11HIGH FIDELITY"

ITMUST, BE. RECOGNIZED that the prob-lem of fidelity is a relative one, asconsideration must be given to the

quality of the program being transmit-ted as well as the conditions underwhich it is being reproduced. Highfidelity in its true sense implies the faith-ful reproduction of the original trans-mission both as to frequency range andvolume level, although the volume levelneed not necessarily correspond with that·ofthe original. The term "high fidelity"has so often been associated with systemshaving an extended high frequencyrange that it no longer conveys its truemeaning. Probably a new term shouldbe en1ployed which more nearly de-scribes the practical attainable goal ofreproduction with the systems availableat present. Such a name would becomeinadequate, however, as soon as newsystems were perfected, so for this rea-son it might be better to define "high, fidelity" as understood at present.In this article high fidelity shall be

considered as the degree of perfection,for an average listener in an averageresidential room, which presents the il-lusion of hearing the program as itexists at its point of origin, insofaras frequency range and amplitude v<tria-tions are concerned, but not insofar asbeing able to associate the source ofthe sound. It is assumed that distortioneffects will be negligible and that thespatial' distribution will be nearly con-stant over at least a 45 degree angleeach side of the speaker system.Before discussing what degree of per-

fection is economically practicable itmight be well to review the character-istics of music and speech and deter-mine the overall requirements for sucha system.

Frequency and Power Range ofSpeech and MusicIt is generally conceded·by all engi-

neers that a frequency range of 20 to15,000 cycles is fully adequate to repro-duce faithfully all fundamentals andovertones of any musical instrument orspeech which are audible to the ear.Such a range is also sufficient to permitready identification of such non-musicalsounds as keys jangling, resin squeakson bowed instruments and numerousothlH mechanical noises which are by-products of the music. Not that thesesounds are desirable, nevertheless theyare discernible on a high fidelity systemunless precautions are taken to minimizetheir pickup.As can be seen from Fig. 1, there is

little to be gained by reproducing fre-quencies above 11,000 cycles exceptperhaps in the naturalness of the cym-bals, triangles, snare drums and 'similarinstruments. Likewise the low freauency

response need not extend below 65 cyclesexcept for the organ, bass viol and someunusual sound effects. So much for therequired frequency range;. no attemptwill be made to establish the most prac-tical bandwidth at this time, since thereare many other factors to be considered.Instead, the variations in amplitude orintensity of the sound to be reproducedwill be discussed.Measurements by Fletcher' show peak

powers as high as 66 watts, with aver-age powers of approximatelYJO.1 wattas representative of large sy,$phony or-chestras. Conservatively, t~n, a 70 dbrange would be satisfactory for repro-ducing an orchestra fltits original soundlevel. This of. course is the extreme;other programs of speech, dance musicor instrumental solos obviously do notrequire such a range, probably some-thing of the order of 20 db being en-tirely satisfactory. Except for studioprograms being transmitted by FM sta-

r-------------------- PRACTICAL HIGH FIDELfTY--------------------,Y t

r----------------------FULL RANGE HIGH FIDELfTY---------------------,t t

---------------------------------- BASS TUBA

------------------------- TRUMPET

-----------:~----------- CYMBALS--- ------,...------------------------ SNARE DRUM

------------------------- FLUTE

-------------.------------ VIOLIN

--------- -------------------- TROMBONE

---------------------------- SPEECH------------------------------- BASS VIOL

I I I Ii II40 60 100

I20,000

I I I I I I 'I200 400 600 1000 2000

CYCLES PER SECOND------- FUNDAMENTAL

I I I III I4000 6000 10,000

Reprinted by Consumers' Research, Inc:, Washington, N. ].;from RADIO, November 1945, publishedby RADIO MAGAZINES; Inc., 342 Madison Avenue, New York 17, N. Y. Copyright 1945 by RadioMagazines, Inc. Subscription $3.00 per year.

tions, the dynamic range is limited .by The ear is a pressure-actuated non-the quality of the wire program lme linear device capable of actually supply-which may vary from 25 to 45 db accord- ing deficiencies (under certain. cOI1.di-ing to the quality of the line employe? tions) in the quality of the musIc bemgAlthough lines are limiting factors 111 heard. Furthermore, the frequency rangemany instances, they do not affect the and sensitivity vary appreciably withoverall maximum requirements and are such factors as the age of the listener,I t· d example of what the intensity of the sound wave, and theon y men lOne as a.n . _

is actually encountered in practice. presence or absence of undesirable .noise.S . . fi d a frequency F Many investigators have contnbutedummanzmg we n . .

range of 20 to 15,000 cycles and a to the knowledge of the charactenstlcsvolume range of 70 db capable of repro-/-. of the e~r,. as can be seen by ref~renceducin high fidelity programs in their to. the ~Ibhography at the conclUSIOn of.. g I f thiS article. However, tests made at the

ongma orm. World's Fair on approximately 500,000FM .Stations-:-5ynthetic Vinyl people of all ages and sexes' h~ve pro-ReSin Recordings vided data which when analyzed resultsUntil FM stations came into existence in a statistical average frequency charac-

very few programs were availab~~ .ca- teristic as shown in Fig. 2. For ~~m-pable of giving even so-called high parison purposes a curve of the cntlcalfidelity" reception. Modulating frequen- listener (best 5%) has been included.cies above 5000 cycles resulted in "mon- Note that the two curves follow eachkey chatter" because of adjacent channel other quite well except for the additionalreception, and as mentioned previously frequency range perceived and the sen-the volume range of most lines is quite sitivity. A seven to ten db difference be-limited because of noise considerations. tween 20 to 8000 cycles, which gradu-FM stations, however, are capable of ally increases to approximately 17 dbtransmitting frequencies up to 15,000 at 12,000 cycles, should be noted. Thesecycles with a dynilmic range o~ 70 db. curves indicate that it .is possible !orThis is due in part to the Wide fre- persons with acute heanng to perceivequency bands available fo: this t~pe frequencies as high as 22,00? cycles a.ndservice and also to the nOlse-reducmg as low as 20 cycles; prOVided the m-capabilities of the system in general. tensity is high enoug~. . .According to the FCC Standards of An interesting family of curves IS

Gooe: Engineering Practice the FM sta- shown in Fig. 3 and are known as equaltion should be capable of transmitting- loudness curves. These show the char-a band of frequencies from 50 to 15,000 acteristic of the ear over the audiblecycles, with certain prescribed pre- spectrum, using a lOOO-cycle pure toneemphasis and with a noise level at le~st as a reference. The curves include the60 db below the program at full devla- entire, range of intensity from barelytion. Total distortion should be limited audible (threshold of audibility) to theto from 2.5 to 3.5 per cent according to point where a sensation of feeling isthe modulating frequency. These sped- experienced (threshold of feeling). Asfications indicate the possibility of a can be seen these curves vary both withtrue high fidelity service. intensity and frequency, reaching a max-The same condition existed in the imum in the vicinity of 3000 cycles.

reproduction of phonograph records. It is evident from the hearing-contourThe ordinary shellac disc had little curves that the response characteristicabove 3500 cycles and the surface noise depend~ to a large degree on the in-was so high that the volume range wasseverely limited. With the introductionof synthetic vinyl resin recordings thesurface noise has been tremendously re-duced, and consequently the volumerange increased accordingly. The re-cording of frefluencies above 7000 cyclesis also practicable. Thus, with these twonew advances a big step has been madein the realization of high fidelity repro-duction.

Characteristics of the EarSince the ear, in the final analysis,

must be the judge of fidelity it most cer-tainly should be taken into considera-tion when planning a true high fidelitysystem. Several of the more importantcharacteristics, such as the frequencyrange, acuity, and the age of the listenerwill be discussed since these have anespecial bearing on the problem.

CRITICAL LISTENER

100 1000 10,000FREQUENCY-CYCLESPER SECOND

Fiq. 2 Hearinq contour curves for averaqeand critical listener - no noise

tensity of the sound source. At lowintensity levels, the low frequency re-sponse is comparatively poor. The highfrequency response is similarly affected.although to a smaller degree. In otherwords, if the intensity of the soundsource is increased a given per cent, theapparent increase will not be uniformover the entire frequency range; theapparent increase at the extreme low andextreme high frequencies audible at thelower level will be less than that offrequencies in the vicinity of 2500 cy-cles.In addition to the above general char-

acteristics, it has been found that thereare variations in response due to theage of the listener. That is, 'as the ageincreases the high frequency resnonsegradually decreases approximately asshown in Fig. 4. This has no importantbearing on determining permissible fidel-ity range, however, since the systemmust be satisfatcory for all age groups.

Masking Effects of NoiseIn actual practice we never have ideal •

lIstening conditions because of noise in-terference. Measurements' show the av-erage residential room noise is approxi-

-THRESHOLD OF FEELING 1-f-tt+t

80

40

.RESHOLD OF AUDIBILITY~

~6011:...J

~40

~ 20

~ 0

~I 0l&J(J)

~-1011.

l:l11::-20

~

100

oi~ 80...Jl&J

~ ~O

~40(J)z~ 20

030 50

Fig. 4. (Above. left) Relative loss in high frequency for average listeners. Curves are for male listeners of ages 25. 35. 45. anq55 years. respectively. Fig. 5 (Above. right) Effective hearing characteristic due to masking effect of noise

mately 43 db.* Noise varies of coursewith conditions outside the home, thenumber of people in the room, etc., butprobably never decreases below 33 db.*Its effect appears to be one of deafen-ing the listener exposed to it and there-fore must be considered in the overallresult. Combining the masking effect ofnoise with the average ear character-istic results in curves as shown in Fig. 5.Other noted effects are auditory fa-

tigue, which may be caused by pro-longed exposure to loud sounds, (anexaggerated example is the feeling ofdeafness after having taken a long air-plane trip), temporary impairment ofhearing due to illness and fatigue causedby listening to "narrow range" fidelitywherein the nerves of the hearing sys-tem attempt to supply those frequenciesnot present.

Determining Frequency RangeIn the determination of the frequency

range for a high fidelity system it isnot enough to decide that since thegreatest audible range extends from 20to 22,000 cycles this should be the goal.From the previous discussion and asso-ciate.d curves it is evident that only it

small percentage of listeners are capableof hearing such a range; furthermorethis range can only be heard under cer-tain conditions (high intensity levels)and granting it can be heard, there isvery little to be gained in enjoymentor perceptibility. Obviously, then, therange can be reduced to at least 11,000cycles at the high frequency end, andprobably to 65 cycles at the low fre-

TABLE 1

POWER OF VARIOUS SOUND EFFECTS

75 pieceo<cheslro 70Bass drum 25Pipe Orvon - 13Snare drum 12c:ymbala 10Plono_,- 0.4Bass voice 0.03Averove apeech 0.000024Violin(aafll O.000003B I

quency end. Another factor, however,enters the picture at this point, that is,tonal balance.A few years ago, designing a high-

fidelity receiver consisted of merely in-creasing the high-frequency responseand extending the upper range a fewthousand cycles on an otherwise nor-mal receiver. This resulted in a verythin-sounding receiver because the bal-ance between bass and treble responsehad been disturbed. Unless both ex-tremes of the frequency pass-band areextended, the overall result is disap-pointing. The main problem is to obtaina balance between the high and low fre-quency cutoffs. Of course other factol'ssuch as practicability. psychological andphysiological effects are important butunless the tonal balanc'e is correct truehigh fidelity will not be achieved.Balance requirements are not too

critical; in fact, if the product of thelow frequency and high frequency cut-offs fall within the range of 500,000 to650,000 the results will be quite accept-able. This is shown graphically in Fig. 6.Statistics' show that a range of 75 to8000 cycles will suffice for the averagelistener; it is then a matter of economicswhether this should be the acceptedrange or whether the additional costinvolved in obtaining a range of 65·to11,000 cycles is justifiable. This latterrange is considered adequate for criticallisteners in very quiet surroundings. Somuch for the frequency range.The power in watts of sound radi-

ated, varies over a tremendous range, asshown in Table 1. Although the actualrange in power between a full orchestraand a soft passage on the violin amountsto approximately 20 million to one, thesensation of loudness, being logarithmic,is only about 70 db. In other words adynamic volume range of 70 db is re-quired to reproduce fully an orchestraof 75 pieces.In the average home a volume level

of 85 db is about as high as can betolerated with any degree of comfort.Quiet operation, or the other extremewould be a level of 40 db, or a dynamicrange of 45 db. This takes into accountthe noise level of the room, and of course

is subject to some variation, but it isrepresentative of a range extendingfrom very loud to very quiet (butlistenable) volume. It is questionable asto whether levels much below 50 db areof any consequence for fidelity, althoughthey are important so far as enjoymentof dynamic range is concerned.

Listening HabitsBecause of shortcomings, either

through lack of training or appreciation, the average li~tenei does not fullyrealize the capabilities of high-fidelityreception. It must be remembered, how-ever, that the general public is becomingmore conscious of high fidelity throughthe medium of present-day advertising,and that the desire for tuning-in distantstations is gradually being replaced bythe desire to obtain noise-free realisticreception.The use of narrow-band receivers

over a period of years has made themappear more preferable than many highfidelity units. There are several possiblereasons for this preference : (1) Distor-tion is more prevalent in a wide rangereceiver unless special precautions havebeen taken to minimize its effect. (Thisis not always due to the receiver but isvery frequently caused by the broad-casting station using poor recordings.)(2) The permissible amount of distortion

fi1400<fl~::I...~a 100~ 80z~ 60fil 50ll: 40~9 30~ ~ ~

HIGH- FREQUENCY CUTOFFCYCLES/SECOND

Fiq. 6. Low and high frequency cutoffgraph for tonal balance showlDq the com·parative values both for average condl·

tions and for high f1deUty.

ai'oI

~ +10

~~ 0cr~f: -10~cr

and noise decreases as the width of thefrequency pass-band increases, (3) Thesignal strength of the station is also afactor. If this is not strong enough tooverride the noise induced into the re-ceiver, the program fidelity is very defi-nitely reduced, Many listeners havegrown accustomed to a so-called "mel-low" tone because it has been necessaryto retard the tone control in order toeliminate unwanted noise and distortion,Items such as these can and are gradu-ally being corrected as the appreciationfor high fidelity increases. .-\t presentonly a few programs require high-fidelity receivers because, unless thequality is transmitted, it cannot be re-ceived.High fidelity must be demonstrated to

the listener under ideal conditions forfull appreciation and the naturalnessmust definitely be pointed out. A simpletest for high fidelity is the ease withwhich it can be listened to for longperiods of time. Inferior quality puts astrain on the nervous system because theear attempts to supply the deficienciesand gradually becomes fatigued.

General ConsiderationsHaving discussed the characteristics

of the ear, the desired frequency andvolume range, and the listening habitsof the public, the next step is to deter-mine what technical problems are in-

volved in obtaining high fidelity repro-duction.The requirements are obviously as

follows:1. The overall reproducing system(transmission is assumed to be ideal)must be free of amplitude distortionor limiting at the required outputlevel.2. Spurious harmonics must not beintroduced at any operating level.3. ~oise, phase distortion, and cross- ...modulation products should be mini-mized.4. The acoustics of the listening roomshould be satisfactory (without ex-cesssive damping or reverberation),and the spatial reproduction should heas nearly uniform as possible.Measurements can be made of all

characteristics including overall soundpressure curves, but the latter can bevery misleading if not properly inter-preted. Sound pressure curves indicatethe acoustic response under a given setof conditions and. unless these condi-tions are exactly the same when thelistener is using the unit (the listenerbeing in the same position as the micro-phone) the measured curve will not bea true picture of the acoustic output.Even under the above conditions somediscrepancies are bound to occur be-cause of the fact that the listener hastwo ears which makes possible the local-

~I

~t10z~ +5~~ 0>~ -5•••go. cr

100 1000 10,000FREQUENCY-CYCLES PER SECOND

SOUND PRESSURE CHARACTERIS1;IC

ization of the sound source. Due to re-verberation in the ordinary room thesound is reflected from the walls, ceil-ing and floor and therefore reaches thelistener from many points with differentphases and amplitudes and, due to thebinaural effect, results in a quite differ-ent response from that measured with asingle microphone. This is particularlytrue when listening to high frequencies.It is possible to walk around a roomwhen listening to a single high fre-quency note and be able to distinguishreadily nodes and anti-nodes due to re-flections. A typical sound pressure curveshowing the effects of reflections in asmall room is seen in Fig. 7. Out ofdoors the same condition exists exceptthe time lag is longer; resulting inechos. This of course assumes there arereflecting surfaces nearby.

COAXIAL MOUNTINGWITH

DIFFUSING ELEMENT

VERT1CALMOUNTING

All these phenomena clearly indicatethat the final checking of a high fidelitysystem must be made by exhaustivelistening tests.

Acoustic ConsiderationsVery little can be said about the

acoustic characteristics of the residential I

room, since the average listener will donothing about it anyway, except perhapsto locate the receiver in a corner or toavoid putting it where a hard wall sur-face close by might cause severe reflec-tions. Other than a few minor changessuch as these the room acoustics are-beyond control. Therefore, it is up tothe designer to make the unit flexibleenough to compensate for these faults .This can be accomplished by adequatetone controls and uniform spatial dis-tribution of the sound.Unfortunately, sound waves emanat-

ing from an ordinary speaker do not

have a uniform spatial characteristic. Inother words, the listener cannot move todifferent places in the room without not-ing a difference in the quality of thereproduction. A typical polar character-istic is shown in Fig. 8 together withsound pressure curves. These curveshave been smoothed up a bit for clarity.It is evident then that steps must hetaken to provide a better polar charac-teristic, otherwise it would be necessaryto sit directly in front of the speaker ifthe high frequencies are to be heard illtheir right proportion.This brings up another point; it IS

impossible' to cover a range of 65 to11,000 cycles with one speaker. Instead,dual speakers are required with an ap-propriate network which divides theelectrical power at approximately 1000cycles. Several arrangements are pos-sible with dual speakers (commonlycalled woofer and tweeter) as shO\nlin Fig. 9. One method necessitatesthe mounting of the high frequencyt\\-eeter coaxially within the low-frequency speaker. The tweeter thenacts as a diffuser for the large speaker.:\ small diffusing unit may then beplaced in front of the tweeter for betterhigli frequency distribution. Anothermethod uses separate mountings for eachspeaker with the tweeter being of themulticellular horn type which has animproved polar characteristic. Sounddiffusers, when properly designed. aremore practical since they simplify thedesign problems and are less expensive.Of equal importance to the acoustic

properties of the room is the speakerenclosure or cabinet. This can be al-!ered to a certain extent b~ the designer..-\ square, flat, symmetrical speaker baf-fle, for instance, should never be usedbecause its symmetry results in the can-cellation of frequencies at its cutoffpoint as shown in Fig. 1(f. This iscaused by the radiation from the frontand rear of the speaker cone arrivingat the listener. ~xactly out of phase witheach other. The simple expedient ofmounting the speaker on an irregularlyshaped baffle will, for all practical pur-poses, eliminate this fault. Another

,I

I I-t- IRREGULAR BAFFL£SQUARE BAFFLE f-r-

a :~<YJ1

I

20

ai0I 10....l/l

~Ina:....0>

5....a:

is desirable to design the system tomuch better standards; -70 db belowmaximum output being a more practicalvalue. since it can be attained, and re-sults in a good safety factor.Beam power tubes are the logical

choice for the output stage. These pro-vide high power output with good powersensitivity, but unfortunately the dis-tortion generated by this type tube(principally third harmonic) is ratherhigh. Of course, when used in a push-pull circuit and properly balanced, theeven harmonics are effectively elimi-nated. The third harmonic remains, how-ever, and this is very objectionable tothe ear; much mo're so than the secondharmonic distortion.The logical solution to the problem

lies in the use of degeneration.· In facta degenerated amplifier is the only prac-tical solution to the problem. The degreeof degeneration of course determines themagnitude of the distortion. Not onlydoes degeneration reduce distortion butit effectively minimizes hum and noisecomponents; it also reduces amplitudedistortion resulting from saturation ofiron-cored components which often pro-duce false harmonics when working at.full capacity. In addition to these almostUtopian qualities, degeneration is notedfor the improvement it makes in thefrequency characteristic without objec-

.30:g

~~ iOIiiz~ 0~ H-++~'VERY LOUD RADIO

Fiq. 11. (Above left) AmpWler havinq- desirable sharp cutoffs. Fi'1. 12. (Above rlqht) CompeJ1Satlon required O'f'er yolume limits Inthe home. Ustener on axis of speaker

common error, when two similar speak-ers are used. is to mount them equallyspaced from the edges of the baffle andas far apart as possible. The two speak-ers should be located close together andsomewhat off center. The cones, whichare in phase of course, then load eachother more effectively and a more effi-cient coupling to the air is obtainedwhich results in better low frequencyresponse.Because of the physical size required,

open baffles do not function .well muchbelow 125 cycles and it was for thisreason that acoustic labyrinths and bass-reflex enclosures were developed. Ade-quate descriptions of these are to befound elsewhere and therefore will notbe discussed at this time. (See bibliog-raphy for furthe!; details.)

Electrical ConsiderationsGranting that there is no particular

difficulty in obtaining a frequency rangeof 6S to 11,000 cycles, the major prob-lems are those of distortion, hum, noiseand amplitude relations.The presence of hum or noise in the

output of a high fidelity system is par-ticularly objectionable because of thelarge volume range required. Assuminga dynamic range of 4S db is required,the hum or noise level must never exceed-45 db compared to maximum output. It

UNNECESSARYDISTORTION

I N

tiunable phase shifts so prevalent withether circuits.A further improvement in amplifier

design results when balanced feedback'is employed. The advantages of theusual degenerative feedback are thenmaintained without at:"!appreciable lo~sin amplification and with a much sharpercutoff characteristic at the ends of thedesired frequency range. Abrupt cutoffsas shown in Fig. 11 are particularly de-sirable for both radio and phonographreproduction. A sharp cutoff at the highfrequency end of the range eliminatesdistortion that might be present dueto higher order harmonics, while a sharpcutoff at the low frequency end willmaterially reduce turntable rumble.The sound waves of music and speech

are not pure tones but consist of funda-mentals and often Illany harmonics orovertones. These harmonics are themeans by which the various instrumentsare distinguished from one .another andprovide the individuality and quality ofthe music or speech. Such complexwaves, which are not often harmonicallyrelated, give rise to intermodulationproducts if at any point In the systema non-linearity exists. The result isquite harsh and disagreeable to the ear.Distortion of this type (often calledtwo-frequency distortion) may notshow up in the ordinary harmonic analy-sis; in fact, higher harmonics of theusual amplitude type are often outsidethe range of audibility and are there-fore not important. But mtermodulationproducts may lie anywhere in the band,and for this reason it is imperative thatthey be measured and minimized if nec-essary. A common complaint when lis-tening to a system having appreciableintermodulation products is that wheaan entire orchestra is being reproducedthe sound seems mixed up or run to"gether, yet when a solo part is heardthe quality suddenly becomes clear. Thisis because fewer intermodulation prod-ucts are formed. A common source dtwo-frequency distortion is often due toinadequate low frequency carrying ca-pacity, so the power capacity of theunit should be of the order of 20 ormore watts.If it were not· for the non-linear

characteristic of the ear, especially whenthe volume level is changed, the prob-lem of designing a high fidelity systemwould be comparatively simple. A com-mon fault with many high fidelity re-ceivers lies in the fact that proper com-pensation for the ear characteristicshave not been taken into account at allvolume levels. The ear has a compara-tively flat response only at extremelyhigh volume levels; this condition is

reached at approximately the same pointwhere the sensation of feeling beginsand the sensation of hearing stops.Referring to the equal intensity con-

tour of the ear it is apparent that asystem having a frequency character-istic which is flat is not to be desired.At volume levels corresponding to thatof a very loud radio, sound reproductionuSHally seems to be quite good. This istrue for two reasons; tlie level morenearly represents that of the original,and the ear characteristic is more nearlyflat. However, even at levels of 85 dbabove threshold an increase of approxi-mately 6 db is required at 65 cycles tomake it equivalent to that at the middlefrequencies. In the region of 3000 cycles(most sensitive region of the ear) adecrease of 3 to 5 db is desirable. Asthe frequency is increased up to 8000cycles the amplitude must again be in-creased, as at the low frequencies, byapproximately 10 db. These correctionsare requi red to make up for deficienciesof the ear at high intensity levels. Asthe sound intensity is decreased to veryquiet, additional compensation is re-quired. As much as 30 db at 65 cvclesand 12 db at 8000 cycles being req~ired.lt is obvious that with such a non-

linear characteristic as this some sortof compensated volume control must beused, which automatically adjusts thefrequency response as the sound outputis changed, Without compensation, thequality becomes unnatural at low vol-ume levels and, since more persons listenat these levels, it is important that com-pensationbe correct. Bass compensationhas long been more or less standardpractice but little attention had beengiven to correlating the compensationwith the characteristic of the ear. Auto-matic treble compensation has not beenso widely used. Fig. 12 shows how thefrequency response of ·a high-fidelitysystem should vary over the volumerange ordinarily used in the home.Summarizing we find the following

points of' interest:

1. A frequency range of 65 to 11.000cycles having a relatively sharp cutoff ateach end of the range is the most desir-able. Practically a balan:ed range of 70to. 8000 cycles will be entirely adequatefor most programs.2. The amplitude characterisdc shouldfollow the equal loudness contour curvesof the ear at their respective levels.3. Amplitude and intermodulation distor-tion should not exceed two per:ent.4. The acoustic response should be asnearly uniform as possible, at least overan angle of 90 degrees.5. ,Power capacity to handle peaks ofpower with a dynamic volume range ofat least 45 db.Recommendations for accomplishing

these reqUIrements are:

1. Beam power amplifier of at least 20watts output.. 2. Degeneration or balanced-feedbackcircuit in the amplifier.3. Two or more speakers with suitablesound diffusing elements.4. Compensated frequency. response forall volume levels.It must be recognized that the ear is

the final judge of the faithfulness ofreproduction and that psychological,physiological and physical (room acou-stics) effects must be considered.The goal for the designer of high

fidelity equipment is reached when thereproduction sounds like the originaland not like a radio. True high fidelityhas then been achieved.

Referencesl1. Fletcher, H.-Speech and Hearing-D. Van Nostrand, N. Y.

2. Steinberg, Montgomery & Gardner-World's Fair ,Headng Tests-BellSystem Tech. Jour.-Oct. 1940.

3. Fletcher, H.-Hearing, the Determin-ing Factor for High Fidelity Trans-mission-Proc. IRE, June 1942.

4. Seacord-Room Noise at Subscriber'sTelephone Locations-Jour. AcousticSociety of America-July 1940.

5. Olsen & Massa-Applied Acoustics-B1akiston. Phila. 1934.

6. Terman"":'Feedback Amplifier Design-Electronics-Jan. 1937.

7. Ginzton-Balanced Feed-Back Ampli-fiers-Proc. IRE, Nov. 1938.

8. Hilliard-Distortion Tests by the In-termodulation Method-Proc. IRE,Dec. 1941.

Bibliography for AdditionalReading1. Sivian, Dunn &White-Absolute Am-plitudes & Spectra of Certain MusicalInstruments & Orchestras - Jour.Acoustic Society of America-Jan.1931.

2. Fletcher, H. - Audi~ory Patterns-Rev. Mod. Physics, Jan. 1940.

3. Snow-Audible Frequency Rangesof Music, Speech & Noise-Jour.Amer. Standard Association-July,1931.

4. Hanson-Down to Earth on HighFidelity-RADI6, Oct. 1944.

5. Wood-Physics of Music-SherwoodPress, Cleveland, O. 1944.

6. Wolf & Malter-Directional Radia-tion of Sound-Jour. Acoustic So.ci-ety of America-Oct., 1930.

7. Fletcher & Munson-Relation be-tween Loudness & Masking-Jour.Acoustic Society of Americlk-July,1937.

8. Fletcher & Steinberg-ArticulationTesting Methods-Bell System Tech-nical Journal-Oct., 1929.

9. Chinn & Eisenberg-Tonal-Range &Sound-Intensity Preferences ofBroadcast Listeners-Proc. IRE-Sept., 1945.

10. Olson & Hackley-Horn & DirectRadiator Louli Speaker-Proc. IRE-Dec., 1936.

11. Massa-Horn Type Loud Speakers-Proc. IRE-June, 1938.

Consumers' Research is a technical and scientific research organization testing and reporting on ultimate )consumers' goods; it is operated as a non-profit institution, for scientific and educational purposes.~Consumers' Research provides a wide variety of special services to ultimate consumer subscribers,likewise to teachers and students of consumer problems and of high school and college courses in thesciences and technical arts.