The Bell System Technical Journal : devoted to the Scientific and Engineering aspects of Electrical Communication, Vol. 19, No. 4

/ V x> i p

*,

VOLUME XIX

OCTOBER, 1940

• ? '9 / ^ 0

THE BELL SYSTEM

TECHNICAL JOURNAL

DEVOTED TO THE SCIENTIFIC AND ENGINEERING ASPECTS OF ELECTRICAL COMMUNICATION

(>»J

The Carrier itfhfafla of Speech—Homer D u d l e y ...495 Manufacture of Quartz Crystal Filters—G. K . Burns . . . 516 Results of the World’s Fair Hearing Tests

—J. C. Steinberg, H. C. Montgomery, and M. B. Gardner 533 The Subjective Sharpness of Simulated Television Images

—Millard W. Baldwin, Jr. 563

Cross-Modulation Requirements on Multichannel Amplifiers Below Overload— W. R. B e n n ett... 587 Radio Extension Links to the Telephone System

—R. A. Heising 611 Abstracts of Technical P a p e r s ...647 Contributors to this I s s u e ... 651

AMERICAN TELEPHONE AND TELEGRAPH COMPANY NEW YORK

50c p e r Copy 1.50 p e r Year

Published quarterly by the

American Telephone and Telegraph Company 195 Broadway, New York, N. Y.

EDITORS

R. W. King J. O. Perrine

F. B. Jewett A. B. Clark S. Bracken

EDITORIAL BOARD H. P. Charlesworth

O. E. Buckley M. J. Kelly W. Wilson

W. H. Harrison O. B. Blackwell G. Ireland

■ llilllllllllllllilllllllllllllllllilib

SUBSCRIPTIONS

Subscriptions are accepted at $1.50 per year. Single copies are fifty cents each»

The foreign postage is 35 cents per year or 9 cents per copy.

llllllllllll llll llll lllll llll lllll llll l'

Copyright, 1940

American Telephone and Telegraph Company

P R I N T E D I N U . S . A .

The Bell System Technical Journal

Vol. X I X

The

Speech s y n th e siz in g is h e re d iscu ssed in th e te rm in o lo g y of c a rrie r c irc u its. T h e sp e a k e r is p ic tu re d a s a s o r t of ra d io b ro a d ­ c a s t tr a n s m i tt e r w ith th e m essag e to be s e n t o u t o rig in a tin g in th e s tu d io of th e ta l k e r ’s b ra in a n d m a n ife stin g itse lf in m u s c u la r w av e m o tio n s in th e v o cal t r a c t. A lth o u g h th e s e m o tio n s c o n ta in th e m essage, th e y a re in a u d ib le b ec a u se th e y o ccu r a t sy lla b ic ra te s . A n a u d ib le so u n d is need ed to p a ss th e m essage in to th e lis te n e r’s ea r. T h is is p ro v id e d b y th e c a rr ie r in th e form of a g ro u p of h ig h e r fre q u e n c y w av es in th e a u d ib le ra n g e s e t u p b y o sc illa to ry a c tio n a t th e v o c a l co rd s o r elsew h ere in th e vocal tr a c t. T h e se c a rr ie r w av es e ith e r in th e i r g e n e ra tio n o r th e ir tra n s m is s io n a re m o d u la te d b y th e m essag e w av es to fo rm th e sp eech w aves. A s th e sp eech w av es c o n ta in th e m essage in fo rm a tio n on a n a u d ib le c a rr ie r th e y a re a d a p te d to b ro a d c a s t re c e p tio n b y re c e iv in g sets in th e fo rm of lis te n e rs ’ ears. T h e m essag e is th e r r reco v ered b y th e lis te n e rs ’ m in d s.

O P E E C H is like a radio wave in th a t inform ation is tran sm itted over

^ a su itab ly chosen carrier. In fact the modern radio broadcast system is b u t an electrical analogue of m a n ’s acoustic broadcast sys­

tem supplied by natu re. Com m unication by speech consists in a sending by one m ind and th e receiving by ano ther of a succession of phonetic symbols w ith some em otional content added. Such m aterial of itself changes gradually a t syllabic ra te s and so is inaudible. Ac­

cordingly, an audible sound stream is interposed between the ta lk e r’s brain and th e listener. On th is sound stream there is molded an im ­ p rin t of the message. T he listener receives th e molded sound stream and unravels th e im printed message.

In th e p a st this carrier n atu re has been obscured by the complexity of speech.1 However, in developing electrical speech synthesizers

1 Speech-making processes are here explained in th e term s of th e carrier engineer to give a clearer insight into the physical nature of speech. The point of view is essen­

tially th a t of th e philologist who associates a message of tongue and lip positions w ith each sound he hears. This aspect also underlies th e gesture theory of speech by Paget and others and th e visible speech ideas of Alexander Melville Bell. The au th o r has been assisted in expressing speech fundam entals in carrier engineering term s by numerous associates in th e Bell Telephone Laboratories experienced in carrier circuit theory. Acknowledgment is made in particular of th e contributions of Mr. Lloyd Espenschied.

495

Carrier Nature of Spe

By HOMER DUDLEY

O ctober’1940

copying th e hum an m echanism in principle, it was soon a p p a re n t th a t carrier circuits were being set up. T racing th e carrier idea back to the voice m echanism there was unfolded, a little a t a tim e, th e carrier n atu re of speech. U ltim ately th e speech m echanism was revealed in its sim plest term s as a m echanical sender of acoustic waves analogous to th e electrical sender of electrom agnetic waves in th e form of the radio tran sm itte r. E ach of these senders em bodies a m odulating de­

vice for molding message inform ation on a carrier wave suitable for propagation of energy thro ugh a transm ission m edium betw een the sending and receiving points.

Th e Ca r r i e r El e m e n t s o f Sp e e c h

T his carrier basis of speech will be illustrated by simple speech examples selected to show separately th e th ree carrier elem ents of speech, nam ely, th e carrier wave, th e message wave, and th eir com ­ bining by a m odulating m echanism . These illustrations serve the purpose of broad definitions of th e carrier elem ents in speech.

T h e illustration chosen for th e carrier wave of speech is a ta lk e r ’s sustained tone such as th e sound “ a h .” In th e idealized case there is no variatio n of intensity, spectrum or frequency. T his carrier then is audible b u t contains no inform ation, for inform ation is dynam ic,2 ever changing. T he carrier provides th e connecting link to th e liste n e r’s ear over which inform ation can be carried. T h u s th e ta lk e r m ay pass inform ation over this link by sta rtin g and stopping in a prearranged code th e vocal tone as in im itatin g a telegraph buzzer. F o r tra n s ­ m itting inform ation it is necessary to m odulate this carrier w ith the message to be tran sm itted .

F or the second illustration, message waves are produced as m uscular m otions in th e vocal tra c t of a “ silent ta lk e r ” as he goes th ro ug h all th e vocal effort of talking except th a t he holds his b re ath . T h e mes­

sage is inaudible because th e m otions are a t slow syllabic ra te s lim ited by th e relatively sluggish m uscular actions in th e vocal tra c t. N ever­

theless these m otions contain the dynam ic speech inform ation as is proved b y their in terp re tatio n b y lip readers to th e ex ten t visibility perm its. A nother m ethod of d em onstrating th e inform ation con ten t of certain of these m otions is th e artificial injection of a sound stream into the back of th e m outh for a “ c a rrie r” w hereby intelligible speech

2 T he inform ation referred to is th a t in th e com m unication of intelligence. T here is, however, static inform ation in th e carrier itself. T his serves for “ station id en ti­

fication” in radio and m ay similarly help in telling w hether it was Uncle Bill or A unt Sue who said “ a h .”

C A R R I E R N A T U R E OF S P E E C H 497 can be produced from alm ost any sound stream .3 T he need of an audible “ c a rrie r” to tran sm it this inaudible “ m essage” is obvious.

The final example, to illustrate the m odulating mechanism in speech production, is from a person talking in a norm al fashion. In this example are present the message and carrier waves of the previous

SUPPLY

Fig. 1—The vocal system as a carrier circuit.

examples, for both are needed if th e form er is to m odulate the latter.

However, the mere presence of th e carrier and message waves will not m ake speech for if they are supplied separately, one by a silent talker and the other by an intoner, no speech is heard b u t only the audible intoned

3 R. R. Riesz, “ Description and Dem onstration of an Artificial L arynx,” Jour.

Acous: Soc. Amer., Vol. 1, p. 273 (1930); F. A. Firestone, “ An Artificial Larynx for Speaking and Choral Singing by One Person,” Jour. Acous. Soc. Amer., Vol. 11.

p. 357 (1940).

carrier. O rdinary speech results from a single person producing th e message waves and th e carrier waves sim ultaneously in his vocal trac t, for then th e carrier of speech receives an im p rin t of th e message by m odulation.

Th e Sp e e c h Me c h a n i s m a s a Ci r c u i t

T he foregoing three illustrations by segregating th e basic elem ents in speech production reveal th e underlying principles. T h e present paper tre a ts of these elem ents as functioning p a rts of a circuit. In Fig. 1 is shown a cross-section of th e vocal system . T he idea to be expressed originates in th e ta lk e r ’s brain a t th e left top. Thence, impulses pass through th e nerves to th e vocal tra c t w ith the com plete inform ation of th e “ m essage,” th a t is to say, w h at carrier should be used, w h at fundam ental frequency if th e carrier is of th e voiced type and w hat transm ission through the vocal tra c t as a function of fre­

quency. T he carrier w hether voiced or unvoiced is shown for sim ­ plicity as arising a t th e ta lk e r ’s vocal cords. T his carrier is m odulated to form speech having th e com plete message im printed on it p re p ara­

to ry to radiation from the ta lk e r’s m outh to the ear of th e listener, who recognizes the im printed message.

In discussing the speech mechanism as a circuit, it is clearer to s ta rt with a block schematic. Figure 2 has th u s been draw n to sketch the

Fig. 2—T he basic plan of synthesizing speech.

basic plan of speech synthesizing. As in Fig. 1, th e idea gives rise to th e message which m odulates th e voice carrier to produce th e speech rad iated from th e ta lk e r ’s m outh. One can follow th e p a th of th e message from its inception in th e ta lk e r ’s brain to its ra d iatio n from his m outh as an im p rin t on th e issuing sound stream . T h e progress of th e sound stream is also seen from its origin as an oscillatory carrier to its radiation from th e ta lk e r ’s m outh carrying th e message im p rin t.4 T he light arrow heads indicate direction of flow while th e h eavy ones indicate a m odulatory control of the carrier by th e message. T his

4 H ere th e carrier p ath is stressed to show th e alteratio n of th e carrier sound stream as it proceeds on its way from th e point of origin to th e point of radiation.

This also accords w ith the im portance of th e voice carrier which is received and used by th e ear, and th u s differs from the tre a tm e n t of the carrier in simple radio broadcast reception.

C A R R I E R N A T U R E OF S P E E C H 499 m odulatory control is exerted on th e carrier wave in p a rt as th e carrier is generated and in p a rt as it is tran sm itted after generation.

Re l e v a n t Ca r r i e r Th e o r y

T he h eart of the speech-synthesizing circuit of Fig. 2 is the p a rt in which the group of waves m aking up th e message m odulate th e com­

ponent waves of the carrier. In any one of these m odulations, there is the simple carrier process blocked out in Fig. 3. H ere a message 5

Fig. 3—The elements of a carrier sender.

containing the inform ation m odulates a carrier determ ining the fre­

quency range so th a t the end product in the form of the message- m odulated carrier contains th e inform ation of the message translated to frequencies in the neighborhood of th e carrier. In this way the carrier sound stream of speech is im printed w ith the message.

The prerequisites of th e carrier system sender are, as indicated in Fig. 3, first, a carrier wave source; second, a message wave source; and third, a m odulating circuit of variable im pedance by which the message controls the carrier. T he carrier wave is for the sim plest case a single sine wave function of tim e characterized by an am plitude, a frequency and a phase. The message wave as a rule is more complex b u t m ay be analyzed as the sum of com ponent sine waves each of which is ch ar­

acterized by its own am plitude, frequency, and phase. In m ost carrier circuits the frequency range of the message is below th a t of the carrier.

This is tru e of speech production.

The function of the m odulating circuit is supplying a means for the message wave to modify a characteristic of th e carrier. If th e carrier wave am plitude is modified by the message wave am plitude the process is known as am plitude m odulation; if the carrier wave frequency is so modified the process is called frequency m odulation while if the carrier wave phase is so modified th e process is called phase m odulation. No distinction is m ade as to w hether th e modification occurs during or

6 The word “ message” has been substituted for th e usual carrier term “ signal”

to avoid confusion since the input signal is commonly speech whereas here th e output wave is speech. “ M essage” seems particularly appropriate w ith its suggestion of code as in telegraph.

afte r th e generation of th e carrier. M odification of th e carrier wave characteristics b y o ther th a n the am plitude of th e message need not be considered here. In th e voice m echanism significant am plitude and frequency m odulations of th e carrier occur. Phase m odulation takes place also b u t will n o t be discussed because th e liste n e r’s ear is n o t very sensitive to these phase changes in th e carrier.

In a tte m p tin g to segregate th e carrier elem ents of speech we run into one serious difficulty. In an idealized carrier circuit as shown in Fig. 3 connections can be cu t between th e two energy sources and th e m odulator so th a t each boxed elem ent can be studied independently.

W ith th e hum an flesh of th e voice m echanism this is no longer feasible;

th e use of cadavers would help very little because norm al energizing is then impossible. T he sam e difficulty often appears in electrical m odulators as, for example, w ithin a m odulating vacuum tub e where a grid voltage m odulates a plate current. In such a case of common p a rts it is necessary to discuss th e action of each of th e three elem ents in th e presence of th e o ther two.

W ith this carrier theory review as a background we are in a position to analyze th e three elem ents m aking up th e carrier tran sm ittin g system of th e hum an voice. W hile th e picture presented is over­

simplified in details the principles hold and aid in applying carrier m ethodology to explain th e m echanism of speech.

Th e Vo i c e Ca r r i e r

In electrical circuits the carrier is obtained from an oscillatory energy source. T he same holds for speech. In th e electrical circuit th e os­

cillatory waves (a-c.) are ordinarily generated from a supply of d-c.

energy.6 T he sam e is tru e in speech w ith th e com pressed air in th e lungs furnishing the stead y supply. Confusion m ust be avoided, for in speech th e conversion of stead y to oscillatory energy is often de­

scribed as modulation. H ere this conversion of energy form will be considered as an oscillatory action so th a t th e term modulation can be reserved for th e low-frequency syllabic control of th is oscillatory energy to produce th e desired speech. Oscillatory th en will refer to au to m atic n atu ra l responses while modulatory will refer to forced responses which are controlled volitionally. T his distinction is consistent w ith carrier term inology.

In the sim plest electrical m odulating circuits th e carrier is a sine 6 In th e usual electrical circuit th e carrier is cu t off b y tu rn in g off th e o u tp u t b u t leaving th e carrier oscillator energized as, for exam ple, in voice frequency telegraphy.

In th e voice mechanism, however, th e oscillator is stopped a t the source. T he difference between th e electrical on-off sw itching an d th e sta rt-sto p sw itching of speech is n o t fundam ental b u t results from th e use of th e m ost suitable action in each case in view of th e conditions prevailing.

C A R R I E R N A T U R E OF S P E E C H 501 wave although this is not tru e of the dam ped wave carriers of m ulti­

frequency type once commonly used in spark wave radio telegraphy.

T he carrier wave in speech is not a simple sine wave. Such a sound would be like a whistle and so too lim ited for th e rich flexibility of speech. Instead the voice carrier is a com pound tone having a m ulti­

plicity of com ponents of different frequencies which together cover the audible range fairly com pletely. W hile these com ponents m ay be considered as a m ultiplicity of separate carriers it is simpler to th ink of th e ensemble as a single complex carrier; so this term inology has been used in the earlier carrier illustration and elsewhere in this paper.

Aside from this compound n atu re of the voice carrier, th e voice has two distinct types of carrier, one for voiced and one for unvoiced sounds.

Some sounds such as “z " have both types present a t th e same time b u t this case m ay be treated as th e superposition of one carrier on the other. F or voiced sounds the carrier is the vocal cord tone, an acoustic wave produced by th e vibration of the vocal cords consisting of a fundam ental frequency com ponent and th e upper harmonics thereof.

These decrease in am plitude w ith increasing frequency. For unvoiced sounds the carrier is the breath tone, a complex tone resulting from a constriction formed somewhere in th e vocal tra c t through which the breath is forced turbulently to produce a continuous spectrum of fre­

quency com ponents in th e audible range.

These carrier waves m ust be dissociated from any effects of resonant vocal chambers, for such characterize the speech message rath er than the carrier. F urtherm ore, these carrier waves m ust be m entally pic­

tured as sustained indefinitely w ith the startin g and stopping of them also characterizing th e message wave. Pauses for breath, due to in­

cidental hum an lim itations, do not invalidate the fundam ental theory.

Th e Sp e e c h Me s s a g e

Since a sustained voice carrier has no dynam ic flow of inform ation there is need for a source of message waves and a m odulating m echa­

nism for im printing the message on the carrier. Conversely, any v aria­

tion from th e sustained carrier infers the presence of a message wave molding the carrier. T he message consists of those articulating, phonating and inflecting m otions of the vocal p arts which im print th e inform ation on the carrier sound stream . The im portance of the message waves cannot be stressed too much. Any im pairm ent of them is an im pairm ent of the message.

T he message waves include the motions producing speech changes a t infra-syllabic rates, such as the effect of anger when a talker m ay be high-pitched for m any m inutes. W hen the carrier is thus altered over

a long period of tim e th e question arises w hether to use a long- or sh o rt­

term value of th e carrier. T h e answ er m ay well be th e sam e as in th e analogous radio problem . If w eather causes a carrier frequency to be slightly high all day, this higher value is tak en as th e norm al carrier in studying short-term effects such as th e degree of m odulation.

B u t in long-term studies of carrier stab ility th e deviations from the m ean represent a frequency m odulation which is observed as a “ m es­

sa g e ” effect.

D ue to th e inseparability of th e message wave m otion and its asso­

ciated wave of im pedance change in th e m odulating m echanism there m ay be confusion in distinguishing betw een th e m odulating elem ents and th e source of th e message waves. T h e rule followed here is simple.

From th e stan d p o in t of th e hum an flesh lining th e vocal tra c t, the message source is internal, th e m odulating elem ents, external. T he message consists of those m uscular m otions (or pressures or displace­

m ents) in th e vocal tra c t which are present in th e “ silent t a lk e r ” and are volitional in natu re. T his definition excludes th e oscillatory m otions which m ake up the carrier. T he m odulating elem ents are acoustic in n atu re since th e carrier s ta rts as a sound stream and ends as a m odulated sound stream .

There are three im p o rtan t v ariations of th e voice carrier and so three types of message and of associated m odulation. These v a ria ­ tions are: first, selecting th e carrier; second, settin g th e fundam ental frequency of th e voiced carrier; and th ird , controlling th e selective transm ission of th e vocal tr a c t.7 T h e message waves in th e three cases will be discussed w ith th e corresponding m odulation reserved for consideration under th e next heading.

Selecting the carrier appears as a simple sta rt-sto p message, com ­ plicated som ew hat by th e presence of two types of carrier and by lo­

cating the constriction for th e unvoiced ty pe a t several places in the vocal tra c t. W e m ay th in k of a sta rt-sto p ty p e of message for each point where constrictions are formed, including th e vocal cords for the voiced type of carrier. A constriction message m ay be plotted as the opening between vocal p a rts a t th e constriction w ith critical values for th e onset of audible carrier. T he constrictions are to a certain ex ten t independent. T hus w ith th e vocal cords v ibrating, a constric­

tion from th e tongue tip to the upper teeth m ay also be formed, as in m aking th e “ z ” sound. Again, in whispering, th ere m ay be simul-

7 A fourth message characteristic prescribes th e intensity of th e speech. This message m ay be included in the carrier selection if th e carrier is selected for intensity as well as type. T he m a tte r of intensity is passed over ra th e r lightly here because a comparison is being developed betw een th e hum an and electrical speech synthesizers w ith th e final intensity in th e la tte r under control of an amplifier setting.

C A R R I E R N A T U R E OF S P E E C H 503 taneous constrictions, both of the unvoiced type, one a t the vocal cords and one in th e m outh. As th e voice has two d istinct types of carrier, the vocal cord tone and the breath tone, th e selection sets up one of four carrier conditions a t an y in sta n t: no carrier, vocal cord tone only, breath tone only, or a com bination of vocal cord tone and breath tone. This start-sto p message resembles th e on-off type of telegraph where switching controlled by other m uscular m otions sets up speech inform ation in anoth er code, th a t of telegraph. As m en­

tioned earlier a com m unication system can be m ade w ith the vocal system by startin g and stopping a voice carrier in a vocal im itation of a telegraph buzzer. W hile this would be a clum sy way of com m uni­

cating inform ation it m arks the start-sto p control of the voice carrier as a speech message and no t p a rt of the voice carrier. A nother check is th a t the “ silent ta lk e r” does form such constrictions.

T he second type of message wave specifies th e fundam ental fre­

quency w ith any related voice changes for th e voiced type of carrier.

This message, in a m echanical form, m ay be the tim e variation of the tension of the vocal cords. As the frequency of each upper harm onic is changed in the same ratio as th e fundam ental frequency, a single param eter suffices for all of th e carrier com ponents. T he unvoiced carrier has no message of this type impressed since th e unvoiced sounds are no t characterized by pitch.

T he third and final type of message wave controls the selective transm ission in th e vocal tra c t. By com parison, th e first two types of message are simple, w ith the selecting of carriers ideally changing all com ponents of the carrier by the sam e am plitude factor and th e fundam ental frequency control changing them by a uniform frequency factor. T he vocal transm ission, however, results from a m ulti-reso­

nance condition w ith more th an one degree of freedom. There follows a selective am plitude m odulation w ith some carrier com ponents de­

creasing in am plitude a t th e same in stan t th a t others are increasing.

M axim um transm ission occurs when a com ponent coincides w ith an overall resonance, m inim um transm ission when it coincides w ith an anti-resonance and interm ediate transm ission for other cases. T he voice message for transm ission appears in mechanical form as th e dis­

placem ents of lips, teeth; tongue, etc., w ith as m any such displacem ents considered as are needed for adequately expressing th e speech content.

This infers finding th e sim plest lum ped im pedance stru ctu re equivalent to the d istribu ted im pedance stru ctu re of th e vocal tra c t to th e neces­

sary degree of approxim ation.

All these mechanical displacem ents of vocal p a rts th a t together constitute th e voice message lead to corresponding displacem ents of

air in th e vocal system , resulting in a set of air waves th a t likewise contain all the inform ation of speech. These airborne message waves, however, are a t syllabic ra te s and so below th e frequency range of audibility.

Th e Vo i c e Mo d u l a t o r s

T he three voice m odulators associated w ith th e three speech messages are th e m echanism s of (a) selecting th e carrier, (b) settin g th e fu n d a­

m ental frequency and (c) controlling th e selective transm ission. T he mechanism for sta rtin g and stopping a voice carrier is simple. Assume a sustained carrier of either th e voiced or unvoiced type. I t can be stopped by opening th e constriction a t which it is form ed. T his alters the acoustic im pedance of th e opening which is th en th e m odulating elem ent in this case.

T he m odulating m echanism for controlling th e fundam ental fre­

quency appears in th e vib ratin g portions of air a t th e glottis. T he exact m echanism is of no im portance here so long as th e message wave a t th e vocal cords finds m eans for altering th e fu ndam ental frequency under th e control of the will.8 T his is a case of frequency m odulation of m ultiple carriers harm onically related.

T he m odulating m echanism for controlling th e transm ission through th e vocal tra c t as a function of frequency consists of th e masses and stiffnesses of air cham bers and openings in th e vocal tra c t. These are varied under control of th e message in th e form of m uscular displace­

m ents of vocal tra c t p arts. T here is a m ore com plicated m odulation in the vocal tra c t th an in th e usual electrical circuit for am plitude m odulation because th e varying im pedances are reactive in th e voice mechanism b u t resistive in th e electrical circuit an d also because several independent m odulator elem ents are used in th e voice m echa­

nism as against either a single one or a group functioning as a u n it in th e simple electrical m odulator. T he reactive n atu re of th e vocal im pedances leads to th e selective control of th e am plitudes of th e various harm onics of th e voice carrier. T he am plitude m odulation of each carrier com ponent by th e combined message waves produces an o u tp u t containing th e carrier and sideband frequencies.

Co m p a r i s o n o f Sp e e c h Sy n t h e s i z i n g Ci r c u i t s

T he fundam ental processes in hum an speech production are th u s analogous to those of electrical carrier circuits. T here is a switching of voice carrier energy com parable to th a t in voice frequency te le g ra p h ; 8 For a simplified theory of the larynx vibration see R. L. Wegel, Bell Sys. Tech.

Jour., Vol. 9, p. 207 (1930) and Jour. Acous. Soc. Am er., Vol. 1, Supp. p. 1, April 1930.

T he analogy of th e larynx to a vacuum tube oscillator is described in an ab stract, Jour. Acous. Soc. Am er., Vol. 1, p. 33 (1929).

C A R R I E R N A T U R E OF S P E E C H 505 there is an altering of speech frequencies as in frequency m odulating circuits; and finally, there is an am plitude m odulation to yield a se­

lective transm ission of the various carrier com ponents of the voice.

However, the voice mechanism differs from th e usual carrier circuit m arkedly as regards com plexity. In th e voice mechanism there are two types of carrier each with a m ultiplicity of p artial carrier com­

ponents. T he incoming message has a m ultiple nature. Finally, several m odulations tak e place including both am plitude and frequency types. T his m ultiplicity of carrier relations indicates the wide range of voice phenom ena possible.

Any electrical speech synthesizer m ust be a functional copy of the hum an speech synthesizer in providing th e essential speech character­

istics sketched in the preceding paragraph. There have been devel­

oped two such electrical synthesizers referred to in the introduction.

A brief description of these will be given followed by some circuit comparisons.

These electrical synthesizers are known as the vocoder and the voder. T he vocoder was so nam ed because it handles the speech in a coded form; the voder, because it serves as a Voice Operation DEm on- strato R . Considerable interest has been m anifested a t the public showings of each of these synthesizers, the vocoder in a limited num ber of lecture dem onstrations and the voder a t the San Francisco and New Y ork W o rld ’s Fairs. C ircuit details have been published elsewhere.9 Of these two speech synthesizers the vocoder was constructed first.

I t works on the principle of autom atically rem aking speech under control of spoken speech instantaneously analyzed to derive th e code currents for th e control. T he vocoder as set up for dem onstration is shown in Fig. 4.

The voder was derived from the vocoder by sub stituting m anipula­

tive for autom atic controls. T he resulting voder as displayed a t the New Y ork W orld’s F air is shown in Fig. 5. In the F air dem onstration, repeated continuously a t intervals of abo ut five m inutes, the male announcer gives a simple running discussion of the circuit w ith the girl operator replying to his questions by forming sounds on the voder and connecting them into words and sentences. She does this by m anipulating fourteen keys w ith her fingers, a bar with her left w rist and a pedal w ith her right foot. This requires considerable skill by the operators. T he vocoder, autom atic in nature, presents no problem of operating technique.

9 The vocoder in the Jour. Acous. Soc. Amer., Vol. 11, pp. 169-177, October 1939,

“ Remaking Speech,” Dudley; the voder in the Journal of the Franklin Institute, Vol.

227, pp. 739-764, June 1939, “ A Synthetic Speaker,” Dudley, Riesz and W atkins.

Fig, 4— The vocoder as demonstrated.

C ircuit diagram s supply a sh o rth an d for expressing th e salient fea­

tu res of electrical circuits. In th e next three figures com parative block circuits will be shown for th e hum an and th e two electrical speech synthesizers, tracing th e com m unication from th e origin of an idea in th e com m unicator’s b rain to final expression as speech. In each d r -

C A R R I E R N A T U R E OF S P E E C H 507

Fig. 5— T he voder being dem onstrated a t the New Y ork W orld’s F air.

cuit, th e arrangem ent in Fig. 2 will be followed w ith sufficient detail to show the functional relations of th e p a rts discussed in this paper.

Figure 6 gives a block diagram of th e voice m echanism of Fig. 1 w ith approxim ating electrical circuit symbols. T he same com m unica­

tion p ath s can be traced. T hus from th e ta lk e r’s brain are sent nerve impulses t h a t set up th e message as a set of m uscular displacements containing inform ation as to th e voice carrier to use, th e fundam ental frequency for th e voiced carrier, and th e selective transm ission of the vocal trac t. T he air expelled from the lungs sets up as carriers the

Fig. 6— Block diagram of th e voice mechanism .

b re ath tone for unvoiced and th e vocal cord tone for voiced sounds.

F or sim plicity th e carrier selection is shown a fte r instead of before the carrier generation. These carriers are m odulated by th e m essage w ave to produce the o u tp u t of speech in th e form of the m essage-m odulated carrier in th e audible range of frequencies.

Figures 7 and 8 show sim ilar block schem atics for th e vocoder and th e voder. T he voder circuit has been simplified by th e omission of a few controls for easier operation. In these electrical synthesizers, the carrier is provided by a buzzer-like tone from a relaxation oscillator for th e voiced sounds and from a hiss-like sound from a gas-filled tube for

Fig. 7— Schem atic circuit of the vocoder.

C A R R I E R N A T U R E OF S P E E C H 509 th e unvoiced sounds. In the vocoder, for sim plicity’s sake, one or the other of these energy sources is used according to w hether the sound is voiced or unvoiced, w ith no provision for the mixed types of sounds found in th e hum an voice. T he analyzer of the vocoder derives the

original speech message in term s of a modified set of param eters. This analyzer suppresses the original carrier of the talker and so resembles the dem odulator in radio reception. T he analyzer acts as an electrical ear to tell th e artificial vocal system of the vocoder w hat to say,’the whole vocoder acting as a synthetic mimicker.

The basic sim ilarity of the electrical and hum an speech synthesizers is seen in these figures. In all three cases the message is originated by the brain of the sender of th e speech inform ation. There is in each case a transm ission of control impulses by the ta lk e r’s nervous system to the appropriate muscles. T he muscles produce displacem ents of body p arts form ulating the speech inform ation as a set of mechanical waves. These waves appear in th e vocal tra c t in the case of normal speech; in the fingers, w rist and foot in the case of the voder, b u t in the case of the vocoder use is made of electrical currents derived from and equivalent to the vocal tra c t displacem ents in ordinary speech.

In each case the message contains the speech inform ation in syllabic waves. In all cases the message waves control the choice of carrier, the fundam ental frequency of the voiced type carrier and the spectrum of power distribution in the speech ou tp u t. Differences arise in the details ra th e r th an in the principles.

TIME IN SECONDS

o

Fig. 9—Oscillogramofthe sound “sa.

C A R R I E R N A T U R E OF S P E E C H 511 Sp e e c h Ch a r a c t e r is t ic s f r o m t h e Ca r r i e r Po i n t o f Vi e w

Now th a t the mechanism of speech has been described in carrier term s it is of interest to observe carrier features as they m anifest th em ­ selves in the characteristics of speech. Some of these can be seen by the eye in speech oscillograms. Some can be dem onstrated to the ear w ith a speech synthesizer such as the vocoder.

For a visual illustration there is shown in Fig. 9 a high quality oscillogram taken from C randall 10 of the sound “ s a ” (P late No. 160.

Spoken by M. B.) for a m edium -pitched male talker. T he carrier shown by the oscillogram is of the unvoiced type for the earlier and of the voiced type for the later p art. As one looks a t the oscillogram he sees a great mass of the high-frequency com ponents of the carrier.

Scrutiny, however, reveals m odulated on the carrier the message in­

form ation in term s of switched energy sources, controlled fundam ental frequency and varied transm ission characteristic. Shortly after .17 second the switching off of the unvoiced carrier begins. R em nants of the unvoiced carrier can be seen in the voice period ju s t before .19 second and the one starting a t ab o u t .19 second. T he switching on of the voiced carrier appears ju st after .18 second and seems to be rea­

sonably well completed a t the end of the second voice period ju st before .20 second. This switching was not instantaneous. However, the ear probably does not observe the duration tim e of the switching.

The fundam ental frequency falls rapidly a t the beginning followed by a leveling out and then a final slight fall in the last few periods. It sta rts a t 140 cycles per second, dropping to around 110 in the level portion, and then to 101 a t the end. T he resonance conditions cannot be followed too well by eye. However, around .20 second there is a m ajor lower-frequency resonance of ab o u t 800 cycles. A t .33 second this resonance appears to have increased to 1100 cycles or so. A sim ilar alteration of resonance conditions m ay be observed if the little shoulder on the rear side of the peak ju st in front of th e .25 second m ark is traced in adjacent periods. I t can readily be followed back to the third period ju st before .20 second and can still be seen in the last dis­

tin ct voicing period starting before .39 second. T he dynam ic variation of the speech a t syllabic rates in accordance with the message content is th u s revealed.

F or another visual illustration of th e speech message Fig. 10 shows a set of oscillograms 11 from the vocoder analyzer for the words “ She saw M ary .” The oscillogram of the in p u t speech is the trace next to

10 Bell Sys. Tech. Jour., Vol. 4, p. 586, 1925.

11 This figure is a copy of Fig. 3 in th e paper “ The A utom atic Synthesis of Speech,”

Dudley, Proc. Nat. Acad. Sci., Vol. 25, pp. 377-383, July 1939.

SPECTRUM-DEFIN ING CURRENTS FOR INDICATED FREQUENCY BANDS

5 3

TIME IN SECONDS Fig.10—Codedspeech(derivedmessage currents) for the words “She saw Mary.”

C A R R I E R N A T U R E OF S P E E C H 513 th e bottom . T he trace below shows the defining current for the funda­

m ental frequency, while the ten traces above show currents indicating th e rectified power in ten frequency bands of 300 cycles w idth except th a t the lowest one extends from 0 to 250 cycles. T he slow rates of change are noted in the message currents when com pared to the original speech wave.

D em onstrations of the vocoder indicate to the ear the carrier n ature of speech. T hus th e carrier used for rem aking speech, w hether a m onotone or a hiss sound, is observed to have no intelligibility when heard alone. T he message currents derived from spoken speech are not audible. However, intelligible “ speech ” is produced by the m odu­

lation of either type of carrier by the message currents of selective transm ission. Similarly, there can be used for the carrier a wide v ariety of sound from th e puffs of a locomotive to instrum ental music.

Upon im print of the transm ission message currents from spoken speech, new forms of odd sounding b u t nevertheless intelligible “ speech” are produced.

T he carrier conception of speech reveals w hat is im p o rtan t and not im p o rtan t in evaluating speech characteristics. An example of in­

terest is the m a tte r of phase. I t has long been known th a t phase was u n im p o rtan t to the ear a t reasonably low listening levels. From th e carrier point of view this is natural, for the phase changes referred to are those in the carrier and so, u nim portant. W hen the phases of the message com ponents are altered, there is a very noticeable effect on the ear, for phonetic units are now being shifted.

T he g reat advance in recent years in th e application of carrier circuits has been guided b y m athem atical theory. Since in electrical speech synthesizers the carrier and message currents are separated physically, it is possible to use carrier equations expressing the m odu­

lation phenomenon. Similar equations m ay be w ritten for th e voice mechanism as represented by Fig. 6. T his has been done in the a t ­ tached appendix, thus separating speech into syllabic and carrier factors.

A P P E N D IX M a t h e m a t i c a l R e l a t i o n s

T he speech concepts developed in the body of the paper m ay be expressed in m athem atical term s which no t only give th e fundam ental relations in sim plest form b u t also aid in the application of the well- established carrier technique to speech. F or voiced sounds, periodic

by n atu re, th e carrier Cv m ay be w ritten as a function of th e tim e t thu s:

n

Cv = E A k COS (kP t + 9k). (1)

4=1

H ere Cv is composed of n audible harm onics of relatively high fre­

quencies w ith th e kth. of am plitude A k, frequency k P radians per second, and phase dk- T h e choice of fun dam en tal frequency P is som ew hat a rb itra ry b u t m ay well represent th e average of the talk er over th e period of interest.

By m odulation processes, there is molded on to this carrier th e to tal message inform ation a t th e relatively low syllabic frequencies. T h e message is divided into three p a rts: (a) th e s ta rtin g and stopping of the carrier; (b) the in stantaneous fundam ental frequency; an d (c) th e selective transm ission through the reson ant vocal tra c t.12 T hese three message functions as they m anifest them selves in varying th e carrier will be represented b y s, p , an d r, respectively. E q u atio n (1) will be modified to indicate th e effect on th e carrier of each of these m odula­

tions separately, after which th e equation will be rew ritten to show th e effect of all three acting sim ultaneously.

T h e effect of startin g an d stopping th e carrier is described m ath e­

m atically as a function of tim e by m ultiplying Cv b y th e switching function s(t), giving:

n

Switched Cv = s[t) E Ak cos (kP t + dk). (2)

4=1

F or sim ple on-off switching, s(t) altern ately equals zero an d un ity, although it m ay in general represent m ore gradual changes or even an y variations of inten sity over th e frequency range.

T he instantaneous fundam ental frequency is obtained by m ultiply­

ing P by th e inflecting factor p (t). T h e effect of th e frequency m odu­

lation 13 is represented by su b stitu tin g for P t th e in teg rated q u a n tity

P p(t)dt = P I p(t)dt.

J o

W riting th is value for P t in equation (1) gives th e inflected carrier w ave:

Inflected Cv =

E

i

f

1

4 = 1 L Jo J

12 As in th e body of the paper, the effect of phase m odulation is neglected here.

13 “ Variable Frequency E lectric C ircuit T heory w ith A pplication to th e T heory of Frequency M odulation,” J. R. Carson and T . C. F ry , Bell Sys. Tech. Jour., Vol. 16, p. 513 (1937).

C A R R I E R N A T U R E OF S P E E C H 515 T he effect of the selective transm ission is allowed for by m ultiplying C* by th e tran sm itting factor r(u, t), indicating th a t the tran sm itting factor is a function of frequency a t any instan t. Applying this factor in equation (1) gives:

n

T ransm itted Cv = Yi r(w, t)A k cos (kP t + 0k). (4) i=i

T he r factor is placed inside the sum m ation to indicate th a t as k changes th e different frequencies have different values of the m ultiply­

ing factor r. If a m ultiplicity of carrier waves is assumed, th e tra n s­

m itting factor would be rk(t), individual to the kth com ponent.

In norm al voiced speech, Sv, these three m odulations are all present sim ultaneously, so th a t:

Sv = s(t) r(u, t)A k cos I kP f p{t)dt + 6k 1 . (5)

k=i L >'o J

Equation (5) shows how the message in the form of the s, r, and p functions has im printed its characteristics on th e original carrier Cv of equation (1).

T he derivation of (5) was for voiced speech. Unvoiced speech, however, is also covered by (5) as a degenerate case. Nevertheless, furth er inform ation is presented by w riting o u t th e unvoiced carrier separately. F or unvoiced speech, th e frequency P approaches zero and the num ber of term s, n, approaches infinity, giving an integral instead of a finite sum of com ponents in equations (1) and (5). T he unvoiced carrier Cu is th e n :

Cu — I A (u ) cos [w/ + (1')

and the unvoiced speech:

P*(jJ

S u = s(t) I r(co, t)A (a;) cos [cd + 0(<y)]dw (5')

c/OJj

w ith the continuously variable frequency w (radians per second) v ary ­ ing over the audible range of energy contribution from wi to w2 and the unvoiced carrier spectrum defined by am plitude A («) and phase 0(y). T he unvoiced speech has no inflecting factor b u t does have switching and transm itting factors to m ake up the message impressed on the carrier.

By G. K. BURNS

Q u a rtz c r y s ta l filte rs u sed in m o d e rn c a rr ie r s y s te m s p r e s e n t new p ro b le m s in m a n u f a c tu r in g te c h n iq u e . In th e a s s e m b ly a n d t e s t ­ in g of th e filte rs a n d in th e p r o d u c tio n of c o m p o n e n t c ry s ta ls , coils a n d c o n d e n se rs, sp ecial fa c to r y fa c ilitie s a re r e q u ire d fo r a c ­ c u r a te m e a s u re m e n t of fre q u e n c y a n d c o n tro l of a tm o s p h e r ic c o n d i­

tio n s . T h e m a n u f a c tu r e of q u a r tz c ry s ta l p la te s in p a r t ic u l a r c o m b in e s se v e ra l fields of a p p lie d science, in c lu d in g c ry s ta llo g ra p h y , p re c isio n g rin d in g , v a c u u m te c h n iq u e a n d h ig h fr e q u e n c y e le c tric a l m e a s u re m e n t. I n d u c ta n c e coils a n d fixed a n d v a ria b le c o n d e n s e rs fo r u se in c ry s ta l filte rs m u s t c o n s is te n tly m e e t a d v a n c e d re q u ir e ­ m e n ts , e sp e c ia lly in re g a r d to s ta b ility . T h e a s s e m b ly of th e s e c o m p o n e n ts in to filte rs re se m b le s th e m a n u f a c tu r e of ra d io r e ­ c eiv ers, d iffe rin g m a in ly b e c a u se of s m a lle r q u a n ti t y re q u ire m e n ts . T e s tin g e q u ip m e n t m u s t p e r m it ra p id s h o p a d ju s tm e n t a n d t e s t of th e c o m p le te d filte rs w ith la b o r a to r y p re c isio n .

In t r o d u c t i o n

E

electricity; th a t is, an electrical voltage applied to th e term inals of a crystal causes a m echanical distortion of th e quartz, and vice versa.

Because of this interrelation a p late of qu artz, a t frequencies near its m echanical resonance, behaves electrically like th e coil and condenser com bination shown in Fig. 1. T he series inductance and capacitance

Fig. 1—E quivalent circuit of a q uartz crystal plate. E lem ents L , C\ and R are associated w ith th e piezo-electric property and m echanical resonance of th e crystal, while Co represents capacitance betw een th e electrodes.

represent th e m ass and elasticity of th e plate, respectively, while the sh u n t condenser represents th e capacitance betw een faces of th e crystal. T he dam ping of such a p late m ay be m ade v ery low, giving a ratio of reactance to resistance (com m only term ed Q) of 15,000 or

1 N um bered references are listed at end of paper.

516

M A N U F A C T U R E OF Q U A R T Z C R Y S T A L F I L T E R S 517 higher, as com pared w ith a practical lim it of 300 for coils. S tability of resonance frequency and com pactness of dimensions are two fu rther respects in which quartz crystals surpass th e best coils and condensers available.

Filters designed to utilize these properties generally consist of one or more crystal plates, plus such condensers, inductance coils and resistances as m ay be required to give th e desired overall performance.

T he principal types used in the Bell System operate a t frequencies ranging from 40 to 600 kilocycles and tran sm it bands varying from 5 cycles to 6 kilocycles in w idth. Physical dimensions range up to 3 X 5 X 1 6 inches.

Unusual m anufacturing requirem ents are imposed by th e n atu re of these filters and of th e system s in which th ey are used. A djusting tolerances and stability requirem ents, for example, range from ± 20 to ± 200 p arts per million on crystals and on coil-and-condenser circuits used in crystal filters. Transm ission losses m ust be m easured to accuracies of the order of ± .03 db a t 100 KC. T o insure stability of ad ju stm en t during service life, com ponent ap p aratu s m u st be protected against d u st and excessive hum idity. M ethods of assembly and testing m ust be adaptable to a variety of types of filters, one of which, the channel filter,4 is m anufactured by the W estern Electric Com pany in quantities of 1500 to 5000 per year, while th e others range from 10 to 1000 per year. Long service life m ust be assured by proper choice of m aterials and technique.

In order to satisfy such requirem ents special m anufacturing pro­

cedures are necessary. In reviewing these features it will be convenient to consider first those m ethods or facilities which are used in several or all stages of th e m anufacture of crystal filters, second the m ethods employed in producing com ponent ap p a ratu s for such filters—p articu­

larly crystals, coils and condensers— and finally th e technique of assembling and testing the com plete filters.

Ge n e r a l Fa c i l i t i e s

A prim ary requisite in the adjusting and testing of both crystals and crystal filters is the precise m easurem ent of frequency. T he equipm ent used for this purpose includes a stand ard frequency gener­

ato r containing a 100 KC crystal oscillator. T his generator norm ally m aintains a frequency accuracy of ab o u t 1 p a rt in 2,500,000 operating under th e control of the Bell System m aster frequency stan dard in New York, b u t will rem ain accurate w ithin 1 p a rt in 1,000,000 even though the m aster signal is interru p ted for as much as 24 hours.

T hree sub-harm onics of 100 KC, nam ely, 100 c.p.s., 1000 c.p.s. and

10,000 c.p.s., are distrib u ted to all te s t positions. Oscillators supplying the individual te st sets are provided w ith cathode ra y oscilloscopes, by m eans of which th ey can be synchronized w ith any m ultiple of th e three stan d ard frequencies. T o set up an odd frequency n o t coinciding w ith any m ultiple it is necessary to in terpolate dial readings between two synchronized points.

C ontrol of atm ospheric conditions also plays an im p o rta n t p a r t in th e m anufacture of crystal filters. T h e tem p eratu re coefficient of frequency of th e crystals m ost com m only used is ab o u t 15 p a rts per million per degree F ahrenheit. For some filters, in order to secure uniform perform ance th roug hout th e tem p eratu re and frequency ranges encountered in service, these crystals m u st be ad ju sted w ithin tolerances as small as 40 p a rts per million. F lu ctu atio n s of as little as 2° F ., in such cases, m ust be tak en into account during th e a d ju s t­

m ent of th e crystals. In addition, crystals, coils and condensers are all sensitive to th e effects of excessive h um id ity . T o m inim ize such difficulties, the assembly and testing of these com ponents and of the filters in which they are used are carried o u t in air conditioned room s controlled a t 75° ± 2° F. and approxim ately 40 per cent relative hum idity.

Cr y s t a l s

Of the several com ponent p a rts used in crystal filters, th e first to be considered in detail are logically the q u artz crystals them selves. T heir properties of low loss and high stab ility are prim arily responsible for th e unusual perform ance of filters in which th ey are em ployed.

N atu ral deposits in th e ea rth co n stitu te th e sole source of supply of qu artz crystals, since no practical m ethod of producing them sy n ­ thetically has been developed. “ R aw ” crystals suitable for use in filter m anufacture m u st be unusually large and free from flaws. T he principal source is Brazil, th e bulk of the q u artz being b ro u g h t in by nativ e prospectors to trad in g posts and shipped to th is co u n try via Rio de Janerio and other coastal cities. T h e cry stals usually range between 3 and 10 pounds in weight, w ith occasional pieces reaching 100 pounds.

T h e raw q u artz passes through successive stages of inspection and selection, commencing a t th e trad in g p o st and culm inating in careful exam inations before and during th e cu ttin g operations. A concen­

tra te d beam of light from an arc lam p (see Fig. 2) is used in locating internal flaws, which generally ap pear as sm all bubbles an d inclusions of foreign m atter. Q uartz takes tw o d istin c t forms, left-hand and right-hand, having opposite piezo-electric polarities. P ortions of raw crystals containing both forms are n o t usable. T his condition, called

M A N U F A C T U R E OF Q U A R T Z C R Y S T A L F I L T E R S 519

Fig. 2— Inspection of quartz crystals. An arc light beam aids in th e detection of internal flaws.

“ tw inning,” appears as shown in Fig. 3 when observed w ith polarized light.

F or use in filters, quartz m ust be cu t into rectangular plates properly oriented w ith respect to th e electrical, mechanical and optical axes of the crystal, as shown in Fig. 4. A polariscope and an X -ray spectroscope are used in locating these axes to an accuracy of ± 0.25 degree. F o r the m ajority of applications th e p late is cu t in th e plane of th e m e­

chanical and optical axes, w ith th e long dimension set a t an angle of 18.5° from the m echanical axis. T his orientation elim inates secondary resonances in th e completed crystal and m akes th e prim ary resonance frequency relatively independent of slight errors in orientation. F o r applications requiring a low coefficient of resonance frequency versus

Fig. 3— R ight and left-hand tw inning in q u artz as seen b y polarized light.

tem perature, plates are cu t w ith their long dim ension 5° from the m echanical axis. Tolerances in cu ttin g and grinding to thickness, length and w idth prior to calibrating are of the order of .001 m m., requiring th e use of technique sim ilar to t h a t em ployed in th e m an u ­ facture of gage blocks. A few stan d ard thicknesses, ranging from .020 to .060 inch, are used for m ost crystal plates. L engths v a ry from 0.5 to 2.0 inches while w idths range from 0.15 to 1.5 inches. Because of unavoidable w aste in th e cu ttin g and grinding operations an d th e rejection of q u artz containing flaws, only a sm all portion of th e m aterial entering th e cu ttin g room finds its w ay into finished plates.

U p to this p oint th e cu ttin g and grinding are purely m echanical operations, directed tow ard securing prescribed physical dim ensions.

D uring final ad ju stm en t and in service, however, th e cry stal p late m ust be connected as an electrical elem ent. E lectrodes are p ro ­ vided by coating th e m ajor surfaces of th e plate w ith alum inum , using a process of evaporation and condensation in a vacuum , sim ilar to th a t employed in th e silvering of telescope m irrors. If th e p late is to be used in a balanced filter section which requires a p air of crystal elem ents of th e sam e frequency, as is frequen tly th e case, th e p latin g on each face is then divided in half along th e longitudinal axis. T his division, one-hundredth of an inch wide, m u st have a d.c. insulation resistance of a t least 100 megohms to insure proper operation in some types of crystal filters.

M A N U F A C T U R E OF Q U A R T Z C R Y S T A L F I L T E R S 521 P relim inary tuning is accomplished with a fixture, sim ulating the final holder, which grips the plate a t the center by four con tact points, one on either side of the division in the plating of each face. These

2

A X IS

Fig. 4— O rientation of a typical quartz plate with respect to its electrical, mechanical and optical axes.

contacts introduce very little dam ping, since the mode of vibration norm ally em ployed is longitudinal, with maxim um am plitude a t the ends of the plate and a node a t th e center. T he te st set-up norm ally used consists of two oscillators w ith a m eter arranged to read the

difference betw een their frequencies. One oscillator is controlled by th e crystal plate being tu n ed and operates a t its resonance frequency;

th e o ther is controlled by a sta n d a rd crystal of th e desired frequency.

S ta rtin g 100 to 200 cycles low, th e p late is ground on the ends un til its frequency approaches th a t of th e stan dard.

T h e plate is then transferred from th e fixture to its final holder, shown in Fig. 5. T his m ounting norm ally accom m odates two p lates

Fig. 5— C rystal plates m ounted in holder. T he four points a t th e center of each plate provide electrical contact and mechanical support.

of different frequencies, each supported a t its nodal p o in t b y contacts projecting from ceram ic blocks. T he entire assem bly is held tog eth er by a spring suspension in order to apply uniform pressure a t all con­

tacts. T o minimize dam ping, the con tacts m u st be accu rately aligned and the q u artz plates m u st be carefully centered upon them .

A final ad ju stm en t of frequency is now perform ed, as shown in Fig. 6. Perm issible tolerances v ary from ± 2 0 to ± 1 5 0 p a rts per million for different types of crystals. C rystals having th e bro ad er tolerances and substantial q u a n tity requirem ents are ad ju sted by com parison w ith a stan d ard crystal, as in th e case of prelim inary tuning. T h e te st set shown a t th e left in Fig. 6 is being used for this purpose. T he upper and lower panels are th e oscillators controlled by

M A N U F A C T U R E OF Q U A R T Z C R Y S T A L F I L T E R S 523 th e stan d ard and the test crystals, respectively, while the center panel indicates th e frequency difference between them . F or very accurate work and for periodic checks of the standard crystals it is necessary to use a precision oscillator, shown a t th e right.

Fig. 6—Final tuning of a crystal plate using a standard crystal (in small box a t upper left) for comparison.

Occasionally, in the course of ad justm ent, plates are carried too high in frequency. In such instances, as a result of the standardization of thicknesses m entioned previously, the plate norm ally can be salvaged by grinding it to th e dimensions of th e next higher frequency p late of the same thickness.

Aging occurs in both resonance frequency and effective resistance, as slight strains created in the quartz and in th e contacts during ad ju stm en t relieve themselves. T he greater p a rt of th e aging takes place during the first few hours after calibration and nearly all of it during the first week. In general, th e frequency rises a few cycles and th e resistance drops slightly. C rystals on which the frequency toler­

ance is approxim ately equal to the shift due to aging are stabilized by one or more tem perature cycles, prior to final m easurem ent of frequency and resistance.