposed by Dunn (25). This suggests that the larynx can be

represented as a sound source feeding a series of cylindrical

tubes which have similar dimensions to the throat and other

S7

cavities in the human vocal tract. This mechanical model can

then be expressed in electrical terms For ease of analysis. The

writer proposes to add one additional cylinder as a termination

to the model tract, to represent the case where a diver is

speaking into a faconask underwater. The reason for assuming that

a mask can be considered as a closed cavity when used underwater,

but not when used in air, is that in this latter case the walls of

the mask can vibrate as a membrane and transmit some sound.

However, underwater the air—rubber—water aC01.12'6iC impedance

mismatch will prevent appreciable tranmnission and the mask will

act as a closed cavity.

After this model, both with and without mask, had been

reduced to simple electrical circuits, these were constructed and

direct electrical measurements made on the effect of the mask.

Dunn (25) has shown that the formation of vowel sounds by the

vocal tract can be raDresented n

Fig 4.3

length

larynx

a transmission line (fig 4.3)

lips

area Ai

The throaty tongue constriction, mouth and lip constriction are

represented by four tubular transmission lines with lengths 11

to 1 and cross—sectional area Al to A

4 respectively. The vocal

chords are assumed to be an acoustic source of constant volume

(generally referred to as a high impedance source in analon-v to

the electrical case). In this model the resonances of the

transmission line will play the same important role as the

resonances of the vocal tract; that is, they will define the voice

fonnants. No nasal cavities are considered because Dunn suggests

that these are normally closed in good English vowels. In diving,

it has been suggested by `;'ebb (20) that as the mouth and nose

are often in different 'masks' at slightly different pressures.,

this will tend to close the ulvula.

Now if dissapative (i.e. resistive) terms arc ignored,

Dunn showed that each of t':e four small sections of his model

could be reduced to the following electrical equivalent:

Fig 4.4

L L 1

A C

L tan(2.1. C = oc)

/De cosec (wl A c /

58

59

The complete vocal tract of fig 4.3 can now be replaced by the

network below. The output current (representing a volume velocity

from the lips) passes through an analogue of the radiation

impedance of free air given by:—

Fig 4.5

cox 2 c

L, = ,§4 3 V A4 TT

Duml suggests that as C2 and C4 both depend on the volumes of small

constrictions they can be ignored. Further as L1 is in series

with a high impedance it too can be ignored. This reduces the

electrical model to the following:—

Fig 4.6

L,+ L3+ 2Lz

tb= L3+ 2L*

C,= C,

C4= C3

60

This is the electrical model of the vocal tract after

Dunn. Before using this with problems concerning facanasks

there is one further simplification that can be made. This is

the assumption that the dimensions of the vocal tract are small

compared with the wavelengths involved and hence to replace the

tangents in fig 4.4 with their arguments and the cosecants with the

reciprocals of their arguments. Although Dunn does not like this

"lumped constant" assumption but prefers to solve the relations

that include trigonometric functions by graphical means, he does

use this approximation latter in his work and produces the first

two fonnants to within 10'10 of the more accurate method.

The writer proposes to use the lumped constant approximation

because it is then possible to construct the model from relatively

few components in the laboratory. A check on the accuracy of these

approximations is provided by comparing the experimental values

with the formant frequencies calculated by Dunn using both the

more accurate "distributed model" and the lumped constant mojel.

The writer has calculated the electrical components .

corresponding to the three Znglish vowels in the words EAT, LOST

and BOOT. These are the sane three vowels that Dunn analyses. The

physical dimensions of the real vocal tract as obtained from X,--ray

photographs are also taken from this reference.

To sum ups the writer has taken a published model of the vocal

tract and deduced the values of electrical components corresponding

to three vowels.

61

Before constructing three electrical circuits, the writer

"rationalised" the component values by multiplying the inductances

and resistances by 100 and dividing the capacitors by the some amount.

This has the advantage of keeping the same frequency scale but

enabling readily available components to be used. Fig 4.7

tabulates the vocal tract dimensions obtained from Dunn (25) and

lists the electrical values appropriate to the lumped constant

model discussed above.

The electrical model or the facemask has been calculated in

the same way and the component values are listed in fig 4.8 along

with the dimensions of the masks.

In the test circuits that were constructed (see fig 4.9) the

laryn:: current generator has been replaced by a Bruel a Ejaer beat

frequency oscillator with a 40 Liillohm series resistance. The

value of the radiation resistance is not over significant (Dunn

chooses to ignore it) and has been aporo:;.inated to a fi::ed 10 ohm

resistance which would be the correct value at about 800 Hz. !lien

the inductive and resistive components corresponding to the

radiation impedance of free air are replaced by the eciuivalent

circuit of the facemask, one should correctly measure -che potential

across the caoacitor C . This would correspond to the acoustic

pressure inside the mask. liol:ever for convenience, the current

passing through C was measured by monitoring the potential across m

a suall series resistance (10 ohm). Strictly speaking we are now

401L(2. .4211 .21511

1 0-0-

0

401L9- .42H .17H

1.4uF

To level recorder.

T3F3PF T°' 1 0 -n-

0

Fig 4.7

Mechanical and electrical values for vocal tract model.

Vowel 1,

Mechanical

lz 13

dimensions

lg

(from Dunn)

A3

Electrical (see Cb

fig La.

equivalents 4.6)

Lb L,

EAT 7. 1.5 4.5 0 7.7 0.9 1.5 1.5 38 4.3 4.2 1.7 .45

LC ST 3. 4. 8. 1. 2.3 .75 10.9 6.3 5.0 61 7.12 .6 .22

BOOT 6. 3. 7. 6.351 1.14 5.0 .52 27 25 4.3 3.0 .76

cms 2 cms pF mII

Fig 4.8

Mechanical and electrical models for the facemask.

L. L,

Fig 4.0. Examples of two of the test circuits (using rationalised values)

62

Mask Volume Length C, L,,,

Full face 1200cc 10cm 8451.1F .045m11

Mouth mask 200cc 5cm 140T1 .07mH

Bit 20cc 2cm 14r .28m11

'EAT' in free air

'EAT' with mouth mask

measuring the velocity at the lips and not the pressure in the

mask. As these will bear a simple relation to one another and

either will indicate resonances, this substitution is of little

importance.

Twelve electrical circuits were constructed. These

correspond to the three vowels in air and into each of three types

of mask. (The mouth bit of a conventional demand valve has been

treated as if it were a very small mask). The response of these

circuits is shown in fig 4.10. The position of the peaks, or

resonances, which represent the voice formants are of greatest

interest. Before looking at the effects of the masks it is worth

comparing the position of the fonnant in free air obtained by the

writer with those calculated by Dunn.

Vowel "acact" distributed Lumped constant writer's constant model, fonnants model, fonnants eN.nerimental calculated by Dunn using calculated by results taken graphical methods Dunn from fig 4.10

1113 325 2300 322 1650 300 2000

LOJT 640 930 625 873 Broad hump 600-800

BOOT 310 850 .305 794 300 800

The agreement is sufficiently close to give confidence in

the method. The only critisn would be that the 2 of the electrical

circuits is somewhat low. As this was determined by the components

available there is no simple wa: of overcoming this.

Perhaps the most noticeable effect of the mask is a

63

1\iouth bit.

_

cr,

fivimP4'"

N.N.4, 444

CJ CD rn

0

rn

0 1,1

0

0 0

CD

0 CD

0

ti

1 °Batt- -'Lost'

11.10•••• =him*

100 200 500 1K

2K. Free air. frequency Hz.

Full face mash.

Mouth cup.

N

65

negative one. The somewhat drastic change from the small inductive

impedance of air to the larger capacitative ones of the facemask

appears to have had remarkably little effect. To take one exampled

at 300 Hz the change is from about 1 ohm inductive to nearly

4 ohms capacitative in the case of the mouth mask. The only effect

of this appears to be a rise of between 50 and 100 Hz in the first

fonnant of ELT and LOOT. However, in all cases the very small

volume associated with the mouth bit has drastically effected the

formants. No further explanation need by sought to the extreme

difficulty in forminc, words in this case.

It must be remembered that this model is in one way very

different from the true method of voice production. This model

does not include the feedback between the mouth and ears. i!or

this reason real subjects would not be expected to proCuce these

vowel sounds as they would try to compensate for the distortions

that they hear.

To conclude, although the electrical model of the vocal

tract would indicate that the formation of vowels into a mouth bit

or directly into the water is difficult it does not explain why

the larger facemasks produce such poor articulation.

If is unfortunately not possible to produce a simple model

for the production of consonant sounds. For this reason the above

analysis must only be considered as representing part of voice

production.

66

4.7 The range of communication

In theory'. one might expect that a communicator

which relied on the unaided ear for reception, would be

limited in range only by the distance over which the signal

falls below the threshold of hearing. In general this is

not so. It is the steady deterioration of the signal to

noise ratio as one moves away from the transmitter which

sets the maximum range. As the range was increased it was

found necessary to remain motionless to hear the signals. Now

the outer ears are closed by the presence of water when diving

and this condition will give rise to an improvement in hearing

by bone conduction (between 15 25 db, Zwislocki 34). Hence

body and equipment movement noises which may be carried through

the human body become important underwater. These sounds,

along with the exhalation bubble noise, rere found to be an

important limitation to the range of cannunication with the

working diver. A particularly Jimpledemonstration of this is

to scratch the back of one's neck. In air this is not a

particularly loud sound, whereas unaerwater it will appear

far louder and will probably mush other hearing.

4.8 Directional hearing

Binaural hearin-4 can be described as the ability to use

both ears to receive acoustic signals and to respond to small

differences between these two signals. Probably the best

known exanple of binaural hearing is directional hearing.

By detenuining the time, or phase, difference between the

arrival of a signal at the two ears it is possible to

calculate a possible direction for the origin of the sound

in the horizontal plane.

Time difference = d cosy

The diagram shows how this can be deduced. In order to

determine which of the two possible directions is the true

one, it is helpful to move the head. Only the true image

will remain stable.

Before proceeding it should be clear from the diagram

that if we expose the head to a steady state sound field with

a wavelength which is less than the interaural distance, there

rill be ambiguities as to the direction of incidence. This

arises because the two ears may measure their time scale from

different cycles in the wavetrain. In air this would correspond

to frequencies higher than about 1600 Hz. It is for this

reason that a sound source that is used for directional hearing

67

68

tests is often a wide-band interupted one, such as a series

of clicks. Most natural sounds, including speech, fall into

this category.

This picture of directional hearing is oversimplified.

There are two other mechanisms which play a part. The first

concerns differences in amplitude. The head will act a shield

and the ear which is closer to the source will receive a louder

sir-,:nal. This becomes increasingly effective at short wavelengths

where the head becomes a proportionally larger barrier. In air

this conveniently becomes useful at frequencies where time or

phase differences are ambiguous. The mechanisms discussed so

far would preclude a person who is deaf in one ear from

directional hearing. It is certainly true that directional

hearing is very much impaired by deafness in one ear and that

person cannot be said to possess binaural hearing. however, the

fact that such a person may have some directional abilities

requires an el%:planation.

Reflector

a

a

b

•

_ , ;_ --- - b -

Microphone

time

time-3

69

In the diaA:rmn, consider the nature of the signal received

by the microphone when a single pulse is emitted from 'AI

and 'B'. The single microphone with a reflector is capable

of differentiating between these two directions. It appears

that the human pinna is capable of acting as such a reflector.

It has been shown by Datteau (35) that reflections from the

pinna are observed as a change in timbre. If such a change

is made artificially, the subject will perceive it as a change

in source direction.

Directional hearing has an important role to play in

everyday life. It enables persons to become aware of danger

such as the approach of a motor vehicle. It has also been

shown to be important in the discrimination of a speech signal

in the presence of noise (36, 37).

Possibly the most well known example of this is the

"cochtail party effect" (Cherry 3S). A person "listening" to

several conversations simultaneously has little difficulty in

isolating an c' following any one his chooses. It is generally

not possihle to do this when listening to a "monophonic"

tape recording of the same conversations. (This can sometimes

be demonstrated with a stereophonic recording and this point

has been used to advertise -CAP, type of equipment). The abilit

to localise the source of wanted signal as a point in space

is not a necessary requirement and this effect can be

demonstrated where no "real" sound ima, es exist*.

70

Early workers in the field of underwater communication

dismissed directional hearing as not possible (Wainright

Interaural time delays would be shortened due to the increased

velocity of sound and the shielding effect of the head will be

quite different. However, if directional hearing is possible

underwater it would provide an explanation for the superiority

of the direct—audio communication system in noisy, reverberant,

shallow water conditions. After all, reception in these

conditions should be analogous to that in the above "cocktail

party".

One might suppose that this hypothesis could be tested

by covering one ear and looking for a drop in intelligibility.

Unfortunately very little is known about the mechanism by

which sound reaches the submerged ear. If this is by some

mechanism similar to bone conduction, a cover would be ineffect—

ive. A further possibility is to test directional hearing by

asking the subjects to indicate the direction of a sound source.

This is the avenue that the writer has chosen to explore. The

results obtained in 10GG (presented in paragraph 4.4) are

inconclusive. This problem rill be examined further.

* During the last war, messages were transmitted to the French Resistance 13: sending a cor:mon voice signal from two radio transmitters. If the enemy "jammed" these tranmissions then each would be jammed separately. To receive the message, two radios were tuned in, one to each transmitter, and the listener sat bet,:Teel: the radios. The effect was that the speech a:)peared to separate from the jamming noise and could be understood. The 1313c have a reconstruction of this in their archives.)

71

Chapter 5

7.3.-fr2E:LE.1=S (1)

5.1 Introduction

The importance of binaural hearing in everyday life

and particularly in speech communication has been discussed

already. lowever it remains to be seen as to whether

binaural hearing is possible underwater. The reaction

obtained from (mestioning divers is all too often that it is

not.

There are two main methods used for testing binaural

hearing on the surface. Both of these are discussed by Nordlund

(10). The first method is to provide the subject with a pair

of headphones and to insert a time delay or amplitude

difference in the sound that is being played through one

earpiece. The subject is asked to localise the sound as a

"phantom ina.ge" inside his head. .il. common test is to

introduce a time delay anL to ask the subject to recentralise

the image by increasing the amplitude of the delayed channel.

In this manner information as to the relative merits' of time

and amplitude differences can be explored. This headphone

method has been reported as being particularly useful in the

diagnosis of brain lesions (Nordlun:: 40).

The alternative is to position the subject in an

anechoic chamber and to ask him to indicate the direction of

a real sound source such as a loudspeaker. The chief problem

72

with this method is the provision of anechoic surroundings

although there is the subsidiary problem in the selection of

a method for the subject to use to indicate his estimate of

source direction. It was proposed to adopt this second

approach underwater as it should yield results that are

directly applicable to the real environment.

Anechoic conditions are virtually impossible to obtain

at audio frequencies underwater. The best that could be asked

for would be a sandy bottom in the open sea, far from the shore.

(Mud would be better as a sound absorber, but would introduce

safety and logistic problems). If visual or photographic

methods were to be used to record the subject's reaction, it

would be essential to use clear water. Finally the water

temperature must be sufficiently warm to allow subjects, who

would not be wearing protective foam rubber around the head,

to remain practically motionless for at least 33 minutes.

To meet these requirements the directional hearing

tests were organised as part of an undergraduate expedition

to L:alta in a similar way to the cgmmunication experiments

described earlier. This took place in 1908 under the

sponsorship of Imperial College Exploration Board.

The decision on the form of the directional hearing

tests was made after some preliminary experiments off the

French 1:editerranean coast earlier that year. Two main methods

were used. In the first the subject diver was suspended in

73

mid—water and asked to point to an audio frequency source.

In the second series of tess the subject was seated on a rigid

box and asked to judge from which of two known positions a

sound vas originating.

5.2 The round L;ource

The requirements for the source were laid down in

section 4.5. The source must be capable of producing a wide—

band, non—continuous sound of sufficient amplitude to be heard

by a diver under normal conditions. Initially the writer

considered using an electrical tone generator coupled to

a transducer. However, such emiipment tenus to be cumbersome

for a diver and lacks the degree of reliability that would

be preferred before embarking on field trials. To overcome

these disadvantages a simple electromechanical unit was

designed. It consisted of a small container, one wall of which

was mace from sheet metal. Inside this air—filled container

a small hammer was arranged to strike the metal which acted

as a diaphro4la. The repetition frequency of the hammer, about

20 Hz, is too low to be radiated efficiently by a device of this

size, but the signal is rich in harmonics and these) along

with resonances of the case, are transmitted through the water.

Figs 5.1 & 5.2 show how these devices were made.

In use the '0' rings are gre sed, the batteries connected

and the lid pushed on. It is necessary to slacken the bleed

screw to do this, otherwise air pressure will force the lid

"(Yrmg.olil

Battery

R„

Magnet Bell mechanism

Perspex case

ti Bleed valve

Reed switch

74

Fig 5.1 Photograph of the 'buzzer' sound source. (with lid and '0' ring removed)

Fig 5.2 Construction of the 'buzzer' sound source.

Metal lid

The metal diaphragm is a lid taken from the type of tin that is commonly used for packing bulk unexposed 35mm film. It was chosen to be a loose fit over a standard diameter sample of -21- inch wall Perspex tube. An open ended Perspex container is constructed from this tube with an '0' ring seal for the metal lid to fit over. Inside the container is mounted an ordinary electric bell mechanism without the gong and with the hammer bent to strike the tin lid, a battery holder with four "high power" size AA (penlight) batteries and the insert from a magnetic reed switch. On the exterior a small circular magnet is mounted opposite the reed switch and a 4BA hole is tapped through the case and counter—sunk on the outside. A SBA cheesehead bolt with a small '0' ring under the head completes the instrument.

75

off again. The first time the instrument is taken underwater,

the hydrostatic pressure will defoni the metal diaphragla to

dish shape. This is of little consequence as the bell hammer

can be re—bent to accommodate this. Two sources have been

made to this design; they have been used to depths of GO feet

and can be heard by a diver, who is listening,ovei- ranges

or up to a hundred metres.

The response of one of these devices, taken from a

tape recording made "on site" is reproduced below, fig 5.3.

The recording was made on a Philips Cassette recorder and the

response on this machine is also shown. The hydrophone used

was a piezoelectric sphere of the type described in 3.5.

For analysis the tapes were re—recorded on a laboratory

machine, formed into loops and replayed through a Druel

ijaer octave filter. The mean level at eacll frequency was

estimated from observations of the meter incorporated in the

filter unit.

5.3 'J'ree choice experimonts

In these experiments the subject, who would be wearing

about eight pounds less weight than for normal diving, was

tied by one ankle to a fixed ballast on the sea bed. This

weight was normally in 8 metres of water on a sandy bottom.

Apart from a facemask with a hinged metal blackout flap and

the absence of any form of rubber hood, the subjects used

conventional aqualung e,:juipment. The sound source was held

76

Fig 5.3. Response of 'buzzer' sound source (Taken from recordings

made at 5 m in the open sea. Corrected for the tape recorder)

0

10db.

50 oo 200 500

frequency Hz.

. 1K 2K

• 5K

Overall response of cassette tape recorder

Sdb

So 100 /00 500 law 2e 5.e 10; Zc'k

77

by an accompanying diver who remained at the same depth as

at the constant distance from the subject (fig 5.4). A

surface swimmer wearing snorkelling equipment, was instructed

to position himself directly above the subject using the

subject's exhaust bubbles as a guide. The surface swimmer

held a "Calypsophot" underwater 35mm camera with a viewfinder

attached which was marked with "cross-hairs".

The method of conducting these tests was for the

accompanying diver to switch the sound source on and for the

subject, in his own time, to point to where he estimated it

was situated. The subject was permitted to move his head and

to rotate his body in his efforts to localise the source. Once

the subject hid chosen, he was instructed to hold his arm out

until the source was switched off. ',;hen the subject had

indicated his choice the cameraman would line his "vertical"

cross-hair on the direction of the source and take a "plan-

view" photograph of the subject. Aftet the exposure the

cameraman would si,gnal for the source to be switched off. In

order to allow for any confusion that may be caused by features

in the area reflecting the sound, the source operator swam to

a new position after every test. AlthoUgh the aim of the

team was to expose 36 frames at a time, the experiments were

often terminated after a shorter run due to the subject

suffering from cold. Under good conditions 36 exposures could

be recorded in about one hour.

-.7

Tape measure

Fig 5.4 Diagram showing the position of the divers in the free choice experiments.

Cameraman on surface

78

Cmneraman uses subject's air bubbles to position himself vertically above subject.

Sound source

Subject anchored to sea—bed about Sm below surface

79

5.4 Two choice experiments

In these experiments a triangle was marked with tapes

on the sea bed at a depth of 30 feet. 11 heavy steel box

provided a firm seat for the subject at the apex of this

triangle and he was allowed to study the layout before

closing the mask black—out. The sound source was positioned

at one of the two distant vertices and the subject had merely

to raise his left or right arm to indicate his judgement of

the source position. It was found convenient to have a third

diver to record the results.

In operation the diver holding the source would

either swim across to the opposite position on the triangle

or swim to the mid—point of the baseline and back again after

each test in order that neither the time intervals nor his

breathing rhythm should give additional information to the

subject. It was normally found possible to take GO judgements

within one hour.

5.5 nesults of Free choice ex,,eriments

The exi)osed film was processed and printed. The angle

the subject had indicated was measured from the photographs

with a protractor, 0 degrees representing the direction of

the source. The temptation to use these angular measurements

as a direct means of evaluating a mean CLirection and standard

deviation was avoided. (,,uch a proceL,s woulc: allow two angles

of 1750 cala — 175° to produce a mean of 0° with a large

80

deviation when in fact the angles are only 10 degrees apart).

The method used was to represent each measured direction as a

unit vector and to evaluate the mean vector for each subject.

The magnitude of this mean vector will be a measure of the

subjects consistancy and the direction, that or the subjects

average choice. The author considers that far more emphasis

should be placed on the magnitude than on the direction of

this vector, as under operational conditions a diver would

ex-.3ect to collect additional information with which to correct

overall angular displacement.

Subject

FiT 5.5

Mean Magnitude

Results

significance Jtandard Deviation

No. of Photos

Vector - Direction

M .L.

J.;:;.1::.*

J.W.

D. R.

R.L.

P.N.

P.Y .

18

14

24

22

33

35

11

21

0.80

0.84

0.75

0.87

0.52

0.23

0.63

0.72 0.01 —112°

0°

— 1° ,

+ 3

— 34°

— 23° o + 3 —< 53°

4'

<0.001

< 0.001

< 0.001

< 0.001

< 0.01

Not. Signif.

0.001

410

55o

46°

53o

36° 74°

500

Mean of standard deviations on significant occasions = 510

* A different sound source was used in these cases (These will be discussed in G.9

The above table shows the results obtained on eight tests

involving six subjects. The first question to ask is whether

these subjects are pointing in a random direction. The

results in column five are an evaluation of the probability

81

of this null hypothesis. (i.e. this is the probability that

the chance addition of this number of random vectors would

give this or a greater mean). Fig 5.6 shows the relationship

between the mean vector, the number of tests and the

significance. The derivation of this relation is considered

in Appendix 2. It can be seen that one test failed to meet

a 50 criteria for significance.

The interpretation put on the direction of the mean

vector is more difficult. From the magnitude of the mean

vector, it is possible to estimate the standard deviation of

the direction. This is shown in column six. Although there

is a tendancy to point to the left and this may be a failing

of riht—handed subject (all the subjects did point with their

right arm), with one exception this cannot be regarded as

significant. The one subject ('.:.L.) who did point to a direction

significantly different from the true one (P<.05), reported

that one ear had failed to clear properly on that occasion.

On a total of 10 of the 178 photographs considered in

fig 5.5, there was an indication of the true direction, of the

source. Either the source itself was visible or a tape running

between the subject's anchor and the source could be seen.

In these cases the angle between the true source position

and the side of the photograph was measured to provide a check

on the overall accuracy of the viewfinder and cameraman. This

angle shows a mean of 1.2° and a standard deviation of 9°.

Fig 5.6 Reliability of the mean vector

35 -

30'

25

Significance

'NI 20 '

Number of

vectors 15 '

5 `

Not significant

• • •

• • • •

82

• •

•

• N. • ▪ •••

• • •• \ • • \

0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Magnitude of mean vector

This figure is obtained from the table in Appendix 2. For small values of 'NI there will tend to be an error in the method that has been used for calculating the significance. l'T.owever, as 'N' tends towards unity, the value of a significant mean vector must approach 1.0. This fact has been used to provide some eNtrapolation for small values of 'N'.

10

a

83

Fib 5.7 Results of two choice triangle tests.

Angle Subject

Total % incorrect

111., ?off, PJ PST JSW JW BR

180° 39/9 45/1 84/10 12%

903 40/13 40/17 40/10 40/1 160/41 25%

60° 26/4 41/5

40/12 28/7 40/5 175/33 19%

45o

41/11 41/8 60/20 60/13 40/13 242/65 27%

20°

60/31 45/12 60/19

00/22 45/13 60/25

60/16 60/20 60/16

40/16 119/36

30/0 60/3

759/242 32%

NOTATIN

No. of judTeTients/::o. incorrect.

Fig 5.8 Consistency of subjects between exoeriments.

Free choice experiments rl1W) choice experiments

Subjects in order of mean vector

1:ean vector

Subjects in4.11 order of score on 20 triangle

% incorrect on 20' T'gle

.37 BR 10% JSW .80 JSW 28.5% JW .79 29.5%

.72 tfl7 30% PN .54 P11 36% RL .52 RL 52%

This list only includes subjects who participated in both series of experiments.

The correlation coefficient between the subject's mean vector and the % correct on the 20

j triangle has been evaluated.

Correlation coefficient = 0.99.

False True position position

Gaussian distribution

<—X

\\\ T

4 10.-t-10(1-2>

84

Although this error is large and it would be desirable

to reduce it on future occasions, it is insufficient to

appreciably alter the results in fig 5.5.

5.6 The results of two choice experiments

The total angle suspended at the subject by the two

source positions was recorded. Fig 5.7 shows the score at

different source angles. The position at the smallest angle

was studied in the greatest detail. At this angle of 20 degrees

750 judgements were recorded on thirteen occasions.

It will be assumed that the scatter in estimation of

source direction follows a gaussian distribution about the

true source position. For the subject to indicate the wrong

position in the 20° triangle test, his estimate would need

to be greater than 100 in error in the direction of the

wrong position . From fig 5.7 it can be seen that this

happened on 32 of occasions.

8S

If the shaded area of this distribution corresponds to

3240 then the area marked as X will be 1840 of the total. It

can be shown from standard tables that 18;; would correspond

to an area lying between 0 and .47d'. hence the standard

deviation is 100,447 = 21°.

Finally the consistancy of the subjects between the

two series of tests was examined. Fig 5.8 ranks the subjects

in the order of their score on the two choice experiments

at an angle of 20 degrees with the magnitude of their mean

vector from the multi—choice experiments. The correlation

is good.

5.7 Conclusions

Most of the divers who helped in these experiments

had the impression that they were not, or at the best only

marginally, able to localise a sound source underwater.

This is sLrane, since the overall conclusion from these

results is that directional hearing is possible underwater.

This conclusion is statistically very reliable' from both

independent series of experiments. In order to throw suspicion

on this result it would be necessary to invoke, either the

subject using his eyes or some form of extra—sensory perception.

its for the former, the heavy copper black—out flap attached

to the mask let in almost no light. It could not be

surreptitiously raised without the use of the hands. In all

blacked—out mask experiments underwater it is suggested that

86

the subject might obtain clues as to his whereabouts by the

action of sunlight shining through imperfections in the

black—out. However, the experiments described above were

designed so that subject awareness of his surroundings would

not be a serious interference. ilfter all, if directional

hearing is possible then the subject may be able to obtain

information about his surroundings by this method. If the

experiments were sensitive to this interference, then although

the overall conclusions may still be valid, they would not

provide a satisfactory quantitive indication. Having said

this, the writer, generally considered to be a proficient diver,

was not aware of gaining visual clues as to his surroundings

on the occasions he acted as subject, and no other subject,

reported visual clues.

87

Chapter 6

7.:112.:.ING (11)

G.1 Introduction

In the previous section the two main experiments

designed to test directional hearing have been described. Both

these experiments reach a clear and statistically reliable

conclusion. That is that a diver can perceive the direction

of a sound source. In this chapter the effect of changes in

either the source or subject are considered. This material is

often derived from a statistically poor sample and must be

regarded as producing pointers to hearing ability underwater

and not unequivocal conclusions.

It is proposed to consider these subsidiary experiments

in turn, dealing first with those arising from the forced

choice exPeriments and later those which are derived from the

free choice ones.

6.2 Different source anles around the subject's head

In all tests of this type considered previously the

subject sat at the apex of an equilateral triangle, facing

the opposite side. In the experiments described below, the two

source positions were not symetrically displaced about the

subject. The table below shows some subject—source positions

and the score obtained. The notation used is that of fig 5 .7

where the total number of judgements is placed before the

number of incorrect answers.

ORIENTATION

-4- 45o

o°

TEST SUBJECT sconE

A J.'I. 40/11

D.n. 40/6

88

B

B.R. 40/12

P.N. 40/19

C

R.L. 40/19

11.L. 40/4

In test A the change in time difference between the

arrival of a sound at the two ears from the two source

Positions will be about 33 pee. This is of the same order

to that experienced in the 200 triangle. 72,oth divers achieved

a score which did not differ significantly from that obtained

on the earlier test.

In the second test, B, there should not have been any

time difference to distinguish the two source positions.

Compared with their previous individual abilities (fig 5.7)

both subjects can be said to have done badly.

Test C would also not be expected to produce any time

difference clues to help the divers. The score of is close

to a null hypothesis, but as this was so on previous tests

score with hood 60/8 One subject — — — D.T. I

200 triangle - score without hood 30/8

89

it cannot be regarded as significant. The good score from

M.L. may have been due to this subject using head movements

to localise the source, although it seems unlikely that this

produce a score that was better than any of the triangle

scores by the saline diver. It is felt rather more likely

that some other mechanism may have helped here. One

possibility (to be considered in a later section) is that

there may be a difference in hearing sensitivity between

sounds arriving from the front and back of the head.

6.3 The effect of a rubber hood

The hood that was used for these experiments was a

conventional wet—suit foam neoprene helmet with small holes

about dianeter adjacent to the diver's ears. The tro

tests completed on the same dive and the helmet was

removed in the water so that the subject would not chan:;e His

depth and "clear his cars" between tests. It was necessary

to remove the facanask and breathing mouthpiece in order to

remove the hood. curing this latter stage there was a

tendancy for the rubber of the hood to become lodged over the

face and for safety reasons further subjects were not invited

to try this experiment. It is suggested that these tests be

repeated with a helmet specially designed to be easily

removed underwater.

90

Although there is the suggestion that a hood may

improve directional hearing underwater it is not possible

to draw firm conclusions from one subject. It can however

be said that this result does not conflict with the reports

of Ide (41) and Feinstein (42) that a hood imoroves directional

hearing.

6.4 The detection of an obstacle

A sheet of foam neoprene about 1 metre square was held

by a diver rouhly 2 metres in front of the subject. As an

alternative the obstacle was held to the side. The buzzer sound

source was operated from a point 7 metres in front of the

subject and the diver was asked to indicate the presence of

the obstacle.

:'.exults p.n. 51/4

J.W. 40/16

One subject appeared to be able to indicate the

Presence of the obstacle, the other diver probably did.

0.5 The effect of the subject hearin7 the source switched 'on

During "debriefing" some subjects reported that they

felt two distinct phases to localisation of the source. There

vas an immediate impression of direction the moment the source

was sritcherl on followed by a longer period when the direction

vas less apparent. On the suggestion of one of the expedition

° me:_lbers a series of tests were desiged using the 20

91

triangle. The subject was asked to breath slowly and.

regularly. The sound source was switched on, either while the

subject was breathing (i.e. the sound or the source was masked)

or when he was holding his breath. The assumption was made that

while the divers were breathing they would be unable to hear

the source switched on. In the breath—holding tests the subject

was asked to come to a quick snap decision whereas he was asked

to consider his judgment carefully in the cases where he did

not hear the source switched on. Atypical run involving GO -

juLgements would contain 30 of each type interleaved in groups

of 15.

Si::: subjects were tested and the score is Clown below:

NOT B=7:LING DP7ATHING

P.J. 30/10 30/12

J.W. GO/10 59/26

P.N. 30/10 30/15

M.L. 30/12 30/7

3.1,. 45/5 45/9

30/S 30/3

TOTAL 225/55 224/77

If one assumes the two distributions are binomial then

the standard deviations will be G.5 and 7.1 respectively. On

this basis the two total scores differ by over three standard

deviations and can be regarded as significantly clifferent.

It appears therefore that the diver should abide by his

initial impression of the source direction.

G.G The effect of a reference on the free choice tests

Two subjects were tested with the free choice

experiment in a similar manner to that described in 5.3 with

the excention that they were seated on a firm steel box. The

divers -:ere at liberty to move around on this seat in their

efforts to localise the source. The results of this test

are shown in rows a a b of fig 6.1 below. Although the

absolute accuracy is not noticeable better than the previous

free choice exneriments, the value of the mean vector is

larger. This represents in both cases a doubling of the

relative accuracy. Their "aim" has now a standard deviation

of about 20° which is similar to that derived from the two

choice experiaents discussed in 5.7.

Fig 6.1

:Airthor Free Choice Tests

Subject No. of nhotos

:.:can Vector 1:agnitude Direction significance

Standard Deviation

a E.L. 12 .936 —12° < .001 10°

b B.J. 14 .053 ...z1.50 -4: .001 22°

c 10 .605 + 2° <:.01 56o

d ' L .: , • • 13 .62 —64o <: .05 65°

e P.N. 17 .784 +24° < .001 48°

f i.L. - 13 .61 4- 1o <..05 65o

}:ethos_ used a L b subject seated on steel box buzzer sound source

c ,2z d subject moored inverted buzzer sound source

c a f subject seated on steel box. Low frequency pneumatic sound source used.

92

93

6.7 albject inverted

On the occasion of three free choice tests, instead

of mooring the blindfolded subjects with a rope around their

ankle, the rope was tied around their shoulders. Under these

conditions the diver's body would hang inverted in the water

(photograph fig 6.2). The first subject to try this kept his

head in a relatively conventional orientation with respect

to his body. After three judgements, 'which included one

vertically upwards, the diver signalled that all was not well

and was helped ashore complaining of vertigo. Two further

subjects were far more successful, they held their heads so as

to face the sea bed and moved their bodies to localise the

source in a line with the top of the head.

In these cases the scores were conmarable to these

obtained in the conventional manner. (rows c d of fig 0.1).

It is worth recording that stringent safety measures were taken

for these tests. Only e:merienced divers were invited to try.

air infla ted lifejacket capable of bring diver and. ballast

to the surface was worn and there was a boat moored close by.

0.8 Libject sinallinE7 the confidence in his judgement

During the course of some of the three choice tests

(recorded in fig 5.5) the subjects were asked to use one of

three hand signals when pointing to their estimate of the

source position. These were indicating with a snorkel, a

clenched fist, or an open hand. The signals were to indicate

94

Fig 6.2

PhotoTraph taken from 'free—choice' film record showing

the subject moored inverted

The sound source was situated beyond the top of this page.

(In order to show the perspective, the complete 35mm frame has

been reproduced)

the subjects confidence in his judgment on a scale of three.

Unfortunately it was only on two occasions that the signals

reproduced unambiguously on the photographs.

shown below:

The score is

Subject Choice No. of Mean Vector Direction Judgments

J.W. Total 24 0.84 — lo

A 3 0.85 — I°

B 12 0.34 — 14°

C 9 0.29 4- 16°

M.L. Total 18 0.72 —112°

(one photo not clear)

A

B

1

12 0.75

_1030

— 96°

C 4 0.31 —1310

Choice A was the most confident, C the least.

No trends are apparent and it appears that the diver is a

poor judge of his own abilities.

0.0 The effect o5-: different sound sources

All previous references to sound sources have referred

to one of two devices described in 5.2. These were sometimes

Imowii as "buzzer" sound sources and their response had a wide

bandwidth covering the centre of the audio band. However, it

was considered useful to investigate the effects of a loT:er or

higher frequency source. For simplicity and reliability

non—electronic devices were examined.

95

96

The low frequency source

The efficiency of a sound source is a function of the

relative dimensions of the working carts ccmpared with the

wavelength of sound (This has been expressed in a more

Quantative way in section 2.3 ). Now as the low frecuency end

of the audio spectrum involves wavelengths of between 3 and 15

metres underwater, the designer of a hand—held source is clearly

in difficulties. The solution adopted is to expend a large

amount of power to overcome the inevitable inefficiency of a

mmall source. The acoustic signal is produced by the release

of bursts of compressed air into the water. To release these

bursts a mechanical oscillator was designed which is analogous

to an electronic "relaxation oscillator". It is small, the

diagram (fig 6.3) is close to actual size, and can produce a

chain of pulses with a repetition frecuency between 1 and 20 Hz.

The setting of the tap on the gas cylinder gave a small

measure of frequency control and larger changes could be

produced by inserting a longer or shorter hose between the tap

and the oscillator. The bleed screw was set so as to' provide

stable operation. In use this source appeared to have two main

disadvantages. It consumed a large amount of compressed air

and it had to be held well clear of the operator's head for

comfort.

Two or the tests listed in 6.1 involved the use of the

pneumatic source. Both of these, e a fy were run with the

Tap

in a similar Response of pneumatic source (taken from recordings way to fig 5.3)

I lE 50 21: 100 200 500

g I 5E

Fig 6.3 Low frequency pneumatic sound source.

Brass tube Aren,,,holes

\\\\\\\\"\\\\ \\\\\\\\A

/ //////

\\\\ \\\\\\\\\\\\\I s>m\\\\\ ii\\\\

Stainless steel Bronze BleedBleed spring ball valve

Retaining plate (with holes t.1 allow a free flow of water)

Steel insert

T 10db

Hz.

97

Flexible air hose

1

air cylinder

98

subject sitting on a steel box in the same way as a b.

However, neither showed the improvement over the original free

choice experiments that was demonstrated by a & b. In the

triangle tests there was one subject who used both the buzzer

and the pneumatic source.

GOo Triangle subject P.N. Buzzer 41/5 Pneumatic 40/12

There is therefore a suggestion that this low

frecuency source is more difficult to localise than the mid—

fremiency buzzer.

The high frecuency source.

A diver propulsion unit was modified by removing the

propeller and substituting a length of slotted aluminium.

It was hoped that the rapidly rotating piece of metal would

proalce cavitation. Alen this was tried the only noise that

vas heard was the noise of the electric motor which sounded

somewhat similar to the "buzzer" scAlrce. No frequency response

is available and the device saw little use. Two of the free

choice experiments in fig 5.5 were made with this source. One

is a good score and the other a bad one.

Tapping a gas cylinder to make a sound.

The tapping of a metal object on the side of a diverts

aqualung cylinder is often suggested as an emergency signal .

To examine the effectiveness of this, it was used as a source in

0 some of the 20 triangle tests.

99

core with Score with Subject rt-linder Buzzer

60/16 60/16

60/20

P.N. 45/13 60/25

M.L. 45/12 60/19

P'2,11C TAGE. 'F23.0 NG 34cifo 33'i;

The cylinder used was small (1 litre water capacity)

and was held in the hand and tapped with a heavy diving

knife. This source produced no noticeable difference in score

from the buzzer and can be recommended as a means of signalling.

6.10 Conclusions

The e:meriments described in this section rill act as

pointers to those areas of this field where future exploration

should. be most fruitful. They also provide the background for

advice that can be given to the we-thing diver.

In order to localise a sound effectively, the diver

should try to provide himself with some reference such as the

sea-bed. The sound should be a mid-freouency one such as the

running of machinery or the banging of metal objects. A series

of thunderflashes of other lor freouency sources may not be the

most useful. The diver should turn to face the direction from

which he feels the sound is coming and he should make up his

mind without undue hesitancy. He should not be put off by a

feeling that he is wrong.

Chapter 7

.AUD I CI. ET ra. ID OD 1V2,..T I 01-;:,.; 101-?-01.. All. UlaM1.1.72,1' at. LA1.1 0 T 0 BY

7.1 Initial e:meriments

Liubjects who were involved in the directional

hearing tests in Malta, 1003, reported the impression that

a sound. source in front of the body appeared louder than

one behind. At the time the author viewed these comments

with considerable sceptisism. However, it was felt desirable

to confirm that this differential was indeed an illusion and

to this end a simple audiometer was constructed. This

instrument was designed to produce a constant audio tone,

the level of which could be accurately adjusted.

At the time the only available components were either

those taken as spares for other work or those obtainable fran

the local 1,:altesse radio shops. In the final device, shorn

in fig 7.1, the sound radiator was a moving coil transducer.

A multi—way switch was let into the rear face of the transducer

and sealed with a combination of '0' rings and candle. wax. A

chain of carbon resistors was wired across the switch, contacts,

the values having been chosen to provide an attenuator with

eleven steps of 4 db. The source of electrical power was the

mechanism from an ultrasonic pinger that was to hand, mo;:ified

to produce a pulse train at about 000 Hz. In the photograph

this is the tubular device strapped above the transducer.

In use the audiometer was positioned firmly on the

100

• r F., .ftje 4"; 4 • . • 4.• ..••• _ ' • ' •

101 Fig 7.1

'Crude' audiometer in

use. (Audiometer can

be seen in foreground

Fig 7.3

The 'imprlvedl audio—

meter complete with

transducer.

102

sea bed 5 metres from the subject diver. The level of the

tone was reduced in 4 db steps until the subject indicated

that he could no longer hear it. The level was then increased

until it was just audible to the subject. This is sometimes

described as making one "excursion" around the threshold.

In conventional surface audiometry, a large number of excursions

can be made in a relatively short time, underwater there are

difficulties imposed by both the inhalation and exhalation

noise. Under (Inlet conditions exhaled bubbles can clearly

be heard until they reach the surface.

The techniue adopted was for the subject and

experimenter to synchronise their breathing, take two or three

good breaths and hold their breaths for up to 30 seconds. The

subject could easily require several such periods before

reaching one decision. 1:eedless—to—say the arrival of any

power boat in bayi or even the operation of a compressor

on shore, 100 metres away, was sufficient to curtail the

exerimons. "Iurthenuo-fe it was foune, that many demand valves

tended to release an occasional bubble and it was necessary

to test a number or these in order to select those which were

satisfactory. The subject did not wear a rubber hood but

was provided with a steel box for a seat and an extra

weightbelt to place across his lap.

There is one further way in which these experiments

103

differ from conventional audiometry. In air it is usual

to conduct the experiments in an anechoic chamber. If it

is desired to change the orientation of the source, then the

subject remains in a fixed position andl the soun generator

is moved around the subject's head. Lt first sight, it might

be supnosed that this method was applicable un.erwater.

However, the sea—bed is far from anechoic and if the transducer

were to be moved, different conditions would ensue and the

sound field at the subject would change. The alternative,

which was adopted by the writer, was to turn the subject

round the keep the source stationary.

the duration of one dive it was found possible

to malt:e two excursions wih the subject facing, the source,

two with his back to the source, followed by two more to the

fc.ce an C. two more to the back. The reason that the directions

were interleaved in this way was to reduce possile error due

to slow changes such as learning or cold. It is important

that the subject remains sea-ced for the whole of the experiment

as any chJ.n7;e in depth might proC.uce a difference in any

residual pressure across the ear Crwn.

Five subjects were tested in this manner. On return

to the US., the Crude attenuator was calibrated and the mean

level for each direction of each subject evaluated. (fig 7.2)

104

Fig 7.2 Results of original audiometer tests (August 1068).

Sub— ject

Front back

Level at which sound could not be heard

Level at which sound could just be heard

Level at which sound could not be heard

Level at which sound could just be heard

Mean front/ back

difference

front 36db 26db 28db 26db

JW back 26 22 26 19

front 32 28 32 28 5.1 db.

back 32 22 26 22

front 32 28 36 28

back 32 26 32 26

front 39 D IJ 28 36 28 5. 4 d b .

back 26 22 26 22.

front 32 28 36 22

back 28 26 28 19 PN front 32 26 36 28

4.8 db.

back 28 22 — —

front 28 26 28 26

back 22 19 22 19 ML

1 front 26 22 26 22 4.5 db.

back 22 19 26 19

front 36 32 36 32

back 32 28 32 26 JSW

front 46 36 46 39 5.1 db.

back 39 32 39 32 •

NOTE. Underwater the position of the attenuator was recorded as a number (1 to 12). On return to the U.1. these numbers were substit—uted for the actual attenuation measured inalaboratory at Imperial College. It is these values th[A, are recorded here. :lence a large figure represents a quiet sound and a small one, a relatively loud sound.

-105

7.2 The construction of an improved audiometer

Even allowing for the restrictions of the crude

instrument described above, there a7)peared to be some

difference in threshold between the front and back of the body.

It seemed essential to confirm these findings in a more

satisfactory manner. The main disadvantages of the simple

au(Aometer were the coarseness of the attenuator, the

ill—defined nature of the tone, and the lack of any method

for determining the absolute sound pressure level at the

subject's head.

In the second design, constructed the following year,

a pure tone is generated by a low distortion Wein Bridge

oscillator (similar to a design published by Eullard Ltd,40).

The output from the oscillator is coupled to an 'L' section

attenuator with a range of 60 db in 2 db steps. A buffer

amplifier (a compound emitter follower) between the attenuator

and the transducer insures that the attenuator is correctly

matched under all conditions. These three items were

assembled in a heavy gauge Perspex box with sealed controls

for the oscillator frecuency and attenuator setting. The

transducer, again a moving coil type, was connected to the

amplifier by about four feet of cable in order that it could

be placed on the sea—bed clear of the operator's body. The

whole instrument was powered by dry batteries and switched on

by moving an external magnet adjacent to a reed switch. This

106

instrument (photo fig 7.3) was rather bulky and required

60 lbs of lead weights to hold it on the bottom.

7.3 The second series of audiometer tests

Plans were made to take the improved audiometer to

L:alta with a group of students in the summer of 1960. The

project was organised in a similar way to those of 1036 and

1963. However, unlike the previous two, the main object of

this expedition was not the acoustic experiments in themselves;

rather it was the testing of a submerged laboratory as a base

for acoustic experiments. For a description of this project

and photograph the reader is referred to Appendix 3. Despite

some very severe weather which destroyed nearly half the

ecuipmcnt, a tro—man underwater laboratory was operational,

7 metres below the surface, towards the end of August 1969.

In the experiments that are to be described, the writer

used this laboratory as a control centre and calibration

facility. The era,; - in which this vas arranged is best described

with reference to fig 7.4. The calibration hydrophone vas

a Clevite CH13 which vas coupled to a Druel Kjaer Precision

Ljound Level meter and octave filter bank inside the laboratory

The filter bank was needed to help extract the signal from

the ambient noise when making a measurement of the sound

pressure level produced by the audiometer at the subject's

head. The person sitting in the laboratory was able to observe

both the subject and the audiometer through the mirror smooth

Level meter

Octave filter

Underwater laboratory

Transducer

10 in.

llydrophone

5 metres

7 m.

107

Fig 7.4 Audiometer experiments with the underwater laboratory

Audiometer

Inside the underwater

laboratory.

The B & K sound level

meter can be seen on

the stool in the centre

of the phAograph. The

'entrance' is in the

foreground.

108

water surface in the entrance. Had this structure been

available earlier, the electronic part of the audiometer

would have been installed in the laboratory, leaving only

the subject and transducer in the water. In any future

trials with this ecluipment, this apl)roach will be adopted.

The method of operation was similar to that used

previously. The attenuator was moved in 2 db steps in the

period when the subject was breathing. This precaution was

taken to preclude any possibility of the subject hearing noise

generated by the switch. 1;orm.Aly only the maxima and minima

of the attenuator settings were recorded, however, fig 7.5

shows the conduct of one particular trial in full. The

audiometer was left in the same position on the sea—bed

between runs as there appeared to be no reason for bringing

it ashore. Unfortunc.tely the weather was still to prove

unfrienly and after five days submergeO, the audiometer

was washed oIL'f the rock on which it was stationed in a storm.

This action somehow cut the sheath of the transLucer cable and

rat,er entered the instrument down the insiac of this wire.

The attenuator was found to be open—circuit when the instrument

was recovered.

7.4 itesults

Five subjects were tested with the uncalibrated source

and three with the pure tone audiometer at 1 I'dz. Of these

eight subjects, seven showed a marked improvanent in hearing

Att

enua

tor

Se

ttin

g

A.

• Time

Subject's Response 4- —Yes

0 — No

B.

Fig 7.5 _::ample of one audiometer test with the subjectl(PS), 'A' facing the source, 'B' back to the source. An attenuator setting of 30db produced a sound pressure of 34db (ref .0002 dynes/cm2) at the subject's head o

COdb'

bound level in the sea (from Clevite CH13 hydrophone)

Sound level in the laboratory (from B cti K condenser mic..)

f 63 250 1K 4K 16K

GOdb Sound pressure rel. .0002 dyne/cm2

40db

20db

110 Fig 7.6 Results of audiometer experiments. (including fig 7.2)

Subject Difference front/back a tone

Calibration Conditions

RL (8) , 5.4db Original wide-band audiometer

no calibration

PN (q) 4.8db EL (0 4.5db JY/ (8) 5.1db

removed aoualung from back JS11 3c) 5.1db

PS (12) 5.2db EFT 54db at liaiz

Held aqualung in hands

NS (6) NS (7)

4.7db 2.6db

EI‘‘`f for both runs 56db at 11. 1z

Blacked-out facemask Both tests I

'ithout any facemask 1 same dive.

MB (23) -0.8db myT 66db at lRiz

Tests performed rapidly. Subject ha4 forced choice every breathing pause

% = Mean front threshold (reference .0002 dynes/cm2 )

Figure in brackets after the subject indicates the total number of threshold 'excursions' involved. (i.e. the total number of pairs of

readings.)

Fig 7.3 Analysis of ambient noise in the sea and in the underwater laboratory (life support system switched off)

Frequency Hz.

n

111

sensitivity when the sound source was in front of the body

(see fig 7.0). In the case of the one subject who showed no

significant difference, the test was conducted faster than

previous ones and the subject was asked to reach a decision

in every breathing pause. It may be sinificant that the

absolute value for the threshold is also somewhat higher

in this case. On the one other occasion that the front—to—

back difference was small, the subject had removed his

facemask and performed part of the test with no protection

over his ayes or nose. With the exception of this test the

subjects could generally see the audiometer when it was in

front of than, but they were instructed to fix their gaze

on some object such as a rock and not to study the instrument.

In any case they would not have been able to gain additional

information from the position of the controls as these were

not visible to the subject.

In the case of the three subjects who were tested with

calibrated equipment, the threshold with the source in front

of the body has a mean of 50 db ref .0002 dynes cm-2

., In

fig 7.7 published values for this threshold are compared.

Fiq 7.7 .Q.12m 7lr1soy, of :lublished uni7 erwater threshold

fi71;ures at 1 KIz

Author

Reference No. Threshold

Ide (1044) 41 75 db

Hamilton (1957) 15 52 db

Imthor neference No. Threshold

'jainwright (1050 IG GO db

nontague a Arichland (1067) 17 60 — 75 db

Brandt C. Horner' (1067) 18 70 db

Sivian (1047) 44 45 — 55 db

:.i:11 tit (1068Y 45 51 db

In addition to these tests, the sound pressure field in

the sea and in the umierwater laboratory were analysed using

the octave filter bank. These measurements are summarised

in Fig 7.8. The noise showed a distinct hump around 8 1;11z

and is believed to be almost entirely biological. It is

normally credited to the "snapping LAibjectively it

appeared that this noise increased just after dush when a

consierable number of small creatures between 1 and 3 mm in

length, chose to climb up the inside wall of the laboratory. No

noise could be detected as arising from the laboratory itself

(all life support emainment switched off) and the airborn sound

pressure level inside annears to be entirely the "snaying

shrimps" on the outside.

It may seem surprising that the absolute values for the

thresholds lie below the ambient noise in octave analysis. This

is ”erfectly reasonable as the human car is an e:xellent tool for

detecting a distinctive sound below a broad—band mashing noise.

The writer will return to this topic in a more ouantative way

later.

112

113

';:lien it became clear that there was real difference

between front and bad: sensitivity, some consideration ras

given as to the cause. The first suspect was the compressed

air cylinder that the diver wears on his back. However,

removing this and asking the subject to hold it above his

head did not seem to effect the situation. One other

possibility is that it is due to the presence of the air

cavity of the face mask. Unfortunately it is not easy to

spend a half hour without any protection for theqyes. The

high salinity of the Mediterranean tends to cause an

uncomfortable stinging sensation.

7.5 Other observations from the underwater laboratory

In section 2.4 the possibility of using an electric

conduction field for communication was discussed. During the

design of the underwater laboratory it was thought desirable

for there to be some means of wireless caaninication to the

shore. Because of its inherant simplicity a conduction system

was used. A block diagram of the this etuipment is shown in

fig 7.9.

The transmitter consisted of a conventional preamplifier

followed by a push—pull output stage capable of driving 1 anp

into a 4 ohm load. The microphone was always "live" and the

unit clerived rower from the main laboratory batteries. In

the event of a power failure a relay automatically changed

over to an internal dry battery. The "earth" of the amplifier

Receiver

Loudspeaker Amplifier

Electrodes

Fig 7.9 Electric conduction communicator used in underrater laboratory.

Transmitter

To laboratory 12 v battery

114

Relay coil

Pre—amp Power amp

Relay

>y 12 v

Lamp

,arth to laboratory ///7///, frame

Electrode

115

was connected directly to the steel base rinL; of the.

laboratory and the output was taken to a disc of perforated

aluminium, 30 cms diameter, rhich was suspended 2 metres to

one side of the structure. On shore some 70 metres away,

two electrodes about 7 metres apart were connected to an

amplifier and loudspeaker.

There is little that can be said as regards the

operation of this device apart from that it worked. No

electrical measurements were taken as priority was given

elsewhere. However, there was one surprising observation that

must he recorded. ?wen when the transmitter was switched off,

there was a sound from the receiving loudspeaker resembling

snapping shrimps. '21ether these creatures are emitting an

electric field or whether the explanation lies elsewhere,

would appear to recmire further investigation.

116

Chapter S

DI:SCUL:SION

S.1 _Articulation underwater

Probably the most serious limitations to present

underwater communication systems is the difficulty in forming

words. It might be suposed that the solution has already

been outlined in section 4.7; that is to use a mash which covers-

the mouth and has a relatively large volume. Unfortunately a

large volume represents a possible "dead space" of gas which

will not be completely flushed on each breath. This can lead to

a build-up of carbon dioxide with serious physiological

repercussions. 2_ solution which has been adopted by one

manufacturer is to insert a lightweight oral-nasal mask within

the main mask. This inner mask can be acoustically transparent

while isolating the main volume of gas from the breathing

systan.

The desiTn of non-return valves in breathing eciuipment

presents a similar problem. -,?or reliability and saf 4-, there

are often two, relatively stiff, non-return valves in. the

exhalation path. These are suspected of causing rapid

fluctuations in back pressure as they operate. It was found

easier to speak into a demand valve incorporating a single,

sesitive, exhalation non-return valve.

If electronic methoC:s can be used to aid articulation,

then they may be of immediate application to existing breathing

117

systems. One such possibility is to use acoustic feedback

inside the mask. If a loudspeaker is inserted into the

facemask and driven such that it will tend to cancel the

original sounds in the mask (i.e. ne2:ative feedback), the

acoustic impedance inside the mask will fall. This idea

is not new and has been proposed for noise cancellation near

machinery in industrial premises (:Aephens Dates 43). An

extension of this idea would be to mount the transducer of

a direct audio c=unicator on the facemask in such a way that

the sound in the water is reinforced while that in the mask

is cancelled. If such a system could be produced it would be,

in effect, an impedance converter or active acoustic transformer.

A further possibility, which seems to have strangely

been ir;nored, is the use of "side—tone". In the early days

of teleDhony it was found that a tele)hone was improved if

a fraction of the voice siznal from the microphone was fed

bac:: to the earpiece. This fraction is known as the side—tone.

If a modern telephone were produced without side—tone it

would sound most unnatural and would be difficult to use.

1. person who is speaking listens to his own voice and

his aticulation is influenced by what he hears. If the voice

is altered then the speaker will attempt some measure of

correction. (This was observed after three clivs in the helium

atmosphere of Jealab 2 (47). Althou::;h this is normally to be

118

desired, it may introduce considerable complications if

a helium speech unscramblor is being used. In this latter

case the unscrw:ibler is asked to deal with the end product

o -' a feedback problem involving the human brain. It is

suggested that a more logical approach is to feed a side—

tone back to the diver after the signal has been unscrambled.

In this case the human feedback should improve the final

signal. • If this technicue proved beneficial it may be

worth considering the possibility of electronic correction

for fomant shift due to the facemask impedance at the same

tine that the helium distortion is corrected.

3.2 Propaation

There is a vast amount of information on the distortion

of :articular signals as a result of transmission through

sea vater. lk)wever, there appears to have been almost no

serious attar is to correlate this with a psychoacoustic

approach. For eNample, the writer has_ already sug!sested

that multiple path fistortionsin an annlitue modulation

ultrasonic system are more easy for the human ear to interpret

thal‘ those in an F.H. system. To confirm this one would need

to construct a model capable of reproducing these types of

distortion and 1,est for intelligibility using samples of

speech recorded in the facemask of a submerged diver. If

a versatile model was available it might be possible to

establish the optimum freQuency, type of modulation, and

119

directivity of the transducers, for any particular situation.

In recent years a number of devices using electric

or magnetic fields have appeared on the market. In some

cases, the claimed range is far greater than tho,t predicted

by the "skin—depth" calculation in section 2.4 and there is

a suggestion that it may be possible to transmit electromagnetic

fields beyond this range. natever the answeri there is an

almost complete lack or any published material on this subject.

Until this omission is rectified, it will be difficult to

estimate the potential use for electromagnetic systems.

8.3 Human hes7.rin underwater — t' le threshold

In 1047 Sivian (44) predicted that the threshold of

human hearing underwater would be in the range 43 — 55 db,

plus any attenuation due to a pressure imbalance across the

ear druri. The value obtained from the tests conducted in

lailta and some of the published figures listed in fig 7.7,

correspond closely to this prediction: Other published values

do not. Very little emphasis has been placed on the second

part of Liiviants prediction and all too frecluently one sees

a stata-aent such as "The divers equalised the pressure

across their ears" and the matter is left there.

The average diver will attempt to "clear his ears" after

a decent of about two metres and would find it difficult to

detect a change in depth of less than metre. Consider the

e:,uivalent chani;c in barometric heirht on the surface. As

120

the density of air is some SOO times less than that or water,

the change of r metre in water becomes 400 metres in air. Most

peoDle would detect a change in their hearing when descending

a hill of only a fraction of this hei7ht. The problem may be

even worse with closed circuit breathing sets where the diver

breaths gas from a bag often worn on the chest. This may well

be an appreciable distance below the cars. To sum Up, it

should not be assumed that pressure differentials across the

ears are nep;ligable. However, in reality, it may be that the

best one can do is to use open circuit scuba where the demand

valve is at a similar height to the ears, and to stress to

the subjects that they should tae plenty of time coughing,

suc!:ing, blowing their nose, etc., before starting the tests.

()lice the tests have canlence the subjects should not change

I:A:mover a threshold is measuredun:_er conditions that

are other than absolutely quiet, one must estimate the errors

introclucca by the background noise. Hawkins C: Uevens (-13)

have investiated the masking effect of a broad—band noise

on a pure tone. Fig 8.1 has been taken from this reference.

It shows the change in threshold (in air) for a pure tone

of llalz against the level of a masking noise in an octave band*

(*Havkins a jtevens refer to the noise in a "critical band" which is 03 Hz wic.e at 1 use in this context it is

more convenient to consier an octave band (a))rox. 730 Hz). For a broad—band noise this will correspond to a level 10 di) higher thn for the critical—bnd.)

Change in threshold of pure tone.

30 "

20 -

10 .

Fig 8.1 Masking of pure tones (After Hawkins Stevens, 48)

40 -.

121

/ (e* ,0

1. Y.J /

or / Sc,/ )c.)

04

No / 4C,

ii / • 944' / 'b-.. oz:'' / 0

.0. N / N

4 o / o N N° o

e / c N> / o-,"7•

/ --,,-

/

1)

0 10 20 30 40 Level of masking noise in octave band (11Cllz)

Fin 8.2 Sterionhonic hearing aid

122

Two lines have been superimposed on this diagram, one

corresponding to the condition that the mashing noise is 10 ab

above the observed threshold and the other that it is equal to it.

From the positions where these lines cut the threshold curve,

it should be clear that if one observes a threshold ecfual to

the noise then this is very nearly the true threshold. On the

otherhand if one finds the threshold more than 10 db below

the noise, then one is observing a mashing phenomina and not

a true threshold. In this latter case the result can only

be considered as an upper limit to the true threshold. Due

to the uncertainty of the use of this masking relation, it

would be unreasonable to expect this to provide an exact

method for correcting for anbient noise. This appears to be

an area worthy of further investigation.

Unfortunately, in the case of two of the subjects tested

the threshold appeared to be more than 10 db below the noise

level (recorded in fig 7.8) and hence these fi2;ures are upner

limits to threshold of underwater hearing. however, if it is

assiuned that the noise is omnidirectional, it will not

affect the validity of the Front/back differential measurements

for any o2 the subjects.

8.4 Human un erwater — Cirectiort7.1 hec,ring.

As has been seen before, most of the divers who helped

in these experiments had the impression that they were not,

123

or at the best only marginally, able to localise a sound

source underwater. These results show that this impression

cannot have been crained throu7h lack of ability to localise

sound. Is this possibility due to a lack of confidence in

interpreting binaural information? This is, after all,

not unreasonable as the binaural information reaching the

subject rill be distorted, mainly because the velocity of

sound in water is some five times greater than in air.

Consequently the time difference between the arrival of sounds

at the two ears will be shorter in rater than in air. Possibly

of greater significance, is the fact that the shift in time

difference as the head is moved will not correspond to that

normally ex-)erienced in air, (Thurlow h :tunge 49) and a

cine—film of some of the free choice tests shows that the

subjects made exagerated head movements before comin:: to their

decision underwater.

Is this lac!: of correspondence between head movements

and the movement of the binaural image that is destrdying

confidence? The probla,1 could be compared with that 'of

following a moving target through a pair of binoculars. The

apparent image dis7)lacattent will not correspond to head

movements and some practice would be desirable before using

binoculars for this work. This suggestion would also explain

why the diver's first impression on hearing the source switched

on tended to be more accurate (at least as regards to the

124

appronriate side of the head) than his considered opinion.

This follows if one assumes that the diver obtains his

first impression before he has had time to move his head.

In air it is possible to localise a wide band sound

source to 2 degrees (Fordlund 40) and on a velocity of sound

hypothesis alone this would correspond to 10 degrees underwater.

However, as underwater one would expect little help from the

pinna and inter—aural amplitude differences will differ from

those normally experienced, it would be unreasonable to

expect this overall accuracy. The figure of 20 degrees from

both the two choice experiments and the free choice ones where

the subjects were seated, appears to be the best available

measurement of underwater sound localisation at the time of

writing.

It is interesting to consider the time difference

between the arrival of a sound at the two ears that corresponds

to an angular displacement of 20°. This will be about 30

microseconds for an interaural distance of 15 ems. :jome

measurements of time difference thresholds (in air) by

Klumn a Eady (51) are reproduced below. These figure's suggest

that under ideal conditions it should be possible to localise

a source to within 70 underwater. They also provide an

explanation as to why the subjects found it more di2ficult

to localise a low frequency source.

Threshold o1 time difference (Nlump Eady 51)

Broad band noise 10

,repeated clicks 11

150 — 1700 Hz 9 Pink nois

e425 — 600 Hz 114

90 Hz 75

250 Hz 27 Pure tone 500 Hz 17

1000 Hz 11

Time

in

microseconds

The accuracy obtained by the subjects who were freely

suspended has no counterpart on land. It represents the case

where the diver has no reference apart from the vertical, on

which to base his directional judgenent. During the period in

which he is making his judgement he must presumable use the

inertia of his body as a reference frame. ihen these problems

are born in mind it seems remarkable that the human system is

capable of localising the sound to within 50 degrees.

This latter ability would be important in using an

acoustic beacon to diver navigation under very low visibility

conditions. If on the eight occasions of the free choice

experiments recorded in fig 5.5 the subjects had been asked

to seek the source under zero visibility conditions, six would

have been expected to find it, one would not (apparently due to

ear clearing trouble) and the outcome of the eighth cannot be

reliably predicted.

125

126

Under normal diving conditions where the subject has

some reference clues one would expect localisation to be in

the range 20 to 50 degrees.

At this stage it would be premature to suggest a model

for the mechanism of underwater hearing. Previous workers have

considered that hearing is either tympanic, that is sound

transmitted along the auditory canal in much the same way as

in air, or bone conduction. In the first case the value for

the threshold is explained as arising from an ilm)edance mismatch

at the druii, and in the second there is a similarity between

the thresholds of bond conduction hearing in air and underwater

hearing. The writer is of the opinion that it may not be

meaningful to consider underwater hearing as being either

bone conduction or tympanic. These ideas apply in air where

the acoustic impedance difference between flesh—and—bone and

air is so large (103 — 10") that one can consider a sounu

wave as travelling in one or the othef. flesh has a similar

acoustic impedance to water and it may be more reasonable to

consider that underwater the two hearing organs exist in an

infinite fluid. This ignors the water—skin boundary completely!

The middle ears, sinus cavities and the facemask along with

other air spaces are impedance discontinuities in the

neighbourhood of the receptors and must be taken into account.

It has been shown that bubbles of air in the vicinity of a

transducer can have a dramatic effect on its performance and

127

produce directionality in an otherwise omnidirectional

device. (Hunter 52). It is hoped that further tests with

and without a facemask may shed additional light on this

hypothesis. If it can be substantiated it may provide an

explanation for the difference in threshold between the front

and back of the head.

8.5 Ap7)lications to co:ummication eouirient desir4n

The most important recommend•Aion stemming from this

work is the adoption of an attitude of mind. If an item of

equipment is designed to help two human beings communicate

underwater, then the most important test is whether it can do

just this. Yet one hears of devices that are ca?able of

providing a mmnunication link between two boats , being

built into waterproof cases and soh. asaivei. telephones".

Yar more consieration should be given to the use of

stereophonic receivers. These have been recommended for use

on the surface and it has even been suggested (W) that

troops be provided with stereophonic radio receivers for use

in battle conditions. further, Pollack C. Picket (54) have

shown that a stereophonic telephone can enhance the apparent

signal to noise ratio by between 6 and 12 db. This type of

equipment has seen almost no use underwater because it has

generally been assumed that binaural hearing is very much

degraded.

A stereophonic receiver produced by the writer for use

128

with direct audio systaas is shown in fig 8.2. The two

hydrophones are spaced apart to produce an interaural time

delay similar to that experienced in air. This particular

design was unsuccessful due to acoustic feedback through

the water between the earphones and hydrophones. However

this configuration could be used as a receiver for an

ultrasonic carrier.

A summary of recosmiendations for the design of

underwater communication enuipment follows:-

1. Design breathing set, mask and microphone to be capable of working as one unit.

2. Reduce the acoustic impedance seen by the lips.

3. ._educe exhalation pressure and in particular pressure fluctuations caused by operation of non-return valves.

4. Consider the application of side-tone.

0 • Helium recordings at atmospheric pressure are no guide to the intelligibility of helium speech un-er high pressure. speech recorded in a compression chamber is only a limited guide to the sounds produced inside the facanask of a working diver.

G. ajuima07A must be checked for the encumberance caused to the cold, exhausted)worl:ing diver.

7. Beware of directional transmitters and receivers where the diver is recluired to adopt a particular orientation for cmalunication.

8. Consider the advantages of a "duplex" system (see 2.2)

9. Consider the use of a stereophonic receiver (or the unaided cars).

10. The ear is tolerant to certain types of propagation distortion, experience based on sonar or telemetry may not be applicable.

S.G Sua-gestions for further research.

At the time of writing the author knows of no comparable

study of hulan communication underwater. For this reason it

is, in general, difficult to compare this work with that of other

people. In consequence further research can be along one of two

lines. The first is to treat this work as a series of pilot

experiments and to use the results for the design of more

exacting and critical experiments. This approach is probably

most suitable for the problems of forming words underwater.

The second is to continue with further research at this sa:ne level.

The writer would argue that the main results from the directional

hearing tests are sufficiently reliable and that further

experiments should investigate the directional threshold to a

similar statistical reliability.

some attempt should be made to eliminate pressure

differentials across the diver's ears. Asking the diver to "clear

his ears" is not good enough and this problem requires further

investigation.

129

ACKNOWLEDGE IENTS

The writer is grateful for the help and encouragement

given by his supervisor, Dr R.W.D. Stephens, and for the

provision of facilities in the Physics Department of Imperial

College. A grant from the Science Research Council followed

by generous help from the 3M's Company enabled the writer to

pursue this study.

The fieldwork in Malta was organised as part of three

expeditions which were sponsored by Imperial College Exploration

Board. The author is grateful to the Board and to the many

other organisations who backed these expeditions.

The safety and operational guidelines involved in the use

of the underwater laboratory were formulated after several

informal discussions with the Director and Staff of the Royal

naval Physiological Laboratory and other sections within the

Royal luyy. In lialta help for this project vas given by various

Government departments, the Biology Departs ent of the Royal

University, and Dry Docks.

The experiments involving human subjects would not have been

possible without the considerable degree of cooperation shown

by the personnel, most of whom were members of Imperial College

Underwater Club.

130

131

Appendix 1.

ME SIGNAL TO NOISE RATIO Frtkl! AN UNDERWATER M.ICROPITONE

It can be shown (30) that the thermal noise in an

incremental bandwidth developed across a network containing

resistive and reactive components is given by :-

2 V.(d0 = 4kTR f df

Where R, is the real component of the network impedance at

frecueney f.

Considering the amplifier (fig 3.5). The input circuit will

be 1 megohm in parallel with the cable and transducer capacitance.

The impedance at the input, Z, will be given by :-

1

j (.0 C IZ

R 7 = J cii c = n(1 — j ,,, pLi)

R + .1

3 to C 1 4- to2 R2 C2

The real part of this will be t—

L;ubstituting this in Nyquists

over the audio band (100Hz t 1012.1z)

formula

:— r W

1

I?.

1 u09

1./ C2

(A.1) and integrating

V2 4kTR ± 4 it 11C` t2 df

I00

10

41:TIL 1 tan 1 2 -n- Ref] 2 TT w 100

Jubstituting R = 106

Ohms C = 10 9 uF

1.38 x 10-23 T = 290° A

V -= 0.85 liVo 1 t

t.1

The noise generated by the mnplifier will add to this.

The amplifier shown in fig 3.5 was one of the more successful

developed specially for this type of source. It has a high

input impedance, a gain accurately defined by feedback (20db),

absence of any input capacitor and a good noise figure with

capacitative transducers. Further the use of a field effect

device in the first stage avoids the low frequency noise that

bipolar transist)rs produce when driven from a high impedance

source. The noise figure frun an amplifier of the type of fig 3.5

would be in the order of 2db. This corresponds to a total

noise at the amplifier input of about 1 microvolt.

The sensitivity of the microphones described in 3.2 is

approximately —110db ref 1 volt/ybar. flence the electrical

noise corresponds to a sound pressure field of —10db to the above

reference or +64db with reference to the usual origin of

.0002 dynes/cm .̀

132

Appendix 2.

SIGNE'ICANCE OF THE LEAN VECTOR

In order to assertain the reliability of the mean vector

it is necessary to evaluate the probability that this (or a

greater) value could occur as the chance addition of random

vectors. Only if this probability is less than 1 in 20) ie Ko)

can the results normally be considered significant.

Consider the addition of random unit vectors. The magnitude

of the sum of 'N' such vectors will he given by :-

2 cos 4 )2

+ ( sin 6 12 I

0 cos 9 cls 0A + S sin 0 sin 0n +

+ 02

2 cos 0

cos(0 — 9) + N

Now if 'N' is large and the vectors random) the summation over

the cosine term will vanish) leaving the familiar result that :—

Now if one makes the assumption that the distribution of

the component of the mean vector along any one axis is a normal

one of the type :—

P (x )dx a 1 A dx (Normalised)

Then if two such orthogonal distributions are combined by the

substitutions :—

133

134

r2

x2 + y2 P(r)dr = P(x) P(y) dx dy

l<

dx dy = r de dr dO = 2 Tr

The distribution of radius vector will be of the form :—

P(r)dr 2 r e dr

The mean of this distribution can be evaluated by

r̀ e 1d r

1.253 (from tables 57)

0

r e --Adr

Consider now a vector of length z such that the probability

of this or a longer one is 5.S. This length will be given by :—

r e --Yar 0.05

z z. z . 2.45 •

For the purpose of this calculation it is proposed to use the

ratio of this vector to the mean as this will be a dimensionless

constant which should be applicable to all similar distributions.

The value of this ratio for a probability of (and also 1`; and

0.1(A has been evaluated.

Probability

2.54

1.06 3.04

2.42 3.72

2.07

Let us assume a null hypothesis that the subjects in the

free choice tests arc pointing in random directions. If N

observations are made on a certain subject then one would expect

the mean vector to be in the order . The probability that

.05

.01 .001

it will exceed 1.96 r: is less than and in this case it is

more plausible to discard the null hypothesis and to assume that

the subject is pointing to a real direction.

To make this procedure easy to apply to the experimental

results, the following table has been drawn to show the

probability that any mean vector is the result of the addition

of random vectors. As the assumption has been made that N is

large, these calculations must only be taken as a guide if the

value of N is small.

Ko probability (1.96/50

1.;: probability (2.42 N)

0.l probability (2.97/1-/T)

10 .618 .765

15 .505 .625 .767

20 .437 .542 .665

25 .391 .484 .595

30 .357 .443 .543

35 .331 .410 .502

Fig 5.6 is a graphical representation of this table.

It is suggested in chapter 5 that the magnitude of a—subject's

mean vector is a good guide to his accuracy. If this is so it

should be possible to relate the mean vector to the standard

deviation of the target error. The cosine-1 of the mean vector

should correspond to the mean target deviation. 1NOw for a normal

distribution the mean deviation can be shown to be 0.8 x standard

deviation (50). The values given in fig 5.5 for the standard

135

deviation were obtainer' in this manner.

Finally, to serve as a check, it is useful to compare the

value for the standard deviation obtained by the above method

with that calculated from the root mean square of the angles

involved. To avoid the +175° —115o anomaly that was discussed

in section 5.5, this comparison was done for a subject who

indicated the direction to within 90° of the true direction on

all occasions. Only one subject managed this WSW fig 5.5).

In this case the standard deviation calculated from the mean

vector was 46° and that obtained from the roA.-mean—square of

the angles, 41°. Considering that N was small, 14, this can

be considered as satisfactory agreement.

136

Appendix 3

:2.71) 01'72;i1:2I1ION Or Ai'?

Introduction

In the Autumn of 106G, the writer beemae aware of

the limitations of conducting experiments on divers in tae

oven sea, from the shore or a boat. The most serious problems

were those associated with the sca surface which presents a

very real barrier between the experimenter and his exleriments.

Cue solution, which had been tried to a limited extent in

other countries, was to construct a submerged, air-filled,

laboratory. This could be operated with the air insiCe at

a pressure equal to that of the surrounding water and allow

direct access to the sea through an open entrance in the floor.

The first proposal submitted by the writer was known

as the "Kralcen" project. sifter two years development it

becal,le clear that the necessary financial backing was not

forthccmin-, an: the plans were reluctantly dropped. In the

Autumn of 1908 the author assembled a joint Em)erialt-LInfield

College group and co-ordinated plans for a far simpler and

less expensive laboratory*

(* lathough the author was in overall charge and was the sole designer of the Life -upport and communication systams, the overall design, construction and operation of this facility represents the work or a large team.)

137

Fig A.1

138

The underwater laboratory in use. Malta 1969.

139

The expense and complexity of Eraken was largely

due to two factors: the problem of placing the habitat on the

sea-bed and the support recluirements once it was operational.

The present project, which grew from the pool of knowledge

gained in the design of Eraken, was an attempt to construct

an inflatable dwelling sufficiently light to be handled by

swimmers in the water, yet comnrehensive enough to be

completely independent of the surface with no umbilical cable.

,:iunplies would be brought in by divers and the only boat

available an inflatable dingy.

There have been previous inflatable habitats, notably that

of Edwin A. Link (L:aclnnis,55) and more recently one used by

the Moscow "Dolphins" Club (Barton,56). Although little is

known about the latter, the former, being a particularly deen

dive, re(aired considerable surface support and relied on

pol:er, gas and communication cables to the surface.

The Desin

A two-man size was adopted as this represent4 the

recluirement oj: the smallest safe working team. Alter some tests

which involved students living in a scaled polythene enclosure,

a size of 8 ft by 6 ft by 6 ft high appeared adecluate. Allowing

one foot clearance between the floor and the water surface

raises the overall height to seven feet, and-gives a submerged

bouyancy oL around eight tons. Al ter a survey of some possible

flexible materials, a nylon-neoprene fabric, manufactured by

140

the Avon :lubber Company for the construction of inflatable

boats, was selected.

Although it was always realised that, if the fabric

stress was to be minimised, the final shape of the inflated

house should resaable an upturned water—drop, initially it

was by no means clear how one could achieve this: A family

of theoretical cross—sections was produced by an iterative

process involving the relations linking the curvature of a

membrane and the pressure differential across it. The cross—

section that appeared to offer the most suitable height—to—

width ratio was chosen. The next starre was to construct a

straight sided approximation to this cross—section curve. When

this process was extended to three dimensions a 48 faced (plus

th figure evolved.

The reason for translating what should be a smooth shape

into one with a nuuber of flat faces, was that all the seams

in the fabric could now be cemented along straight wooden

jigs. Previously it had been founL1 that a strai,rht semi could

be made with a strength nearly equal to that of the virgin

material, whereas it was found difficult to produce a satisfact—

ory curved seam. _Before cutting the material and commencing

conLruction, one final requirement had to be fulfilled. Under

the calculated working stress the fabric would stretch by

between 5 and Furthermore, the stress—strain relationship

for this fabric varied markedly with the angle to the weave.

141

It is difficult to foresee any accurate methoa of allowing

for this anisotropic behaviour. The technique adopted in

this design was to reduce each of the flat pannels by an

estimated fraction.

The base frame was a ring of 3 inch steam pipe with

12 attachment points for moorings. The fabric was taken

round the pipe and cemented back on itself. In use, the

floor area was Sft N ilft and the walls leant out to provide

Oft N ilft Gins at worhing height. Estimates indicated that

the maxi :um stress would not exceed 1/5 the breaking stress

of the fabric.

It should be realised that although the structure was

constructed as a series of flat panels, in operation the

forces involved would distort the fabric into a smooth shape.

The degree to which this has happened can be seen from the

accompanying photograph.

ballast in the fora of natural rocks was

Used to nrovi(le an anchorage for the ha'Jitat. The initial

preparation inclui,ed surveying the site and attachin four

wire ropes to the ehosen rock. A series of wires were

attached to the 12 points on the house ring: and taken through

four large shackles. In operation, those shackles lay about

four feet below the ring (see fig A.1).

142

The habitat was lashed as a flat package and manhandled

into the water. At this stage the air that was trapped under

the fabric tended to keep it afloat. After being towed into

position above the ballast, the package was untied and a

vent value opened tc allow it to sink. 'our divers guided

it onto the selected rock. Once the four mooring wires had

been attached to the shackles, a small volume of air was

introduced into the house. At this stage the length of the

mooring wires could be adjusted by hand.

Once all appeared to he in order, the structure was

partially inflated and the final levelling operation started.

This was accomplished by attaching a Tirfor pulling machine

in parallel with each of the mooring wires in turn. The

tension ras taLen up by the Tirfor and it was then a relatively

easy matter to ajust; the length of the cable ar.t release the

machine. A.:ter some practice this whole levelling operation,

involving all four wires, could be accOmnlished by three divers

in just over an hour. Finally the house was fully inflated

with an air line from the surface.

The wooden floor, supported on a fexion frame, was

bolted to the main ring as one complete sub—assembly. A total

of about 20 man—hours were spent by personnel, working inside

the inflm.ed house, constructing all interior fittings. These

included the bunks, table and storage lockers and were made

from wood and DeNion with the aid of a hacksaw, drill,

Fig A.2

143

STIO-;-2

Main Camp

Two vehicles Tent accodation for up to 12 personnel.

Telephone connected to jt. Pauls Bay G.P.O.

40 watt fluorescent floodlight. 0 lead acid acctrlulators. 240 volt converter for fluorescent

light. Kyclrogen, oxygen and nitrogen store. 613(„a lime store (5 cwts). First aid kit.

Landin7 Place

Compressor • 830 cu ft 3300 psi air bank. Normal diving equipment — aqualungs

etc. incl-eing fluorescent

and quartz iodine spot lmnps. Derrich — 5 cwt lifting capacity. Diving ladder. ;Lyon led shank and 5 1p outboard. Under water tug. Various watertirht cases for transport of items to the house.

H011;f_l

Domestic

Two bunks, top bunk converts to table.

Life sup,)ort unit. Fuel cell (on sea, bed).

ervo Vie: 0.L150 oxygen meter. :an rose carbon dioxide meter.

-;rae-fer detector tubes for ;as analysis.

Hot coffee heater. Food store. Low voltage fluorescent light. Porta—shower fresh water sl:ower. Gaprleter for metering o::yr4en

input. on r c inr; valve and cylinder.

Tele:)houe. ConC_uction communicator. First Aid kit.

Jcientific

Clevite CH13 hy:Iro?hone. :gruel a 1.jaer precision sound level meter and octave filter bank. 1 inch condenser microphone. ,Aibmersible arZiometer. :-,Arominor test meter. • 2 Calynsonhot ca eras, one with Clash.

2 -:: 20 watt fluorescent lights for photography.

144

screwdriver and spanners.

The Modular Life larpport system

Once the habitat was inflated and level, the first

item to be installed was the life support module, These

units, of which three were constructed, consisted of a case

17" by 13" high with a sealed lid and a all:2,11 air cylinder

with valves to automatically pressurise the LSM.to ambient

sea pressure. Lead acid cells provided a 24 hour supply of

power and their weight gave the unit an overall negative

bouyancy (lo lb s) and insured that the bo:: had a sufficiently

low centre of p_-ravity to be stable when carried by a swimmer.

Mounted above the batteries, a single integral unit housed the

blower was drawn from the inside of the case so as to prevent

the possibility of a build—up of hydrogen or acidic vapours.

The waste heat from the motor and the work done compressing

the air by the single. stage cen-f_rifu.7a1 blower, produced a

slight temvierature rise and ensured that the air meeting the

absorber was not quite saturated with water vapour. sls is

common European practice, a soda—lime charcoal absorber was

used. This material has the advantage that it is most effective

at the high levels of humidity that were bound to be present.

The exhaust from the absorber was not ducted away, instead it

was released through an orifice at a relatively high velocity.

This "jet" of gas was deflected around the habitat by the

curved roof. This method a)nears highly successful as no

145

pockets of carbon dioxide could be detected.

plan proof 12 volt outlets were provided for lighting

a minature immersion heater (for hot drinks) and communication

equipment. An inlet was available also for coupling to an

external power source. -111 outlets were switched and protected

by circuit breakers and the battery voltage was monitored. The

main illumination was with low voltage fluourescent lights.

,llthough a conventional telephone was used in the

habitat for some of the time, including the use for one call

from the house to London, a "wireless" telephone was available

for use in emergency. This communicator used the ionic

conduction field generated by two spaced electrodes places

beneath the house. Although the communicator was normally

powe:'ce from the it was arranged to switch over

autom:Aically to an internal battery in the event of a power

failure. A pilor light in the communicator circuit indicated

this conition provided some illumination as it was

assumed that the main ligh',ing would have been inoperative

in this situation.

The low pressure oxygen line in the habitat was fed

from a cylinder and reducing valve situated below the house.

A gauge was used to set and measure the flow into the atmosphere.

The level of carbon dioxide was checked with a "Itingrose" meter

and oxygen with a paramagnetic meter, every four hours. Once

146

a day these were in turn checked against chalical indicator

tubes. This type of tube was also used for adaily check

of carbon monoNide, hydrogen, stibine and arsnine. Although

the former was present in a concentration of five parts per

million, no trace of the latter impurities was detected.

The louel Cell

_It a late stage in the development of this habitat, a

submersible fuel battering manufactured by the -2:Aectric Power

Storage Co., Ltd., became available. This was self contained

unit using compressed hydrogen and or.sygen. This 16 cell

battery was housed in a light metal cylinder and was automatically

pressurised to ambient sea—bed pressure with nitrogen. The

battery, which could be handled by two divers, was placed on

the sea—bed below the habitat and was capable of supplying

electrical power up to 100 watts. The gas cylinders were

replaced weekly. Because of the ease of using this battery,

the L2: was normally operated from the fuelcell in preference

to its in.ternal lead battery.

Operation

The site chosen was lAarfa Point in the north of

The shore facility was eivided into the main camp situated

about 100 yards from the sea and the lan:ing place where the

divers entered the water. The accompanying table shows the

main items of specialist ecluipment that were eAployed in this

operation. The first habitat was moored at 30 feet in

70 feet of rater on August 5th. However, when preparations

were in hand for the first diver to be nut under saturation

conditions a sudden storm destroyed the habitat complete

with fittings and life support module and damaged the shore

camp.

It was not until August 31st that the first pair of

divers were ready to take un occupation in a second habitat.

._'ter the first storm it was realised that 30 feet

was an awkward depth to face unsettled weather; it was not

deep enouh to ride out a storm, yet the length of decompression

required made a ranici escane impossible. The choice now

open was to operate much deeper at 00 feet (the was

desined for a manimum of 70 feet with nitrogen) where the

effect of wave motion woul be much less or to reduce the

operating denth to 20 feet where a rapid a;ort would be possible

leaving the support party time to deflrte the haAitat and

stow it safely on the sea—bed. The lack of information on

decompression after saturation with nitrogen mintures detennined

that the latter course be adopted.

;Jomewhat to the surprise of the occupants, and the

author who spent four days living in the habitat, it was

comparitively easy to work and sleep underwater. The imowled2,:e

147

148

that the breathing atmosphere and safety was dependant on

oneself and not on a third party whom may not have been seen

since leaving the surface several days previous, appeared to

instil a sense of security.

The inside of the fabric was generally wet with

condensation but the scientific apparatus and life support

equipment did not collect moisture. After a freshwater shower

the human body could be dried with a towel and would remain

dry. No trouble vas experienced when using writing paper or

books. However, as has been reported on previous experiments

of this nature, cold was the main problem. Although the

water temperature was 25° C and air temperature 23°C, the

divers found that it could take several hours to ware up

after a relatively short dive. This was somewhat puzzling

as on shore, where the air temperature was the same on cloudy

days, there was no similar problem.

Throughout the experiment the partial pressure of carbon

dioxide was around and that of oxygen 20 of one atmosphere.

:11 divers spent one hour free orinning at 10 feet for

decompression after one or more days in the habitat.

Acknowledcralent

This appendix refers to the work of the joint

Imperial/Enfield College expedition of 1969. The Imperial

College team was led by lir. P. Newman who was also the

expedition diving officer. Mr. D. Baume led the Enfield

College team. Financial backing came from many bodies and

included the Royal Geogra9hicalsociety- and Imperial College

Exploration I3oard. Consi:lerable assistance to this project

was also given by many organisations in the u.r., 1:alta

and the continent of Europe.

149

150

REFERENCES

1. The British sub—aqua Club diving manual, Eaton publications 1966.

2. Bennett,P.B. ix Elliott,D.N. "The physiology and medicine of

diving and compressed air work" Bailliere Tindall a Cassell 1969

3. Miles,S "Underwater medicine" Staples Press 196G.

4. Williams,S "Underwater breathing apparatus" The physiology and

medicine of diving and compressed air work, Ed. Dennett,P.B.

Elliott,D.H. Bailliere Tindall & Cassell 1969.

5. Baddeley,A.j. "Diver performance and the interaction of stresses"

Underwater Assoc. Report 1966-67, Iliffe.

6. Dugan,J eNplores the sea" 'Tarnish Hamilton 1956.

7. Sims, Rear Admiral W. "Victory at sea" J. Murray 1920.

S. Derktay,H.O. Gazey,B.K. « Teer,C.A. "Underwater communication

past present and future" J. of sound allibr. 7,1. G4 (1968)

9. Mason,D. "U—boat the secret menace" Macdonald u Co 1965.

10. Tucker,D.G. & Gazey,D.L. "Applied underwater acoustics"

PerTamon 1066.

11. Derktay,11.0. a Gazey,D.K. "Communication aspects of underwater

telemetry" iladio a: Electronic Eng. 33,295 (1967)

12. Gazey,B.E. & Morris,J.C. "An underwater acoustic telephone for

free swimming divers" Electronic Eng. June 1964.

13. This is the U.S. Navy AN/UPC-1 system and is used by the NATO

Navies.

14. This equipment is produced by the Aquasonics Co in the U.S.A.

15. Hamilton,P.M. "Underwater hearing thresholds" J.A.S.A.

29,792 (1957)

151

16. Wainwright,W.N. "On comparison of hearing thresholds in air

and in water" J.A.S.A. 30,1025 (1958)

17. Montague,W.E. & Strickland,J.F. "Sensitivity of the water

immersed ear to high and low level tones" J.A.S.A. 33,1376 (1961)

18. Brandt,J.F. & Nollien,11 "Underwater hearing thresholds in man

as a function of depth" J.A.S.A. 46,803 (1969)

19. Bauer,B.B. & Torick,E.L. "Experimental studies in underwater

directional communication" J.A.S.A. 40,25 (1966)

20. Webb,H.J. Webb,J.R. "An underwater audio communicator"

I.E.E.E. Trans. AU-14, 127 (1966)

21. Sims,C.C. "Development of the USRL type J.9. transducer" USTI!,

research report 49, March 20 1959. (Also J.A.S.A. 32,1305 (1960))

22. Liartelli,L. Reinberg,C. "Electronic self contained apparatus

for sound or voice communication" Italian Patent No 617081 (1961)

23. I, orse,P.M. "Vibration and sound" McGraw Hill 1948 sections

27.15 & 27.4.

24. Ldater,J.C. Yrank,N.H. "Electromagnetism" :.1cGraw :ill 1947,

nap. 10.

25. Dunn,H.L. "The calculation of vowel resonances and an electrical

vocal tract" J.A.S.A. 22,740 (1950)

26. Stover,.. "Technique for correcting helium speech distortion"

J.A.S.A. 41,70 (1967)

27. Golden,11.11. "Improving naturalness and intelligibility of

helium-oxygen speech using vocoder techniques" J.A.S.A.

40,621 (1966)

28. Annon "Underwater radio communication" wireless World Feb. 1966

152

29. Needy,K.K. "Divers communication improved" science 158,321 (1066)

30. Nyquist,J "Thermal agitation of electric charge in conductors"

Phys. fev. 32,110 (1928)

31. A throat microphone is used with the 'Aquaphone' made by

Aquaphone Ltd of Poole Dorset.

32. Toby,P. & Dinsdale,J. "Transistor audio power amplifier"

:'tireless World November 1961.

33. Report on enquiry- into Sealab 3 accident. Ocean Industry May 1969

34. Zwislocki,J. "Ear protectors" Handboidi of noise control Ed.

C.M. Harris, McGraw Hill

35. Batteau,D.`11. "The role of the pinny in human localization" Proc.

Roy. _;oc. 168,158 (1967)

36. Nordlund,B Fritzell,B "The influence of azimuth on speech

signals" Acta Oto—laryngologica 56,632 (1963)

37. Hirsh, I.J. "Relation between localization and intelligibility"

J.A.S.A. 22,196 (1950)

08. Cherry, C. "On human communication" Science Editions 1961

Chap. 7 4.3.

39. Hollien,H Doherty,2—T. "Speech intelligibility of diving masks

and mouthcups" 47,127 (1970)

40. Nordlund,B. "studies of steriophonic hearing" Univ. of

Goteborg 1963.

41. LAgnalling and homing by underwater sound; for small

craft and commando swimmers" Classified paper, data taken from

':Iainwrifr,ht (16)

42. Feinstein,S.H. "ihunan hearing underwater, are things as bad

as they seem?" J.A.S.A.40,1561 (1966)

43. Reference manual of transistor circuits, Mullard Ltd 1961.

44. Sivian,L.J. "On hearing in water versas hearing in air"

J.A.S.A. 19,461 (1947)

45. Smith,P.F. conduction, bone conduction and underwater

hearing thresholds in man" J.A.S.A. 44,389 (1068)

46. Stephens,R.W.B. & Dates,A.E. "Acoustics and Vibrational

Physics" Arnold 1966.

47. "The sealab 2 human behaviour program" Project Sealab report

Office of Naval Research report ACR-124.

4S. Hawkins,J.E. Stevens,S.S. "J,Lasking of pure tones and speech

by white noise" J.A.S.A. 22,6 (1950)

49. Thurlow,W.R. & Runge,P.S. "Effect of induced head movements on

localization of direction of sound" J.A.S.A. 42,480 (1967)

50. Topping,J. "Errors of observation and their treatment"

Institute of Physics 1950.

51. Lllump,1!..G. :i;ady,71.R. "Some measurements of interaural time

difference thresholds" J.A.S.A. 28,860 (1956)

59. :;ureter, 7. "The influence of gas bubbles on the generation

of underwater sound" Phi) Thesis, University of London 1007.

53. Eauer,D.B. a Di:Javtia,11..L. "Transmission of directional

perception" I.E.E.E. Trans. AU-13 5. (1965)

54. Pollack, I. fz. Picket,J.M. "Sterionhonic listening and speech

intelligibility against voice babble" J.A.S.A. 30,131 (1950

153

55. MacInnis,J.D. "Living under the sea" Scientific American

214 (3)1 24 (March 1966)

56. Barton,it. "International Oceanics" Hydrospace 2 (4),10

(December 1969)

57. Dwight,;1.3. "Tables of Integrals and other Mathematical Data"

:Macmillan 1961

154

LIST OF SYMBOLS.

A a Area

C Capacitance

c Velocity of sound

d Interaural distance, Skin depth for E.M. radiation

db decibel

✓ (10 Farad 0Jicrofarad)

f frequency

H (mH) Henry (Millihenry)

I Acoustic intensity

k Boltzmanns constant (1.38 x 10-23

Joule/degree)

1 length

L Inductance

O Volume flow of fluid

R Resistance

✓ ILadius vector

t Time

LT Particle velocity of fluid

✓ Voltage

1:ensity

CP :tesistivity standard deviation

Ls-) Angular frequency = 21-rf

(Ks.) Ohm (hillohm)

Permeability

155

156

GLOSSARY 0? DPTING TEIES

Equipment

Aqualung Apparatus for providing the diver with

compressed air for breathing .

Closed-circuit A breathing system where all the exhaled gases

are purified and breathed again

Counterlung A gas-tight bag into which the exhaled gases

pass in a closed or semi-closed circuit breathing

set.

Demand valve The component of an aqualung which regulates the

supply of air to the diver's breathing

requirements. (See fig 1.1)

Demand valve, A breathing system where the demand valve is

single-hose positioned close to the diver's mouth. It is

coupled to the air cylinder by a single high-

pressure flexible hose.

T:emand valve, A breathing system where the demand valve is

twin-hose positioned behind the diver's head and coupled

to a rubber mouthpiece by way two large diameter

low pressure rubber hoses. (une for inhalation,

the other for exhalation)

Foam rubber Cellular neoprene sheet, generally about gun

thick, used for making protective clothing

Full-face-mask A single diving mask covering the whole face.

Helmet Either a rigid gas-tight dome covering the

157

Mouth—mask,

or mouth cup

Mouth—bit

entire head, or a close fitting neoprene hood

covering the head but not the face.

Rubber cup covering the mouth, designed to allow

the diver to articulate underwater. Used in

conjunction with a mask covering the eyes and

nose.

Rubber 'bit' held between the diver's teeth and

coupled to his breathing apparatus. Used in

conjunction with a mask covering the eyes and

nose.

Non—return Device for allowing the passage of gas in one

valve. direction only

Scuba Self Contained Underwater Breathing apparatus.

Normally used to refer to aqualung equipment.

Semi--closed Breathing system where part of the exhaled gases

circuit. are purified and breathed again.

Physiology

Clear the ears The conscious act of equalising the pressure

in the middle ear. (Normally by swallowing or

blowing into the nose)

Decompression Reducing the absolute pressure on the human body

slowly. Failure to observe a safe "decompression

schedule" may result in decompression sickness

often called 'the diver's bends'

Eustachion tubes The connection between the miaile car and back

of mouth. Although normally closed, the

eustachion tubes open during the act of 'clear—

ing the earst

Hypothermia Loss of body heat.

Inert, or The gas that is used to dilute oxygen in a

dilutant gas. breathing mixture. Nitrogen and helium are the

most common.

Narcosis Many inert gases and in particular, nitrogen,

(inert gas) have a narcotic effect when breathed under

pressure. (Helium is often said to be free from

this trouble)

158

ADDITIONAL MA= ;11:1, P71-23ENTED WITH

M,TTQ 11.„

"Voice communication between divers" Underwater Association

report 1966-67.

"Communication between divers" Oceanology International 69

(Proceedings of the S.U.T. conference Brighton 1069)

"Audio communication between free divers" 'Underwater Acoustics'

(Ed. Stephens 11.W.B.) Wiley (in press)

Noise Message

Underwater Association Report 1966-67 47

Voice communication between divers

B. RAY Physics Department, Imperial College, London

SUM MARY The report describes work on various problems

connected with communication between divers. The recordings and tests were made at Marfa Quay, Malta. The distortions introduced by speaking into a face mask are considered and a model proposed to explain them. Apparatus that was used to enable divers to communicate over ranges in excess of 100 yards is described. The relative performance of systems using only the unaided ear for reception are compared with carrier wave methods. Finally, the paper considers the problem of directional bearing underwater.

INTRODUCTION There are two possible means of "wire-less" speech

communication through water. The simplest is to amplify the human voice and send it through the water in such a way that the unaided human ear can be used for reception. This is, of course, just the way a public address system works in air. Alternatively, the speech signal can be used to modulate some carrier. Possible carriers include ultrasonics, laser

beams, and electric and magnetic fields. The carrier system requires some form of receiver, and one loses the advantage of interruption or two people talking at once. A send-receive switch is normally required.

Except where otherwise stated this paper refers to the former "direct speech" method of communication. A block diagram of the system used is shown in Fig. 1. Fig. 2 shows a "direct speech" communi-cator developed at Imperial College.

FORMATION OF WORDS A problem common to all kinds of communication

systems, with or without wires, is the physics of speech inside a small closed volume. Test recordings were made with various kinds of masks and mouth-pieces, as it has been reported (Webb, 1966) that some kinds have a distinct advantage in communication work. In our tests the microphone inside the mask was in each case directiy connected to a surface tape recorder. The diver read a test paragraph as distinctly as possible.

Table I overleaf provides a subjective impression of the quality of the recordings.

DIRECT

VOICE

1 Perception

i 1

Voice Microphone Prop ag ation in Amplifier through Ear

Mask Transducer Water

Fig. I Block diagram of the direct speech communication system.

sec 2n f11 gm cm

48 B. RAY

s •

• • ----- • •••

r

ti

Fig. 2 A "direct speech• - communicator developed at Imperial College.

A suggested explanation of these observations involves the acoustic properties of the mask and the concept of acoustic inductive (mass-controlled) and capacitative (stiffness-controlled) impedance. In air

Table I.

Conditions Comments

All masks in air Complt!tely intelligible

Full face mask underwater

Occasional words missed, meaning clear

Mouth mask underwater

Many mtences had to be repeated to under-stand them

Speaking directly into water by removing a conventional mouth "bit"

Some divers could con-vey simple words but most could achieve no communication

J

the walls of the mask will vibrate as a membrane and transmit some sound whereas, in water, the air-rubber-water impedance mismatch will prevent appreciable transmission and the mask will act as a small closed cavity. To a first approximation the acoustic impedance of such a cavity will be capaci-tative and of the form:—

PCI —4 -- Xe

Where for body temperature and atmospheric pressure :—

Density p =1.14 x /0-' 3 gm/cc Velocity c = 3.53 x 104 cm/.sec Volume of mask = V cc

Hence X, is proportional to 1/V (I) Dunn (1950) has shown how the formation of

vowels by the throat, mouth and lips can be repre-sented by a transmission line (Fig. 3).

Dunn reduced this to a simple electrical lumped

l

i

t

(a.)

Voice communication between divers 49

Throat

Mouth Lips

Fig. 3 Mechanised model of the vocal tract (Dunn, 1950).

Air

(115) Fig. 4 (a) Electrical model of the vocal tract (Dunn, 1950). (b) Modified model of the vocal tract including the effect of the mask. circuit (Fig. 4a). However, this model only refers to speaking into free air. It is proposed to modify the model by replacing the mainly inductive loading of the air by the mainly capacitative loading provided by the mask (Fig. 4b).

It can be shown that the number of resonances in the system does not change from the original four. The capacitative load of the mask does not add an extra resonance; it merely modifies the ones already there. This will be true whatever the values of the components.

With this modification to the model it is possible to account for the better intelligibility achieved with the full-face mask when computed with the smaller ones. In order to comnare the free air and under-water face mask electrical models we can, to a first approximation. add the capacity of the mask to that of the mouth to form a new "ficticious" mouth capacity. Hence from equation (1) the new mouth cavity thus generated will be given by:-

1 = 1 4. 1 V

fic. V V

mask mouth For the English language typical mouth volumes

for the vowels in the words "eat". "lost", and "boots" are 7,90, and 35 cc respectively (From X-ray analysis,

Dunn, 1950). If these volumes are compared with 1200 cc for the full-face mask, 200 cc for the mouth mask, and only a few cc when speaking directly into water, we have a possible explanation for the differing effects of these masks on speech.

A more detailed analysis shows that the low frequency components of the voice will be affected most. Supporting evidence was obtained when underwater speech was filtered. In normal speech there is a frequency such that if one electrically removes all lower or all higher frequencies, then the intelligibility remains the same. This frequency is usually found between 1500 and 1800 c/s. When recordings from the mouth mask were filtered this frequency was estimated to be between 2000 and 2500 c/s. This suggests that the mask has caused more distortion to the lower frequencies.

At the time of writing this is being further tested with recordings made under laboratory conditions.

MICROPHONE The choice of microphone was relatively simple.

In a voice frequency system both throat and bone conduction microphones, being themselves in contact with the water, are prone to acoustic feedback through the water from the transmitter. Lip microphones, isolated by an air cavity from the amplified signal, do not suffer this defect and hence were used throughout the experiments.

AMPLIFIER A transistorised amplifier of conventional class B

design was employed. It was housed in a case 12 in. long made from in. wall "Perspex" tube. "0" ring seals were used throughout, and the design proved entirely satisfactory. The mean electrical output v'hilct the diver was speaking was about two watts.

TRANSDUCERS Several types of transducer were tried; the most

successful was a moving coil design. Pressure com-pensation was achieved by an internal bladder open to the water.

Both fel- diver to surface communication and for subsequent analysis of acoustic signals in water. a smali hydrophone was employed. This gave a very low output, and it was found convenient to anchor the hydrophone to the bottom with a preamplifier. The preamplifier was encapsulated in epoxy resin. and power was fed to it down the cable used to convey the amplified signals to the surfac... re:order.

NOISE Under very quiet cl;nditions the range of a com-

municator, relying only on the unaided ear for reception, may only be limited by the signal Calling below the threshold cf hearing. However, in general,

mask

Surface

Bottom

Fig. 5 Propogation of sound between one transmitter and two receivers at (A) and (B)

50

B. kart

Table II

Origin of noise Type of noise Masking effect Effect on

MHZ carrier system at 25 yds.

Biological "Snapping shrimps"

Impulsive noise

Very little Very considerable

Sea state Low frequency continuous

Very little None

Thermal High frequency None (below threshold)

_Theoretical only

Receiving diver's demand valve

Wide band Complete masking

Very considerable

Receiving diver's exhalation

Low frequency intermittent

Masks weak signals

Small

Other human noises and movements

Bone conduction continuous

Masks weak signals

Small

As one moved to the limit of the range of the transmitter it was found necessary to remain motion-less to hear the signals. The auditory canals of the diver are closed by the presence of water, and this condition gives rise to an im-provement in hearing by bone conduction (between 15 and 25 db.. Zwislocki, 1957). Hence body and equipment movement noises are carried through the body and be-come important underwater. These sounds, along with exhalation bub-ble noises, were found to be the main limitation to the ranee of direct voice communication with the working diver. Although a commercial system was found to be limited to about 10 ft., the amplifier and transducer described already, allowed reliable communi-cation over ranees in r.xcess of 100 ft. Used between surface and diver clarity of signals enabled them to be heard beyond 300 ft.

the steady deterioration of the signal to noise ratio as one moves away from the transmitter will set a maximum range. The ability of different noises to mask speech differs. A continuous noise is worse than an impulsive one. Column three of Table II shows the effect of various noises on a direct audio system. The effect on an ultrasonic carrier communicator can be seen from the last column. This system con-tained both transmitter and receiver and had to be operated rather like an ordinary walkie-talkie.

As can be seen, the three serious noise sources are the demand valve hiss, low and medium frequency noises gen-erated by movement in the water and exhalation noise. It would seem unlikely that there is any way of silencing the conventional cit-mand valve sufficiently to avoid masking incoming signals It follows that for reliable com-munication both parties must synchronize their breathing rate. Ironically this is aided by the valve noise received from the other party. It was for this reason that no electronic means were used to prevent breathing noises being trans-mitted.

RECEPTION Fig. 5 illustrates one transmitter and two possible

receiving positions. Two main propagation paths are drawn between the transmitter and the diver in position (A). One of these is a surface reflection. Surface reflections suffer very little attenuation but are reversed in phase. It follows from this that there will be a zone close to the surface where there is almost complete cancellation of reception or transmission. In practice it was found useless for a

Voice communication between divers 51

diver to attempt communication within two feet of the surface. Another result of this was that music heard within a few feet of the surface lacked bass frequencies compared with the same test piece heard at a deeper depth. The paths from the transmitter to position (B) differ considerably in length. Signals arriving at (B) will arrive at different times and cause "multiple path distortion". This gave rise to con-siderable distortion of signals recorded close inshore over a rocky bottom, whereas in 70 ft. of water over a sandy bottom, signals were far less distorted.

It is generally impractical to design an electronic circuit to remove multiple path distortion. However, it should be clear from the figure that if the diver (B) possessed some form of directional receiver he would be able to differentiate to some extent the direct path from the others. On land man is able to tell the direction of a sound and ignore echoes. The extent to which man can tell the direction of a sound underwater is important in a communication link.

To test directional hearing a wide-band sound source—music was used as nodal patterns would have precluded the use of a narrow-band source such as a tone—was placed 15 ft. below the surface. The subject divers were placed about 30 ft. away at the same depth. With their eyes closed the subjects were first spun around and then asked to point to the direction of the source. No protective clothing was worn on the head. The divers were requested to hold their heads still while performing this test. The results of this were not conclusive; the seven subjects pointed to the correct hemisphere on 32 out of the 50 tests. Better tests are at present being planned for the future.

CONCLUSIONS The difficulties in evaluating communication systems

in the underwater environment are considerable. Tests involving the human element will be affected by a whole range of unusual problems.

There was a tendency for divers to face each ()thee

at close range and discuss simple topics such as depth, weather, etc. This gave an over-favourable first impression that did not stand up to more critical analysis. Likewise, tests performed in swimming baths or with surface operators hanging equipment from boats should not be directly compared with tests using divers in the open sea.

REFERENCES DUNN, H. K. (1950) The calculation of vowel

resonances in an electrical vocal tract. J. Acous. Soc. Am. 22. 740.

ZWISLOCKI, J. (1957) Ear protectors, Handbook of Noise Control (McGraw Hill), 8-12, 8-13. (Ed. C. M. Harris).

WEBB, H. J., and WEBB, J. R. (1966) An under-water audio communicator. LE.R.E. Transac. on Audio and Electroacoustics, AU-14, 127.

ACKNOWLEDGEMENTS We would like to thank the Imperial College

Exploration Board, The Royal Geographical Society and the many other organizations who gave us financial and material support. We would particularly like to thank Dr. R. W. B. Stephens for his valuable help both in London and in Malta.

MEMBERS OF 1966 IMPERIAL COLLEGE MALTA EXPEDITION

A. D. Baume P. Jenkins. B.Sc.. A.R.C.S. A. P. Kingwell, B.Sc., A.C.G.I. R. Leonard J. Love, B.Sc., A.C.G.I. B. Ray, B.Sc., A.R.C.S. R. Wharton, B.Sc., A.C.G.I.

Reprinted from, "Oceanology International CO" (Proceedings of

the Society for Underwater Technology ..Collference, Brighton 1969.)

C C12. TUN I CAT ICN B ETWEEN D rirms B. Ray (Imperial College)

INTRODUCTION.

In normal circumstances the task of a communication engineer commences at the microphone and finishes at the human ear. Underwater the environment is so hostile that the act of speaking or for that matter hearing cannct be assumed. Communication travels frcm thought to mind and any system requires judging on this basis. There are many occasions underwater when the best solution today to a co=munieation problem lies with visual signals or tugs on a lifeline. However it is proposed to limit this discussion to voice communication. In general this can be represented by Fig.l.

BEFORE EVEN SPEECH.

The diver is subject to many physiological conditions such as cold, ahx-iety, narcosis, carbon dioxide build-up, and anoxia, that have the effect, in the first place, of slowing down and distorting mental processes. The communication equipment must not aggravate these troubles by encumbering the diver, nor must a significant mental effort be required to operate it. No record appears to exist of an equipment actually leading to the drown-ing of a diver, but there are examples of systems that went some way in this direction (1).

THE FORMATION OF WORDS.

The main differences between speaking into a face-mask or helmet under-water and speaking in free air, are tabulated in Fig.2. Of these,nuinbers one and six are the most serious. Many people have heard demonstrations of 'helium speech', which will be dealt with in a later paragraph, out the problems associated with the volume of.the mask are less well known. It is best demonstrated by listening to the signal received from a 'fully equipped diver who is standing with his breathing set submerged but with his head above the surface of the water.. As he submerges completely there is a noticeable drop in intelligibility. In air there is appreciable transmission through the relatively thin walls of the face-mask, which reduces the acoustic impedance of the mask cavity, as seen from the lips, below that of an infinitely rigid mask. When,the mask is completely sub-merged, traasmission through the walls is negligible: and the mask can be regarded as a closed cavity, pro--tucing a high acoustic impedance at the lower speech frequencies.

It has been sunested that this high impedance will produce speech distort-ions, the magnitude of which vary inversely with the volume of the mask (2). Two possible solutions to the problem are proposed. The obvious one is to increase the size of the mask. A lightweight inner-mash can be inserted covering the nose and mouth to prevent a large 'dead-space' as the latter leads to possible carbon dioxide build-up (3). The second is to add an acoustic feedback path by placing an amplifier and loudspeaker inside the mask. This, if designed properly, should reduce the acoustic impedance

Fig 1. The communj.cation chain.

PsyFholog!Fal + Thc4nt .1 a noise i

+ Articulation g 1 Ecrs

t /

4

Speech in mask Acoustic noise, 4. I demand voive,etc. 4,- ._,

4,

Microphone

Trons/ miss:cn

tlese in sea

Propagation ; -

Dover L

Rece; ver

Ecrphone Acoustic noise, bubbles, breathing, etc.

Eor

4- Percepticn ro se

If necessary

'Diver IX

Fig. 3. View seen by cameraman in free choice experiments.

The sound source is beyond the top left hand corner of

the photograph.

1.7rWr.' 7, • 7prreqt,":rr.<9',64.:,

Fig. 21 Comparison between spea underwater and speaking in free air

Difference I Effect Contrary factors Possible solutions

Mask has a finite volume coust;c impedance differs from air

III

Larger volume can cause problems :with CO! build up

(a) Inner mask (b) Electronic feedback

The nose is generally blocked, or in a separate cavity

Uvula may be closed, giving rise to a chaitge in vowel sounds

Requirement to 'clear' the ears Mask that allows nose to be pinched for car clearing, but with nose normally free in main cavity

The air path between mouth and cars does not exist (except with a helmet)

The feedback between voice and hearing is radically altered

1

(a) Helmet covering whole head (b) 'Side-tone' fed to cars

Water (or rubber) surrounding the chest and throat

from throat or chest

Change in damping of the throat cavity and lack of radiation • - .

Mechanical restriction of mask on face muscles

Difficulty in producing certain speech sounds 1

Better masks

Increased velocity of sound in breathing gas when using helium—oxygen mixtures nitrogen

'Donald Duck.' effect on voice

•

At deeper depths air cannot be used (inert gas narcosis)

(a) Within limits it may be pos-sible to add small amounts of

(b) Unscrambler Increase in ambient pressure Believed to cause a lack of

fricatives and inability to whistle

Small pressure fluctuations caused by breathingequipment (Generally about I to 2 in. water gauge)

'Difficulty' in articulation Reliability of breathing equipment sets a limit on the sensitivity of the non-return valves used in the construction

Better design of breathing apparatus

Fig. 4.

Masking effect Masking effect Origin of Noise Description on Direct Audio on Carrier System

r

Sea state,and waves on shore

Continuous Low frequency

Very little None

4 - Receiving Divert Diverb Wide band iComplete Complete Demand Valve i iMasking iMasking

4- -4- ..t. "Snapping iImpulsive Small Considerable Shrimps" Biological + I-- L. Human movement & Continuous Masks weak Small bubble noise Bone Conduct—

ion Signals

Reverberation iRelated to Not obvious * Very * Signal Noticable

4.

I-

L

* As reverberation is related to the signal it is difficult to assess the masking effect.

seen from the lips. Although this technique has been used for sound absorb-tlon in air (4), there are practical difficulties in applying it underwater.

The problem of 'helium speech' has become a familiar one in recent years. The velocity of sound in helium is nearly three times greater than in air and in mixtures of helium and oxygen (heliox) it lies between the two. The -effect--of the changed velocity of sound in the breathing gas is that the frequencies of the speech foments are higher in heliox. However, this increase in frequency is in itself proportional to the original frequency, and here occurs the first difficulty in unscrambling helium speech. It is of no help to lower the absolute frequencies by a fixed number of Hertz (a relatively straight forward operation). One popular method (5) at present is to break the speech signal up into a number of frequency bands and to reduce the frequency of each by a different, fixed amount. It has been stat-ed that in saturation diving there is some change in the characteristics of a diver's speech after two or three days, leading to some improvement in intelligibility.

One practical problem in diver communication is that the effect of adding two distorting factors (fig 2), which taken individually may be acceptable, can be disastrous. One example of this is a simple wire comunication sys- • tem between divers and the Sealab 11 habitat based at 205 feet. The report (6) stated, "The transmission was completely garbled and unintelligible. The same type of intercom had been tested at 10 ft at the U.S. Navy Mine Defense Laboratory, with very good results."

This particular system used a bone conduction microphone that relied on sound transmitted through the bone structure of the head. Although this is .the simplest type to operate underwater (7), a more conventional air micro-phone inside the mask will give better reproduction, especially of the high-er voice frequencies. The only other alternative, a throat microphone, normally provides far too low a standard of reproduction.

TRANSMISSION & RECEPTION.

The simplest, and often most efficient, method of transmission is to use a pair of wires between the diver and receiver. If it is convenient to combine the telephone line with an air- or life-line, then this method is the obvious choice (8). However, when the diver is not attached to any form of line, a 'wire-less' system is required. The simplest method of wire-less communic-ation is to amplify the diver's voice and transmit it through ehe water in such a way that the unaided human ear can be used for reception. This will be described by the term 'direct audio', and is, of course, the equivalent to an ordinary public address system.

Alternatively the speech signal can be used to modulate some carrier. Possible carriers include ultrasonic radiation, optical beams, electric and magnetic fields, and very low frequency radio waves. Although this latter has been used for some time for communication with submarines, aerial requ-irements render this quite impractical for divers. A system using the conduction field generated by two electrodes on the divers back is available (9). Optical sytems will, in areas where there is diving interest, be

limited in distance by the amount of suspended matter in the sea water. In many cases this will be a matter of a few feet. The majority of diver comm-unication equipment available uses an ultrasonic carrier at a frequency of between 8 and 150 KHz. Most forms of modulation familiar to the radio engineer are to be found. The range over which communication is possible varies from a few hundred metres to several kilometres. It is not proposed to enlarge on ultrasonic communication here as this subject has been well covered elsewhere. (10),(11).

NOISE.

The ability of different noises to mask speech differs. Fig.4. shows the effect of various noises that were experienced by the author while using compressed air diving equipment in the Mediterranean. Their effect on a 'direct audio' communicator is compared with that on a carrier system, (3 KHz, SSB modulation). The equipment was tested over a range of 25 metres on a rocky shore in sea state 2 to 3. As can be seen, the three serious noise sources were the breathing apparatus, the diver's equipment and cloth-ing and reverberation. It would seem unlikely that there is any way of silencing the conventional demand valve sufficiently to avoid masking in-coming signals. It follows that for reliable communication both parties must synchronize their breathing. Ironically this is aided by the valve noise received from the other party. It was often found necessary to remain motionless to hear the signals. The auditory canals of a diver are closed by the presence of water, and this condition gives rise to an im-provement in hearing by bone conduction (between 15 & 20 db (12)). Hence, in addition to external noises, body and equipment movement sounds are trans-mitted through the body and become important underwater. It was concluded that whereas noise generated by the receiving diver, set the limit to the range of a 'direct audio' system, externalnoise and reverberation were the limiting factors with the carrier equipment. The fact that the 'direct audio' system was superior over short ranges in noisy reverberant conditions leads one to the possibility that its advantages lie with the use of the unaided human ears underwater.

DIVER HEARING.

The free diver usually has his ears open to the water and they normally fill with water, leaving only a small bubble of air on the drum. Pressure on the inside of the ear must be equalised to the external pressure by the diver opening his eustachian tubes periodically (swallowing or blowing into the nose are the usual methods). Failure to equalise properly can cause loss in hearing sensitivity and complete neglect can lead to a rupture of the drum (13).

It is observed that in everyday life there are many cases where.the human brain and ears are required to understand a speech sizmal in the presence of considerable noise. The brain appears to be able to perform this function partly from the nature of language and partly through the use of two ears. This ability is well demonstrated by the 'cocktail party' effect (14), A person 'listening' to several conversations simultaneously has little diff-iculty in isolating them .and understanding which ever one he pleases:

However, this ability is lost when, for example, listening to a recording of the same sound. If this same ability exists underwater then there is good reason for using a binaural receiver or a method utilising both of the unaided ears.

EXPERIMENTS IN BINAURAL PERCEPTION UNDERWATER.

A series of experiments on binaural hearing were carried out by a team from Imperial College in 1968. Although two main experiments were planned, it was found necessary to design a third.while in the field.

The accuracy with which a diver can perceive the direction of a sound source (in the horizontal plane) was measured. The subject diver was suspended in free water by adjusting his weight so that he was bouyant and tying him, by means of a rope around his ankle, to the sea-bed. He was blindfolded and asked to point to the direct- ion of a wide-band source. A plan-view photograph was used to record the indication. A preliminary survey of the results suggests that the diver was correct on most occasions. (Photograph - Fig.3.)

b/ A seated diver was asked to judge from which of two positions a sound originated. The seat and the two possible positions formed a triangle on the sea-bed. The diver had merely to raise his right or left hand to indicate his choice. When the angle between the source positions was greater than 200 few divers had difficulty in indicating correctly. While performing these tests several subjects reported there being a marked difference in the level of sound between facing towards and away from the source. To test this impression a simple audiometer was constructed. It consisted.of a square wave source of 400 Hz, the amplitude of which could be varied in 3.5 db.steps. Five.subjects were tested with this instrument and all showed a difference of between 3.5 db and 7 db between front and back, the forward direction being the more sensitive. The result was not affected by removing the air cylinder and demand valve from the diver's back.

Quantative aspects of these experiments will be available at the time of the conference.

CONCLUSIONS.

At the present timethe most serious problem in voice communication between divers is the act of speaking underwater. Far more research is needed on an interdisciplinary front.

Binaural hearing underwater has been largely overlooked. This is puzzling when one considers that often the diver's sense of touch and vision can be reduced to a very low level of efficiency. The use of the unaided ears,,or possibly some form of binaural receiver (15), appears to offer advantages particularly in the reverberant conditions that are likely to be met in- shore or in the vicinity of wrecks, construction work, etc. Binaural hearing offers a simple way of navigating underwater. It was found that after some practice divers were able to find a hidden acoustic marker. It remains to be seen whether this can be put to use under cowercial conditions.

ACKNOWLEDOENTS.

The author wishes to acknowledge the help and guidance of Dr R.W.B. Stephens. Financial support from the Minnesota Mining & Manufacturing Co.Ltd. enabled the author to pursue this study. The trials that were carried out in the Mediterranean were supported by the IMperial College Exploration'Board.

REFERENCES.

1. O.N.R. Report A.C.R.-124, "Project Sealab Report" p143 'Swimmer intercom' Office cf Naval Research (1967)

2. Ray B. "Voice Communication Between Divers" Underwater Assoc. Report 1966-7, p47, Pub. Iliffe Science &*Technology Publications Ltd.

3. A mask of this type is made by, Draeger Normalair Ltd. 4. Stephens R.W.B. & Bate A.E. "Acoustics & Vibrational Physics" 15.18

Arnold 1966. 5. Golden R.M. "Improving Naturalness & Intelligibility of Helium-

Oxygen Speech Using Vocoder Techniques" J. Acoust..Soc. Amer.40,621. (1966)

6. O.N.R. Report A.C.R.-124, "Project Sealab Report" p144 'Swimmer intercom' Office of Naval Research (1967)

7. Needy K.K. "Divers Communication Improved" Science 153,321 (1966). 8. An example of a combined life-line/telephone is, Siebe Gorman/McMurdo

Duck-Set. 9. An electric conduction communicator was marketed by Andrew & Dalton

Ltd. 10. Tucker D.G. & Gazey B.K. "Applied Underwater Acoustics" Pergamon 1966. 11. Berktay H.O. & Gazey B.K. "Communication Aspects of Underwater

Telemetry" Proc. I.E.R.E. Conference on Electronic Engineering in Oceanography 1966.

12. Zwislocki J. "Ear protectors" Handbook of Noise Control 8-12 (Ed C.M.Harris) McGraw Hill 1957.

13. Miles S."Underwater Medicine" p85 Staples Press 1966. Cherry C. "On Human Communication" Chap.7,4.3. Science Editions 1961.

15. Bauer B.B.,DiMattia A.L.,& Resenheck A.J. "Transmission of Directional Perception" I.E.E.E. Trans. on Audio AU-13, 5, (1965).

'Underwater Acoustics' (Ed Stephens R.W.B.) Wiley 1970.

8

Audio Communication between Free Divers

B. Ray Physics Department, Imperial College of Science and Technology, London

.•

8.1 Introduction .

8.2 The problem . . . .

241

243 8.2.1 Before words arc farmed 243 8.2.2 !Vold formation tender water • . . • 244 8.2.3 Problems associated with breathing mixture . 244 8.2.4 Acoustic impedance of the facemask . . 247

8.3 Microphones . . . . • 247 8.3./ Transmission through water • 248

8.4 Reception and noise 250 8.4.1 Hearing 251

8.5 Conclusions 252

References 253

8.1 Introduction Before looking at the communication problems encountered by men

working under the sea, it is necessary to examine the equipment and tech-niques that are used. This introductory section will be devoted to a brief examination of diving technology.

Until the present decade nearly all serious underwater work was per-formed using a system in which the diver was connected to the surface at all times. The diver wears a 'Hard-Hat' diving dress, which comprises a heavy copper helmet attached to a rubber suit of generous proportions. together with the necessary lead weights and heavy lead boots to keep the diver stable on the sea floor. In Hard-Hat. as with other diving systems, the human body is exposed to the ambient water pressure: however it all the cavities within the human body are 'equalized' to this pressure then the diver may be quite unaware of the absolute pressure. Since the body will not tolerate a pressure differential and if the diver has failed to equalize the pressure in his lungs, ears or sinuses by more than a few feet water gauge,

241

242 Underwater Acoustics

the results can be very serious. It is imperative, therefore, in diving equip-ment to provide breathing gas at the ambient water pressure; in the Hard-Hat system this requirement is met by an air line to a surface compressor which continually supplies the diver with excess air. The correct air pressure is then maintained in the diving dress by expelling this air into the water through a valve on the helmet. Communication with the surface is maintained by a telephone cable alongside the air line. There is ample space inside the copper helmet to fit an intercom, using a conventional moving-coil loudspeaker and microphone.

Since World War 11 compressed air breathing apparatus has become available. The heart of a compressed-air diving set is the 'demand-valve'. This is usually a one or two stage pressure reducer attached to a high-pressure air cylinder. It is designed to provide air at a pressure that will remain within about 2 in water gauge of the pressure at its sensing dia-phragm. Because of the sensitivity of the demand valve, it is necessary to return the exhaled air to a point close to the sensing diaphragm before releasing it through a non-return valve. The self-contained compressed-air diver may wear a mask covering his eyes and nose and hold his breathing hoses between his teeth with a mouth bit; he may wear a mask covering his eyes and nose and a second cup-shaped mask over his mouth; or he may use a 'full-face' mask covering the whole of the face. For protection against the cold a close-fitting rubber suit is normally worn and a pair of fins give the free diver far more mobility than his predecessors. The acoustic problems brought about by the use of a mask will be dealt with later. However, at this stage it should be noted that broad-band noise will be produced by the pressure reduction and this will completely mask most hearing. In the exhalation process there will be the operation of several non-return valves and the production and release of bubbles.

The equipment described so far is not suitable for work below about 250 ft. For deep diving it is usual to use some form of closed- or semi-closed-circuit breathing equipment. In this apparatus the breathing gas, which is now probably a mixture of oxygen and helium, is recirculated and used again. The exhaled gases pass over a chemical to remove carbon dioxide. A small bleed of high-pressure oxygen makes up for the amount of oxygen used by the human body, and the gases finally pass into a flexible bag. This bag is known as the counter-lung, from which inhalation takes place. Although the ancillary equipment tends to be similar to that of the com-pressed-air diver, the level of inhalation and exhalation noise is very much less.

Tiiis is often called SCUBA in the U.S. and AQU‘`.1 UNG in the U.K. These terms :will not be used here as they are trade marks in certain countries.

Audio Communication between Free Divers- 243

8.2 The problem The use of free divers and modern diving techniques however, has

brought about some new problems with communication. The most impor-tant are listed below

(1) The free diver cannot use a cable telephone. (2) The free diver's 'helmet' fits his head far closer than does the 'Hard-

Hat', so creating problems in forming words into a confined, or non-existent cavity, and hearing a signal under similar conditions.

(3) The modern diver may well be using a breathing gas that differs markedly from air. The main reason for this is that air, or to be precise, the nitrogen in the air, has a narcotic effect under pressure. It is generally not. practical to use oxygen/nitrogen breathing mixtures at depths below 250;f and either helium or hydrogen are normally used as the inert gas for deeper diving; both helium and hydrogen have a velocity of sound that is greater than that in air.

In the context of these problems it is important not to take too much for granted. The communication engineer must consider all communication methods and must not reject such possibilities as hand signals. However it is proposed to limit this discussion to voice communication. In general this can be represented as in Fig. 8.1.

8.2.1 Before words are formed

The diver has many physiological enemies such as cold, anxiety, the build-up of exhaled carbon dioxide, anoxia due to inefficient breathing equipment, and narcosis; these may all have the effect, in the first place. of slowing down die mental processes of the diver. It is important that the communication equipment does not aggravate these troubles, nor must operating the equipment require considerable mental effort.

One interesting example of this last part is the use of 'mouth-masks', that, is rubber cup-shaped mouldings designed to strap over the mouth and lips. Their object is to remove the requirement for the diver to hold a conven-tional 'bit' between his teeth. Although these devices do, in many cases, offer improved comfort and clearer communication. they can also cause prob-lems in sealing the mask to the face. When a mouth-mask was used in a U.S. Navy saturation diving project' this problem was magnified to a daniag ous degree by the back pressure exerted by the breathing apparatus. Need-less to say, no intelligible communication was received from these divers!

The author has himself experienced conditions under which he was un-able to remember which end of a vertical rope led to the surface; the reader is left to imagine how reliably one could operate a transmitter under these circumstances.

Psychological noise

Acoustic noise, demand valve, etc.

► Thought

Articulation

f Speechiin mask

Microphone•

Transmission

4-

1

Ears 4

-J

Receiver

If necessary ;

Earp.hone

Ea▪ r

Psychological ▪ Perception noise"

Acoustic noise, bubbles, breathing, etc.

244 Underwater Acoustics

Propagation

Noise in sea

Diver I

'Diver It

Fig. 8.1

8.2.2 Word formation under water The main differences between speaking underwater and speaking in air are

shown in Table 8.1 which should be largely self-explanatory. The 'Contrary factors' column gives the reasons why the obvious solution may not be a suitable one. For example, one may try to overcome problems caused by the limited volume of the mask by increasing the volume, but this un-fortunately can often give rise to pockets of expired air becoming trapped in the additional volume.

It is proposed to limit the discussion to iv:0 of the more important problems.

8.2.3 Problems associated with breathing inixture Before discussing the effects of using breathing gases other than air, it may

be useful to examine the reasons that drive physiologists to look for exotic mixtures. In general it is the partial pressure of the constituents of the

Table 8.1 Comparison between speaking underwater and speaking in free air

Difference

Effect

Contrary factors

Possible solutions

Mask has a finite volume

The nose is generally blocked, or in a separate cavity

The air path between mouth and ears does not exist (except with a helmet)

Water (or rubber) surrounding the chest and throat

Acoustic impedance differs from air

Uvula may be closed, giving rise to a change in vowel sounds

The feedback between voice and hearing is radically altered

Change in damping of the throat cavity and lack of radiation from throat or chest

Larger volume can cause (a) Inner mask problems with CO2 build up (b) Electronic feedback

Requirement to 'clear' the cars Mask that allows nose to be pinched for ear clearing, but with nose normally free-in main cavity

.(a) Helmet covering whole head (b) 'Side-tone' fed to cars

Mechanical restriction of mask oa lace 111LISCICS

Increased velocity of :-,odnd in breathing gas when using helium-oxygen mixtures

Difficulty in producing certain speech sounds

'Donald Duck' effect on voice

Better masks

At deeper depths air cannot be (a) Within limits it may be pos• used (inert gas narcosis) sible to add small amounts of

nitrogen (b) Unscrambler

Increase in ambient pressure Believed to cause a lack of fricatives and inability to whistle

Small pressure Iluctuations 'Difficulty' in articulation Reliability of breathing Bet.:r design of breathing caused by breathing equipment. equipment sets a limit on the apparatus (Generally about I to 2 in sensitivity of the non-return water gauge) valves used in the construction

ul

246 Underwater Acoustics

inhaled gas that define its suitability for breathing. Oxygen is poisonous above 2 atmospheres partial pressure and nitrogen has a narcotic effect above about 7 atmospheres. Hence it should be clear that below about 200 ft it is necessary to reduce the percentage of oxygen and to use a substitute for nitrogen. The most common substitutes are hydrogen and helium. In mixtures containing either of these gases the sound velocity is greater than in air.f

The classical model for the human voice considers the larynx as a wide-band noise generator with the mouth, throat and•nose cavities acting as resonators. The end result is a frequency spectrum whose envelope is defined by the vocal tract and whose fine structure is a function of the larynx waveform. The effect of the increased velocity of sound in helium/ hydrogen mixtures is to shift the envelope shape up the frequency spectrum. The larynx pitch is not affected to a significant degree. The maxima of the frequency spectrum are termed the `formants' of the voice and the most obvious effect of breathing a helium or hydrogen mixture is an increase in the frequency of these formants, the increase being in proportion to the change in the velocity of sound in the exhaled breath.

From the foregoing discussion it should be clear that in order to render helium speech intelligible it is pointless to reduce the component frequencies of speech by a fixed number of cycles per second (a relatively straight-forward operation). What is required is a proportional frequency reduction, which is just what happens when a tape recorder is played back at slow speed. Although under some circumstances it may be useful to use a tape recorder for this purpose, the distortion of the time scale normally negates any advantage gained. One practical method does, in fact, use a time-stretching technique to correct the frequency spectra. However a part of each separate sound is discarded so that the overall time scale remains true.2 At the time of writing the most popular method of unscrambling the speech of divers is to break the speech signal into a number of frequency bands and • to reduce the frequency of each by a different, fixed, amount. The larger the number of frequency bands the closer will be the approximation to the ideal.

Unfortunately, at present the performance of even the best instruments leaves a considerable amount to be desired. It appears that apart from shift- ing the formants. the gas mixture distorts speech in other ways. The role played by the absolute pressure is still in doubt; a mixture of 2! per cent oxygen, 79 per cent helium has the predicted effect at atmospheric pressure and can be unscrambled remarkably well. At 20 atmospheres pressure, a suitable breathing gas would contain less than 5 per cent oxygen and the

t The reduction of the oxygen percentage makes it possible to design a diving system that avoids explosive combinations of oxy-hydrogzn.

Audio Conimunication between Free Divers 247

velocity of sound in this would be higher than in the 21/79 mixture that was used at the surface. ro the untrained observer the surface 21/79 speech would appear strange but understandable; however, the speech from 20 atmospheres would be completely unintelligible, and might not be recog-nized as the human voice.

The problems of making measurements under high pressures are many, for not only does there exist some doubt as to the nature of speech recorded under these conditions, but there are limitations on the gas mixtures that can be used in the case of human subjects. Furthermore, experiments at 20 atmospheres are very expensive to perform because it will take several days to decompress a subject from even a short exposure to this pressure.

8.2.4 Acoustic impedance of the facemask If one listens to the signals received from a fully equipped diver who is

standing with his breathing set submerged but with his head above the surface of the water, as he submerges his body completely, there will be a noticeable drop in intelligibility. In air there is appreciable transmission through the relatively thin walls of the facemask, which reduces the acoustic impedance of the mask cavity, as seen from the lips, below that or an infinitely rigid mask. When the mask is completely submerged, trans-mission through the walls is negligible and the mask can be regarded as a closed cavity having a high acoustic impedance. Although this argument strictly only holds for frequencies below the cavity resonance of the mask, the conclusion, namely that the mask will produce an effect which is. inversely proportional to its volume (i.e. proportional to its impedance), appears to provide a useful guide to the choice of masks for practical communication purposes3.

Two possible solutions to this problem will be considered. The obNious. one is to increase the volume of the mask. If this is done, then some light-weight inner mask can be inserted to prevent a large dead-space as the latter leads to the possibility of carbon dioxide build-up. The second is to add an acoustic feedback path by placing a microphone, amplifier and loudspeaker inside the mask. It should be possible to design such a system to reduce the acoustic impedance seen from tile lips. Although this technique has been used for sound absorption in air there are practical difficulties in applying it underwater.

8.3 Microphones Three types of microphone are in use by divers; they are air, bone-con-

duction and throat. All of them have particular problems. The air micro-phone, positioned in the mask, stands the best chance of receiving

Moulded rubber Peizo- electric

ceramic Reducible air volume, Fixed air

V volume, v

Seal

—Resin

--Electrodynamic microphone insert

Cable

Flexible membrane

248 Underwater Acoustics

intelligible signals. However, it must be designed to withstand a large p: e;-sure change and frequent splashing with water. A design for a pressure-compensated air microphone is given in Fig. 8.2. By contrast the bone-conduction microphone is a vibration pick ups and can more readily be made as an encapsulated pressure-resistant device. Throat microphones are not normally used underwater as the quality of reproduction is seldom sufficient to allow intelligible communication.

The choice between air and bone-conduction may be influenced by their different performance with locally generated noise. The main interference

Examples of underwater microphones

Bone-conduction Air microphone designea for use inside a face mask microphone

Showing the effect of pressure on this microphone : the membrane con move to equalize pressure.

Max;mum depth capability =, v — metres

Lead out wires

Fig. 8.2

experienced when using an air microphone is the sound generated by the breathing set, particularly the inhalation noise. On the other hand it is the noise caused by the exhalation bubbles that is the main limitation of the bone-conduction microphone.

8.3.1 Transmission through water Referring again to Fig. 8. I it can be seen that the stage has now been

reached where an electrical representation exists of the words spoken by the transmitting diver. Some methods that can be used to transmit a speech signal through water are listed in Table 8.2. The simplest, and often most ef9cient of these is to use a pair of wires between the transmitter and the receiver. If it is convenient to combine the telephone line with an air- or life-

Table 8.2 'Telephony transmission systems

Method

Propagation

Radiator

Receiver Main limitation

Typical range

Two wires Electric current

Telephone

Telephone

The inconvenience of No limitation the wires

Direct audio Audia frequency Piston sound

The car Noise generated by 0-30 m sound

(loudspeaker)

the receiving diver and his breathing set

Ultrasonic Modulated high frequency sound 8 -200 kl lz

Resonant piezo-electric transducer

Resonant piezo- Reverberation and electric transducer multi-path

distortion

0.1-5 km

Two spaced Interference due to electrodes some- any electrical what aligned to the installation with transmitting eat th currents electrodes

Electric field

Optical

Magnetic field

(a) Audio frequency Two spaced

electric current electrodes field

(b) Modulated high frequency field

Visible light (green Laser or semi- window) conducting diode

Magnetic induction Large coil

Photodiode or photocell

Coiil somewhat aligned to trans-miting coil

Scattering and absorption of transmitted beam. Transmitters and recekers are generally highly directional

A few metres with early models. Information rather obscure on later ones

0-100 m (clear water) 0-5 m (coastal water)

Very low frequency EM waves Aerial (impractically Aerial radio

large for diver)

250 Underwater Acoustics

line, then this method is the obvious choice. However when the diver is not attached to any form of line a wire-less system is required.

The simplest method of wire-less communication is to amplify the voice of the diver and to transmit it through the water in such a way that the unaided human ear can be used for reception. This has been described by the term 'Direct Audio', and can be considered as the equivalent of an ordinary public-address system. The advantages of a direct audio trans-mission are that it is simple, and that a two-way conversation is possible without the need for a send/receive switch. On the debit side, the range is limited and the performance is dependent on the protective clothing that the diver is wearing around the ears.

The most popular method of transmission is to modulate an ultrasonic carrier. Ordinary amplitude modulation (AM), frequency modulation (FM)7, and single sideband AM systems are all in use, and there have been experiments with pulse modulation. The choice of modulation is governed by its ability to resist multiple-path distortion as this will be the predomi-nant interference in water depths of interest to divers. Frequency modula-tion has a natural resistance to the arrival of spurious multipath signals that are of small amplitude compared with the carrier, high-amplitude signals will cause serious distortion. In contrast an AM system will demodulate all the incoming signals and present the listener with the reverberation. The ear is not unaccustomed to dealing with a reverberant signal and there will be further discussion on this later.

Although some of the first communication sets to be marketed used the electric conduction field set-up by two spaced electrodes on the back of the diver, and electric fields are generated by certain fish, this method is less popular today. The advantage of being free from acoustic limit•itions was offset by limited range and 'dead spots' when the transmitting and receiving `aerials' were at right angles. The introduction of high-frequency carrier conduction field sets with automatic gain control at the receiver may well revive this technique.

8.4 Reception and noise Diver-to-diver communication is inherently a short-range technique.

Range limitations that arc due to poor signal to ambient noise (sea-state. thermal, biological, etc.) raiio are seldom of interest: it is, in general, the acoustic noise locally aenerated at the receiving end that limits the per-formance of these diver commurication systems. The major part of this often stems from the breathing equipment; however, this may not present as serious a problem as might be .-:xpected from acoustic measurements. Because the act of breathing may often completely mask incoming signals,

Audio Conummication between Free Divers 25T

it may be necessary to synchronize the breathing of the two divers; this is often helped by the valve noise received from the other party. With most forms of face mask the auditory canals are closed by the presence of water. This condition gives rise to an improvement in hearing by bone conduction of between 15 to 20 dB8. The effect of this is that body movement sounds and noise generated by movement of clothing and equipment on the body become significant underwater. A simple example of this is the effect of scratching the back of the head; in air this is unlikely to interfere with normal hearing, whereas underwater, this may well mask voice com-munication.

Table 8.3 Threshold of hearing underwater Comparison of published figures at 2 kHz. The divers are not wearing a

protective rubber helmet.

Reference Value of threshold Authors no. ref. 0.0002 dyne

Ide, J. M. (1944) (obtained from Ref. 15) 73 dB Hamilton, P. M. (1957) 14 53 dB Wainwright, W. N. (1958) 15 82 dB Montague, W. E., and Strickland, J. F. (1961) 10 70-80 dB Brandt, J. F., and Hollien, H. (1967) 16 60-70 dB Zwislocki, J. (1957) 17 Threshold of bone

conduction hearing in air 46 dB

There is a further source of noise that is significant in air-filled structures such as sea-bed shelters and underwater laboratories. By nature of their design these structures are often highly reverberant: this is particularly so in the case of some of the smaller flexible ones. Here the 'Q' of the gas-filled shelter may approach that of a bell on land. Excess gas is often continuously vented from such structures and the release of bubbles appears to be an excellent way of exciting their resonances. The noise level inside can be sufficient to mask telephone conversation and to cause considerable annoyance to the occupants.

8.4.1 Hearing The diver must equalize the pressure across the ear 'drum. This is nor-

mally accomplished by allowing the outer ear to be open to the water or helmet cavity, and equalizing the pressure in the middle ear by opening the eustachian tubes periodically (swallowing or blowing into the nose are the usual methods). Failure to equalize properly can cause loss in hearing sensitivity, and complete neglect can lead to rupture of the drums.

252 Underwater Acoustics

Normally when man swims under water there is a bubble of air trapped in the auditory canal; the part played by this is uncertain, as is the mechanism by which sound reaches the inner ear. It is possible that almost all hearing underwater is by bone conduction, and the similarity between the under-water and bone-conduction hearing thresholds have been used as evidence to support this". Alternatively the sound may be transmitted along the auditory canal in much the same way as in air; the ability to hear direc-tional information is the evidence in this case. Table 8.3 compares some of the published values for underwater thresholds at 2 kHz. The author is of the opinion that it is not meaningful to consider underwater hearing as either tympanic or bone conduction. These terms apply to hearing in air where the difference in acoustic properties between human flesh-and-bone, and gas, is such that one can consider a sound wave as travelling in one or the other. Under water it may well be more reasonable to consider the two hearing organs as existing in an infinite fluid. This ignores the water—skin boundary completely. The middle ear, sinuses and other air cavities are impedance discontinuities in the neighbourhood of the receptors and must b3 taken into account. Directional hearing and changes in angular sensi-tivity can be explained while retaining some similarity with bone conduction hearing.

It is possible for an underwater swimmer to tell the direction from which a sound originates11. This suggests that there may be considerable advantages to be gained in underwater communication by using a binaural receiver, as it is well known that binaural hearing in air helps in the under-standing of a speech signal in the presence of noise or reverberation. This ability is well demonstrated by the 'cocktail-party effect'12 ; a person `listening' to several conversations simultaneously has little difficulty in isolating them and understanding whichever one he pleases. This ability, however, is lost when listening to a monophonic recording of the same sound.

Although the use of binaural receivers has been proposed before13, the only available equipment that allows binaural reception are the 'direct audio' communicators. These, of course, use (both) the unaided ear(s) for reception. It has been reported3 that under some conditions communica-ion is possible with 'direct audio' equipment where carrier systems failed

due to reverberation.

8.5 Conclusions In the experience of the author, most equipment designed for diver com-

munication fails through mechanical design, either leakage (poor seal design), corrosion (wrong combinations of metals), or it encumbers the

Audio Con►nunicution between Free .111ver:s 253

diver (bouyancy index, physical size or method of operation). Having eliminated equipment that cannot be reliably tested outside the laboratory, the second most common failing is in the acoustic design at the transmitting end: facemask, microphone and breathing set.

It is only after the mechanical and acoustic design has been considered need one look to the ability of the equipment to transmit the voice signal through the water; yet it is in this area where almost all the evaluation of devices is done. It is of little wonder that a communicator that performs well in the test tank all too often fails to be of practical worth.

Although it has been emphasized that the whole of the diver's equipment must be designed with communication in mind, it is, of course, well known that the human brain is very tolerant to a considerable amount of speech distortion. It is possible to make up for a weak link in the communication chain providing all the other links are capable of passing on the distortion produced in order to give the person at the receiving end the best chance of. 'reading' the signal. For this reason, helium speech plus a pair of wires, a poor facemask plus a pair of wires, or a person speaking in air with a poor submarine telephone, may all be intelligible. Whereas combinations that use two weak links are seldom useful; their encumberance generally out-weighs any communication advantages they may offer the diver.

References I O.N.R. Report A.C.R.-124, Project Sealab Report, p. 143, 'Swimmer

Inrercom' Office of Naval Research, Washington, D.C. (1967). 2 Stover, W. R., 'Technique for Correcting Helium Speech Distortion'.

J. Acoust. Soc. Am. 41, 70 (1967). 3 Ray, B., 'Voice Communication Between Divers', Underwater Association

Report 1966-67, p. 47, Iliffe, London. 4 Stephens, R. W. B., and A. E. Bates, Acoustics and Vibrational Physics 15, 18,

Arnold, London, 1966. 5 Needy, K. K., 'Divers Communication Improved', Science 153, 321 (1966). 6 Webb, H. J., and J. R. Webb, 'An Underwater Audio Communicator',

Inst. Elec. Electron. Engrs Trans. AU-14, 127 (1966). 7 Gazey, B. K., and J. C. Morris, 'An Underwater Acoustic Telephone for

Free-swimming Divers'. Electron. Eng., June, 1964. 8 Zwislocki, J., 'Ear Protectors'. Handbook of Noise Control (C. M. Harris,

ed.), pp. 8-12, McGraw-Hill, New York, 1957. 9 Miles, S., Underwater Medicine, p. 85. Staples, London. 1966.

10 Montague, W. E., and J. F. Strickland, 'Sensitivity of the Water Immersed Ear to High- and Low-level Tones'. J. Acoust. Soc. Am. 33. 1376 (1961).

11 Ray, B., 'Communication Between Divers', Oceanology International 69 (Proceedings of the Society for Underwater Technology Conference) (1969).

12 Cherry, C., On Human Comnwnication, Ch. 7, 4.3, Science Editions, New York, 1961.

254 Underwater Acoustics

13 Bauer, B. B., A. L. DiMattia and A. J. Rosenheck, 'Transmission of Directional Perception', but. Elec. Electron. Engrs Trans. AU-13, 5 (1965).

14

Hamilton, P. M., 'Underwater Hearing Thresholds', J. Acoust. Soc. Am. 29, 792 (1957).

15 Wainwright, W. N., 'On Comparison or Hearing Thresholds in Air and in Water', J. Acoust. Soc. Am. 30, 1025 (1958).

16 Brandt, J. F., and H. Hollien, 'Underwater Hearing Thresholds in Man', J. Acoust. Soc. Am. 42, 966 (1967).

17

Zwislocki, J., 'In Search of the Bone-conduction Threshold in a Free Sound Field', J. Acoust. Soc, Am. 29, 795 (1957).