Wireless World, November 1977 63 Audible is not a mystery

Some things are believed because people feel as if they m ust be true, and in such cases an immense weight of evidence is necessary to dispel the belief. BERTRAND RUSSELL by Peter J. Baxandall, B.Sc.(Eng.), F.l.E.E., F.l.E.R.E.

There is a very widely held belief that all two positions, and the outcome of this time, and the number of switchovers, sound different, and that the was that we found that a reduction of some genuine and some not, should be reasons for this are so subtle and not more than about IdB in the volume large enough for a proper statistical mysterious that no-one has yet properly from the amplifier made it ab­ interpretation of results to be made. understood them. I do not agree with solutely impossible for either of us to tell Guesswork, maybe unconscious, is then these views, and confidently maintain which amplifier was operating! More largely prevented from influencing the that all first-class, competently design­ recently it was found that by choosing results. ed, amplifiers, tested under completely the moment of switchover in relation to An amusing illustration of some of fair and carefully-controlled conditions, the musical phrasing, to coincide with a these psychological ideas arose on an including the avoidance of overloading, change in sonority, one could produce exhibition stand by a well-known firm, sound absolutely indistinguishable on the reaction that either one or the other who had arranged things so that visitors normal programme material no matter of two systems was the better. This sort could listen, at precisely the same how refined the listening tests, or the of thing can, of course, happen volume, to three of their amplifiers, listeners, may be; and that when an spontaneously, without anyone being being invited to identify the most inferior amplifier is compared with a aware of it. Another possible cause of expensive model. In fact it was found very good one and a subjective quality deception is a trace of hum in one that voting for "the best amplifier" was difference is genuinely and reliably system but not in the other, due to about equally distributed between the established, it is always possible, by insufficient care over earthing arrange­ three, so that, naturally, about a third of straightforward scientific investigation, ments in the test set-up - this hum can the visitors picked the right one. When to find a rational explanation for this get misinterpreted as a degradation in told they had been successful, the difference. general quality. almost universal reaction of these With regard to psychological factors, individuals was one of pleasure at their Subjective reactions I think it should be openly recognised evident skill, whereas, of course, an When people claim to have detected a that those of us claiming to have equally logical reaction would have difference in the sound of two "golden ears" in matters of sound been to congratulate themselves on amplifiers, the true explanation for this quality judgement can nevertheless be their good luck! may be any of the following: very easily led astray in various ways. The BBC Research Department is For instance if, without being aware of well aware of the dangers of reaching -the amplifiers actually did produce it, we have listened for a long period to quite wrong conclusions from subjec­ different audible , some equipment with, say, a 6dB dip in tive tests. Very careful precautions -there was a slight difference, the frequency response at 3kHz, but are taken to eliminate as many probably unsuspected, in the test otherwise of first-class performance, psychological and physical disturbing conditions, removal of the dip is very likely to factors as possible, and even to derive, -psychological factors were exerting produce the reaction, at least initially, where appropriate, a quantitative an influence. that the reproduction has become too estimate of the reliability of the results3• It is possible to be quite misled by strident. However, if it was known to It is very evident that in many other some small physical effect, thought to the listeners beforehand that a dip had places such precautions are not be of no consequence at the time. I well been intentionally introduced, removal properly taken. remember a particular case some years of it is then more likely to produce the ago when a friend claimed to be able to reaction "Yes, now the violin tone is Recording systems detect by ear the difference between a more realistic" or something of the sort! and amplifiers good valve amplifier and a good Such pre-conditioning and psychologi­ Unlike amplifiers, conventional tape transistor amplifier. He invited me to his cal influences are quite strong, and and disc recorders, even those of the house and had a changeover switch should be allowed for. Another psycho­ highest professional grade, have dis­ which I was asked to operate, not logical phenomenon, very significant I tortion levels and -to-noise ratio knowing which position was which. I think, is that few of us like to admit that which are only just about good enough soon found I could indeed detect a slight we "just cannot tell the slightest subjectively. A very instructive experi­ difference, one position seeming just a difference" in the presence of others ment is to record the same mono little smoother and less "grainy" than who have professed to hear subtle programme source on both tracks of a the other. I supposed this to be the valve differences. So most people will succeed good stereo tape recorder, with a level position, which was correct, and we in convincing themselves that they difference of, say, lOdB. The replay were both pretty well convinced we really have managed to notice small gains are then adjusted to give outputs were hearing a trace of crossover changes in sound quality. In properly of equal magnitudes, and these are distortion. It then occurred to me to conducted subjective tests, however, subtracted one from the other to give, wonder just how accurately the the participants should not know which ideally, nothing but noise and distor­ volumes had been set to equality in the system they are listening to at any given tion. The distortion is mainly that of the 64 World, November 1977

more heavily recorded track, whereas accompanied by the wanted program­ goodness of amplifiers in a quantitative the noise is mainly that of the weaker me. manner and for establishing criteria track. (In practice a little hJ., and With amplifiers, on the other hand, it that should be met if an amplifier is to be possibly l.f., phase correction may be is comparatively easy to reduce the totally free from audible distortion. The necessary to get fully satisfactory audible distortion and internally-gene­ technique can obviously be implement­ programme cancellation.) With gains rated noise to far lower levels than in ed in various detailed ways, and Fig. 1 set to give normal listening volume any normal recording system, and this shows one arrangement which is when only one track is reproduced, the is what is done in equipment of the suitable when the amplifier under test is distortion heard with both tracks highest grade. Provided such amplifiers of the phase-inverting type. When, as is operating is quite horrible and is loud are tested under sufficiently carefully more usual, there is no phase-inversion, enough to be very easily audible all over controlled and fair conditions, are free a very low-distortion phase inverter the room even in conditions of from faults such as hum and d. must be introduced into the circuit in moderately high ambient noise level. interference susceptibility, have insig­ one of several possible places. This gritty, blasting, distortion is only nificant differences in frequency res­ For setting the circuit up, it is found in somewhere about 40dB below the ponse, and are not overloaded, the quite practice that an audio noise source is un cancelled programme level during inevitable result is that one amplifier is more suitable than normal programme loud passages, yet it is virtually absolutely indistinguishable from ano­ input, since all frequencies are present unnoticeable when accompanied by the ther, on normal programme material, no all the time. Thus SI and S2 are both music. Tests with tone input show that matter how "golden" may be the ears closed, and PI plus the several the distortion is mainly third-harmonic, involved. adjustments in the frequency-response the percentage distortion being propor­ Quad have showns,9 that, with their and phase-balancing network are tional to the square of the output transistor power amplifiers, if the adjusted for minimum output from the voltage and reaching about 2% at peak amplifier distortion, including hum and monitoring system. The potentiometer recording level. The distortion is fairly noise, is reproduced by itself at its P2 should initially be set to a low independent of frequency over most of normal level, without the music, the resistance value, the value being raised the audio band. Thus a first-class result is total silence under ordinary as the balance condition is made more professional tape recorder gives distor­ listening conditions. This is enormously and more nearly perfect. Potentiometer tion of about the same magnitude and better than the result obtained when a P2 should finally be set so that, with SI character as a push-pull class A somewhat similar test, as des­ or � opened, the voltage fed to the amplifier having a distortion figure of cribed above, is done on a monitor system loudspeaker is the same about 2%, assuming this also to be high - grade professional recorder. as that fed to the load circuit of the reasonably frequency-independent. But, to me, the most amazing thing amplifier under test. With both switches Experiments I have done with class A is that Peter Walker tells me that again closed, the distortion alone will push-pull amplifier circuits, involving few of the people who have witnessed then be reproduced by the monitoring balancing out the programme and this experiment seem able to appreciate system loudspeaker at its proper level. listening to the distortion by itself, do its true significance, which is, quite Having thus got the circuit correctly set indeed show that it sounds much the inescapably, that such amplifiers are up - a rather tedious operation because same as that produced by a good tape subjectively perfect with a large margin of the number of adjustments involved recorder, and that 1 or 2% distortion is to spare and give an audible perform­ - a little thought will show that a low enough for results of the highest ance which can never be improved variety of interesting and very informa­ quality, provided the amplifier perform­ upon. Quad do not maintain, however, tive tests may then readily be done, ance is clean enough in all other that theirs are the only amplifiers about such as: respects. which this may truly be said. Of course • The of the monitoring system Similar experiments with class B if, during the above experiment, such may be increased until the distortion push-pull circuits, adjusted to give amplifiers are allowed to overload, even does become audible by itself, thus considerable crossover distortion, momentarily, the silence is broken and obtaining a measure of the margin by show, not surprisingly, that the the distortion fairly cracks forth. But which it was previously inaudible. distortion is rougher and more unplea­ amplifiers should not be allowed to • The effect on the audible distortion of sant sounding, and tends to be nearly as overload, and if they do, the only proper loading the amplifier under test with loud during fairly quiet parts of the solutions are to turn the volume down loudspeakers and/or dummy loads programme as during the loud parts - it or employ more powerful amplifiers. having various different impedance appears as an almost continuous A few people have raised the characteristics may be investigated. background fuzz. For absolutely objection to the above experiment that (When a loudspeaker load is used, it is first-class quality, distortion of this type though the distortion may be inaudible necessary, of course, to prevent the must be reduced to much less than 1 % at on its own, the ear and brain are sound from this loudspeaker from all output levels and over most of the exceedingly complex and subtle, and reaching the person listening to the audio spectrum. This topic will be the effect of the distortion might distortion on the monitoring system considered in greater detail later on. conceivably be perceived when it is loudspeaker. Rather than use well-se­ In recording systems, unless very accompanied by the music. This, parated rooms and very long leads, a refined and expensive digital techniques however, is quite contrary to what is more convenient procedure is to tape are used, there is always the need for a found to happen in the tape-recorder record the distortion and listen to it careful compromise between sig­ experiment referred to earlier, where later on.) nal-to-noise ratio and distortion. the distortion is easily heard on its own • The two loudspeakers of Fig. 1 may Compandor systems, of which 'dbx' is but is very well masked when accom­ be placed next to each other, P; then the latest, and very welcome, develop­ panied by the music. Experiments I being adjusted to determine by how ment4,S, can achieve an impressive have done involving crossover distor­ much the distortion may be increased improvement in subjective sig­ tion show that it too is fortunately above its "natural" level before a nal-to-noise ratio, together with some subject to a considerable degree of just-detectable degradation in music reduction in peak distortion level, but masking in the presence of the quality begins to become evident. they do not actually affect very greatly associated programme. • With S2 only closed, and then SI only the signal-to-noise ratio existing during closed, P2 being set for a suitable loud passages. Thus reliance is still A diagnostic tool listening volume from the monitor­ being placed on the masking effect6,7, The technique employed by Quads,9 for ing-system loudspeaker, reproduction whereby unwanted sounds, which listening to amplifier distortion by itself, via the amplifier under test may be would be very easily audible on their on programme input, provides a very compared with that via the passive own, become virtually inaudible when useful tool for assessing the subjective network. With a first-rate amplifier, Wireless World, November 1977 65

Program me source ",---",C.R.0

Monitoring system Input Out put

Load circuit

Am plifier under test (Phase-inver ting type e.g. Quad 3030r405)

absolutely no difference whatsoever Fig. 1. The diagnostic set-up. A manufacturers still devoting a lot of should be detectable on any kind of modified frequency-response and research and development effort to music programme input, provided that phase-balancing network may making better amplifiers?". The cynical no overloading of the amplifier under sometimes be required. reply might be "to produce even more test is allowed to occur. The test may be impressive-looking figures for reviews"! extended to assessing the degree of But in fact the enlightened designer is unpleasantness of various degrees of probably spending most of his time overloading, with and without protecti­ that the true waveform of the amplifier struggling with far more difficult ve circuits in operation, etc. distortion is displayed, unaffected by problems, such as how to achieve • When two amplifiers are found to oscillator distortion or by slight greater reliability, how to simplify the sound genuinely different in ordinary harmonic phase shifts contributed by design and hence reduce the manufac­ subjective tests, they may then be the notch filter that would normally turing costs, how to eliminate the need tested in a circuit of the Fig. 1 type to see have to be used. for preset adjustments, how to increase whether the distortion is audible when The Fig. 1 type of arrangement can the maximum available output, how to reproduced by itself. It may be found also be made the basis for an accurate improve the ability of the amplifier to that the distortion is of an overloading and very satisfactory technique for cope with a wider range of load type, though perhaps happening at a harmonic and distor­ impedances, how to eliminate lower output level than the expected tion measurement, which has the "switch-on plonks" etc. None of these level because of the operation advantage of not demanding a high problems directly involves the concept of protective circuitry within the degree of oscillator waveform purity. of subjective listening quality. amplifier - or it may be that the It is when problems such as those just amplifier has been badly designed with Some conclusions mentioned are considered that the true regard to its slew-rate capability. Such One of the conclusions to be drawn nature of the enormous advances made possibilities may then be looked into in from tests such as those just outlined is in audio-amplifier technology becomes detail. On the other hand, if both that amplifiers do tend to differ evident. In 1938, a British 14-watt amplifiers give inaudible, or very somewhat in the degree of unpleasant­ high-quality amplifier sold for about unobtrusive, distortion, it is worth ness of the distortion they produce £19, yielding a figure of 0.74 watt per £1. testing one amplifier in the Fig. 1 circuit when allowed to overload, but, apart A recent 200-watt stereo amplifier, of with the frequency-response and from this I feel sure that nobody who smaller size and weight, sells at £115 and phase-balancing values adopted for the has actually himself used these largely gives 1.7 watts per £1. Allowing for other amplifier in place. Then, if there is subjective investigational techniques inflation, it is clear that the true cost per a noticeable difference in quality when could possibly continue to believe that watt using modern solid-state techni­ only SI or only S2 is closed, the all amplifiers sound different or that the ques is down by a factor of the order of mid-frequency gains having been set to subjectively perfect amplifier has yet to ten on what could be achieved in the 2 precise equality, it is likely to be because be designed. This is why Quad have valve era • This is undoubtedly a great of the slightly different frequency been prepared to stake their reputation engineering achievement. responses - in particular, the response and say without reservation that they As a designer of audio amplifiers and below the audio spectrum may be would be prepared to accept a challenge other equipment, of which some is important in influencing the amount of to have their 303 or 405 amplifiers currently in use in BBC studios and rumble or other sub-audio-frequency compared effectively, using the Fig. 1 elsewhere, I must have spent many signal reaching the loudspeaker, where type of set-up, with what they have thousands of hours inventing, thinking it may cause large cone movements and called a "straight wire with gain"lO about and experimenting with audio thereby affect the loudspeaker distor­ Provided certain quite reasonable test amplifier circuits, but I cannot recollect tion. conditions are satisfied, their claim is ever having carried out subjective • By using an with the Fig. simply that no-one will genuinely be quality-assessment tests as a direct part 1 set-up, much can be learnt about the able to detect the slightest difference in of the design and development process. relationship between the type of sound quality on any programme input Subjective tests have been done distortion waveform observed and the derived from normal sources. separately from the design work and for corresponding subjective nature of the The unconvinced may well say "if the purpose of helping to establish distortion. The system also has the subjective perfection has already been criteria which need to be satisfied by the great virtue, when used with tone input, achieved, then why are amplifier equipment designed. 66 Wireless World. November 1977

Without knowledge of such subjecti­ ambience and "seeming to make the References P. vely-derived criteria, it is natural to music sound amateur"! Enquiries l. Walker, J., Current dumping audio "play for safety" and make the revealed that the unit in question was amplifier, Wireless World. Dec. 1975, pp. 560-562. www.keith-snook.info performance far better than it actually subsequently retested by the manufac­ 2. Baxandall, P. J .• High-fidelity amplifiers, needs to be. This is particularly the case turers and found to be in perfect order. , TV & Audio Technical Reference with or control-unit When descriptions such as the above. Book. Chapter 14 (Newnes-Butterworths, design, where the non-linearity distor­ which could only properly apply to 1977). tion is usually of the simple equipment with quite gross faults, are 3. Jones. D. K., A subjective investigation smooth-curvature type, which does not used in relation to items known to be into preferred balances. BBC need to be reduced anything like as far first-rate, it is clear that either Eng. Mono. No. 68, July 1967. F., as it is possible to reduce it in order to something was wrong with the test 4. Ogden. dbx. Hi-Fi News, Oct. 1976, pp. become quite inaudible. To elaborate set-up or that the reviewers - not to 92-94. 5. Blakely, L .• dbx for disc. Studio Sound. the design, with consequent increase in question their sincerity - had fallen Sept. 1976. pp. 32-34. cost, to the point where the distortion is, prey to their own imaginations. 6. Olson, H. F., Elements of acoustical say, a hundred times, or more, below the Since the belief that all amplifiers engineering, 2nd. Ed., pp. 469-471 (Van subjective detection limit - which it is sound different has become so widely Nostrand, 1947). quite possible to do - is surely not in accepted, it is natural for people to want 7. Beranek, L. L., Acoustics. pp. 404-406 the true interests of the customer. to find technical explanations for it. (McGraw-Hill, 1954). Needless to say, very great care indeed Since little correlation with perform­ 8. Jones, F., Dynamic testing of audio must nevertheless be taken with things ance as ordinarily measured can be amplifiers. Hi-Fi News, Nov. 1970, pp. that really do matter, such as leaving found, the notion has built up that 1655-1657. kelm@Jsna:Jk-eu 9. Collins, A. R., Testing amplifiers with a sufficient "headroom" to accommodate something extremely subtle and elusive bridge. Audio, March 1972. all pickup sensitivities2, achieving very is involved. To explain these supposed 10. Walker. P. J., Rational amplifier testing. low hum and interference susceptibility, subtleties, those with more imagination Hi-Fi News, July 1977, pk. 135. etc. than scientific understanding proceed 1l. Reviews of 17 power amplifiers. Hi Fi for Once the designer has freed himself to evolve a series of wilder and yet Pleasure, March 1976, pp. 56-63. from various quite irrational and wilder pseudo-scientific hypotheses. 12. Reviews of 11 . Hi Fi for unfounded beliefs, e.g. that there is an New jargon is created - "musicality", Pleasure, June 1977. pp. 86-107. inherent subtle difference between "loss of information", etc. An article of 13. Hiraga, J., Can we hear connecting wires?, valve and transistor sound, that French origin which has recently Hi-Fi News, August 1977, pp. 35-4l. transformers always produce detect­ appeared in Hi-Fi News 13 - accom­ able subjective distortion, that class B panied, however, by an expression of amplifiers can never sound quite as editorial neutrality and non-commit­ clean as class A ones. that feedback ment - says the quality of copper used Harold W. Bamard should only be used in small amounts, in loudspeaker leads influences the etc., he can then proceed in a proper quality of the information transmission, Many people in the electronics industry will scientific manner to develop designs of the best wires having a purity as high as be saddened to hear of the death of Harold W. good economy and reliability, and 99.99995%. The alternating magnetic Barnard. editor of Wireless World from 1965 immaculate subjective performance. field generated by a loudspeaker cable is to 1973. Although he held this post for only He will appreciate that there are said to represent a significant loss of eight years he had in fact given a lifetime of countless ways of designing equally information. Even in the wiring of devoted service to the journal. Starting in good-sounding amplifiers, and concen­ electric-bell circuits, the use of Litz wire 1925 as an assistant to the production manager, he transferred in 1936 to the news trate his efforts largely on seeking the is claimed to give "tintinabular side of the (then weekly) journal to become optimum engineering solution. superiority". How silly can we get? All what was known as a "leg-man" - getting this sort of thing, which seems to be news the hard way without the assistance of encouraged by some of the hi-fi today's information services and pUblicity Amplifier reviews magazines, for whom it no doubt organizations - and eventually took The belief that all amplifiers sound provides easy material for filling their complete charge of the news section. During different seems to be even more deeply pages, is surely not good for the future the 1939-45 war he was a member of a small rooted with the popular hi-fi press and of the audio industry. being liable to team that kept the journal going under extremely difficult conditions. In 1959 he was their reviewing teams than it is with bring it to a state of disrepute with appointed assistant editor, a fitting tribute to designers. I feel that a great disservice is intelligent people. his journalistic abilities. being done both to the buying public Admittedly the subtleties and diffi­ When he retired in 1973 we wrote this of and to some manufacturers by reports culties of many aspects of good sound him: "Kindness, courtesy and dedication are on amplifiers and control units of the reproduction are enormous, but it three qualities not very much in evidence in typewhich have appeared, for example, seems a pity that an atmosphere of quite the modem industrial scene. They are the in "Hi Fi for Pleasure"ll,12, The irrational mysticism should be encou­ three qualities which one would most likely reviewers claim to have been able to raged to invade even those parts of the pick if one were asked to characterize in a detect by ear specific deficiencies in field where things are properly under­ few words the retiring Editor of Wireless World. virtually all units submitted to them, stood and quite straightforward. Harold W. Barnard. Readers may wonder what such things have to do with including differences between "cancel" Finally, lest some readers may feel technical journalism: they don't seem to be and "tone controls flat" in all cases that the views here expressed are relevant to the business of turning out good where such a comparison was possible. representative only of an engineering articles and news on radio and electronics. But ones incredulity is surely stretched outlook, it may, perhaps, be relevant to But technical journalism, like many other beyond the limit when one finds a add that I have a passionate interest in professional and industrial activities, runs on well-known control unit. widely adopt­ music. that I frequently go to concerts, the fuel of human contacts. What is printed ed by discriminating professional users, do a good deal of recording of live in each issue is the final result of much described as having a mid-range that is music, and that much music making, talking. listening, letter writing, discussion, forward yet lacking in detail, with some some professional. goes on in my argument, persuasion. joking. threatening, criticizing, and praising. All these are compression of peaks and an unstable household. necessary functions, but it is the personal image, and a top end performance that qualities an editor brings to exercising them is thin and rounded off, but with a that makes all the difference. It would not be splashy character imparted on cymbals, fanciful to claim that kindness, courtesy and and similar explosive sounds, the The next article will discuss some dedication have been significant factors in overall performance being summarized detailed technical matters relating to the making of Wireless World during the as dull, with a great loss of presence and amplifier design. eight years of Haroid Barnard's editorship."