66

A New Stereophonic Amplifier*

KORkIAN H. CROWHURST?!
Summary-A central feature of the new designof a stereo amplifier is an output transformer with original features that makes possible reduced cost and improved performance at the same time. This paper discusses a varied possibility of design objectives for a stereo system, and explains the in which the new output transway former functions. By variation in its method of use, or in choice of parameters,a whole range of amplifiers can apply advantages in different proportions or degrees. The basic design of an output transformer, which is essentially inexpensive to make,provides for separationbetween"left"and "right" a s well ascrossover,and combining networksformixed lows, if desired, without additional external circuits. It makes possible a newtype of tone control, achievinghigh performance economically, using feedback, and/or improved matching between amplifier and loudspeakers over the entire frequency range as well as better separation and efficiency than the single-ended and push-pull transformer matrix cangive. One particular amplifier is discussed in detail, while a more generaldiscussionshows possibleapplicationto morediversedesign objectives.
SESIOR NEMBER, IRE

each channel has onl>- flow through one primarywindto ing and olle secondaryn-inding,wherethetmo-transformer matrix uses two n-indings in series for both primary and secondary (Fig. I ) . aWso the maintenance of good separation is less dependent on precise control of the number of turns in the .\-arious windings.

LEFT CHANNEL OUTPUT

si

DESIGSOI~JECTIVES

HILE theprimarJ-objecti.vebehindthisdevelopment was econom\,-nlal<ing i t possible t o achieve stereo of reasonable quality a t considerablq- reduced cost-it was also agreed that realistic standards of qualit\- must be met; perhaps it would be possible to achieve superior quality at lower cost. This can often happen, where a simplified, more logical approach replaces older, an more complex one. I th a s happened in the group sJ-stems to be described in this of paper. IVhile there is one central feature that characterizeseachamplifier of thegroupdeveloped,its application is so flexible t h a t a n-hole range of amplifiers has been designed to suit the entire range needs of a phonograph line. The most expensive part of any audio amplifier has always been the output transformer. Some economy was effected in an earlier development1 that used a circuit configurationsimilartosingle-channelpush-pull,but carried one stereo channel i n each side of the push and pull. This also effected some economy in output transformer b\- utilizing a closed-cored, high-efficiency unit for the push-pull,monophonic or element, Ivith a gapped, lower-efficiency unitforthesingle-ended,or stereoelement. T h e new development carries econonly this much further. Efficient!. is improved because the current for

Fig. ]-The t\\-o-transforn1er matrix limits the economy that can be achieved because the signal current path for each channel flon.s through a Jvinding on both transforlners in series, both primary qnd secondary; current path for one channel is shown by heavy hes.

REQUKIXENTS
111 deciding what is acceptable performance, the question of separation and 1 1 0 ~ well i t is maintained at extreme frequencies, low and high, must be considerctl.2 Thecharacter of separation is also important:most measurements of separation do not determine what the signal is t h a t leaks from one channel to the other; i t is assumed to be the left program leaking into the right channel, or vice versa; what is more important is the leakage of distortion components from one channel to the other. T o illustrate this difference: if the leakage between channels is pure program, some 1 2 t o 15 db is probably adequate for almost all purposes : improvement beyond this would not noticeabl\r improve the stereo effect. set a minimum of 20 d b a s a target to insure a good margin. But if the leakage consists of distortion components,then20-dbcrosstalkrepresents 10 percent distortion! Separation, at the low irequenciesparticularl).,resolves into two kinds from the practical program viewpoint. The test method usuall\- employed assumes the signal is present in only one channel, and must not be allowed to leak into the other. T h e practical program, except type has the that been "doctored for super

* Receivedbythe I'G.A, January 20, 1961; revised nmnuscript received, IIarch 27. 1961. f. CBS Labs., Stamford, Conn. B. B. Bailer, et a / . , "-1 two-n.ay stereophonic amplifier," A u d i o , vol. 42,pp. 19-20, 92; October, 1958.

* iY, H. Beaubien and H.B. AIoore, "Perception of the ,Lereophonic effect as a function of frequency,'' J . .4zldio Engyg. Soc.. voi. 8 , pp. 76-86; .lpril, 1960: also in I R E TR.ASS. OY .\I:DIO, vol. -1L.4, pp. 111-153; September-October, 1960.

1961

Crowhurst: A hew Stereophonic Amplifier

67

linking the with coil [Fig. 2(a)]. An iron-cored instereo, has some of the signal present in both left and of right channels. Aural separation is achieved by differ- ductance occupies this path with magnetic material high permeability [Fig. 2(b) This greatly increases the ences in the intensity and phase with which the signal correspondingtodifferentprogramsourcesorinstruQ of the inductor, but also introduces a nonlinear comments is contained in the two channels. ponent, due to the nonlinear relationship between and B Theconventionalmethod of testing for separation H i n the magnetic core. merely determines b,, how much a system will degrade Leakage inductance is usually defined as a measure of the proper amountof intensity separation. For example, the unshared magnetic flux due to imperfect coupling if a piece of program has the intensity in one channel of betweenthe coils and is quantitatively dependent on 14 db above thatn the other, and the system separ the primary flux t h a t does not link the secondary, and i a t Ion vice versa. While this is a correct definition, the concept is 20 d b : t h e leakage represented by signal content is l,/Sth, and that added by the system (assumed in phase) i t as imperfectcouplingleadstothenotionthat of is l/lOth, making a total 3 1 1 0 t h ~so the 20-db system leakage flux is a small percentage of the main flux, and of ; separation may degrade a program separation of 14 d b possesses thesamenonlinearitiesthemain flux does to only 10 db. which is not true. But program can differ in timing, or phase, as well as in intensity, between the two channels. In a normally \ recorded program, both differences contribute a t lower \ I frequencies because intensity sound throughout a I / studio mill not vaq- appreciably at these frequencies, / and timing can only change by a fraction of a period from point to point. T h e effect of incorrect timing can well be illustrated by reversing the phase one channelat the loudspeaker. of MAIN BELD TO SUSTAIN VOLTAGE TRANSFER On some programs, all sense of stereo location is lost, CQRE butanexaggeratedsense of spaciousnessremains. In LEAKAGE ffELD DUE TO CURRENT IN Born W I N D I ~ others, where a different recording technique was used, there may be little difference. But, however the program is recorded, the reversing phase of one speaker always givesanimpression of bassdeficiency. T h e only case Fig. 2-Comparison of properties of (a) air cored, (b) iron cored a!ltl (c) leakage inductances, for use in filter design; the leakage 111where this may not occur is in the super-stereo type ductance arrangement is shown i n part section with the paths of program, if all the bass is in one channel. This indicates main and leakagefield indicated. t h a t i n most program material, both correct phase or time relationship intensit)and differences imare An alternative to way define leakage inductance portant at these frequencies. avoids this difficulty. All the flux that stays in the core Of course,couplingbetweenchannels will degrade couples both primary and secondary in most designs. phase as well as intensity difference, but in a different (In special designs magnetic material may be inserted way. The distinction is mentioned here because i t can intheleakage flux path,butthese will not beconexplain some of the discrepancies between experimental sideredinthepresentpaper.) T h e leakage flux is i n tests conducted to determine the need for maintaining duced to pass between the coils [Fig. 2(c) ] by the comseparation at the extreme low and high frequencies. bined current in both windings (principally due to load current drawn from the secondary and reflected to the THEOCTPUT TRANSFORMER primary) and is responsible for a voltage difference t h a t X new departure in output transformer design forms is added vectorially to the voltage that would beinthe central feature of the newline of amplifiers,with duced in the windings by the main flux. So leakage indifferent aspects of its flexible range of attributes ductance is the inductance between two coils occupying utilizedfordifferentapplications. T o understandthe the same core due t o a magnetic path between them, functioning of this transformer, i t will be necessary to which path,usually, is whollyinair.Leakage flux is explain the properties of the quantity called leakage in- only suchwhen i t leaves the surrounding core. From ductance, and we can illustrate the development of the this understanding,theproperties of leakageinductnew unit i n terms of earlier applications. ance may be seen to differ from conventional inducta.nce i n two respects. LEAKAGE INDUCTANCE Although i t has the properties of an inductance, i t

1.

PPIYAW

1)

For filter design, leakage inductance combines properties of iron-cored and air-cored inductances. An aircored inductance uses a very long magnetic path in air,

provides circuit isolation because two coils are involved, which may or may not be connected externally. The inductance is defined i n terms of the

68
signyielded a n efficient andcompact lo\v-pass f ~ l t e r , using theparametersgenerally know-n as nz derived." Later,thesameprinciple was extendedtoproduce a transfornler n-ith built-in crossover, for feeding low- and high-frequenqloudspeakers their with respective ranges of frequency.4 This use was the first to demonstrate a way of achieving high-pass as well as low-pass action through the use of leakage inductance (Fig. 4). Fig. 4(a) represents a condition where the capacitor in shunt with the output winding operates in conjunction with the leakage inductance, which is effectively in series with the the transmission path to provide lowpass action. In Fig. 4(b), the position of L and C are transposed. This is achievedby using windings with the same numbers of turns, phased so the output voltage The magnitude of leakage inductance can be varied in can be the drop across the leakage inductance between severalways. 111 consideringitsmagnitude,reference the two windings. must be made to a specific winding. The same leakage IT'hen stereofirstbecamepopular,the use of tn-o inductance will have two distinct values,if the -\lindings windingassembliesonthesamecore as a means of between which i t appears do not have the same number econom). was e11visaged.j In this case, the leakage inof turns.Leakageinductanceisproportionaltothe ductance between windings disposed on the two limbs square of the number of turns in the reference winding of a core-type construction n-ould provide potential and is governed also b\- the geometry of the windings. separation between channels, and by correct phasing of Increasing turn length, the spacing between ~vil~dings, thesignalinthetwochannels,single-endedoutputs or dimension the of thewindingstransversetothe could be used in such a way that the core loop has 110 leakage path, will increase lealage illductance [Fig. resultantpolarized(dc)magnetization. 3(a) 1. Increasing the dimensions parallel to the leakage of path, or multipl~-ing the number winding sections by division made parallel with the leakage path [Fig. \T pp: 129-134; - ~ p r i l 1949. inductance," Electronic Engrg., 3(b) vel, H Crowhurst, "Leakage , and (c)] will leal<age reduce inductance. N. H. Crowhurst ('Tannov, Ltd.), "Improvements to Relating A l l earlJ7 Llse of leakage ~ l l d L l c ~ a l l c e audio filter deior Electrical Transformers," Brit.' Patent No. 734,346; July, 1955.
N. H. Crowhurst,"Stereophonic Sound,'' john F. Rider 1'\1l>lisher, Inc., New York, S . Y., 1st ed.;1957.

voltage induced i n both coils, referred to the turns in one of them, by the rate of current change in both coils. 2) I t has a Q considerab1)- superior to an air-cored coil because the magnetic path length, althoughi n air, is considerabl). reduced. AAt the same time, i t does not, in itself, have the distortion-generating property of a n iron-cored inductor. The leakage inductance is entirely due to magnetic pathin ai?. The part of themagneticpathwithinthecore material is not active in the leakage inductance element, but is part of themagnetizingcurrent characteristic of the transformer of which leakage inductance is another element.

ur

T
EQUIVALENT

/NDUCTANCE

LEAKAGE

CIRCUIT

EQUIVALENT

CIRCUIT

(3)

(b)

Fig. 3-Dependence of leakageinductance on the geonlett-J. of the transformer into which it is built: ( a ) dimensionsshownhere will eachincreaseleakageinductance i n proportion tothemselves; (b) inverseproportion on thisdimension; (c) the firstsimple leakage step in sectionalizing to reduce inductance.

Fig. 4-Connections with double-~-ound transformer to utilizeleakage inductance in (a) low-pass and (b) high-passfilter. Xote that in (a) no electricalconnectionisnecessarybetween input and output, while in (b) suchconnection is necessary and theturns i n the windings must be equal.

1961

Amplijfer Stereophonic A New Crowhurst: THENEW DEVELOPXENT

69

the leakage inductance between each main winding and I n a sense, the basic element of the new design cotn- the third winding, serves to couple the left and right bines the features represented i n t h e l i t e r a t ~ r e . ~ ~it B uspeakers directly to their respective windings. t Immediately above crossover, to up a frequency does more than this; it gains some advantages due to where the leakage inductance between the left and right the particular method of conlbination that do not perwindingsisdirectlyadequatetoprovideseparation, tain to either feature individually. the presence of the third winding also serves to improve T h e basicconfigurationforthe new transformeris separation. The sum signal appearing across i t is split shown Fig. in 5. Theoutputs of a push-pull stereo i n two and subtracted from the partially-coupled signals amplifier are accommodated on symmetrically disposed appearing across the main windings. The result is t h a t windings of a core-type transformer. \h7ithout the third eachoutputismuchmorefullJ-separatedfromthe winding, which incorporates the crossover function, the other(Fig. 6). separation progressively improves frequency as rises I t should be noted that the main windings are not, in due to the rising impedance of the leakage inductance themselves, treble windings, although when separate between the two winding assemblies. common bass is used, their nlain function is to supply Basically, the third winding provides low-pass filter load current a t treble frequencies. But the voltage across action between a mixture from both the main windings, is tightly coupled to its which can be used to feed a common-bass unit. A ca- each main-minding secondary respective primary winding. By grounding one end of pacitor connected across this third winding completes a eachmain-windingsecondary,andthecentertap of two-element low-pass filter action. A full equivalent the third winding, the filter action, both low pass and circuit of thisjointleakageinductanceactioncannot high pass, can be achieved, while the main windings give convenientlybedrawnwithconventionalcircuitelefull-range voltage for feedback purposes. ments, cansimple nor a equivalent circuit conbe fiotice that the entire load current for the commonstructed. leakage The inductance measured as from bass output is deliveredby- the third winding. Except in either main minding to the third winding, in conjuncthe region just above crossover, the current for left and tion with the leakage inductance as measured between right is delivered directly from the main windings, althe two main windings, do not conform to either loop or thoughthethirdwindingisinseries.Thecapacitor meshconfiguration of a three-terminal network. across i t bypasses the higher-frequency currents from Below the crossover frequency, which is determined the main windings, so voltage drop due to third-winding by the referred leakage inductance value for the turns resistance is avoided. The od>- currents that cause voltused i n conjunction the with capacitor, chosen to belowcrosssynthesize a filter having constant resistance properties, age drop in the third winding are those over, and this is then the active winding, feeding the most of the energy is tightly coupled to the third windefficiency canbemaining. Above crossover, progressively less of the energy is common-bassspeaker.Thus, tained easily with two-transformer more than the coupled to this winding. But what there is, is strictly matrix. sum or monophonic signal. T h e circuit is moreefficient,as well as costingless Thesecondarywindings on themainwindingasthan either two separate transformers or a two-transsemblies have the same number of turns as one half of the third winding. By appropriate phasing, the output voltages oppose a t low frequencies, so that the left- and right-speaker terminals receive less and less signal below crossover. Above crossover, the b>-passing action of the capacitor across the third winding, i n conjunction with

Left main

windinps

Frequency

- Cycldstcond

Secondary

Fig. 5-Physical configuration of the singleassemblytransformer.

Fig. 6--Effect of the third winding and its capacitor on separation: A (dashed line), response in active channel without use of third winding; B (dashed line), response into inactive cha~mel (breakthrough or crosstalk) without use of third winding; C (solid line), and high-pass response intoactivechannelwiththirdwinding filter action; D, response into inactive channel (breakthrough or crosstalk); E, response into cornlnon-bass circuit.

former matrix, followed b?. separate LC lorn-pass filters to provide common bass. The common-bass filtering is provided in the magnetic circuit n-ithout resistive loss additional to normal transformer operation.

Spmrovs EFIWCTS
I t proved important to avoid an^. internal resonance in third the winding of the transformer above the criticallyloadedone used for crossover. A n earlyattempt used a simple winding with a center tap. Such a windinghas a lealtageinductancebetweenitshalves. which is veq- much smaller than the leakage inductance Fig. 'i-LYhen thethirdwinding is center-tapped i n conventioml from the other \\indings (Fig. 7 ) . manner (shown here physically), the leakage inductance between itshalves is considerablysmallerthanthatfromeithermain However capacitor the is connected--even if tn-o \\-inding to the third winding. separatecapacitorsare used tobypasseach half separatelS--a resonanceoccursbetweenthe leal<age inductance from one half, regarded as the exciting winding, and the other, across n-hich the capacitor appearsa s a virtual shunt load. This is aseriesresonantcircuit that builds up a peak, injected into the opposite channel i n the region of 1 2 kc. The disadvantage of this is t h a t a t 6 kc a signal fro111 onesingle-endedamplifier contains a 1 2 kc distortion component that resonates i u the opposite neutralizing winding to inject a high-amplitude, double-frequencl- cross-talk component. The remedy w a s simple--bifilar winding with oppoFig. 8-Schematic of amplifier utilizing the single-assembly output site ends of the two sections connected together to form transformer to provide comtnon bass, left- and right-treble outthe center tap. This maltes the coupling so tight that puts, and full-range feedback to drive stage. no resonance occurs within the audio range, and where the ne\$-resonauce might occur, the Q has deteriorated t o much less than unit:..
'rHE ~ ~ I I ~ I , l l ~ l E l ~

Variations of the transformer have alreadS- been applied in several different anlplifiers, but one application i n particular s h o w how this unique circuit can produce advantages i n severaldirections,some of whichwere not envisaged at the outset of this project. Firstdeveloped w a s astraightamplifier,towhich straight feedback v a s applied, J-ielding lower distortion and improved separation (Fig. sj. This amplifier yielded enough gain to give full output with 0111~- voltage one stage, using a ceramic pickup, and with sufficient margin to allow for turning up the wick on weak records, or to acconlmodate the people n-ho do not think it's loud unless it's distorted.
r \

Fig. Y---Partial schematic to show the six extra components needed to add a very effertive two-channel-bass control to the circuit of Fig. 8.

1C ) S E

C'C)YTKC)I,

T o provide tone-control facilities would require either an extra stage to provide for the loss needed to obtain boosts as well as cuts, or the insertion of reverse-type controls in theover-allfeedback.Economysuggested trial of the latter method. Rut inverting a bass tone control, be terminated b,, to the input-stage cathode resistor,is not ex+-. Then w a s it realized that transformer the contains its o 7 1 bass m1 filter,already used for thecommon-bassoutput. BJ. augnlellting or attenuating the feedback obtained from

the appropriate half of the third winding, and attenuating the low-frequency feedback from the main winding by using a suitable value of capacitor i n series n-ith its feedback resistor, we have a full-bass control tone (Fig. 9). This is where the circuit gives a very useful bonus. 'I'he low-pass filter that separates bass from mid-rallge uses a two-reactancefilter,enabling a sharper slope boost and cut than is possible with conventional circuits using on1~- reactance (usually a capacitor) for this one purpose.Thisenablesthe level of frequencies below crossover (chosen a s 250 cycles, which is the center of the musical scale) to be varied quite drastically without ]loticeable change i n gain above this point in mid-range. The tone-control actionis superior to many circuits that

1961

Crowhurst: A New Stereophonic ilmplifier

71

cost 'very much more in terms of components and gain. To complete the tone control, using a similar method for separating the higher frequencies, first a maximum lift is inserted by connecting a resistor and capacitor across the cathode resistor (Fig. 10). Then a high-pass resistorandcapacitorpick off frequenciesabovethe pointwherethe lift commences,toprovidetheadjustable element of feedback above this point. A simple variableresistance,bypassingthe fixed oneformidrange gain, completes the treble tone control. The resulting amplifier produces frequency response, distortion and separation characteristics that represent Fig. 10-Circuit of a two-channel treble control, using the additional considerable improvement over predecessors its in a components marked with asterisks: this circuit is partial, added to the basic circuits of Figs. 8 and 9. comparable price range; it also has extremely effective tone controls, without the need for extra stages, and be0 cause of the fewer components, it effects quite a costLUH savingtoo.The new developmentmakesiteasyto improve performance and cut cost at the same time.

OTHEKVARIATIONS
This is only one of a group of amplifiers developed, using the same central principle, with variations. A less expensive version omitted the third common-bass winding, but retained feedback, using it for volume control functionincludingcompensationthatmakes it effec- Fig. 11-Change in output connections needed when two full-range speakers are to be used instead of common bass. tively a loudness control. This produces lower distortion and better bass than was possible i n this cost bracket previously, without any increase in cost. For the really high-quality applications, many variations are possible. If the tone control function is separated to its more conventional location in a preamplifier a greater amount of feedback is used in the power amplifier section to reduce distortion almost to vanishing point. If the common-bass feature is not desired, a 0 change in connections enables full-range two loudI speakers to be used (Fig. 11). Fig. 12--A further change in connections that can be used to improve There are other possible advantages of the new transmatching of the amplifier totheloudspeakerimpedanceover the whole frequency range. formeranditsassociatedcircuitry.Forexample,by usingthecommon-basswindinginreversedpolarity, with full-range units, the impedance matching from the CONCLUSION tubes can be made to suit the impedance characterisThe new transformer principle which forms the heart tic of a loudspeakerwithitsrisingvalueinthebass of the new development provides a number of ad(Fig. 1 2 ) . vantages, which may be used in various combinations or Above crossover, the main windings only provide separated left and right outputs. Below crossover, the fairly degrees i n individual amplifier designs: suddencoupling of thethirdwindingaddsturnsin An efficient means of combining mixed lows and series with each output so that the impedance match is retaining separate and left right a t frequencies above crossover a built the into design of the for a higher value. This enables an amplifier to deliver transformer. greater power a t t h e low frequencies where i t is sometimes needed without sacrificing the damping that feedImproved separation for the degree of circuit back can give, or limiting the properly matched power complexityprecision or involved, in boththe available for the mid-range and higher frequencies. rangeimmediatelyabovecrossoverandtheexUse of this development is notconfined to tubes or to treme high frequencies. a single-ended operation. It may provide its most effecT h e provision of convenient take-off points which tive economy to single-ended circuits, but it retains its may be used for full-range feedback, or for the inother advantages, including economic, with circuits usclusion of feedback bass tone control, with superior ingtwoseparatechannelswithpush-pulloutputs,or performance;theaddition of circuittoinclude using transistors instead of tubes, treble control is relatively simple.
1

72

I R E ?RANXACTIOAVS O X -4 LDIO

JIay-Jwne
ACKNOM-LEDGMEXT

4) \Yhere the design is applied to other than

a mixed lows system, the built-in crossover may be used t o provide more efficient power matching to dynamic loudspeakers at both bass and treble frequencies. 5) \Tiith any or all of the foregoing advantages, considerableeconom-)- i n application because of the extent to jvhich free elements within the single transforlner are utilized.

The author wishes to express thanks to CBS Laboratories forpermission t o publishthispaper,thework havingbeendoneundertheirsponsorship.Hewould also like t o expresssatisfaction i n worltingwith the Laboratorq-personnel,whoprovidedanidealatmosphere for this development contributed and helpful suggestions along the way.

A n Improvement in Simulated Three. Channel Stereo*

Summary-Some two-channel stereo systems have employed a third full-range speaker system in the center, reproducing an equal in-phase mixture of the signals in the two channels. Advantages of this arrangement over the usual two-speaker array are better reproduction of the location and size of central sound sources. A disadvantage is the sizeable reduction in the apparent spread, or distance between flanking sources. The reasons for these effects are discussed, and it is indicated that this disadvantage can be largely overcome by electrically reducing the ratio of sum to difference of the two channels, which ratio was effectively increased by the addition of the center speaker. It is shown that the signals to the three speakers may be regarded as three. independent channels with certain signal-to-crosstalk ratios, which are derived as a function the level of the center speaker and of theamount of electricalreduction of the sum-to-differenceratio. The choice of optimum parameter values and appropriate circuits is discussed.

AIETHOD hasbeendevised for improving the geometric fidelitl-of reproduction of two-channel stereo. In the sections thatfollow, the defects of present sl-stems described are and i t is shown how these may be partl?. overcome.

A.

DEFE(,TS T\TO-SPE.UER OF STEREO

There is no doubt that a conventional two-speaker system can provide good reproduction of the location and, to a certain extent, the size of sound sources. In order t o accomplish this, however, the listener must be satisfied with a narrow speaker spacing and a precisely centered listening position. Ifthespeakersarefarenoughaparttosubtend a

desirably large angle at the listener, and the listener is on the center line or axis of the system, central sources appearlargerandtheirlocationmoreindistinctthan similarsources attheextreme left or right.M7hereas thisphemmenonmaybeaggravatedby poormicrophoning, it cannot be blamed entirely the microphone 011 technique, for i t occursevenwhenthetechniqueis such that these central sources are recorded purely monophonically, ;.e., n-ith identical signals i n both channels.Thephenomenon is probablycausedbyat least the three following factors : 1) E f e c t of head rotation: IIhen a small sound source is directlJ-infront of anobserver,andtheobserver rotates his head slightly to the left, the sound strikes the right ear more perpendicularll- and grows louder in that ear, particularly high frequencies. Sound reaches at the left ear more obliquely than before, and its loudness in that ear, therefore, decreases. If the rotation is sufficient, the left ear is shadowed by the head ant1 the consequentattenuationatthatear is even greater. Also, sound reaches the right ear sooner than the left. Theobserver uses therelationbetweenrotationand amplitude and time differences to pinpoint the source and determine its size. Obviously, if the source is very large, or consists of tn-o widely separated points emitting identical sounds, a given rotation of the head produces less amplitude and time differences at the ears than i t would if thesourcewereasinglepoint. The
F. -1. Firestone, The phase difference and amplitude ratio at the ears due toa sourceof pure tone, J . Acoust. SOL. flm.,vol. 2 , pp. 260-270; October, 1930.

* Received by the PGX, April 5 , 1961.

i\larwiclr Manufacturing Co., Chicago, Ill.