Please note: Because of the poor quality of the preserving microfiche, and poor OCR, "[…]" in the sentences below, unfortunately, denote some missing content.
Editor's Note: In this article, Martin Colloms summarizes the issue of how to choose the “right” amplifier for a particular listening environment. We at TAA began to struggle with this issue in 1978 and were so impressed with Mr. Colloms's article in the May 1985 issue of Hi-Fi News & Record Review that we permission to reprint it here.
John Atkinson, HFN/RR's editor, introduces Mr. Colloms as follows:
Martin Colloms has the widest experience of amplifiers of anyone I know, due to his multiple test work first for Hi Fi for Pleasure in the 1970s and then for Hi-Fi Choice and HFN/RR in the 1980s.
Despite this long intimacy of 'blind' listening, or perhaps because of it, Mar tin, an engineer and a specialist, not a scientific generalist like myself, finds himself reaching similar conclusions on amplifiers. I asked him, therefore, to put his most recent thinking down on paper. This article is the result. We welcome your comments.
ONCE UPON A TIME, amplifiers were regarded as one the more stable and well-specified components in the audio chain.
They appear to conform to our known rules of electrical theory, now highly refined and via the art of modern electronics.
Designers were confident that they fully understood the basic statements of amplifier performance such as linearity (which controls distortion); output voltage and current (which together with a specified load determines the power output); gain (the input-output ratio or amplification); and finally, gain/ phase margin (which determines stability and transient response).
Amplifiers are used in a wide variety of nonmusical applications, such as servo control systems and in measuring equipment with a highly accurate numeric output, and in these situations, they behave exactly as described by the designed statements or specifications. Here they serve us very well.
A Conventional Understanding
For a while, many designers and reviewers, including, I must admit, myself, thought that the same rules and specifications were also entirely valid for audio amplifiers. Many de signers still believe this to be true, point, though having said that, it is necessary to add a qualification. For most noncritical audio applications, the general run of amplifiers do a fine job. Their duties range from studio monitoring, stage -- and sound reinforcement to music-center, rack and “normal” hi-fi systems, where models designed to the established rules are giving good service. In truth, few in the industry would quibble with this particular status quo.
The other side of the coin has to do with setting standards. Should amplifier quality remain relatively static, or should it, as I believe, continue to improve as hi-fi itself has improved since its emergence as a specialist interest? For an audio enthusiast, or for that matter a dedicated critic, the goal remains the pursuit of quality. Just good enough is insufficient. The state of the art can only be advanced by those willing to master the present and progress beyond it.
A few years ago, well-organized testing under rigorous conditions suggested that sound-quality differences […] then-accepted “good ” power amplifiers were small enough to defy statistically verified identification.
With hindsight, I must accept that the standards set then for the test in terms of reproduced quality were probably not high enough.
Since that time, the more listening tests I have undertaken, the harder it appears to be to set up a controlled statistical test for sound quality. If amplifier differences are to be readily observed, significant differences must be present in the first place.
Perhaps it would be worth repeating such tests using some new audiophile models versus some old stalwarts.
Amplifiers Do Sound Different
For my part, and despite many jibes by the “establishment, ” all my listening since that original series of amplifier panel tests has only rein my conviction that amplifiers do sound different in many important respects, and my personal spending on equipment has clearly reflected that belief. Even though only a portion of my recent listening has included limited blind or 'double-blind testing, where this has taken place, it has strongly confirmed the results obtained from sighted testing.
I suspect that listeners vary in their sensitivity to subjective amplifier characteristics. While they may be of little consequence for some, for others relatively small differences can assume important dimensions. Depending on one's viewpoint, one could, therefore, arrive at the stage where the amplifier becomes the most important component in the listening chain- certainly this can seem so, for example when an enthusiast auditions a really fine preamp or power amp for the first time.
Since that rigorous test series con ducted some eight years ago, we have been blessed by the availability of improved amplifiers from many sources.
Some of these have been better than the original standards set, and our whole view as to what is possible from an amplifier has undergone a necessary revision. Furthermore, these superior models have been astonishingly diverse on technical grounds, reflecting the various individual approaches of their designers toward achieving good-quality sound via applied audio electronic engineering.
It is precisely this diversity of design and performance specification shown by these amplifiers that provides the basis for a discussion of those qualities, which may or may not determine what makes a good amplifier. As we shall discover in this article, many of the long-accepted features favorable to design and performance are not invariably associated with the most successful examples of the designer's art.
This discussion will also be backed by many examples of proven quality established by careful auditioning and lab testing. For many of these, almost none of the published literature, test standards or lab specifications appear to bear much relevance to sound quality ranking.
In the past, amplifiers have been regarded as the safest area of audio design, but now it would appear that they must be considered as one of the more difficult, with their development fraught with danger. We now seem to be flying blind, and successful amplifier designers talk more of an intuitive feel for the subject rather than a reliance on a cast-iron set of rules.
How can this be, when electronics has given us so much-computers, radio, telephony, television and more? Surely in view of this dazzling display of technical virtuosity, we should be able to specify how an amplifier will perform under real conditions--i.e., reproducing music via loudspeakers.
Of course, in a general sense we can, and to a limited degree certain design rules must not to, = broken. For example, if distortion is excessive (i.e., more than 5 percent), it may become audible, or if the available voltage or current is insufficient to meet the load demand, audible distortion may result. As the state of the art is approached, however, the established rules start to break down. At every price level, you can find amplifiers that happily go against one or more criteria laid down by another manufacturer or by a test standard as essential to a good performance.
Passive Component Sound
Another aspect also has a bearing on amplifier sound and may provide some clues to the present apparently inexplicable differences. Renewed and increased attention is being paid to the effect on sound quality of passive components--transformers, resistors and capacitors--with published material and work in hand indicating that these can have a marked effect on quality-an effect beyond that expected from their prescribed duty in a given circuit. The recent results concerning the sound quality of copper wire with an improved crystalline structure indicates that the wire and circuit board copper must also be taken into consideration.
It should be possible to set up a comparison of two identical high quality amplifiers, fitted into sealed black boxes. Based on the inclusion of just two components to one box, namely a coupling capacitor and a length of wire, and their omission from the other, it should be possible to demonstrate significant sound quality differences. If judged by classical theory, however, their inclusion should have no appreciable effect on amplifier performance.
Taking HFN/RR's current critical standards, suppose the power amplifier without the components scores 70 percent; recent data would indicate that the likely outcome for the other model with the extra passive components is likely to be a reduced rating, probably around the 60 percent level. Yet the amplifiers are basically identical in electrical terms.
The point I am making is that we have reached a situation where the passive parts and the wiring can count for as much as the theoretical electronic circuits themselves. Since our understanding of the sound-quality effects exerted by various components is incomplete, how can we begin to design and then quantify the performance of the next generation of amplifiers? We can, of course, investigate many more obscure parameters--for example, those pertaining to capacitors.
These components are normally specified, as one might expect, as a unit of capacitance, say 1uF (10-6 farad) with a given maximum voltage rating and associated tolerance of, say, 100V and 10 percent. Circuit theory generally assumes that a capacitor is just that and nothing else.
Unfortunately, in the real world, there is no such thing as a pure electrical component. Inevitably, small proportions of other classes of components are mixed in as well. A capacitor will possess some small value of inductance due to its connecting leads and internal construction--e.g., a foil winding. The total wiring will impart some series resistance, while a parallel resistance is also present, this being generally ex pressed as a loss factor, due to the small proportion of energy in the capacitor being dissipated in the dielectric or insulating part. (This e wiring can count for as electronic circuits them parallel resistance could also be diodic.) Dielectrics come in a range of types, qualities and, as is now believed, sounds. All show piezoelectric effects to a varying degree. Imagine the capacitor as slightly elastic; a varying electrical stress or voltage due to an audio signal causes the dielectric to change its physical size. As the audio passes through via the electrical connectors, the body of the capacitor mechanically pulsates, with the levels of this vibration more pronounced than you might expect. Indeed, one researcher suggested using a commercial capacitor as the driving element for a cone in a high-frequency loud speaker drive unit. The piezoelectric effect is generally reversible, so when the capacitor reverts to its normal or rest state after being stressed, it produces an electrical pulse with timing dependent not so much on its electrical capacitance, but on its physical pulsation resonance.
Does this sound like delayed resonance as commonly discussed in con-[…] electric absorption, by which means it may attempt to hold on to a […] later! This sounds like, and indeed is, a hysteresis effect, in a sense related to the coloration and distortion deriving from the piezo effect.
What is required here is some basic research into the properties of electronic components as related to their use in audio equipment, with an at tempt to relate technical aspects to perceived audio quality. Then and only then can a sound basis for perhaps 150 to 200 models. If the amplifier is better than average, a 70 to 75 percent rating can be achieved, which is pretty good. Occasionally, a standard-not only for other amplifiers, […] ancillary equipment. When this hap pens, a chosen musical passage may take on a new look. […] you may hear additional detail and finer nuances of instrumental playing as if for the first time. Such exceptional a 95 just how will […] trick.
Dimensionality judgment, you can identify one factor that assumes increasing importance. This concerns the properties of dimension as related to the perceived stereo image.
The term “dimensionality” (first put to me by Jacob Zelinger) is an apt partner to musicality, with the latter being more difficult to define.
Wrapped up in an understanding of dimensionality are the subjective […] seams to bold the key good sound quality. It forms the vital property of live sound as generated in real acoustics, yet so many audio systems fail to resolve all but the merest hint of it.
If the dimensionality of a system […] also capable of resolving this parameter, then other important subjective factors usually fall into place. For ex ample, tonal balance is often discussed, with some amplifiers de scribed as “pinched ” or hard ” in the midrange. Such an effect is obvious on singing voice, which may become edgy, fatiguing, forced -- unduly forward in presentation. There is also a connection with dimensionality here, for it can be destroyed by a hard and/or forward midrange. Perspectives will be distorted even to the point where the stereo image is wholly two-dimensional and has no apparent depth.
A similar link can be established for the treble, where problems can manifest themselves in several ways.
Higher frequencies can sound different, with a “grainy ” or “spitty ” effect. High-frequency transients also may appear strongly localized in the vicinity of the speaker. This can be heard as a type of clumping in one area of the frequency range. Alter natively, the high frequencies may sound slurred, and this slurring may also be evident as a smearing of positional focus, which often extends right across the frontal sound stage.
These specific defects are easily heard in the context of dimensionality. Clumped treble prevents the treble component of wide-range sounds from joining the fundamental tones in their correct place in the image plane. Furthermore, treble “smeared” in width destroys good focus and makes the upper range sound flat and two-dimensional. Distorted or grainy-sounding treble also draws attention to itself, this prominence again disturbing the qualities of depth perspective in the stereo image.
In terms of dimensionality, good bass appears well attached, in both space and time, to the harmonic transients of a bass instrument, and the combined sound localizes at definite positions in the image. An amplifier with poor bass may sound 'out of time’ with the main sound. The lowest frequencies can lose their connection with their transient edges- for example, the “smack ” immediately preceding the “thunder ” of a large drum. Such bass may appear to float around, and the overall ambience in the stereo range can suffer as far as the qualities of solidity and foundation are concerned. Only in very rare cases can an amplifier pro vide an essentially neutral, tidy and well-balanced sound, yet fail to offer a satisfying degree of dimensionality.
My experience of audio components, from microphone to disk, both vinyl and CD; from turntable, tone arm and cartridge to connecting leads; and from preamplifier to loudspeaker shows that with general improvements in audio quality, there is a commensurate increase in life, interest, satisfaction and most of all, dimensionality.
For a reviewer or electronics engineer, the frustrating aspect of dimensionality is that no direct laboratory measurement exists to quantify it. Indeed, many of the other qualities discussed previously, such as mid range tonal hardness, treble slurring and low-frequency imprecision, likewise defy normal measurement.
The Technology of Amplifier Design
I assert that there is no absolute truth at present as regards amplifier de sign-this is amply borne out by the huge variety of ideas and techniques oo have all been successfully brought to bear on the question. It is self-evident that a majority of well informed persons will agree on the truth. Amplifier design is rife with disagreement and conflict. In S-- at some of these conflicting ideas, I shall first take the notion of band width, frequency response and its specification.
We have often been told that the bandwidth should be as wide as possible, with the amplitude-frequency response in the pass band, or audible range, as uniform or as flat as possible. But just how wide and how flat? For convenience, a 20Hz to 20kHz audible frequency range is often quoted, which sounds nice and tidy. In reality, few listeners can hear or possess program with significant energy outside a 40Hz to 15kHz range. Indeed, very few speakers work well outside a 50Hz to 12kHz range, yet it has been suggested that we need a response from 0Hz (DC) to 250kHz or more in the case of an amplifier to render the amp's sonic signature inaudible. Because very small deviations in level or frequency response occurring over broad areas may be audible in critical A/B tests, it has also been necessary that frequency responses should be very flat and accurate to +0.1dB or even better. Yet much greater limitations and deviations are accepted from a pickup cartridge or loudspeaker.
I know of at least two highly rated amplifiers where on the one hand, for well- reasons, the designer has subscribed to a relatively narrow bandwidth, while on the other, he has plumped for a very wide bandwidth.
The former could be represented by a Naim such as the NAC32/NAP250 combination, and the latter by an Audio Research SP-8 preamp and Robertson Forty-Ten power amplifier.
In the case of the Naim, the -3dB points are typically 12Hz and 40kHz, but for the Robertson/Audio Research combination, 2Hz and 200kHz limits are nearer the mark. The bandwidth question is thus neither proved nor disproved, as both are well-regarded combinations.
Class A or A/B
The class of an output stage for a power amplifier is often cited as a vital factor, with common ones including class A/B, quasi class A and its sliding bias variants, and pure class A. Class D also occasionally appears, while agreement as to what constitutes further classes is poor. For example, class G has several possibilities. The word “class' actually defines the relation ship between the steady voltage and current in the output stage and the audio signal passing through.
In class A, for example, the standing current remains at a constant maximum value, with variations due to the audio output fully contained within that maximum. In class A/B, the standing or quiescent current in the output stage is set at some nominally low value, say 0.02A, instead of the 5 to 10A that is likely to be required for class-A operation. This low standing power is, in fact, the “A ” part of the A/B description. Large audio signals demand a greater current flow according to their need; this is the “B ” part.
Pages of arguments could be produced extolling one view as opposed to the other and touching on questions such as thermal stability, gain variation with current, bandwidth changes and several other matters.
(Class D involves pulse-modulated or switching techniques, and at present the method is generally classed as in adequate for top-quality sound.) Class G relates to an amplifier that cruises at one voltage supply level, but can switch to a second larger supply on momentary peaks. Here the argument is that the dynamic range is improved and that large transients can be re produced without the amplifier clipping or overheating.
Critics often dismiss the G techniques as valueless, yet the Carver Cube did pretty well in context and exploited the method to the full Others have said that only high pro portions of class A or even full class A operation can produce top-quality sound. The Krell amplifiers are outstanding examples or class-A design, but others-the PS Audio, Naim and Robertson-have managed to do very well with an absolute mini mum of standing current, running virtually in class B.
Output Coupling
The type of output coupling also has long been a topic of discussion. Tube amps are generally transformer coupled, while early transistor models were connected to the loudspeaker via a large capacitor. Modern amplifiers are generally direct coupled, which is said to confer a high damping factor, as well as wider bandwidth and lower. In my view, neither technique appears an outright winner. While the Krell, the Robertson and many others are direct coupled, I have been assured by Dennis Morecroft of DNM that a new capacitor-coupled power amplifier of his is sounding good at the prototype stage. The highly rated Futterman tube models are also capacitor coupled. Finally, some of the finest sounding models available at present--Audio Research and Conrad Johnson--are all transformer coupled.
These latter tube products represent particularly deep thorns in the sides of designers of semiconductor amplifiers. I have also tested many poor sounding power amps, which variously feature all three coupling methods.
Type of Active Device
Another technological aspect overused commercially to promote one model over another is the type of amplifying device used, be it transistor, FET, MOSFET or tube-or even combinations of all three. When transistors were first introduced, the tube device was declared obsolete overnight, and many virtues were claimed for the transistor designs. Of these, very few have been of proven worth. Indeed, tube amps built prior to the onset of the transistor have maintained a reasonable value, while most of the early transistor models have justifiably sunk without a trace. Many reviewers and designers fooled themselves or in some cases were insufficiently experienced to know any better.
The introduction of FETs in recent years was hailed as a solution to transistor ills by combining the virtues of both the tube and the transistor with the defects of neither. Integrated circuits using transistors and/or FETs appeared and again were subject to an inconsistent mixture of abuse when used in amplifier design.
I can, however, heap range of good and bad equipment that incorporates every possible combination of the aforementioned active devices. In the case of Audio Research, tubes are good, while for Krell and Naim, bipolar transistors rule the day.
In a more modest price category that includes products from Mission, Perreaux and Hafler, MOSFETs are a success. Examples of poor technique are legion and are probably best omitted here. At present, I cannot see a clear winner in the active device context; probably the ancient tube holds the ultimate honors for now.
Power Supplies
Much technical justification is given for the importance of good design when it comes to amplifier power supplies. Since it can be shown that all the signal power is derived from, and also passes through, the power supply, this section should be considered as important as the amplifier itself. To this end, designers and copywriters have attempted with some justification to promote the need for bigger and better reservoir capacitors, as well as larger power transformers, double-mono construction involving a separate supply for each channel, expensive regulated or stabilized power supplies and finally, separate supplies for every identifiable amplifier stage. Once again, I could take considerable space to outline the technical pros and cons of these various configurations. Instead, I shall offer a few amplifier examples, which in a general sense, neatly contradict most of the stances.
Consider the Naim 250 and Robert son Forty-Ten, both fine semiconductor power amplifiers of a similar power rating. The Naim has regulated supplies, while the Robertson does not. Or take the case of the Robert son and the Krell. The former has a single, shared power transformer of modest size, while the latter has two enormous power transformers arranged in a double-mono configuration. The Naim NAP135 is a fully regulated, double-mono system, yet arguably it is no more effective in subjective terms than the Audio Research D-70 II with its normal shared power supply. The Naim NAC32 preamp continues to sound better and better with increasingly gross power transformers, culminating in the 250VA model at present featured in the Hi Cap power supply unit. While the and consequently afterthoughts; in such […] is expected.
Counterpoint SA-7 survives com parison with the Naim well, it uses only a puny 20VA transformer.
Many designers actively decry the use of integrated circuits, but both Ben Duncan, in his AMP-01 HFN/ RR preamp project (May-November 1984) and West German designer Dieter Burmester have shown that their use can produce fine results.
These apparent contradictions are partly due to circumstance. An idea that works with some degree of success in one model may prove ineffective in another. These various ideas are rarely for general application.
Separate Integrated Amps Even the generalization that pre/power combinations are better than integrated amplifiers can be contradicted given the appropriate evidence. Ultimately, it is probably true that 'separates' are better than integrated amplifiers, but in the case of the majority of models currently avail able, this statement must be regarded as not proved.
In recent years, a cozy establishment of middle-to-upper-rank separate pre/power combinations has emerged.
Among these is the Musical Fidelity pairing, The Preamp and the Doctor Thomas power amp. Naim has several offerings, such as ~ 42/160, while Meridian has the 101/105, Hafler the 110/220, Sugden the C128/P128, and A&R Cambridge the C200/P200.
These products have, however, recently come under considerable pressure from a number of integrated amplifiers offering a price saving of around 25 to 50 percent.
Naim led the way with its NAIT, followed by Rotel and its B and now BX versions of the RA820. Musical Fidelity has almost scored its own goal with the fine Synthesis integrated model, and Audiolab almost achieved the same result with the 8000A integrated even before the company's separates had been released. Mission has also put its massive pre/power amplifier in jeopardy, first with the MOSFET 778 and later, and even more convincingly, with the diminutive Cyrus One and Cyrus Two, both integrated designs.
Many of these new models, costing in the region of $168 to $420, equal and occasionally better the performance of accredited pre/power combinations costing up to four times as much. Where, then, is the claimed clear advantage of the separate form of construction, never mind the justification for the far higher cost?
Negative Feedback
The level of feedback is another debatable point. At one time, it was regarded as the panacea for all ills, but now high levels of negative feedback appear to be going out of fashion. The use of 70dB of overall loop feedback was common with many transistor power amplifiers and in accordance with basic operational amplifier theory, offered many benefits, including very low harmonic distortion and a high input and low output impedance. The latter represents another claimed “plus ” parameter, namely a high damping factor.
As a pure generalization, it is possible to establish a link between poor sound quality and high negative feed back, while neglecting the question of […] treble took on an added “'grain ” and “edge. ” Both these effects were associated with a reduction in depth model, some intrinsic “softness ” had been identified, and interestingly, a reduction in feedback made this soft ness too obvious, resulting in a decline in subjective quality.
Tone Controls
Tone controls and related facilities such as filters and the like have also come under close scrutiny. These have long been exploited by the mass market as well as audiophile groups.
Yet the present amplifier situation […] such as those at Audiolab, have, however, shown that provided care is taken, the inclusion of tone controls can be a virtually inaudible addition, as exemplified by both the 8000A and 8000C amplifiers. Good filters can be expensive to execute and consequently are often no more than after thoughts; in such cases, a loss of Simple Short Signal Paths Considering the tone control section as a potentially redundant stage brings one to the view that simplicity and directness are important qualities in an audio chain. While this may seem too obvious to be worth stating, my experience suggests that the shorter and simpler the path from cartridge to - (including all the elec and interconnecting cables), […]
The reproduction is more 'immediate' and easier to listen to, while at the same time it is also more involving.
In my discussions and tests on cable, I introduced the concept of cables sounding “short ” on the basis that the best-sounding cable was also physically as short as possible. Such shortness is equally applicable to any part of the audio chain; for example, in some listening tests, I have felt that equivalent of a long length of […].
Some sections of amplifiers commonly have a surprising effect on sound quality and would justify an entire article. Space restrictions permitting, I shall attempt to outline the major ones.
Recently, a team of skilled listeners examined the effect of the small out put inductor present in almost all value and rolls off the amplifier out put above 100kHz or so, thereby improving the stability margin with reactive loads. In the case of the test amplifier, the design was such that the inductor was not essential. When in place as normal, a given sound quality was established and noted by the team. The inductor was then deleted and the unit re-auditioned.
Our expectation was that either the result would be inaudible, as per theory, or that the sound might appear a trifle brighter and coarser due to the wider treble power bandwidth and possibly worsened stability in view of the complex speaker load used.
In fact, the outcome provided an extraordinary contradiction. Not only did the sound differ significantly, it was also improved. The treble appeared slightly sweeter with a greater purity, while the midrange register showed more stereo depth.
In another example, a very fine preamplifier carried a redundant passive section for stereo-mono selection as well as channel balance in its signal path. Furthermore, 20cm of printed circuit signal track was used to link the arrangement. A bypass was fitted to the preamp using selected cable, and this proved that the circuit board track, balance potentiometer and switches noticeably degraded the sound, although they had a negligible measured effect on response or Many more such examples can be found to confirm or refute arguments concerning head amplifiers and pre amplifier front ends, as well as any other aspect of audio technology one might care to bring up.
Amplifier Tests & Specifications
There is now a good body of evidence to indicate that the majority of commonly used tests and specifications are of limited value in quantifying sound quality. Despite the increasing power of laboratory test equipment,
the association between measurement and sound quality seems to be on the decline rather than the ascent.
Distortion is a good case in point.
With the appropriate equipment, it is possible to measure the distortion at decent signal levels to as low as -140dB or 0.00001 percent--one part in 10 million! The general view of distortion is y simplistic and is based on the ps that high distortion levels are bad. Certainly at 3 per cent or more, distortion may become audible on speed and music, but at one tenth that value, this point is doubtful, since the asymmetry and attendant distortion produced in the ear itself are rather higher than this, though highly dependent on the sound level involved.
We remain reasonably happy with the subjective distortion of good analog pickup cartridges and loud speakers, these measuring in the 0.2 to 1 percent range. A figure for amplifier distortion of 0.1 percent also seems reasonable enough, but despite this, their designers have often promoted low distortion as a major quality factor, to the degree that some modern amplifiers can require all the resources of a costly set of lab equipment to quantify them.
Does distortion lower than 0.1 per cent result in better sound? In general, the answer would appear to be no.
While many good-sounding, low-distortion amps have been built, many other amps show comparatively high measured distortion and yet produce a still better sound. Indeed, given the association between high negative feedback and low distortion, aiming for the latter could impair overall quality.
Power Output and Loudness
Even something as obvious as the specification for power output can be highly misleading. Power output is generally given as a linear power, for example watts into a standard load- usually an 80 resistor. Printed figures and literature for commercial amplifiers suggest that these figures matter a great deal. A given international manufacturer might offer a range of amplifiers of different powers graded much like car engine sizes: 25, 35, 50, 80 and 100W, or perhaps 1.3, 1.6, 2.0 and 2.3 liters for the car. Hearing is a nearly logarithmic function, how ever, and linear amplifier powers have little subjective value.
When an amplifier power is rated logarithmically in decibels relative to a standard level, conveniently 1W, 80 set as 0dBW as in my reviews, then a rather different picture emerges.
(Remember also that 2 to 3dB is a just noticeable change in loudness, corresponding to 1 or 1% notches of a volume control, and that the ear quickly adjusts to a given loudness. In so doing, it largely loses its perception as to how loud or soft that sound level actually is.) Those amplifier power ratings may now be specified log arithmically as 14, 15.4, 17, 19 and 20dBW. Notice that the difference between the smallest and the largest amplifier is just 6dB, represented by only 2 or 3 notches on a volume control, and quickly forgotten.
Further argument can show in other ways exactly how little value should be placed on the power output specification. For example, there is the matter of dynamics and dynamic range. The term “dynamics' is another of those difficult subjectivisms that attempt to describe the feeling of attack and impact in the sound, or perhaps whether the amplifier is capable of accurately conveying the contrasts between loud and soft passages or between different music sections, say percussive, where one appears to be played with energy and one with restraint.
A low-power amplifier with good dynamics may sound satisfactorily “loud ” at appropriate points in the music due to its ability to resolve these dynamic contrasts in the music.
The state of stasis generally represents boredom. The brain essentially responds to changes or contrasts, which is, of course, true of all our senses.
The better those transient changes are resolved, the greater is the impact in terms of the listening experience.
What we really wish to know about an amplifier from a power specification viewpoint is how loud it can get.
Of course, loudness itself is a highly subjective phenomenon. Pure constant-level loudness of the 'roaring' variety is vaguely associated with the amplifier's maximum power rating.
Even in this case, the amplifier's loudness may be complicated by other subjective effects. For example, a medium-power amplifier with a hard, aggressive tonal balance may well sound louder than an alternative, naturally balanced example of perhaps two or three times the power rating.
Conversely, a small amplifier with good dynamics may appear more powerful in a “lifelike ” way than a much larger design with a flat, lifeless rendering.
The ear is surprisingly tolerant of short-term limiting and clipping, provided that no long-term spurious effects accompany the limiting. With digital and direct-cut program, the peak-to-mean power ratio may exceed 20dB, and mild 3 to 6dB clipping at the edge of the full program power may not be too important, subjectively. Thus, a 50W rated model of good clipping performance played 6dB into clip can sound louder on a mean volume assessment than a 100W model that has begun to sound unpleasant using as little as 1.5dB above clipping.
Tube amplifiers often clip well and generally sound several decibels louder than transistor equivalents of the same rated power. Low feedback levels also improve the subjective clip ping performance.
These variations mean that a specified 80 power, expressed in watts, is a very weak indicator of subjective, life-associated loudness. This conclusion can be reached without recourse to further valuable arguments concerning adverse load power delivery, the latter relating to a recently popularized parameter-peak output current.
Peak Current
In company with several reviewers and designers, I find the concept of peak current useful in the case of transistor amplifiers. The peak current demands of complex loudspeaker loads have been investigated and found to be more serious than anticipated.
Commercial speaker systems in the 4 […] recent batch of 15 such designs, no obvious improvement in sound quality has been obtained when compared sors. In these cases, output current was clearly not the major limitation.
Alternatively, take the case of a the audio industry to rely more on enced ears and less on e The Audio Research D-70 is a case in Driven hard, the D-70 must current clip into many loads, yet it sounds little the worse for this. Its high inherent quality is clearly sufficient to outweigh such a difficulty.
Perhaps its current clipping is so good that given the present state of the art, the moderate current capacity does not represent much of a limitation?
[...] speaker cable, 0.5 to 1 ohm of inductor wire and 4 to 6 ohm of motor coil wire.) i damping is really considered necessary, the Audio Research D-115 tube amplifier provides neat proof.
Despite its poor damping factor of around 15, it offers some of the finest subjective bass-much better than the vast majority of bipolar amplifiers with much higher damping factors.
The key to good bass is locked within the circuitry and does not significantly relate to an externally measured damping factor.
High separation is considered a virtue, but just how high? CD players […] as a whole must learn n the evidence of experi […] expensive lab equipment.
[…] can provide 100dB, yet few analog cartridges better 30dB, even in the mid band. Several recent outstanding amplifier introductions (namely the Audio Research SP-8 preamp and the Naim NAIT and Mission Cyrus integrated models) display what an advertising copywriter would describe as poor separation. Very probably, 45dB or better is a good enough figure for channel separation.
CD itself is a fascinating example of nearly perfect specifications allied to a continuing spectrum of sound quality yo shown between various models. Measuring CD players is an in activity, but so far the correlation between measurement and sound quality is weak in the extreme. We are presented with the technically absurd result that a top class analog turntable playing a top quality disk can still sound marginally better than the CD equivalent. By all the amplifier-related specifications, the CD reproducing chain is infinitely superior to the vinyl alternative, yet our ears may tell us otherwise.
The Need for New Research
It is clear that audio technology has outgrown itself. The rules and specifications are not appearing to work anymore. Most audio designers are groping in the dark and just guessing at the likely requirements for good specifications as well as good sound.
Take the case of something as apparently innocuous as a piece of wire.
How can a meter or two of ordinary audio cable affect sound quality more than the insertion of a complete pre amplifier line stage in a given audio chain?
The published specification of an amplifier is of little value in determining its sound and even fails to tell us how loud it is going to be. Indeed, many aspects of established specification may actually be harmful to the production of good sound. This is because those now-unknown parameters that we should be emphasizing may well be unwittingly sacrificed to obtain a good accepted specification.
Much recent evidence suggests that this is what happens.
None of the wide variety of contradicting audio technologies seems to solve the problem, either. All have their own costs, strengths and weak nesses, and their success or failure rests in the hands of the designers who apply them. In this context, it seems that the designer's personality, taste and character matter more than his choice of technology. Good hi-fi is more the product of an experienced craftsman than the intellectual creation of a theoretician.
Established specifications do not cover the sonic effects of passive components, cables, switches and internal wiring, yet these can all influence the sound as much as the circuits themselves.
Fourier Does Not Rule
Established technology and measurement is based on the steady state, generally on the use of a steady-tone stimulus. Electronics theory tells us that the time and domains are simply alternative views of the same phenomena, and for convenience sake, we use time set to “infinity” with frequency set to a fixed value to carry out most tests. Fourier theory tells us that any repetitive impulse or transient may be broken down into a series of superimposed discrete frequencies of specific amplitude and phase relationship, and we can use a Fourier analyzer to capture an impulse, assume it is repetitive and transform it into a frequency response, thus illustrating its transfer function.
Both Malcolm Hawksford and John Atkinson have pointed out that speech and music are discontinuous random signals, having no clear relationship with any of the assumptions on which amplifier measurements are based. If the subjective qualities of dimensionality, life, attack, tonality, etc., are primarily related to non repetitive transient phenomena, then our established steady-state analysis may be of little value.
More research into the psycho acoustics of hearing is urgently required, with relevance to the technical aspects of audio equipment design and specification, particularly for music transients. The audio industry as a whole must learn to rely more on the evidence of experienced ears and less on expensive lab equipment.
A slavish compliance with the apparent commercial needs for better and better specifications must be resisted. New tests are needed to better define the subjective areas of amplifier performance-those including passive components. New technology for its own sake must also be resisted until its merits are proven, and at this stage in our understanding, this means proven by critical listening tests.
During some 40 years of very expensive development, the transistor has not yet beaten the tube in reproducing music when judged at the highest quality level. To twist the knife even further, the active tubes in such noted examples are themselves 30 years old in design terms.
All this does not imply that all established specifications are valueless.
The basics of power, level, impedance, linearity and amplitude frequency response are certainly necessary, but need only be taken to realistic limits and thresholds. Such tests also help to control model consistency once in production, as any fault that alters a basic parameter is also likely to affect the sound, provided that me first and specification passive components remain unaltered.
In our present state of ignorance, however, I can only reiterate that sound must come first and specification second.
----
Also see:
PREEMINENT PHONO PREAMP, By John H. Roberts
avahifi's Omega 260 (1996 model): MOS-FET Amplifier 130 watts/channel
LOW-DISTORTION, LOW-FEEDBACK POWER AMPLIFIERS, by R. N. Marsh