electro acoustic - cieri
TRANSCRIPT
ELECTROACOUSTICMeasurements
Bruel & Kjser
Q
X
X
X
(X)
X(X)
XX
X
X
XX
aX
X
X
X
X
X
X
X
X
XX
o o
X
X
X
X
XXX
XX
X
XX
X
X
X
X
XXX
X
X
XX
0X
X
X
XX
X
XXX
X
X
XXXX
Page
33344
4667
8QO
101112121213141414
151515171818182020
2121
23242626262727
283132343536
374041
Why Make Electro-acoustic Measurements?What Measurements are relevant?Frequency ResponsePhase ResponseDistortionElectrical Frequency Response,Signal to NoiseAcoustical Frequency Response in a Free FieldLoudspeakersMicrophones"Free FieSd" Response in Ordinary Roomsusing Gating TechniquesBasic PrinciplesEarly Reflection — or "Box Sound"Mechanical Resonances of CabinetDirectional CharacteristicsHarmonic DistortionTotal HarmonicIndividual Harmonic DistortionDistortion of Phonograph Pick-upsAlternate methodsOptimum Distortion MeasurementsDifference Frequency and IntermodulationDistortionDiscrete FrequenciesSwept MeasurementTransient Intermodulation Distortion (TIP)Impulse ResponsePhase Response and Related Transient ResponseBasic PrinciplesPractical ExampleSystems with Long DelayMechanical Stability of Tape Recordersand TurntablesComplex ImpedanceInstrumentation Microphones forMusic RecordingTest RecordsFrequency response of CD-4 systemsTrackabihtyWow and flutterRumbleTone Arm resonanceFrequency Response in the Actual Listening Roomusing 1 / 3 Octave Pmk Weighted Random NoiseReverberation TimeSound PowerVisual Analysis of Speaker MotionFrequency Analysis of Noise and CrosstalkProduction Control
1011,
12,
13,14
15
16
17,18192021
APPENDIX
1 Basic Acoustic Properties and Formulae2 Typical Instrument Combinations3 Electroacoustic Standards
By Henmng Meller, Bruel & Kjaer
1. Why Make Electro-acoustic Measurements?Frequency response, phase re-
sponse, harmonic distortion, inter-modulation distortion, transient in-termodulation distortion, directivityindex, efficiency, sensitivity thelist of possible measurements is al-most endless Too often the mass ofdata assembled leads to confusion,and not insight about the quality ofthe product being measured
Therefore, the fundamental prob-lem for the electro-acousticengineer is to select the appropriatemeasurements and evaluate the re-sults by giving the proper weight tothe various measurements Hemust then translate the objective re-sults into the subjective domain us-ing understandable and relevantterms However, it is an art to dothis It requires a great deal of ta-lent, insight, instrumentation, andtime Only after a wide variety of
measurements have been made is itpossible to obtain "the complete pic-ture"
In practice, the objective and sub-jective evaluations must work to-gether to improve the final productObjective measurements alone arenot enough, subjective judgmentsalone do not suffice They must becombined Then the objective meas-urements will lead to an improvedproduct, and the subjective judg-ments will suggest improved, per-haps even new objective measurement techniques
The purpose of all these measure-ments and subjective judgments isan improved product in practice,that means higher fidelity, a betterloudspeaker, a superior micro-phone, an improved tape recorderAnd in the final analysis, whether
or not the product has been im-proved will be judged subjectivelyBut objective measurements willhave played an important role inachieving the desired end result
These notes will concentrate on abroad range of measurements andwill emphasize the dual applicationof most techniques, to acousticaland electrical measurements Thenotes are also structured along aline beginning with simple and thenmoving to more complex instrumen-tation The Bruel & Kjaer system isa building block system, where theaddition of one or more instrumentsto the basic combination permitsnew measurements This avoids du-plication of equipment and permitsadvanced measurements to be ob-tained for a relatively small margi-nal cost
2. What Measurements are Relevant?Only those measurements that in
some way, directly or indirectly affeet the subjective quality evalua-tion can be considered relevantHere we will deal with the primaryones in brief before discussing de-tailed instrumentation techniques
Frequency ResponseThe most common electrical and
acoustical measurement is fre-
quency response The pure sine fre-quency response is relevant for theelectrical part of the system Butwhen loudspeakers or other acousti-cal transducers are measured in ananechoic room, the pure sine excita-tion is not sufficient since the influ-ence of the acoustical load, the lis-tening room, is ignored Therefore,additional measurements such assound power and measurements in
a "typical" room should be made
Recent investigations have shownthat frequency response measure-ments in the actual listening roomusing third octave, pink weighted,random noise, along with power re-sponse and phase response give thebest correlation to subjective judg-ments
The main problem is defining a ty-pical room, and it is doubtfulwhether this will ever be solved be-cause of the wide variety in room di-mensions depending on buildingtype and also the country. However,if a typical room could be agreedupon, it would permit comparison ofmeasurements from different manu-facturers, and more important,would make it easier to predict theperformance of a loudspeaker in agiven consumer's room. Lackingthis standard room, it would stili bevaluable for a manufacturer tomake his own standard room withwell-defined acoustical characteris-tics.
However, anechoic facilities arestill necessary to evaluate the manydifferent phases of loudspeaker de-sign. Fortunately, it has recently be-come possible to synthesize ane-choic conditions in a normal reflec-tive room by the use of gating tech-niques (see section 5). Thus thecombination of anechoic and nor-mal room measurements shouldyield valuable results.
Phase ResponseFrequency response alone tells us
little about the transient response
of the system. But when combinedwith the phase response, the tran-sient response can be predicted. Tra-ditionally, transient response hasbeen measured — or rather evalu-ated — using tone bursts, but nowa more compiete picture can be ob-tained since phase can be mea-sured as easily as frequency re-sponse. The more linear the phaseresponse, the better the response ofthe system to transients such asfound in percussion instruments,the attack of a trumpet, the pizzi-cato of violins, etc.
DistortionThese notes will also describe
measuring techniques for varioustypes of distortion, all of whichhave audible relevance. There is to-tal harmonic distortion, individualharmonics measured separately, in-termodulation and difference fre-quency distortion, and transient in-termodulation distortion (TIM).Since distortion measurement at asingle or at spot frequencies maygive non-representative results, allof the measurement techniques de-scribed here will permit swept meas-urements of most types of distortionover the 2 Hz to 2OOkHz range.
Frequency response, phase re-sponse, and distortion: these areprobably the three most importantcategories of measurements. But inreality, all three categories are var-ious types of distortion. If the fre-quency response is not flat, the am-plitude will be distorted, and if thephase response is not linear, a timedistortion will result. In fact, virtu-ally all measurements seek to de-tect distortion of the original wave-form in some manner or another.For example, wow and flutter re-sults in frequency distortion, andnoise adds to the waveform, thusdistorting it.
As the various techniques formeasuring these types of distortionare described in the following chap-ters it is hoped that insight will begained, not only in better measure-ment technique, but in the art ofcombining the mass of objectivemeasurements in a sensible man-ner and transforming them to thesubjective domain. If this goal is at-tained, the ultimate purpose of ob-taining better high fidelity can be ap-proached.
3. Electrical Frequency Response,Signal to Noise
The basic system which will bethe foundation of most measure-ments discussed in this course con-sists of three instruments, a sinegenerator, a voltmeter, and a levelrecorder. In the Bruel & Kjaer sys-tem, these three instruments havea number of unique features thatgive them great flexibility and per-mit them to be the heart of a morecomplex measuring system. Thegenerator used is the Sine Genera-tor Type 1023 which provides athree decade linear or logarithmicsweep in synchronization with theLevel Recorder Type 2307. The2307 records the results of themeasurements on pre-printed cali-brated paper. The Measuring Ampli-fier Type 2606 is a true RMS
Fig. 1. instrument set-up for electrical frequency response
voltmeter providing a calibrated ACor DC output for connection to a Le-vel Recorder. With this basic sys-tem, frequency response, signal tonoise ratio, and power can be mea-sured. Later in the notes, other in-struments will be added to this sys-
tem to permit a wide variety ofother measurements.
A typical instrument set-up forthese measurements is shown inFig 1 and a typical curve is shownin Fig.2.
To give greater flexibility, threedifferent generators are available
1023 10 Hz to 20 kHz, hn or logsweep, 0 ,1% distortion, war-ble tone
1027: 2 Hz to 200 kHz, lin and logsweep, 0,01% distortion,narrow band random noise,pink noise, white noise
2010: 2 Hz to 200 kHz, lin and logsweep, 0,01% distortion,plus a narrow band analyzerwith the same frequency anddynamic range
g.2. Frequency response curves measured with set-up from Fig.1 influence of tone controlsand filter of an audio amplifier is shown
fined at the point where clipping be-gins, or the distortion reaches agiven percent (for example 1 or3%). The signal is then removedand the input is terminated with itscharacteristic impedance The gainof the measuring amplifier is thenincreased to read the noise voltage(N). The ratio of the two voltages(S/N) in dB is the signal to noise ra-tio
When measuring noise, the band-width or weighting network usedmust be specified. If the measure-ment is made with linear response,the upper and lower frequency li-mits must still be specified. The Aweighting or a psophometer filtermay also be used to give a closer ap-proximation to audible levels Usu-ally, measurements with a pso-phometer to DIN 45 405 will giveabout a 7dB higher noise level thanwith the A weighting network.
Often, frequency response curvesare made using a simple RC oscilla-tor and a voltmeter. However, thisonly permits measurements of onefrequency at a time.
The advantage of the Briiel &Kjaer system is that the frequencyof the oscillator is automaticallyswept in synchronization with theLevel Recorder, giving permanentdocumentation on the preprintedpaper. The generator covers threedecades in an accurate logarithmicsweep, and unlike most RC oscilla-tors, has a stable output amplitudewhile sweeping. And since the sig-nal is generated using two high fre-quency signals, these may be usedfor easy synchronization of add-onslave filters for distortion measure-ments. The only disadvantage of thebeat frequency oscillator is that theultimate distortion figure currentlyattainable is about 0,01% while afew of the very best RC generatorsmay be slightly better.
For overall record/playback re-sponse measurements of tape recor-ders, the 1023/2307 combinationmay be used provided that thesweep is slow enough so the delaybetween heads is not significant. Orthe paper can be off-set to corre-spond to the delay, thus permittingfast sweeps. When creating a stand-ard test tape to use for playback re-sponse measurements, the Re-sponse Test Unit Type 441 6 is usedto provide the proper automatic syn-chronization to the sweep of theLevel Recorder (Fig.3).
Signal to WoiseThe above system of Beat Fre-
quency Oscillator and MeasuringAmplifier is also used for signal tonoise ratio measurements The max-imum output voltage (S) of the sys-tem is measured when operatinginto a typical load. This is measuredusing a pure sine in the mid-fre-quency range of the system, andthe maximum point is usually de-
Fig.3. Frequency response measurements of a tape recorder
in
Fig 4 Set up for acoustical frequency response measurements
LoudspeakersBy adding a Power Amplifier Type
2706 to the output of the generatesand a Condenser Microphone Type4133 with Preamplifier Type 2619to the basic system described abovefor electrical measurements acous-tical response measurements maybe made The Condenser Micro-phone has flat response within 2 dBover the entire 4 Hz to 40 kHz fre-quency range Fig B Sine response curve of a medium quality two way loudspeaker system
This measurement requires a freesound field — either an anechoicroom or an outdoor location (seeAppendix) This makes this type ofmeasurement somewhat inconvenient and expensive if free field facili-ties are not already available (Insection 5 we will show how to simu-late free field conditions using gat-ing techniques) A typical instru-ment set-up and response curve isshown in Figs 4 and 5
This measurement may also beperformed using bands of noiseTwo types of noise bandwidths areavailable With constant bandwidthnoise, the bandwidth is independ-ent of the center frequency There-fore, this is most often used with alinear frequency scale For constantpercentage bandwidth noise, thebandwidth is always equal to afixed percentage of the center fre-quency, such as 23% for third oc-tave noise (Fig 6) Therefore the ab-solute bandwidth increases with in-creasing frequency, but when pre-sented on a logarithmic scale thebandwidth will appear to be con-stant at all frequencies
Logarithmic Frequency Scale
Fig 6 Constant bandwidth and constant percentage bandwidth filters
Fig.7. Set-up for response measurements using narrow? band noiseSine Random Generator Type1027 may be used to generate boththese types of noise. It provides con-stant bandwidth noise with band-widths selectable from 3,16 Hz to1000 Hz. It also provides pink noisewhich then may be filtered by an ex-ternal third octave filter such asType 1615. If only constant percen-tage bandwidth noise is required,Noise Generator Type 1 405 may beused instead. A typical instrumentset-up is shown in Fig.7 and the re-sulting curve is shown in Fig.8. Inthis set-up the switching of the fi l-ters of the 1615 from one centerfrequency to the next is automati-cally synchronized with the LevelRecorder. Fig.8. Response of loudspeaker to third octave noise
achieved by using a reference micro-phone (Type 4133 is generally rec-ommended) whose signal is fedback to a compressor circuit in theSine Generator 1023 (Fig.9). Theoutput level is then automatically ad-justed to give flat frequency re-sponse at the point of the referencemicrophone. The microphone undertest is then placed relatively close
to the reference microphone andthe frequency response at that loca-tion is assumed to be flat. If the twomicrophones are placed too close toeach other, however, the reflectionsfrom the microphone under test willinterfere with the sound field at thereference microphone, and viceversa.
MicrophonesMicrophone measurements are
slightly more difficult since they re-quire the establishment of a soundfield with a constant sound pres-sure independent of frequency.This, of course, implies a loud-speaker with flat frequency re-sponse, which is not attainable.However, this can be artificially
Fig.9. Compressor loop ho!ds sound pressure constant for measurement on microphone
Fig.11. Response displayed on screen ofFrequency Response Tracer
Fig.10. Frequency response of an inexpensive dynamic microphone
The resulting response curve 4712 (Fig.11). The response tracer(Fig. 10) maybe plotted on Level Re- is especially of use in productioncorder paper, or may be displayed testing since upper and lower limitson Frequency Response Tracer Type may be drawn on the screen. How-
ever, the trace on the 4712 is notretained permanently.
5. "Free Field" Response in Ordinary Roomsusing Gating Techniques
Basic PrinciplesBy adding the Gating System
Type 4440 to the above mentionedbasic system of Oscillator, Measur-ing Amplifier, and Level Recorder,free field conditions may be simu-lated in ordinary reflective rooms.The restriction on lower limiting fre-quency is related to the size of theroom, as is the case for anechoicrooms.,With the gating system, fre-quency response, directional charac-teristics, harmonic distortion, andphase of the loudspeaker may bemeasured. In addition, the GatingSystem may be used as a tone
burst generator for traditional toneburst evaluations of transient re-sponse. The system may also beused for measurement of early re-flections, indicating the response ofthe loudspeaker a few millisecondsafter the stop of the tone burst.
Since these measurements are de-scribed in detail in Application Note15—107, "Electro-acoustic FreeField Measurements in OrdinaryRooms — using Gating Techniques"(Fig. 12), only a short descriptionwill be given here.
The principle of the gated meas-urements is shown in Fig. 13. Atoneburst (A) is fed to a loudspeakerplaced in an ordinary room. The sig-nal received by the measuring mi-crophone contains not only the origi-nal burst, but also several reflec-tions from walls, floor, ceiling, etc.(B).
In addition, the shape of the in-itial burst is distorted. The first partof the burst contains the overshootof the loudspeaker and the last partcontains the overhang of the loud-
Fig. 13. Principle of gated measurementsFig. 12. B & K Application Note 15-107
speaker But somewhere in the mid-dle is a steady state portion whoseamplitude is equal to the free fieldresponse This section is selectedby a measuring gate (C) with appro-priate width and time delay Thelevel of the gated signal (D) canthen be measured
A key idea in the Gating Systemis that the received signal can be"edited" using the measuring gateThus all parts of the signal may beexamined individually and interest-ing phenomena such as early reflec-tions may be studied
Fig 14 Set-up ysing Gating System
A typical instrument set up isshown in Fig 14 The output of theSine Generator 1023 is passedthrough the transmitting section ofthe Gating System Type 4440 andtransformed into a toneburst of ad-justable length (0,3 ms to 1s)which then is fed through thepower amplifier to the loudspeakerThe received signal from the Mea-suring Amplifier passes through thereceiving section of the Gating Sys-tem and is gated by the measuringgate, whose width and delay arealso adjustable over a wide rangeA peak detector measures the ampli-tude of the desired signal and feedsa DC voltage proportional to thisvalue to the Level Recorder for auto-matic recording of frequency response The peak detector containsa hold circuit which is reset foreach new toneburst to permit cap-turing the new amplitude as the frequency changes To monitor the ad-justment of the Gating System, atwo channel oscilloscope is essen-tial (Fig 15)
Fig 16 Frequency response of loudspeaker measured in anechoic chamber
Fig 15 Proper adjustment of gate seen onoscilloscope Fig 17 Frequency response of same loudspeaker measured with the Gating System
Typical results of gated measure-ments are shown in Figs 16 to 19and show good correlation betweenanechoic measurements and GatingSystem measurements
Directional characteristics of loudspeakermeasured with pure sine in anechoic cham-ber
Directional characteristics of same loud-speaker measured with the Gating System inan ordinary room Note the 10 kHz curveshows the background noise level of theroom
Early Reflections — or"Box Sound"
If the measuring gate of the Gat-ing System 4440 is adjusted just af-ter the end of the tone burst, earlyreflections inside the box which of-ten are predominant at cabinet res-onance frequencies, may be mea-sured as a function of frequency.Such typical curves are shown inFig.20 It is seen that the steadystate curve contains only a slightpeak at 3 kHz which is much morepronouced in the early reflectioncurves Thus the subjective soundcoloration at 3 kHz might be greaterthan the steady state curve sug-gests
Directional characteristics of loudspeaker Directional characteristics of loudspeakermeasured with Gating System in anechoic measured with pure sine in an ordinary roomchamber. The sharp dips and peaks are notas pronounced as for pure sine
A more sophisticated set-up forthis type of measurement is shownin Fig.21 Here, white noise isgated in bursts and the transmittingsection of a second gating system isused to process the received signaland pass it on to Real Time NarrowBand Analyzer Type 3348 whichdisplays the entire frequency spec-trum with 400 lines resolution
Fig.20. frequency response curves as a function of time after the stop of the tone burst indi-cating the influence of early reflections
10
Fig 21 Set-up for detailed investigation of early reflections using two Gating Systems, and Real Time Narrow Band Analyzer Type 3348 to exam-ine qated white noise
Recent tests in Denmark haveshown the value of these types ofmeasurements By use of appropri-ate damping materials, the ampli-tude of resonance peaks was re-duced approximately 1OdB and anaudible improvement was noted
the Level Recorder This will clearlypoint out the resonances (Fig 22)Any section of this curve may be ex-panded for a more detailed examina-tion, and the frequency marking pro-vided by the Heterodyne AnalyzerType 2010 still permits frequencycalibration of the paper as alsoshown in the figure
Mechanical Resonances ofCabinet
One possible reason for early re-flections is mechanical resonancesof the loudspeaker cabinet Thesecan readily be measured by placingan accelerometer on the cabinetand sweeping the generator andplotting the acceleration level on
Fig 22 Measurement of loudspeaker cabinet resonances
11
6. DirectionalCharacteristics
The basic measurement systemconsisting of the Sine GeneratorType 1023, Level Recorder Type2307, and Condenser MicrophoneType 4133 (with preamp ) can eas-ily be expanded to perform direc-tional characteristic measurementsby adding Turntable Type 3922 Apractical set-up and typical result isshown in Fig 23 and Fig 24
These measurements are espe-cially of importance in determininghigh frequency dispersion and canalso be used for calculating soundpower The loudspeaker (or micro-phone) under test is placed on theturntable which rotates in synchroni-zation with the Level Recorderwhich is fitted with polar paper.Power to the loudspeaker is pro-vided through slip rings to preventcables from getting tangled. If com-plete directional characteristics arenot desired, the system may be setto only sweep through a restrictedangle, say 60° while still maintain-ing Level Recorder synchronization
Fig 23. Typical directional characteristics of a loudspeaker
Fig.24. Turntable is used to measure directional characteristics
7. Harmonic Distortion
Total HarmonicUnfortunately amplifier and loud-
speaker systems introduce other fre-quencies into the output of the sys-tem than were present at the input.This is primarily due to unlmearitiesin, for example, transistors, or themagnetic field of the loudspeakerFor a single input frequency this re-sults in harmonic distortion, forseveral input frequencies, intermod-ulation distortion.
Harmonic distortion is illustratedin Fig.25 where a 1 kHz tone is in-
Fig.25. Illustration of harmonic distortion. Band stop filter rejects fundamental for measure-ment of total harmonic distortion. Band pass filter can measure distortion componentsindividually
12
troduced in the system which thengenerates distortion components at2 kHz (second harmonic), 3 kHz,4 kHz and so on. The total harmonicdistortion for this signal is the RMSsum of the distortion components re-lated to the RMS value of the totaloutput signal.
This measurement as a functionof frequency (which is required forexample, by DiN 45500} is madeby simply adding Heterodyne SlaveFilter Type 2020 to the basic instru-mentation package mentioned. A ty-pical set-up and result is shown inFig.26. The filter is connected tothe high frequency tuning signalsprovided by the Sine GeneratorType 1023 thus ensuring automaticsynchronization
Individual Harmonic DistortionAlthough total harmonic distor-
tion measurements are useful froma convenience standpoint, more de-tailed information, especially of in-terest to the designer, is availableby measuring the amplitude of eachof the harmonics separately as afunction of frequency. This is simplydone by expanding the system by ad-ding Tracking Frequency MultiplierType 1901 as shown in Fig.27. The1901 locks onto the low frequencysine output of the 1023 and gener-ates the tuning signals necessary to
tune the 2020 to the required har-monic. This harmonic may be se-lected by a simple thumbwheel set-ting on the 1 901 . Even subharmon-ics can be measured since the1301 has a resolution in steps of0,1 — i.e. from the 0,1 to the99,9th harmonic. This instrumenta-tion set-up has a dynamic range of70 dB (0,03% distortion) in theband-pass mode, but is restricted to55 dB (0,2%) in the rejection mode.If greater dynamic range is re-quired, the instrumentation shownin Fig.29 or Fig 32 is suggested
x 100%d =a, 2 +
where ax is the amplitude of thex'th harmonic.
This measurement is readilymade by a rejection filter which re-jects the fundamental while passingall harmonics (including noise andhum). The output of the rejection fi l-ter is then the distortion level,which when plotted on Level Recor-der paper will be in dB below thereference level (— 20 dB = 10%,—40 dB = 1%, — 60 dB = 0 ,1%etc.).
Fig.26. Total harmonic distortion measurement
Fig.27, Individual harmonic distortion is measured by adding Tracking Frequency Multiplier 1901 to set-up in Fig 26.
13
Distortion of PhonographPick-ups
Distortion of pick-up cartridgesmay be measured using a similarset-up. However, instead of the gen-erator, Bruel & Kjasr Test RecordQR 2009 or QR 2010 is usedwhich contains a sine sweep thatcan be synchronized with the LevelRecorder using Response Test UnitType 4416. Such a set-up is shownin Fig.28.
Alternate MethodsHarmonic distortion measure-
ments may also be performed usingFrequency Analyzers Type 2120 orType 2121. These analyzers con-tain constant percentage bandwidthfilters of bandwidths of 1%, 3%,10% and 23% (1/3 octave) plus aband stop filter, high pass and lowpass filter. All of these filters arecontinuously tunable over the 2 Hzto 20 kHz range, although they can-not be synchronized with the gener-ator. Hence measurements can onlybe made at discrete frequencies.However, for measurements of dis-tortion at very low frequencies (be-low 20 Hz), these filters are supe-rior to constant bandwidth filterssince they have a narrower abso-lute bandwidth at these frequen-cies.
Real Time Analyzers Types 3347and 3348 can also be used to viewdistortion, and have the advantagethat all distortion components areseen simultaneously. The nominaldynamic range of 50dB (0,3% dis-tortion) of these analyzers can be ex-tended by using a high pass filter(such as that in Type 21 20) to atten-uate the fundamental.
Fig.28. Distortion measurement of pick-ups
Fig.29. To obtain maximum dynamic range. Heterodyne Analyzer 2010 is used for distortionmeasurementsOptimum Distortion
MeasurementsWhen a wider dynamic and fre-
quency range is required than possi-ble with the previous instrument set-ups, Heterodyne Analyzer Type2010 can be used to extend the fre-quency range from 2 Hz to 200 kHzand the dynamic range to 80 dB(0,01%) or better. The 2010 con-
measured while the 1902 permitsintermodulation and difference fre-quency distortion measurements. Atypical set-up using the 2010 fordistortion measurements on a com-plete tele-communications link isshown in Fig.29.
tains both a low distortion oscillatorand a narrow band filter. The Filtercan be tuned to the chosen distor-tion component using either Track-ing Frequency Multiplier Type 1901or Distortion Measurement ControlUnit Type 1902. With the 1901,only harmonic distortion can be
14
Fig.30. Intermodulation distortion measurement using two fixed frequency tones
Discrete FrequenciesWhen two or more single frequen-
cies are present in a system, the un-lineanties of the system will resultnot only in harmonic distortion, butwill also generate sum and differ-ence components called intermodu-lation distortion. This measurement
can be made using two generatorstuned to different frequencies andthen analyzing the output of the ob-ject under test using Heterodyne An-alyzer Type 2010. The plot (Fig.30)then shows the various distortioncomponents and their amplitudes.In the figure, test tones of 200 Hz
and 1,5 kHz are used. However,this method only yields informationabout the system at these two fre-quencies, so many curves of thistype need to be run to get a morecomplete picture.
Swept MeasurementsHowever, by using Distortion
Measurement Control Unit Type1902 with the 2010, a completesystem is available that permits dis-tortion measurements while thetest frequencies are swept. The sys-tem serves as a two-tone generator,thus permitting intermodulation anddifference frequency, as well as har-monic distortion measurements.The different capabilities of the1902/2010 combination areshown in Fig.31 and a typical instru-ment set-up and distortion curveare shown in Figs. 32 and 33.
In the harmonic mode, a singletone output is provided by the 1 902which is in the 2 Hz to 200 kHzrange. The analyzer may be tunedautomatically to any harmonic up tothe 5th.
In the difference frequency mode,the 1902 generates two frequen-cies which are swept together with
Fig.31. Relative positions of test frequencies generated by the 1902 and the analyzer frequen-cies of the 2010 for three types of distortion measurements
15
a constant frequency difference be-tween the two. This frequency dif-ference is adjustable from 20 Hz to2 kHz. The analyzer is then set totrack any distortion sidebands up tothe fifth order and the result is auto-matically plotted on the Level Recor-der. Fig.33 shows third order minusdistortion measurements on a typi-cal high quality tape recorder, it isseen that the distortion levelreaches approximately 10% at thehigher frequencies where tape satu-ration occurs. A harmonic distortionmeasurement would not be as sensi-tive to this because of the inherenthigh frequency roll off of the tape re-corder which would reduce the lev-el of the harmonics. This distortionis also of great importance audiblysince the distortion components{the difference between the two fre-quencies) fall in the audible range,whereas harmonic distortion compo-nents at these frequencies mightnot be considered important sincethey fall outside the audible range.
in the intermodulation mode the1902 also generates two frequen-cies. However, this time the lower
Fig.32. Set-up for distortion measurements using the 1902/2010
Fig.33. Difference frequency distortion of a tape recorder
frequency is fixed, but selectablefrom 20 Hz to 2 kHz while the upperis swept and the analyzer is tunedto the plus or minus sidebands of theupper tone. The amplitude ratio ofthe two tones is 4 to 1 (12 dB) asspecified in various standards. How-ever, any other amplitude ratio mayalso be obtained.
test tape must first be prepared byrecording the twin tone test signalon one channel of the tape while si-multaneously recording the higherof the test tones (available at theBFO output of the 201 0) on the sec-ond channel. This second channelis then used in playback to triggerTracking Frequency Multiplier Type1901 which then controls the1 902 and permits automatic track-ing of the distortion components.
A similar test can also be devisedfor test records, although it requiresthe cutting of a special test record.
If the difference frequency and in-termodulation distortion measure-ments must be performed on a taperecorder that does not have simul-taneous record/reproduce facility, a
16
9. Transient Intermodulation Distortion (TIM)
Transient intermodulation distor-tion arises due to the time delay inthe feedback loop of an amplifier,and the resulting clipping of thefeedback signal. Measurement ofthe phenomenon is difficult andseveral different techniques havebeen suggested. One technique sug-gests using a broadband pink noisesignal with one frequency bandmissing, for example a 1/3 octave.The receiving side (Fig.34) wouldthen have a band pass filter whichwould measure the amount of ener-gy in this band, which in a perfectsystem should be zero. This meas-urement should be sensitive to har-monic, intermodulation and TSMsince the test signal is randomnoise which might be considered tohave a transient-like nature.
Otala in Finland has indicatedthat there is a relationship be-tween TIM and audible quality. Hepoints out that TIM depends on therelation of bandwidth of the pream-plifier to the bandwidth of thepower amplifier, and also theamount of feedback.
Another approach to measuringTIM is to use a low frequencysquare wave with a high frequencysine wave superimposed. The influ-ence of transient intermodulationdistortion can then be seen by exa-mining the frequency spectra. If de-sired, this measurement can alsobe performed using the1902/2010 combination simply byadding a Schmitt trigger to the1902 output to convert f 1 to asquare wave.
The disadvantages of all these
Fig.34. TIM measurement using broad
methods is that none of them yieldresults in numbers, or well-defineddistortion characteristics. However,a twin tone test using the 1902and 2010 gives a well defined char-acteristic and is also related to TIM.This is because TIM is primarily ahigh frequency phenomenon relatedto the slew rate of amplifiers andtheir bandwidths. Traditional inter-modulation distortion measure-ments will not disclose these prob-lems since they only cover the audi-ble frequency range. However, withthe 1902/2010 system, swept in-termodulation measurements canbe made up to 200 kHz and it is atthe frequencies above 20 kHz thatmany amplifiers show a dramaticrise in distortion due to their re-stricted slew rates etc. Hence theuse of high frequency swept inter-modulation measurements is sug-gested as a standardized measureof TIM.
From Fourier analysis we knowthat if the performance of the sys-tem at all frequencies is defined,then its transient performance isdefined. Unfortunately the Fouriertheory is only applicable for linear
band noise with a missing narrow band
systems, therefore not for distor-tion.
High frequency distortion meas-urements are not the complete an-swer to TIM but it seems in practiceto be the best indication. Thus aproblem with many current at-tempts at TIM measurements is thata wide enough frequency range hasnot been explored.
As indicated in Fig.35, the steady-state distortion of an amplifier withamplification A and feedback p is re-duced by 1/(1—78A). But, what willhappen if a transient or a step func-tion is introduced? Then, the morefeedback the greater the transientdistortion. This is illustrated Fig.35where amplifier (a) typically willhave less feedback, hence highersteady state distortion but lowertransient distortion than amplifier(b).
More details on this method ofdistortion measurement is given inthe application note 15—098"Swept Measurement of harmonic,difference frequency, and intermod-ulation distortion".
Fig.35. Increased feedback gives lower steady state distortion but higher transient distortion
17
10. Impulse Response
As indicated in Fig.36, the GatingSystem 4440 might be used forthis application since it, with a nar-row pulse width, will produce onecycle. By using a diode in series
with the output, a sin2 * function isapproximated. Since the burst widthcan be as ahort as 0,1 ms it will im-ply pulses with a frequency contentup to 10 kHz.
For investigations of impulse re-sponse, the Digital Event RecorderType 7502 is an ideal instrument.The digital memory holds up to10000 samples, each of 8 bits. Asignal captured in this memory canbe reproduced indefinitely and maybe transformed up or down in timeto fit the requirements of oscillo-scopes, external analyzers, or com-puters. The set-up shown (Fig.36)is used to record the response of aloudspeaker when excited by asingle, squared sine pulse.
The advantage of a squared sinepulse is that it contains the entirefrequency spectrum, and if suitableprocessing facilities are available,the frequency and phase responseof the loudspeaker can be calcu-lated using the Fast Fourier Trans-form. The pulse may also be viewedon an oscilloscope for subjectiveevaluation. Fig.36. Set-up for impulse response measurements
11. Phase Response and Related Transient ResponseBasic Principles
The most complete way of charac-terizing the transient response of asystem is to measure its phase andamplitude response. Then its entiretransfer function is known and thetransient response is defined. Themore linear the phase response, thebetter the transient reproduction,which especially is noticeable onmusical transients such as the at-tacks of notes, the various percus-sion instrument and string pizzicato.
Phase response measurementsare outlined in great detail in theBruel & Kjaer Application Note No.15—090 "Loudspeaker PhaseMeasurements, transient responseand audible quality". Therefore,only the main points wii! briefly besummarized here.
Loudspeaker phase measure-ments are diffucult to make be-cause the time delay between loud-speaker and microphone results ina large continually increasing phasechange. However, this problem isovercome by introducing a corre-sponding delay in the referencechannel of the Phase Meter usingPhase Delay Unit Type 6202. The6202 corrects both for the time de-lay and for the phase response er-ror of Heterodyne Slave Filter Type2020, if this filter is used. A typicalinstrument set-up for phase meas-urements is shown in Fig.37. It isseen that this builds on the same in-struments previously used with theonly addition being the Phase MeterType 2871 and the Phase DelayUnit Type 6202.
18
Fig 37 Set-up for loudspeaker phase measurements
Fig 38 Simplified block diagram of Phase Meter
The principle of operation of thePhase Meter is shown in Fig 38From the diagram it is seen that thelength of time that the flip-flop isset is directly related to the phaseangle When the output of the flip-flop is low pass filtered, the remain-ing DC component will be propor-tional to the duty cycle, and hencethe phase angle The Phase DelayUnit Type 6202 (Fig 39) is seen toconsists of two separate sectionssix shift registers which can beclocked internally or externally to de-termine the delay time, and a lowpass filter to compensate for thephase response of HeterodyneSlave Filter Type 2020 Fig 39 Principle of Phase Delay Unit Type 6202
19
Fig.40. Repositioning individual units gives flat phase response
Fig.41 Non-ideal phase response for "normal" positions of speaker units
Practical ExampleFig.41 shows a typical three-way
loudspeaker system in which eachof the individual drivers is assumedto have linear phase response. Theacoustic centers of the drivers are in-dicated by an x and are assumed tobe independent of frequency. The re-sulting phase response curveclearly indicates the relative offsetof the acoustic centers of the threedrivers. From this curve, it can becalculated how much to repositionthe drivers to obtain flat phase re-sponse as shown in Fig.40. Withthe acoustic centers of the speakersin line, the transient performancewill be improved since the respec-tive frequency components will nowarrive at the listener with theproper time relationship.
Phase response seems especiallyrelevant to audible quality at thehigher frequencies, while third oc-tave response seems more relevantat low to mid frequencies. Phase re-sponse at the very low frequenciesnear the speaker cut-off frequencywill also reveal its low frequencytransient response, which if good,
ing on the maximum frequency re-quired. Thus it can be used formeasurements of audio delay Sines,be they of digital, acoustic, or an-alog bucket brigade design. Thetechnique can also be applied totape recorders (Fig.42). However,wow and flutter may make thephase reading unstable at higherfrequencies.
results in a tight sound, or if poor,gives a muddy, and poorly definedbass.
Systems with Long DelaySince the Phase Delay Unit Type
6202 may be used with an externalclock it can be used for phase meas-urements of systems with delays upto approximately 200 ms, depend-
Fig.42. Phase response measurements on tape recorders
20
12. Mechanical Stability of Tape Recorders
Fig.43. Phase measurements between channels of tape recorder
The Phase Meter Type 2971 mayalso be used as a very sensitive testof the mechanical stability of me-chanical reproducers. If for exam-ple, the tape skews slightly, thiswill show up in the phase betweentwo channels of a tape recorder. Aninstrument set-up for this measure-ment and a resulting curve isshown in Fig.43. The curve indi-cates the phase variations as a func-tion of time.
The same measurment may alsobe made between the two channelsof a stereo record. This is especiallyimportant for the socalled "matrix"four-channel systems (such as theSQ and QS systems) where the four-channel information is encodedboth as amplitude and phase varia-tions. In addition, the phase be-tween channels is of great import-ance in all multi-channel recording,reproducing, and transmitting linksto ensure monaural compatabilityand a precise stereo image.
of interest. Therefore the delay canbe compensated for by setting thefrequency of the generator to give apen deflection around mid-paper. Itis important to note that the higherthe test frequency, the more sensi-tive will be the test of mechanicalstability.
The overall stability and wow andflutter of a tape recorder can alsobe evaluated by comparing the in-put and output phase of the tape re-corder (Fig.44). Any speed varia-tions or tape stretching will then beseen as a phase change. The PhaseDeiay unit is not necessary for thistest, since only the relative phase is
Fig.44. Phase measurement of tape recorder to evaluate mechanical stability
13. Complex ImpedanceTo measure impedance of a loud- sistor placed in series with the loud- to the loudspeaker. The voltage ac-
speaker and feeding this voltageback to the Compressor input ofthe oscillator (Fig.45) This voltageacross the resistor will then be heldconstant giving a constant current
speaker requires a generator whichIs constant current source Thismay readily be accomplished withthe Beat Frequency Oscillator bymeasuring the voltage across a re-
ross the loudspeaker will then be di-rectly proportional to the imped-ance. The results of such a meas-urement are shown in Fig.46.
21
However, impedance is a complexquantity and hence also requiresthe measurement of the phaseangle. This may be done using thePhase Meter 2971 in the set-up in-dicated in Fig.47.
The complex impedance can bedescribed as the vector | Z| Z® fromthe origin (0,0) to a point on the cir-cle (Fig.48). The point closest to theorigin corresponds to a frequency of0 Hz (DC) where the amplitude isminimum and the phase is zero.When passing clockwise around thecircle it is seen that both the ampli-tude |Z| and the phase angle l<t> in-crease until the vector is tangent tothe circle. The magnitude |Z| thencontinues to increase unti! reson-ance is reached where the imped-ance is purely resistive since thephase angle is zero. Above reson-ance, the magnitude decreases andthe vector swings around the bot-tom of the circle until it reaches apoint close to the DC value.
Fig.45. Use of Beat Frequency Oscillator as constant current generator
Fig.46. Typical loudspeaker impedance curve
Fig.48. Veetor representation of compleximpedance Fig.47. Set-up for measuring phase eomponent of impedance
22
14. Instrumentation Microphonesfor Music Recording
and transparent sound. This differ-ence in audible quality is most likelydue to the difference in phase re-sponse of the microphones.
In addition, the microphones areeminently suited for music record-ing in ail respects, with perfor-mance unequalled by any other om-nidirectional microphone commer-cially available. The one-inch micro-phones provide state-of-the-art lownoise performance. This micro-phone has a noise floor of only10dB(A) while still being able tohandle a maximum SPL of 140dS.However, for the majority of applica-tions, a higher noise level will makeno difference since the recordedlevel is so high. In these cases, thehalf-inch microphone offers the
A striking example of the import-ance of phase response is seenwhen comparing the audible qualityof Briie! & Kjaer 1 " and 1/2" con-denser microphones. The two micro-phones both have flat response inthe audible region, but the only dif-ference is the phase response,which is far superior for the half-Inch microphone. The half-inch mi-crophone {Type 4133) has a 90°phase shift (indicating resonance) at25 kHz, well outside the audible re-gion. However, the one-inch micro-phone (Type 4145) has a resonancefrequency of only 10 kHz, still in-side the audible range. When com-paring the two microphones on tran-sient filled material, the 4133 (half-inch! gives an audible superior tran-sient reproduction — a more clean
best choice due to its superior tran-sient reproduction. Still Its noisefloor is only about ?.5dB(A) and themaximum SPL is 1 60 dB — thus en-suring that aiS transients and peakswili be reproduced distortion free.When studio microphones areplaced near or inside musical instru-ments, peak sound levels begin toapproach this figure, and overloadof most microphones occurs onpeaks. The arrtpie overload marginof the 4133 assures that this willnot occur.
The free-field response of variousBriie! & Kjae*' microphones is shownin Fig.49. The sensitivity differ-ences between the microphones isalso seen.
23
Fig.49. Frequency response of B-uel & Kjaer free-field microphones
15. Test Records
Briiel & Kjaer produces three dif-ferent test records whose main fea-tures are indicated in Fig.50. Withthese records, frequency response,wow and flutter, rumble and othermeasurements may be made.Fig.51 shows the program con-tained on record QR 2009. To syn-
Fig.52. Frequency response measurements using test record
Bands 1 to 8-20 Hz-20 kHz4/15 octave per rev.
470007
(45 rpm)
Fig.51. Test record QR 2009Fig.53. Recording characteristics of test records
chronize the test records with theLevel Recorder for automatic plot-ting of frequency response, the Re-sponse Test Unit is required. A1 kHz tone at the beginning of thetest sweep on the record is detectedby the Response Test Unit whichthen starts the Level Recorder(Fig.52).
All three records are recorded ac-cording to IEC Recommendation 98.However, QR 2009 and QR 2010are recorded with constant velocityabove 1 kHz to prevent too high ac-celerations at high frequencies andto avoid too low a reference level.This is compensated for in the Re-sponse Test Unit. However, QR
2011, is recorded with the full nor-mal RIAA correction above 1 kHz be-cause this record is designed foruse on normal hi-fi equipment avail-able in a typical listening room. Thismay result in some mistracking atthe higher test frequencies. The var-ious recording characteristics areshown in Fig.53.
Fig.54. Principle of the CD-4 system
24
ooCM
a
oo
a
<0CM
QC
a
m,=,0CM
QC
a
BandNo
1 &5
2 & 6
3 & 7
4 & 8
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1
2
3
4
1
2
3
4
5
6
7
Type ofModulation
Left
Right
L + R
L— R
Left
Right
L + R
L + R
L + R
L + R
L + R
L+ R
Left
Right
L + R
L—R
L + R
Left
Right
Right
Left
L + R
Empty
Left
Right
Right
Right
Left
Left
Left
Right
L + R
Left
Left
Right
Right
L + R
L + R
L + R
L + R
L— R
L + R
L + R
L + R
L + R
L + R
Level*)
— 10dB
I —10dB
+ 8dB
+ 6dB
+ 4dB
+ 2dB
OdB
— 12dB
OdB
—20 dB
— 10dB
—20dB
— 10dB
— 1 1,3dB
—20 dB
OdB
OdB
—20 dB
OdB
OdB
— 10dB
— 10dB
—20 dB
—22 dB
1 — 22 dB
—22 dB
1 —24dB
J
—22 dB
— 10dB
—24 dB
Accuracy
± 1 dB
+ 1 dB
±0,5dB
± 1 dB
±0,5dB
±2dB
+ 1 dB
± 1 dB
1 + 1 dB
±2dB
± 1 dB
±1 dB
± 1 dB
± 1 dB
±1 dB
±1 dB
± 1 dB
Frequency
20 Hz to 20 kHz
log sweep
20 Hz to 45 kHzlog sweep
1 kHz
3150 Hz
1 kHz
30 kHz
315Hz
1 kHz
1 kHz
400 Hz to 10 kHzlog sweep1 kHz
1 kHz
400 Hz to 10 kHzlog sweep
20 Hz to 20 kHzlog sweep
5 Hz to 20 Hzlog sweep
1 kHz, 1/3 oct Noise
20 Hz to 20 kHz,1/3 oct Noise
1 kHz, 1/3 oct Noise20 Hz to 20 kHz,1/3 oct Noise
! 1 kHz 1/3 oct Noisej 20 Hz to 20 kHz,I 1/3 oct Noise
20 Hz to 20 kHz1/3 oct Noise
20 Hz to 20 kHz,Wideband Noise
20 Hz to 1 kHz, Noise
1 kHz to 4 kHz, Noise
4 kHz to 20 kHz, Noise
20 Hz to 1 kHz, Sine
20 Hz to 20 kHz,Wideband Noise
Duration
16,7 s/dec ,
total 50 s
5 s/dectotal 16,7s
1 5 s each
6 0 s
3 s
3 s
1 s
1 s
1 s
5 s each
1 5 s
6 0 s
3 s
2 s
5 s/dectotal 7 s3 s
2 s
5 s/dectotal 7 s
5 s/dectotal 1 5 s
50 s/dectotal 30 s
6 0 s
500 s
60s
500 s
60s
500 s
150 s
30 s each
1 5 s each
85 s
240 s
SyncBurst
Yes
Yes
No
No
No
No
No
No
No
Yes
Yes
No
Yes
No
Yes
No
Yes
Yes
No j
No
Yes
No
Remarks
Distortion < 4%
Distortion < 3%
Distortion < 1%
Distortion < 1%
Distortion < 1%
Max Wow < ±0 06%(peak-weighted)
Crosstalk < 20 dB
Rumble according tolEC-methodA Weighting < 50 dBB-Weighting < 65 dB
Crosstalk < 20 dB< 30dB at 1 kHz
Constant velocity
At the beginning of eachband a voice states centrefrequency
At the beginning of eachband a voice states centrefrequency
At the beginning of eachband a voice states centrefrequencyNo voice comments
Correct phase
Reversed phase
Applications
Frequency response, cross-talk and balance measure-ments
Frequency response and cross-talk measurements
Determination of maximumtracking ability
Wow measurements on turn-tables
Quick polarity check
Crosstalk measurements at30 kHz in 4-channel transdu-cers
Rumble measurements onturntables
Crosstalk measurement re-quiring only an AC-vottmeter
Measurement of responseand crosstalk at small me-chanical wavelengths
Investigations of arm reso-nances
Calibration
Manual response measure-ment in listening room
Calibration
Manual response measure-ment in listening room
Calibration
Manual response measure-ment in listening room
Automatic response measure-ment in listening room
Phase check of the entiresystem
Individual phase check
Tracing resonating parts
Room distribution of widerange signal
') Relative to 10cm/s RMS, Lateral (L + R) at 1000Hz
Fig.SO. Specifications of B 8t K test records
25
Frequency Response of
Test record QR 2010 contains asine sweep from 20 Hz to 45 kHzwhich makes it ideal for tests of sys-tems for reproduction of four-chan-nel sound encoded in a high fre-quency sub-carrier (Fig.54). Such atypical pick-up frequency responseis shown in Fig.55.
Tracks on this test record are alsoprovided where only one channel isswept at a time, thus permittingchannel separation measurements.
Fig.SB. Frequency response of a pick-up suited for CD-4
TrackabilityTo determine the maximum track-
ing ability of a pick-up, test recordQR 2010 contains a track with a1 kHz signal recorded at very highlevels from 0 dB up to + 8 dB (ref.1Qcm/s RMS) in steps of 2dB. Bymonitoring the signal on an oscillo-scope or by a frequency analyzer,the maximum highest level withoutmistracking can be determined(Fig.56).
Fig.88. Monitoring distortion level to check trackability
Wow and FlutterSmall speed variations are termed
wow and flutter. Variations below10 Hz are called wow, and above10 Hz, flutter. This can be mea-sured as the frequency modulationof a 3,15 kHz signal recorded ontest record QR 2010 by use of aflutter meter (not yet available fromB&K)(Fig.57).
Fig. 57. Wow and flutter measurements
26
HumbleRumble is a signal to noise meas-
urement using either the A or B fil-ters shown in Fig.58 Track 11 oftest record QR 201 0 contains a 1 5second 31 5 Hz reference tone fol-lowed by a 60 second silent groovefor rumble measurements. The Aand 8 weighting curves are builtinto the Response Test Unit Type4-416. A typical instrument set-upis shown in Fig 58.
Fig.SS. Set-up for rumble measurements
Tone Arm ResonanceTest Record QR 2010 can also be
used to measure tone arm reson-ances. These resonances can be solarge that the movement of the armcan sometimes be seen by thenaked eye. The measurement ismade using track 15 which con-tains a slow sweep from 5 Hz to20 Hz. The output of the pick-up car-tridge is recorded on the Level Re-corder which clearly shows the re-sonance frequency (Fig.59) Notethat the frequency scale is multi-plied by 0,1
Fig.59. Response curve showing tone arm resonance
DistortionA set-up for swept measurements
of individual harmonic distortioncomponents was shown in Fig 28.
However, this can be expanded toDifference-Frequency and Intermod-ulation measurements as indicatedin Fig.60. The BFO frequency from
2010 is used as a tracking signal inone channel while the two-tone testsignal is recorded synchronously inthe other channel.
Fig.60. Set-up for Difference-Frequency and Intermodulation measurements
27
1 6. Frequency Response in the Actual Listening RoomUsing 1/3 Octave Pink Weighted, Random Noise
One objective measuring methodwhich gives good correlation to audi-ble quality is the measurement offrequency response in the actual lis-tening room using 1/3 octave pinkweighted random noise. Since thissubject has been treated in greaterdetail in Application Note No.15—067 (Fig.61) only the highpoints will be discussed here.
Listening and measuring testswere performed using five differentloudspeakera each in three differentrooms (Fig.82). The results of theobjective third octave measure-ments are shewn in Fig.63. Herethe vertical columns represent thethree rooms and the horizontalrows represent the five speakers.From these, the influence of therooms on the same loudspeakershows some very significant differ-ences. This underlines the import-ance of including the influence ofthe room in loudspeaker measure-ments.
The results of the listening evalua-tions by five experienced listenersshowed that the subjective evalua-tions of the speakers in the variousrooms corresponds very closely tothe third octave measurements.
L1 L2 L3Fig.62. The three rooms and five loydspeakere used in the third octave and listening tests
Fig.63. Third octave measurements of the five loudspeakers in three different rooms. Threhorizontal charts are for the same loudspeaker test in the three different roomsFig.61. Applieation Note No 15—067
28
Fig.64. Use of Real Time Analyzer for third octave loudspeaker measurements
Fig.65. Sequential measurement of third octave response
The third octave measurementsmay be made in three differentways of various degrees of conveni-ence and expense The most sophis-ticated technique uses a Digital Fre-quency Analyzer Type 2131 whichdisplays the level of all third octavebands simultaneously (Fig.64). Asimpler technique involves the trans-mission of one third octave band ata time and plotting the results onthe Level Recorder (Fig 65). Thesimplest and least expensive tech-nique uses Pink Noise Test RecordQR 2011 which contains the thirdoctave bands of noise The resultsare measured with Precision SoundLevel Meter Type 2206 (Fig.66) andare manually noted on the graphpaper provided with the test record.The more sophisticated techniquesare especially relevant for soundsystem equalization of cinemas,theaters, concert halls, and record-ing studios. However, all threenethods show excellent agreementwith each other and with the re-sults of subjective tests.
Fig.66. Portable third octave measurements using special test record
Fig,68, Portable equipment for third octave measurements with automatic recording of resultson 2306
29
If more portable instrumentationis required on the receiving end.Precision Sound Level Meters Type2203 or 2209 fitted with Third Oc-tave Filter Type 1616 can also beused to measure the response ofthe system to the pink noise excita-tion (Fig.67).
The portable instrument set-up in-dicated in Fig.66 may be expandedby use of Portable Level RecorderType 2306. This may be synchron-ized with band B-3 of the test re-cord QR 201 1 by manually startingthe Level Recorder at the end of the1 kHz tone and using a paper speedof 1 mm/s.
Another portable set-up is shownin Fig.68. Here the output of NoiseGenerator Type 1405 is filteredthrough Tunable Band Pass FilterType 1-621 set in the third octavemode. The filter is swept manually,but pulses from the filter sent to theLevel Recorder 2306 ensure syncro-nization of the two instruments.
Filter characteristics of Type 1616.Fig.67. Sound level meters with portable filters
30
] 7. Reverberation Time
Fig.69. Set-yps for measurements of reverberation time
The reverberation time of a roomis defined as the time it takes thesound level in a room to decrease60 dB when a sound source is sud-denly switched off The reverbera-tion time of a room is one of theacoustic characteristics that hasfound most widespread use when aroom's acoustic qualities must bedetermined.
To measure reverberation time, asound source and a sound measur-ing set-up connected to the level re-corder are used. When the soundsource is switched off, a decaycurve is recorded on the level recor-der and the reverberation time isfound from the slope of the curve.A, special protractor, SC 2361, is
available for this purpose. As the re-verberation time usually changeswith frequency, either the transmit-ting or receiving instrumentationcan be made frequency selective.Selective reception increases the dy-namic range by attenuating back-ground noise, while selective trans-mission reduces the power require-ment. Fig.69 shows two set-ups fordetermining reverberation time as afunction of frequency and alsoshows typical curves obtained
Narrow bands of noise or a war-bled sine wave signal may be usedfor selective excitation Pure sine isnot recommended due to standingwaves. The noise signal has the ad-vantage that many room reson-
ances are excited simultaneously.Constant bandwidth narrow bandnoise may be generated by SineRandom Generator Type 1027 Con-stant percentage bandwidth noise(1/3 octave, for example) may begenerated using the pink noise fromthe 1027 or from Noise GeneratorType 1405 and filtering it throughFilter Set Type 1615. An instru-ment set-up using both selectivetransmission and reception isshown in Fig.70 All the.curves areplotted on the same paper by form-ing the level recorder paper into acontinuous loop. Filter switching isautomatically synchronized with theLevel Recorder.
Fig.70. Automatic measurement of reverberation time
31
Reverberation Processor Type4422 (Fig.71) allows measure-ments of reverberation time accord-ing to the integrated tone burstmethod. This results in very smoothand reproducible decay curves andgives a much more accurate deter-mination of the initial slope thanthe methods using noise or warbledsine. Since the upper 10 to 15dBof the decay curve are of special im-portance when evaluating acousticquality of the room, the 4422 hasbeen designed to give a direct read-ing of "Early Decay Time" (EDT) onits built-in meter.
Portable equipment is often re-quired for reverberation time meas-urement. Two portable set-ups aretherefore shown in Figs.72 and 73.
Fig.71. Measurement of Early Decay Time (EDT)
Fig.72- Portable set-up Fig.73. Optimum portable set-up
1 8. Sound PowerSound power is a loudspeaker par-
ameter that gives reasonably goodcorrelation to listening tests. Soundpower is defined as the total acousti-cal power radiated by the loud-speaker as a function of frequency.
If this measurement is performedin a free field (anechoic chamber) areasonably large number of pointson a sphere around the speakershould be measured to permit calcu-lation of the power. In theory, allpoints in all directions should bemeasured and averaged.
In a reverberant room, the sound
pressure at all points should be thesame and hence only one measure-ment should be necessary at eachfrequency. However, this ideal isnot always achieved, and averagingof several points may be necessary.This can be achieved by placing themicrophone on Rotating Micro-phone Boom Type 3923 (Fig.74) orby using a multiplexed array of mi-crophones and averaging the resultsby Computer (Fig.75). The soundpower can then be calculated basedon the mean sound pressure, the re-verberation time, and the volume ofthe room according to the formulaon page 33.
Fig.74. Rotating Microphone Boom Type3923
32
Fig.75. Use of Computer to calculate sound power from multiplexed microphones
= 20 log10 ( ^ ) - 10 log10 (—) +10 log l 0 (—) - 1 4 d B
where
is the power in wattsis the reference powerof 1 0 - 1 2 Wis the mean sound pres-sure in Pa (N/m2)is the reference soundpressure of 20fjPais the reverberation timeis the reference reverber-ation time of 1 sis the room volumeis the reference volumeof 1 m3
m
V
This computation may be madeautomatically by Computer Type7504 using a software programa/aifable from Briie! & Kjaer, A typi-cal print-out is shown in Fig.76 Fig.76. Typical computer print-out of sound power calculation
33
Motion Analyzer Type 491 i al-lows visual observation of mechani-cal movements by use of the st:o-boscoplc principle. Special featuresof this motion analyzer are a siowmotion position whereby a fast mo-tion can be msdo to appear to moveslowly, and a phase deviation fea-ture whereby a motion can befrozen in any position. Provision forexternal Triggering permits ihe flesh-ing of the light to be synchronizedwith the movements as the speedor frequency changes. A typical set-up is shoveli in Fig.77 and a photo-graph showing wire movements ofs loudspeaker is shown in Fig.78.
Fig.77. Use of Motion Analyzer
Fig.78. Motion Analyzer "freezes" motionof loudspeaker wires
When loudspeakers or other ob-jects are placed on a vibration exci-ter, resonance search is a valuableaid for design and test purposes.The Motion Analyzer can be con-trolled directly from the sweep gen-erator and all resonances can be vis-ually observed without any furtheradjustments of the Motion Ana-lyzer. For the study of particular re-sonances, the generator sweep canbe stopped and the exact resonancefrequency be found by manual tun-ing. The resonance can then bestudied in greater detail using thephase and slow motion adjustmentsof the motion analyzer (Fig.79).
Fig.79. Examining resonances of object on vibration exciter
34
»e and CrosstalkThe various B & K Analyzers are
well suited for frequency analysis ofail kinds of noise Typical measure-ments relevant to the HI-FI industryare shown in Figs 80 to 82
Modulation noiseModulation noise shows up as
sidebands to a pure tone recordedon a tape due to the discrete natureof the magnetization of the individual particles of the tape In practicethis phenomenon will be difficult toseparate from mechanical frequency modulation due to tapespeed variations A typical resultmeasured with the Heterodyne Anal-yzer 2010 is shown in Fig 80
Fig 80 Modulation noise
Qrvd & KjfflsrQ D D O a D a O D Q D D D a O
f»r_BMS_ Uav l > F n _ ! 2 _ H i Wr Sp«J-
C rosstaikA frequency analysis of the other
channel of a stereo system whenthe first channel is swept with apure tone indicates the channel sep-eration or the crosstalk A typical resuit for a tape recorder measuredv/ith the 1/3 Octave Filter Set1615 is shown in Fig 81
Fig 81 Crosstalk
Hum and NosseThe total noise of a system as a
lunction of frequency can easily bemeasured using the various B & KAnalyzers A typical result for atape recorder measured with Frequency Analyzer 2120 is shown inFig 82
Hum and noise from tape recorder50 dB below overload limit
Fig 82 Hum and Nosse
35
Fig 83. Production control using real time analyzer
Fig 84. Production control using stepped filters
read-in digitally, can be made Digi-tal read-out is possible and thestore feature permits read-out tothe Level Recorder while the loud-speaker under test is being replacedby the next speaker
Similar, but slower set-ups maybe used incorporating combinationsof the generators and analyzers al-ready discussed. An example isshown in Fig.84
If a high test speed is required.Frequency Response Tracer Type4712 may be used This instrumentgives a fast and accurate picture ofthe response and thus is well suitedfor production control A set-up isshown in Fig.85. Another methodis to excite the speaker with pinknoise and measure the responsewith the Digital Frequency AnalyzerType 2131 (Fig.83) The nature ofpink noise is more closely related toactual audio material than a sine ex-citation and hence is a realistic test.The Digital Frequency AnalyzerType 2131 is a new octave and1/3 octave analyzer. It is almost en-tirely digital in operation since ituses digital filtering, detection andaveraging techniques The resultsare displayed on an 1 1 " screen.The frequency range is 1,6 Hz to20 kHz and the dynamic range onthe screen is 60 dB There are lin-ear and logarithmic averaging possi-bilities and two memory capabilitiesso comparison with for instance areference spectrum, which can be Fig.85. Set-up for finding frequency response of amplifiers, fiiters, etc
36
1.Properties of Sound
Sound is vibrations in gaseous,liquid or solid media. It manifests it-self as pressure fluctuations in themedium and it is characterized bytie instantaneous amplitudes of thepressure fluctuations, as well as bytheir frequencies and phase relation-ships.
The simplest sound signal is apure tone, or sinusoid, which isfully determined by its peak pres-sure amplitude ppeak a n c ' i*s fre-quency f (equal to I/T, T is the pe-riod of oscillation).
Sound PropagationThe sound is propagated at a
speed c which is characteristic forthe medium and which, thereby, de-termines the wavelength A for agiven frequency from the formula:
A = cT = c/f
The velocity of sound in air atroom temperature is app. 344 m/sand in fresh water app. 1 450 m/s.
In a non-dissipative medium theintensity I of the sound at a dis-tance r is given by the sound powerV emitted by the source divided bythe surface area of the sphere withradius r
WI =
4?rr2
At a sufficient distance r awayfrom the sound source (in the farfield) the intensity is also propor-tional to the square of the soundpressure p whereby p is propor-tional to l/r. This relationship iscalled the "inverse distance law" orthe "inverse square law".
Sound Pressure Magnitude andSound Pressure Level
The sound pressure magnitude isoften characterized by other quanti-ties than Ppeak. The most com-monly used value is the RMS value{root-mean-square value) becauseof its direct relationship to thesound energy, it is defined as
The relationships between theabove magnitude quanties are oftenused to characterize a signal
The Form Factor F., = P R M S
Pav(= 1,11 for pure sinusoid)
The Sound Pressure Level (SPL)in decibel (dB) is defined as 20times the logarithm of the ratio ofthe sound magnitude PRMS t 0 t n e
reference magnitude p0 R M S
RMS
Another important characteristicvalue is the average value
The Crest Factor Fc =
(= v 2 for pure sinusoid)PRMS
~r / oT i P ( t ) | d t
1 Pa= 1 N/m2 = 10/ibar= 10 dynes/cm2 = 94 dB (Ref. 20 j1 Atmosphere = 1,013 bar = 760 mm Hg. = 1,013 x 105 Pa = 194 dB
20 log— =94dB= ( 4 x 2 0 + 1 4 ) d B ^Po
p= 104 x 5,012 p = 104 x 5,012 x 20 x 10~6 Pa ̂ 1 Pa
Frequency Distributionof Sound
All finite signals can be describedas a combination of a number ofsine waves. These signal compo-nents constitute the signal fre-quency spectrum. A single sinewave (a pure tone) is representedas a line in the frequency spectrumand any periodic or quasiperiodicsignal is given by a number of dis-crete sine waves, or lines in the fre-quency spectrum.
For noise signals, however, thefrequency spectrum consists of infin-etely closely spaced frequency com-ponents and the spectrum in usu-ally given as a spectral density. Themost commonly used quantity is thePower Spectral Density (PSD) whichis defined mathematically as
W(f) =
BT Jolim iim
(t.f) dt
where B is the signal bandwidthand T the averaging time. One of
38
frse advantages of power spectraldensity is the fact that from the fre-quency spectrum, the power {andthereby the PR^S value) can be cal-culated for any given bandwidth.
For transients and impulses thereis no steady state power and the fre-quency spectrum may, therefore,better be given as an energy spec-trum
Free-field and Reverberationfield
A free field is a defined as aspace in which the boundaries (ifany) exert a negligible influence onthe sound field. Thus a free fieldlacks any physical object whichmight disturb the sound by reflect-ing it or by diffraction.
Free field conditions can be foundoutdoors at some distance abovethe ground, or inside in roomswhere the walls are lined withhighly absorbent materials therbypreventing reflections. In a free-lield, the sound pressure ievei willdrop 6dB for a doubling of distancefrom the source. This is call the in-verse square law.
A reverberant, or diffuse field is aspace in which reflection causes auniform sound distribution over theentire space. This is achieved bythe use of very hard, reflective sur-
faces which are placed non-parallelto prevent standing waves whichwould destroy the homogeneity ofthe diffuse field.
A normal listening room, ofcourse, is somewhere betweenthese two extremes, and hence istermed a semi-reverberant room.
39
2. Typical Instrument CombinationsFor various reasons, smaller in-
A main point in these notes has existing package enables new and strument combinations may be re-been the "add-on" principle indicat- more advanced measurements. The quired. Therefore the most typicaling that the Briiel & Kjaer instru- ideal situation is therefore to have combinations are listed below,ments are built to work together, so all the instruments. The add-on possibilities are indi-the addition of an instrument to the cated by the vertical lines.
Generator
Measuring Amplifier
Level Recorder
Response Test Unit
Microphone
Preamplifier
Power Amplifier
Response Tracer
Gating System
Turntable
Slave Filter(Constant bandwidth Analyzer)Tracking Frequency MultiplierDist. Measurem. Control Unit
Phase MeterPhase Delay Unit
Digital Event Recorder
Voltmeter
Analyzer(Constant percentage bandwidth)
1/3 Octave Filter
Real Time Analyzer
Motion Analyzer
Reverberation Processor
Rotating Microphone Boom
Test Records
Computer
Tape Read/Punch
SineSystem,(pure tone)
1023
2608
2305
4416
4133
2619
2706
4712
4440
3922
2020
1901
29716202
2425
2121
4911
3923
QR 2009/10
7504
7102/6301
Sine/Noise,(pure tone+ noise).
1027
2606/07
2307
4416
4133
2619
2706
4440
3922
2020
1901
29716202
7502
2426
2120
3348
491 1
4422
3923
QR 2009/10
7504
7102/6301
CompleteSophistic.System(pure tone+ noise).
2010
2307
4416
4133
2619
2706
4440
3922
(2010)
1902
29716202
7502
2427
2120
3347/48
4911
3923
QR 2009/10
7504
7102/6301
Portable1/3 OctaveSystem.
QR 2011
2206
2306
4416
4148
1621
1616
Real Time1/3 OctaveSystem.
1405
2606
2307
4133
2619
2706
4440
3922
2425
1615
3347 or2131
4422
3923
4 0
Above is given a list of the most relevant standards applied to electro acoustics. Many standards in the border area of this subjecthave been omitted and thus the list cannot be considered as being complete. Futhermore it is a known fact that several standardorganizations have relevant subjects under consideration which may outdate the present survey in a relatively short period.
41
Organi-zation
IEC
[EC
ANSI
BS
BS
BS
BS
BS
BS
BS
BS
DIN
DIN
DIN
DIN
DIN
DIN
DIN
DIN
DIN
DIN
DIN
DIN
DIN
DIN
DIN
DIN
Number
94
200
S. 1.5
1568
1927
1928
1988
2498
3499
3499
3860
45 500
45 500
45 500
45 500
45 500sheet 1-3
45 500
45 507
45 511sheet 1-3
45 512sheet 2
45 513sheet 1-6
45 519sheet 2
45 521
45 542to 45
45 573sheet 1
45 573sheet 2
45 575
Date
1968
1968
1963
1960/86
1953
1965
1953
1945
1969
1986
1965
NOV1970
MAY 1971
APR 1966
AUG 1971
FEB 1971
MAY 1971
OCT1966
MAY 1971
FEB 1969
1962-71
MAY 1971
OCT 1963
FEB 1968
JUL1962
JAN 1969
MAY 1962
Sta-tus*)
S
S
s
ssss
s
s
s
s
D
D
S
D
S
ss
s
D
S
D
S
s
s
s
s
Short Description
Tape recorders, recording and reproducing characteristics, speeds and dimensionsof tape, position of tracks
Loudspeakers, frequency response, polar plot, resonant frequency, power handling,test conditions, standard baffle
Loudspeaker, impedance, frequency response, polar plot, distortion, effeciency,power handling, recommended
Tape recording at 38, 19, Th and 33A in/sec. Originally revised and extended 1966
Loudspeaker, physical dimensions, impedance and resonant frequency
Disc records, characteristics of reproducing equipment
Wow and flutter, general recommendation for measurements
Loudspeakers, frequency response, distortion, efficiency polar plot, impedance,transient response, conditions
School music equipment, tape recorders, wow and flutter, frequency response, noise,distortion
Amplifiers for musical instruments, function of tone controls, hum, noise
Amplifiers, distortion, rated power, frequency response, intermodulation, hum andnoise, methods and data presentation
Disc reproducing equipment, wow and flutter, rumble, pick up frequency response,distortion, crosstalk
Tape recorders, wow and flutter, frequency response, distortion, noise, crosstalk,erase damping
Microphones, frequency response, polar characteristic, distortion
Amplifiers frequency range, distortion, intermodulation, crosstalk, noise, outputpower, damping ratio
Loudspeakers, frequency response, power handling, distortion, music power
Combined equipment, frequency response, crosstalk, distortion, intermodulation
Sound recorders, measurement of wow and flutter
Tapes VA", VI" and 1" width. Position of tracks
Tapes, determination of electroacoustic properties.
Test tapes for Vi" tape at 76, 38, 19, 9V2 and &U cm/sec.
Tape, determination of print-through and nonuniformity of recorded flux
Tape recorders, determination of crosstalk
Test records, wow and flutter, rumble, crosstalk, tracking angle 33 and 45 rev/rnin.
Loudspeakers, test conditions and methods for type test
Loudspeakers, power handling and life test
Loudspeakers, standard baffle for measurements
*) S - StandardD - Draft
. first in Sound and Vibration
Briiel & KjaerDK-2850 N/ERUM, DENMARKTelephone + 45 2 80 05 00TELEX: 37316 bruka dk
ORBOE PRINT . COPENHAGEN DENMARK