Vol. 4, Bo. 12354 PHILlPS TECHNICAL REVIEW

omotor itself can relatively easily be rendered freeof interference by the well-known method 2).When however the apparatus is taken in the handnew interferences appear. They' are caused by acapacitative coupling between the collector (whichis under the highest interference voltage) and thehand which is practically earthed for high-fre-quency voltages.This interference can be removed by shielding

the collector. The stator already fulfils this functionto a certaht extent, but this is jiot.sufficient. Ade-quate shielding is obt~ined bi' an extra pair ofplates connected electrically with the stator andforming with it a closed :Q.~iwork around thecollector and armature.

Safety

Since the apparatus is connected with the mainsand brought into contact with the skin,' the greatestcare has been taken that there shall be no dangerin the use of the apparatus, in case certain partsshould become out of order. It might for exampleoccur, due to the breaking of a' wire, that the bodyof the motor was under a tension. Even such a de-fect would not bring the user into contact with themains; because the motor is completely surrounded 'by a "Philite" housing, while the cutting elementitself, which is in contact with the skin, is connectedwith the motor axle only through an insulatingcoupling piece of "Philite".

Another point to which attention has been givenis that of radio interference which might be causedby the motor, as by all other collector motors. The

2) Cf. Combating radio interfcrences, Philips techno Rev. 4,237, 1939.

THE TESTING OF LOUD SPEAKERS

by R. VERMEULEN. 621.395.623.7

For judging the quality and for comparing different loud speakers it is desirable to makeuse of quantitative data, such as the distortion and the frequency characteristic. Themethods used in the Philips laboratory for the rapid determination of these factors ,makeup the subject of the first part of this article. In the second part the fundamental problemof the conditions under. which the measurements on the loud speaker must take placeis dealt with. The measurement of the direct sound alone is discussed as well as the mea-surement of the total sound radiation and a measurement in which a certain combinationof direct and indirect sound is used as a basis. . .

The highest instance in the judging of the qualityof reproduetion of a loud speaker is the ear of thelistener. The most natural method for the testingand comparison of different loud speakers consistssimply in listening to the music or speech re-produced. This method has however the obviousimperfection, that the judgment can only be of aqualitative nature, and that any slight deviationsare either entirely' unnoticed or observed in sucha way that it is difficult to draw any conclusionsabout their cause. It is thereforr desirable to sup-plement the listening tests with measurements

fwhich permit a more accurate ànd objective judg-ment of the quality of the reproduction.

The first question which arises is, what must wemeasure? Or in other words, by what quantitiesis the quality of the reproduetion determined?In the main there are two such factors:1) The. distortion, which constitutes a measure of

the non-linear distortion ID the loud speaker.

(1)where al and a2, 3,4... indicate the amplitudeof the fundamental an'á of the overtones oc-curring in the loud speaker, respectively.

2) The frequency characteristic, which indicatesthe sound intensity produced by the loud speaker .when' alternating voltages of different frequenciesare applied to its terminals.

For ideal r,eproduction FI ,= 0, and the fre-quency charaèteristic must be flat, i.e. the sensitivityof the loud speaker must b~ the' same for all fre-quencies. The deviations from these conditionsfound in' practical cases, while they do not permitthe drawing of any conclusion about their effecton the reproduction without the aid of listeningtests., do however make possible a comparison ofdifferent loud speakers and p,articularly an estima-

DECEMBER 1939 THE TESTING OF LOUD SPEAKERS 355

tion of the influence which may be expected fromany change in the construction.

In this article we shall concern ourselves with themanner in which the factors mentioned are obtained.In addition to the apparatus necessary for thispurpose, the problem of the conditions under whichthe measurements must he carried out also deservesattention. This question will also be discussed indetail.

The measurement of sound

The simplest method and the one most oftenapplied for the determination of sound intensitiesmakes us~ of a microphone introduced into thesound field. If the microphone is calibrated, i.e.if the relation is known between the sound pressureand the microphone voltage generated thereby, themeasurement of sound is reduced to the measure-ment of an electric voltage, which can be done v, ithan amplifier and a voltmeter.A condenscr microphone (fig. 1) is usually used

as measuring microphone, one of the reasons beingthat the calibration can in this case be carried outin a simple way, namely, by allowing an electro-static force instead of the sound pressure to acton the membrane. In an earlier article in thisperiodioal+) this calibration was discussed in detail.It is sufficient now to point out that the calibration

Fig. 1. Diagram of a condenser microphone. Thememhrane 1 is placed at a very small distance(16 [1.) from the rigid plate 3. The separationis det.ei-mined by the thickness of the ring 2lying between. 1 and 3 together form a condenserwhose capacity varies upon vibration of themembrane. With constant charge voltagevariations hereby occur between 1 and 3, whichafter amplification may be measured. Plate3 is provided with a number of holes in orderto prevent the rigidity of the cushion of airbetween 1 and 3 decreasing the sensitivity of themicrophone.

54103"

takes place ID two steps. We first determine thevoltage which the microphone produces when analternating force of a given amplitude and fre-quency acts on the membrane. A correction mustthen however be introduced, due to the fact thatthe microphone itself causes a distortion in thesound field, so that the variations in pressurefound by measuring the voltage, are not exactlythe same as in the undistorted sound field. Thiscorrection depends only on the shape and dimen-sions of the microphone, and need therefore onlybe determined once for a given type of microphone.

1) J. de Boer, Absolute sound pressure measurements,Philips techno Rev. 1, 82, 1936.

The determination may be by calculation or ex-perimentally, by comparison of the results of themeasurement with those obtained with aRay Iei gh disc (sce the article cited in footnote l).Infig. 2 some measuring microphones are shown.

The microphone amplifier is set up at some distancefrom the microphone, in order to avoid disturbanceof the sound by the amplifier cabinet. The amplifiedvoltage of the microphone can be measured witha voltmeter.

Fig. 2. Condenser microphone. Among the parts on the leftmay be seen the rear electrode 3, the intermediate ring 2and the membrane stretched in its housing 1 (see fig. I). Infront of the middle microphone lies an auxiliary electrodein the form of a grid, which is used in the calibration of themicrophone to exert an electrostatic force of known magnitudeon the membrane.

Determination of distortion

The condenser microphone causes no appre-ciable non-linear distortion of the signal. If a puresinusoidal voltage is fed to the loud speaker (fun-damental), any overtones which appear in the mi-crophone voltage, and whose intensity can bemeasured separately with a suitable analysinginstrument, must be ascribed exclusively to thedistortion in the loud speaker. From the individualmeasurement of each overtone the distortion Flcan be calculated according to equation (1). Thefactor Fl can also, however, be measured directlywhen the microphone voltage is divided into twoparts by means of filters, one of which parts con-tains only the fundamental and the other only theovertones. The energy of these two signals can bemeasured with the aid of thermocouples. By meansof a compensation circuit, the principle of whichis given in fig. 3 and is explained in the textbeneath, Fl can be read off directly from a scale.It must be mentioned in passing that the degree

of distortion is indicated not only by the factor Fl'but also by a factor defined by the formula:

356 PIHLIPS TECHNICAL REVIEW Vol. 4., No. 12

F2= 1/2 a2

2 +,3 aa2

~ 4 a42 + ... (2)

(tl

In this formula greater weight is assigned to thehigher harmonics, in agreement with the phy-siological fact that the higher harmonics lead morequickly to an unpleasant lack of purity than tothe low ones: F2 like FI can be determined fromthe separately measured overtones, or directly bymeans of an arrangement like that of fig. 3, whenan amplifier V is used whose amplification increasesproportionally with the frequency.

--------.,L r---------------,~~~

34107

Fig. 3. Sketch showing the principle of the compensation cir-cuit for direct measurement of the distortion factor Fl. The

. filter L which passes only Iow frequencies passes 'only the fun-damental (ul) of the voltage of the microphone Af. The energyof this voltage is measured by the thermocouple Tl" The over-tones of the microphone voltage are sifted out by the filter IIpassing only high frequencies, and they are then broughtto about the same intensity as the fundamental in the ampli-fier V. The potentiometcr P is so adjusted that the outputvoltages of Tl and T2 exactly compensate each other, whichis ascertained by means of the galvanometer G. The setting ofP then provides a measure of the ratio of (U22 + ua2+ U.,2 + ...)and ut2• The filter L can also be omitted. One then actuallymeasures the ratio of . .

(U22 + ua2+ ul + ...) and (U12 + U22 + ua2 + ...),

by which the value of Fl is also determined unambiguously.

Recording the [requency characteristic

In order to record the frequency characteristicthe voltage of the measuring microphone is mea-sured while the frequency of the voltage on theloud speaker is varied. The voltage measured willat any frequency be proportional with the electricalenergy supplied to the loud speaker. One wouldtherefore prefer to keep this energy constant duringthe recording of the characteristic. This, however,requires rather elaborate measures, since the im-pedance of the loud speaker changes with the fre-

quency. In practice therefore either the amplitudeof the voltage or that of the current is kept con-stant during the measurements, and one, then ob-tains two frequency characteristics which will ingeneral be :different. The characteristic recordedat a constant voltage amplitude is important whenthe loud speaker is to be connected with an amplifierwhich gives a constant' output voltage, as forexample for radio distribution. If however thefinal amplifier gives a constant current (as in or-dinary receiving sets with pentode) it is clear thatthe loud speaker to be used must be judged accor-ding to the characteristic. recorded -with c0J?-stantcurrent.The voltage with variable frequency required for

recording the characteristic is produced by a tonegenerator, the frequency of which can be variedwithin very wide limits (25 to 16 000 cis.) bymeans of a rotating condenser. The voltage ofthe ~one generator is supplied to the loud speakervia an amplifier. This amplifier must not of coursefalsify the loud speaker measurement, and musttherefore have a perfectly flat frequency charac-teristic and only very slight distortion. Furthermoreit must be able to give; as desired, a practically'constant output voltage or output current, which, means that the internal resistance must practicallybe made zero or infinite.

The point-by-point determination of a frequencycharacteristic is a time-consuming operation, since,because of the sudden variation which may occurin the curve, it is necessary to measure very manypoints. Since such measurements must be carriedout repeatedly in the Philips Laboratory, an in-stallation is used which aûtomatically records therequired frequency characteristics. The principleof the apparatus is very simple (fig. 4). A syn-chronous motor turns the variable condenser of thetone generator so that the frequency of the loudspeaker- current 'varies continuously, while theamplified voltage of the measuring microphone isrecorded by a recording voltmeter on a band movingat a constant velocity. It is however desirable todraw the frequency characteristic in logarithmiccoordinates in order to do justice to all par~s of the

3'4110

Fig. 4. Diagram of the apparatus for the automatic recording of frequency characteristics,Mo motor, T tone generator, V amplifier, L loud speaker, M microphone, R recordingvoltmeter, S moving band.

DECEMBER 1939 THE TESTING OF LOUD SPEAKERS 357

extensive frequency and intensity range 2). As tothe frequency, a logarithmic scale is obtained byarranging that the driving mechanism of the tonegenerator is such that the angle between any twopositions of the mechanism is always proportional-to the ratio between the two corresponding fre-quencies. In other words, with motion at constantvelocity the frequency passes through an equal'number of octaves in equal time intervals, and anoctave thus always has the same width on themeasuring band .. For the voltages to be me~silred, logarithmic

registration is obtained in the following way 3).'I'he voltage to be measured is supplied to an am-plifier whose amplification can be regulated by aso-called logarithmic potentiometer. This is apotentiometer in which the resistances between thesuccessive contacts form a geometrical progression(see fig. 5); a. rotation of the contact arm througha given angle is always followed by a change of

. the output voltage by the same factor, no matterwhat the initial position may be .. With a giveninput voltage of the amplification there is thereforeproportionality between the angle of. rotation ofthe potentiometer and the logarithm of the outputvoltage delivered. The converse is also true: inorder to obtain a constant output 'voltage the poten-tiometer must. in each case he set, so that its po-

2) R. Vermeulen, Octaves and decibels, Philips teclm.Rev. 2,47,1937.

3) Cf also: E. C. Wente, E. H. Bedell, K. D. Sw ar t jr.J.Acoust. Soc. Am. 6,121,1935; E. Meyer and L. Ke id el,E.N.T., 12, 37, 1935; H. J. von Br au n ràü h l and W.Weber E.N.T. 12, 221, 1935.

sition is proportional to the logarithm of the inputvoltage. This "volume cozi.trol" is automaticallyobtained by causing the shaft of the potentiometer

Fig. 5. Logarithmic potentiometer. At each jump of the con-·tact arm the resistance between A and B increases by the samefactor f (here assumed to be 2). The potentiometer in the ap-paratus described has 60 steps. There are four such potentio-meters with f = 1.122, 1.059, 1.029 and 1.011 respectively.The vo1tage scale then covers 60 X .r = 60 db; 30 db, 15 dband 6 db respectively. Moreover such potentiometers are ingeneral use as volume regulators, in radio sets for example,for the purpose of obtaining a uniform variation of the sub-jective sound intensity according to the law of Web er andFechneL . .

to he driven simultaneously by two magneticcouplings which turn in opposite directions at :aconstant velocity (see fig. 6), one of which is morestrongly magnetized when the output voltage of theamplifier is greater, the> other when it issmaller than a constant comparison voltage serv-ing as zero level. in the first case therefore cou-pling 1. dominates and carries the potentiometershaft along with it, so that the output voltage islowered until it reaches the value of the com-parison voltage. At that moment the two couplings .have the same magnetization, the two couplesare in equilibrium and the potentiom~ter remains

Fig. 6a. Driving mechanism of the logarithmic potentiometer 4) by means of two magneticcouplings 1 and 2. The iron discs Pl are rigidly fastened to the shaft A, which bears the armof the potentiometer P and the recording style R, and lie in slight frictional contact withthe drums T which turn continually in opposite directions, When the magnet in coupling1 is more strongly magnetized than that in 2 the shaft A is carried around in one directiondue to the greater pressure and the consequent increased friction between Tand Plofcoupling 1, and vice versa. On the recording film S the position of the potentiometer isregistered, The mechanism reacts extremely rapidly. When the drums T are allowed torotate at their maximum velocity the whole voltage scale (60 steps of the potentiometer,see fig. 5) is passed over in 1/10 sec. In ordinary cases, in order not to subject the couplingsto unnecessary wear, a velocity of 1/10 of the maximum is used. .

358 PHILIPS TECHNICAL REVIEW Vol. 4, No. 12

Fig. 6b. Photograph of the mechanism sketched in fig. 6a. Onthe right may be seen two of the potentiometers of fig. 5 fora scale of GO and 30 db, respectively.

stationary. Wh~n the output voltage is lower thanthe comparison voltage, coupling 2 takes over andthe potentiometer is taken along in the oppositedirection until once more the output voltage isequal to the comparison voltage. The arrangementworks very rapidly, so that rapid variations ofthe input voltage of the amplifier (steep parts ofthe frequency characteristic to be measured) arefollowed.

A style is attached to the shaft of the potentio-meter, which records tbe position of the potentio-meter on the above mentioned band, which movesin the direction of the shaft (fig. 6). The band is apiece of "Philimil" film 4), in the black coveringlayer of which a transparent line is scratched bythe recording style. The record obtained is repro-duced by photographic printing, and the scale offrequencies and voltages is introduced on thepositive.

In ,fig. 7 is a photograph of the complete instal-lation which clearly shows the different parts herediscussed.

Further consideration of the conditions ofmeasurement

In the above we have continually spoken of"the" distortion and "the" frequency character-istic of a loud speaker. Actually however the deter-mination of these factors for a single loud speakermay produce very varied results. In the first place

4) This film is used in sound recording by the Philips-Millersystem (Philips techno Rev. 1, 230, 1936). The "Philimil"film has for our purpose the advantage that it is very easyto work with.

Fig. 7. The complete installation for automatic recording of frequency characteristics. Onthe left the recording apparatus proper, in duplicate. On the right two tone generatorswhich supply the voltage for the loud speaker, and behind them the amplifier for thisvoltage. On the wall to the right and left of the window a series of contacts and bindingposts which provide connections with other rooms where loud speakers may be set up formeasurement.

DECEMBER 1939 THE TESTING OF LOUD SPEAKERS 359

the sound radiation of the loud speaker is not thesame in both directions, and in the second place,under ordinary circumstances, in addition to thedirect sound of the loud speaker, there is also theindirect sound, which reaches the microphoneonly after one or more reflections against thewalls, etc. How must the measurements be arrangedin order to obtain results which really do form ameasure of the usefulness of the loud speaker '?

a

Fig. 8. Frequency characteristics of two different loud speakersrecorded in an ordinary living room. The combination ofdirect and indirect sound causes numerous irregularities whichrender the interpretation of the curve practically impossible.

It would seem obvious that the measurementsshould be carried out under the same conditionsas those under which the loud speaker is used inpractice, thus, for example in an ordinary livingroom. The result of such a measurement is givenin fig. 8. Due to the interference of direct and in-direct sound numerous irregularities are obtainedin the frequency characteristics, and the resultingdiagram is so little characteristic of the loud speakerused that one could say with some truth that fig. 8represents a measurement' of the properties of theroom rather than of the loud speaker.

Measurement of the direct sound

In the open air

In order to gain some information about theloud speaker itself, it is thus necessary in the firstplace to measure the direct sound alone. All re-flecting walls must be removed from the soundfield, which makes it advisable to carry out themeasurements out of doors. When this is doneonly the ground remains which might still causeunwanted reflections. This difficulty can howeverbe met by placing the loud speaker and microphonevery high, for instance projecting over the edgeof a high roof. Such an arrangement is shown III

fig. 9.Because of the above-mentioned dependence of

the sound radiation on the direction, the frequency

characteristic and the distortion must be measurednot only at the axis of the loud speaker, but alsoin directions at different angles to the axis. Theloud speaker must therefore be placed on a turn-table so that it can be set at different angles withrespect to the fixed connecting line between themicrophone and the centre of the turn-table.

For a complete record the measurements shouldbe done in all possible directions. The investigationwould however thus become extremely elaborate,and moreover the result (a whole series of charac-teristics) would be very difficult to judge. Forthese reasons it is very often considered sufficientto take a measurement on the axis of the loudspeaker and one in a direction at 45° to this axis.Two such frequency characteristics are reproducedin fig. 10. These measurements are supplemented

Fig. 9. Set up of loud speaker and measuring microphone (M)in the open air. The loud speaker (radio set) is placed on aturntable. In front of it is the microphone amplifier. For theavoidance of interferences due to induction the connectionsfrom the microphone to the amplifier are shielded by a tube.The large plate seen in the background is used as an "infinitelylarge" baffle in the investigation of separate loud speakersystems, i.e. when the effect of cabinet or baffle of the loudspeakers must be excluded. In order in such cases also to beable to measure in different directions, the measuring micro-phone with its amplifier is fastened to a rotating arm.

360 PHILIPS TECHNICAL REVIEW Vol. 4, No. 12

by a determination of the sound intensity, keepingthe frequency constant and varying the directioncontinuously. In this way a diagram is obtained

b

Fig. ID. Frequency characteristics of a single loud speakermeasured in the open air at angles of 0° and 45°, respectively.The loud speaker is the same as in the case of fig. Ba, but washere placed in the large baffle of fig. 9.

of the directional distribution of the sound intensitywhich may be compared with light distributioncurves such as are customarily recorded for sourcesof light. The directional diagram must be recordedfor different frequencies (for instance 1 000, 2 000and 4000 cjs), since loud speakers generally showa selective directional effect 5). The directiondiagrams are recorded with the same recordingapparatus as the frequency characteristics, wherebyonly the adjustment of the tone generator remainsunaltered while the loud speaker is turned con-tinuously.

The "soft chamber"

Measurement out of doors has two objections.The results are influenced by extraneous noise, andone IS dependent upon the weather. As for thefirst objection, it may be met by introducing afilter behind the microphone, which only passesthe frequency which is supplied to the loud speaker.There is however no solution to the difficulty of rainand wind. There have therefore been made numerousattempts to imitate out-of-doors conditions In aroom, a so-called "soft chamber". For this purposethe walls of the room to be used must be coveredwith a material which absorbs sound as completelyas possible. The available technica] dampingmaterials have been found inadequate. With thevery best materials of this kind it IS scarcelypossible to attain absorption coefficients higher than70 per cent. The effect of the remaining reflectionson the measurements can be roughly calculatedIn the following way. The variations III pressure

5) J. de Boer, Sound diffusers in loud speakers, Philipstechno Rev. 4, 144., 1939.

in a reflected sound wave, with the absorption

coefficient mentioned, amount to 100Vl-0.7 = 55per cent of those in the original wave. Let us as-sume that the path of a reflected beam which strikesthe microphone is on the average five times as longas the path of the direct beam; as the intensityof the sound decreases with the square of thedistance and the sound pressure proportionally tothis distance, the variation in pressure at the mi-crophone in the reflected wave is Ij 5 X 55 = 11per cent of the direct beam. In an ordinary rectan-gular room there are six reflecting surfaces, andthus the same number of beams reflected once.In the most unfavorable case all reflected rays canbe in phase with the direct ray at the microphone,or all of them will be in the opposite phase atthat point. The variations in pressure measured maythen in the first case amount to 100+ 6 X 11= 166per cent of the undisturbed value, and in the otherto 100 - 6 X 11 = 34 per cent.

One must therefore have a much higher absorp-tion coefficient than 70 per cent. It may be seen inJig. 11 how the wall covering of the "soft chamber"is actually constructed. An absorption coefficientof about 97 per cent has been reached in~this case.

Fig. 11. The "soft chamber" in the Philips laboratory. Inorder to make the walls as sound absorbent as possible theyare first covered with a layer of a technical absorbent material(slag-wool). In front of this layer and perpendicular to thewalls strips of crepe paper 1/" m wide are hung P/~ cm apart.On the floor this covering is replaced by a series of curtainshung close to each other, which can be pushed to one side whenit is necessary to enter the room in order to set up the mea-suring microphone.

DECEMBER 1939 THE TESTIl)'G OF LOUD SPEAKERS 361

If the above estimation is repeated with this valuemaximum deviations of 20 per cent are found,which means therefore about 2 db too much ortoo little. This is still rather high but it is never-theless permissible for many measurements.

p

3'4104

Fig. 12. Decrease of the sound pressure with increasingdistance from a point source of sound measured in the chamberof fig. 11 for frequencies of 500 and 5 000 cjeec, respectively. ,The theoretical decrease (with an absorption by thè walls of100 per cent) is indicated in each case by a broken line.

The' residual influence of the walls can also becontrolled directly in a simple way, by studying thedistribution of pressure around an approximatelypoint source of sound. The sound pressure in thatcase, in the entire absence of indirect sound, wouldbe independent of the angle and would decreaseproportionally with the distance. In fig. 12 is adiagram of the distribution of pressure me~sured.As may be seen the theoretical variation is fairly ,closely approached.We may mention a general consideration for all

measurements . of the direct sound from loudspeakers. In order to rende! the influence of dis-turbances as small as possible, it is desirable tomeasure at high intensities, and it would thereforeseem advisable to place the microphone close to the

, loud speaker. This .is, however, impossible for thefollowing reason. Since an ordinary l{)'udspeaker'is anything but a point source, interferences mayoccur b~tween the waves which 'are radiated bydifferent parts of the cone. If therefore the mi-crophone is shifted along the axis of the loudspeaker, successive minima and maxima of inten-sity are observed. Orily at distances greater thantwo or three times the diameter of the radiatoris this eff~ct diminished to such a degree that the

measurements experience no appreciable, effect.For the investigation of ordinary radio loudspeakers, therefore, it is a general rule that the mi-crophone must stand at least 1 m from the loud'speaker.

Total sound radiation of a loud speaker

In certain respects analogous concepts can beused in the study of sources of sound as in that ofsources of light. This was noted above in' the dis-cussion of directional distribution. Another quan-tity which is always used in light technology andfor which an analogue can be conceived in acousticsis the totallight flux. The analogue is obviously thetotal sound radiation produced by a loud speaker.This quantity is particularly important when it isa question of "filling a room with sound", i;e.when the indirect 'sound contributes a considerableportion to the formation of the sound field. In theextreme case when the indirect sound' dominatescompletely over the direct (i.e. when there is 'avery long reverberation), even the average 'densityEm of the sound energy in the room is still dependentonly on the total sound radiation ..W:

,/

TEm = 0,072 V' W .

rit.:(3)

T and V are the reverberation time and the volumeof the room,' respectively ~1i,The total sound ra-diation can be determined in different ways. Thèmost direct method consists in recording the soundintensity' la in the above-described way as a func-tion of the angle a to the axis, and calculating thetotal sound radiation W by means of the followingformula (see fig. 13 for its derivation):

• :n;

W = 2 :Tt r2 f Ia sin a da .'o

It is unnecessary to state that this method is quite.elaborate. Another method is the following. The'electrical power taken up' by the loud speaker isdetermined a) urider normal conditions, i.e. when'it is radiating sound, and b) when radiation hasbeen made impossible by fixing the loud speakercoil. The difference may be cónsidered as theacoustic power radiated. Apart from the lack ofaccuracy of such a differential measurement (theacoustic power amounts to only a few per cent ,of the total 7)) the~e is also the objection that the

(4)

6) The equation may easily be derived from equations (4)and (5) in the article by A. Th. van Urk,' Auditoriumacoustics and reverberation, Philips technoRev. 3, 65, 1938.J. de Boer, The efficiency of loud speakers, Philipstechno Rev. 4, 301, 1939.

362 PHILIPS TECHNICAL REVIEW Vol. 4., o. 12

mechanicallosses which occur during the movementof the coil and which are not accurately known,are also included in the difference of power found.

34111

Fig. 13. Since we may consider the axis of the loud speaker(a = 0) as the axis of symmetry of the sound field, the samesound intensity la will he found at all points on the shadedsegment of the sphere. The area of the segment amounts to2:n . T sin a . T da. The energy iucident on the segment isla . 2:n . r sin a . r da, and tbe total sound energy radiatedby the loud speaker is found by integrating oyer all valuesof a. Tbe result is formula (4).

The simplest method of determining the totalsound radiation makes use of the relation given byequation (3). The conditions under which thisequation is valid are realized in a so-called "hardchamber" i.e. a room in which the walls absorblittle sound. Due to the repeated reflection of thesound waves, the sound field in this case, as isrequired by the formula, is practically entirelydetermined by the indirect sound, and in theideal case the average energy density Em is thesame at all points in the room. To use again theanalogy with light, the "hard chamber" may becompared to an Ulbricht sphere. In the acousticexample however there is an additional difficultyto be met, which does not occur in the optical case,namely the occurrence of standing waves: althoughthe average energy density Em which is to bemeasured is constant, the sound pressure never-theless exhibits strong local maxima and minima 8)due to interference phenomena. If it is desired toknow the true average, the microphone must bemoved rapidly through the chamber a distance of

B) The uniform distribution of en.ergy forms no contradictionto tbe observation of such maxima and minima in thesound measurements. The energy at every point is com-posed of a potential and a kinetic part. In tbe loop of astanding wave the potential part (tbe pressure amplitude)is at a minimum and the kinetic part (tbe velocityamplitude) at a maximum; at a node the reverse is true.Witb tbe condenser microphone we measure only thepressure amplitude; tbe kinetic part of tbe energy thusremains unobserved.

at least one wave length during the measurement.One mayalso keep the microphone stationaryand cause the maxima and minima (the loopsand nodes of the standing waves) to change theirposition rapidly and continuously. Since the posi-tions of these maxima and minima are determinedby the dimensions and shape of the chamber, bythe frequency and by the directional distributionof the loud speaker, the purpose in view can beachieved in three ways. The dimensions (shape)of the chamber can be varied by means of movingreflecting surfaces (fig. 14); the frequency of theloud speaker current can be varied quickly overa given interval, and the directional distributionof the direct sound in the chamber can be varied,for instance by allowing the loud speaker to rotateabout a vertical axis. Actually several of thepossibilities mentioned are used simultaneously inthe measurement.

Fig. 14. A view of the "bard chamber" during a measurement.On tbe extreme rigbt may be seen the reflecting surfacesin tbe form of sector-shaped partitions S, wbicb rotate in tbeirown planes. Because of its rapid motion in a wide circle themeasuring micropbone is badly blurred. This is also tbe casewith the rotating loud speaker L.

Combination of direct an indirect sound

The material which has been discussed untilnow could be summarized briefly in the followingway. The measurement of a loud speaker in anordinary room gives a result which is too compli-cated due to the incalculable superposition of directand indirect sound. In the extreme cases of a "softchamber" and a "hard chamber" simple resultsare obtained, since in the one case only the directand in the other only the indirect sound deter-mines the result. One may now attempt to ap-proximate a "normal chamber" (ordinary room) tosome extent when the measurements are 'done underout-of-door conditions, while allowing the loud

DECEMBER 1939 THE TESTING OF LOUD SPEAKERS 363

speaker to rotate rapidly about a vertical aXIS.What is then measured is the average value of thesound intensity in all horizontal directions:

2n

1 J'Id = - la da2n

o

Compared with the measurements of 10 alone (softchamber), this measurement differs by the factthat the values of la in directions other than a = 0and n/4, which manifest their presence as indirectsound in an ordinary room, are also included. Withrespect to the measurement of the total soundradiation (hard chamber) the difference is, as maybe seen from equation (4), that the same weightis assigned to all values of la, while in equation (4)the term 10 (the direct sound on the axis of the loudspeaker) was determined by the relation sin a = O.It is therefore clear that the measurement of the

integral (5) represents a certain compromisebetween a measurement of 10 alone and a measure-ment of the total sound. A frequency characteristicmeasured in this way is reproduced in fig. 15.

(5)

a

- %C:-r- es.,

-~.- -

'''±~155

L_ ~= ,-=h'S:1'=;- I ,jt5-:...rr 'u , 0 I

r::. ::::1--+ ~~~ ::--i-,

r : "-~-" .5!L

_:'J.::.7-' -= f=1_: ~.1- -I--'~- 1 1- -140

'co 'u lUll ruuu e 0 10 '"

b

Fig. IS. Frequency characteristic recorded out of doors,a) with a stationary loud speaker at an angle of 0", b) with arapidly rotating loud speaker. (The loud speaker is the sameas in Fig. 8b).

ABSTRACTS OF RECENT SCIENTIFIC PUBLICATIONS OF THEN.V. PHILIPS' GLOEILAMPENFABRIEKEN

1429: Balth. van der Pol and C. C. J. Addink:Orchestral Pitch. A cathode ray methodof measurement during a concert (Wire!.World 44, 441-442, May 1939).

A detailed article has since appeared in thisperiodical on the same subject: Philips techno Rev.4, 205, July 1939).

1430: M. J. O. Stru tt and K. S. Knol: Mcs-sungen von Strömen, Spannungen undImpedanzen bis herab zu 20 cm Wellen-länge (Hochfrequenztechn. U. Elektroakust.53, 187-195, June 1939).

Due to the inductive and capacitive couplingbetween supply lines and electrodes of the measur-ing instruments for currents and voltages - inconnection with the occurrence of skin effect inconnections and finite transit times of the electronsin high vacuum tuhes - great difficulties occur inthe measurement of alternating currents and volt-ages as well as impedances for waves shorter thana few metres. These difficulties are overcome byusing diodes, thermocouples and hot wire ammetersof a special construction working with air expansion.Furthermore these new instruments are employedin measuring impedances of several ohms to severalthousand ohms at wave lengths down to 20 cm.

1431*: A. B ou wer s : Elektrische Höchstspannun-gen (333 pages; Springer, Berlin 1939).

In this first volume of Technische Physik inEinzeldarstellungen subjects are dealt withconnected with thc generation and application ofelectrical voltages of at least several hundred kilo-volts. After a description of the different methodsused for the excitation of high tensions, follows adiscussion of the calculation of the electric fieldswhich occur with various simple forms of electrodes.On page 132 a summarizing table is given of anumber of field formulae hereby obtained. Inaddition the properties of insulating media arediscussed and the different parts of which a hightension installation is composed. In a separatechapter direct and indirect methods arc describedfor the precise measurement of high voltages, andin conclusion the most important applications arediscussed. The bibliography, which consists of324 items, makes no claim to completeness, but itdoes contain all the articles which were consultedin writing the book.

*) An adequate number of reprints for the purpose of dis-tribution is not avaible of those publications marked withan asterisk. Reprints of other publications may be obtainedon application to the Natuurkundig Laboratorium, N.V.Philips' Gleoilampenfabrieken, Eindhoven (Holland),Kastanjelaan.