The Sound of Audio Amplifiers: Can you hear a difference between Amps?

Originally published March 4th, 2013

Introduction

Whether or not amplifiers sound different is a subject of ongoing controversy. There are sensible, well-informed enthusiasts and magazine reviewers who will swear to their many, obvious differences, differences that are—to them, anyway—almost as obvious and important as the differences between speakers.

There are also just as many well-informed enthusiasts and reviewers who say that there are no meaningful differences in the sound of two properly-functioning, properly-performing amplifiers operating within their undistorted performance envelope.  And there is a third contingent that opines that the in-situ application of different amplifiers with different speakers and other associated equipment and connectors can produce differences in the sound even though nothing in the system is misbehaving per se.

We’ll open up the discussion, put forward some factual information, relate a few anecdotal experiences and look forward to your responses.

Rod Elliott of Elliott Sound Products (aka. ESP) has written a great article, called Amplifier Sound and we suggest reading it in its entirety.  We've excerpted some important passages here and we’ll use those for the basis of our discussion.

Please also watch our recently added YouTube video where Gene and Hugo discuss amplifiers and debate whether or not can sound different and the reasons behind it.

Do All Amplifiers Sound The Same?

[Elliott]

The sound of an amplifier is one of those ethereal things that seems to defy description.  I will attempt to cover the influences I know about, and describe the effects as best I can. This is largely hypothesis on my part, since there are so many influences that, although present and audible, are almost impossible to quantify. Especially in combination, some of the effects will make one amp sound better, and another worse. I doubt that I will be able to even think of all the possibilities, but this article might help some of you a little, at least to decipher some of the possibilities.

I don't claim to have all the answers. This entire topic is subject to considerable interpretation, and I will try very hard to be completely objective.

The Components of Sound

When people talk about the sound of an amplifier, there are many different terms used. For a typical (high quality) amplifier, the sound may be described as "smeared", or having "air" or "authoritative" bass. These terms - although describing a listener's experience - have no direct meaning in electrical terms.

Electrically, we can discuss distortion, frequency response, phase shift, current capability, input/output impedance, and numerous other known phenomena. I don't have any real idea as to how we can directly link these to the common terms used by reviewers and listeners.

Some writers have claimed that all amplifiers actually sound the same, and to some extent (comparing apples with apples) this is "proven" in double-blind listening tests.

[Audioholics]

Several years ago, noted reviewer Tom Nosaine conducted a study for Stereo Review magazine in which he attempted to compare the sound of two systems to a listening group through blind A-B testing. One system was made up of the very latest SOTA components and connectors.

The other system was comprised of very “pedestrian” equipment, such as a 1970s-vintage Heathkit power amplifier, a garden-variety CD player, etc.

Both systems were connected via a switcher to the same pair of speakers, so unit-to-unit speaker differences and speaker placement issues didn’t enter the picture at all. An acoustically-transparent but visually-opaque drape was drawn in front of the systems and the room lights were dimmed to assure no visual distraction. Nosaine was well known for his adherence to the scientific process and his rabid attention to detail, so the scenario was well executed.

His results showed that as long as neither system was pushed into distortion there was no statistically-significant preference shown for either system by the group, using a very extensive and varied selection of program material. His listening panel included novice listeners, mild enthusiasts and self-professed “golden eared” audiophiles.

It made no difference. Whether novice or golden ear, no listener or group of listeners scored any better at identifying which system was which.  Supporters of the “no difference in amps’ sound” side were quick to say, “See? Told ya so.”  Believers in the “amps’ sound is easily identifiable” camp were just as quick to point out the myriad aspects and conditions of the test that were faulty in their view and thus rendered Nosaine’s conclusions inaccurate.  So, in the end, his test settled nothing.

 

[Elliott]

There are some differences that cannot be readily explained. An amp that is deemed "identical" to another in a test situation, may sound completely different in a normal listening environment. It is these differences that are the hardest to deal with, since we do not always measure some of the things that can have a big influence on the sound.

For example: It is rare that testing is done on an amplifier's clipping performance - how the amp recovers from a brief transient overload. I have stated that a hi-fi amplifier should never clip in normal usage - nice try, but it IS going to happen, and it is more common than we might think. Use a good clipping indicator on the amp, and this can be eliminated, but at what cost? It might be necessary to reduce the volume (and SPL) to a level that is much lower than you are used to, to eliminate a problem that you were unaware existed.

Different amplifiers react in different ways to these momentary overloads, where their overall performance is otherwise almost identical. I have tested IC power amps, and was dismayed by the overload recovery waveform. My faithful old 60W design measures about the same as the IC in some areas, a little better in some, a little worse in others (as one would expect).

Were these two amps compared in a double blind test (avoiding clipping), it is probable that no one would be able to tell the difference. Advance the level so that transients started clipping, and a fence post would be able to hear the difference between them. What terms would describe the sound? I have no idea. The sound might be "smeared" due to the loss of detail during the recovery time of the IC amp. Imaging might suffer as well, since much of the signal that provides directional cues would be lost for periods of time.

[Audioholics]

Yet it’s not just clipping behavior that distinguishes the sound of one amplifier from another. Many people swear that there is a fundamental difference in the overall tonal quality between amps, that, indeed, their ‘color’ or ‘character’ differs from each other.

Here’s another great anecdotal example: Many years ago we were listening to our latest albums on a very high-quality (for that time) system—a top-of-the-line Kenwood integrated amplifier (rated at 60/60 wpc RMS with vanishingly low distortion), AR-3a speakers and a Dual 1249 turntable with a Shure V15 cartridge. Highly-regarded equipment, operating well within its intended performance environment.

My friend had just purchased a Dynaco ST-120 power amplifier (60/60 wpc RMS) and he wanted to make sure it worked properly, so he brought it over. The Kenwood had pre-out/main-in jacks, so we used the Kenwood as the preamp. The speakers, turntable, cartridge, and speaker wires remained the same. The only change in the system was the power amp.

We weren’t looking to “compare” the power amps’ sound, we were only looking to confirm that the Dyna worked.

We played the original system, then we swapped in the Dyna and played the same material.

The Dyna worked, but the difference in sound character was stunning. Stunning. We raised and lowered the volume. The differences persisted at all levels. We reconnected the Kenwood and re-listened. Then we re-connected the Dyna again.

Now, remember, originally we weren’t consciously looking for differences in their sound, but it was so obvious that it just hit the two of us over the head like a ton of bricks. It was so obvious and apparent that we spent the rest of the night listening to all manner of records, first on the Kenwood, then on the Dyna, and over and over.

There might be several explanations for the differences, but they were real, without question.

Oh, just for the record, the Kenwood (a more modern design) was much “tighter” and “more controlled” in the bass, while the Dyna was “flabby” and “loose” by comparison.  Night and day.

 

Distortion 

[Elliott]

Technically, distortion is any change that takes place to a signal as it travels from source to destination. If some of the signal "goes missing", this is distortion just as much as when additional harmonics are generated.

We tend to classify distortion in different ways - the imperfect frequency response of an amplifier is not generally referred to as distortion, but it is. Instead, we talk about frequency response, phase shift, and various other parameters, but in reality they are all a form of distortion.

The bottom line is that amplifiers all suffer from some degree of distortion, but if two amplifiers were to be compared that had no distortion at all, they must (by definition) be identical in both measured and perceived sound.

Naturally, there is no such thing as a perfect amplifier, but there are quite a few that come perilously close, at least within the audible frequency range. What I shall attempt to do is look at the differences that do exist, and try to determine what effect these differences have on the perceived "sonic quality" of different amplifiers. I will not be the first to try to unravel this mystery, and I doubt that I will be the last. I also doubt that I will succeed, in the sense that success in this particular area would only be achieved if everyone agreed that I was right - and of that there is not a chance! (However, one lives in hope.)

In this article I use the somewhat outdated term "solid state" to differentiate between valve amps (“tubes”), and those built using bipolar transistors, MOSFETs or other non-vacuum tube devices.

Clipping Distortion

How can one amplifier's clipping distortion sound different from that of another? Most of the hi-fi fraternity will tend to think that clipping is undesirable in any form at any time. While this is undeniably true, many of the amps used in a typical high end setup will be found to be clipping during normal programme sessions. I'm not referring to gross overload - this is quite unmistakable and invariably sounds awful - regardless of the amplifier.

There are subtle differences between the way amplifiers clip, that can make a very great impact on the sound. Valve amps are the most respectable of all, having a "soft" clipping characteristic which is comparatively unobtrusive. Low feedback Class-A amplifiers are next, with slightly more "edge", but otherwise are usually free from any really nasty additions to the overall sound.

Then there are the myriad of Class-AB discrete amps. Most of these (but by no means all) are reasonably well behaved, and while the clipping is "hard" it does not have significant overhang - this is to say that once the output signal is lower than the supply voltage again it just carries on as normal. This is the ideal case - when any amp clips, it should add no more nastiness to the sound than is absolutely necessary. Clipping refers to the fact that when the instantaneous value of output signal attempts to exceed the amplifier's power supply voltage, it simply stops, because it cannot be greater than the supply. We know it must stop, but what is of interest is how it stops, and what the amplifier does in the brief period during and immediately after the clipping has occurred.

Figure 1 - Comparison of Basic Clipping Waveforms

In Figure 1, you can see the different clipping waveforms I am referring to, with "A" being representative of typical push-pull valve amps, "B" is the waveform from a conventional discrete Class-AB solid state amp, and "C" shows the overhang that is typical of some IC power amps as well as quite a few discrete designs. This is a most insidious behaviour for an amp, because while the supply is "stuck" to the power rail, any signal that might have been present in the programme material is lost, and a 100Hz (or 120Hz) component is added if the clipping + "stuck to rail" period lasts long enough. This comes from the power supply, and is only avoidable by using a regulated supply or batteries. Neither of these is cheap to implement, and they are rarely found in amplifier designs.

Although Figure 1 shows the signal as a sinewave for ease of identification, in a real music signal it will be a sharp transient that will clip, and if the amp behaves itself, this will be (or should be) more or less inaudible. Should it stick to the supply rail, the resulting description of the effect is unlikely to accurately describe the actual problem, but describe what it has done to the sound - from that listener's perspective. A simple clipped transient should not be audible in isolation, but will have an overall effect on the sound quality. Again, the description of this is unlikely to indicate that the amp was clipping, and regrettably few amps have clipping indicators so most of the time we simply don't know it is happening.

 

[Audioholics]

The possible reasons that amplifiers had specific sound characteristics were becoming a popular topic among audiophiles in the early 1970’s. There was a 1973 meeting of the Boston Audio Society that featured a presentation by a test equipment maker (I forget which) that could display the spectral characteristics of an amplifier’s THD. This was pretty heady stuff in the early 1970s. A very popular receiver at the time, the Marantz 2270, was well-known for sounding harsh if pushed too hard. At this BAS meeting, we found out why:

Even though its actual distortion rating was low (probably 0.3% THD @ rated output), the scope/analyzer showed that when it hit its clipping point, the Marantz exhibited large amounts of higher-order distortion products.

As we know, harmonic distortion is the unintended signal products generated by an audio device such as a speaker or amplifier that are whole number multiples of the original signal. For example, if an audio device is tasked with trying to reproduce a 40 Hz signal, and instead produces 40 Hz and a small amount of 80 Hz, the 80 Hz product is called harmonic distortion. Small amounts close in multiples (lower-order) to the original signal are barely audible; larger amounts of distortion in greater multiples away from the original signal (higher-order) are grossly objectionable to the human ear. (Remember the harmonic structure of Western music is based on octaves and 3rds, so 2^nd and 3^rd-order distortion is harmonically-related to the music in a way that we do not find dissonant or audibly offensive.  Lower-order THD has to be pretty significant before we notice it in a negative way.) The sum total of all harmonic distortion products is usually expressed as a percentage of the original signal, or % Total Harmonic Distortion (% THD).

FFT Distortion Analysis of Emotiva UPA-7 Amplifier at Near Clipping - notice the higher level of harmonic distortion products (ie. 2kHz, 3kHz, etc)

On the scope, we could clearly see that the Marantz produced large amounts of 4th-, 5^th- and 6^th-order harmonics when it was pushed into clipping. A Pioneer receiver there for comparison was better behaved and its distortion was composed mostly of 4^th-order and lower components. An AR receiver showed by far the best distortion behavior, with most of its harmonics being very benign-sounding 2^nd- and 3^rd-order products. Now, finally, there was a concrete, repeatable, tangible explanation for why some amplifiers sounded better than others when pushed into clipping.

The Loudspeaker & Amplifier Interface

[Elliott]

Many claim that the ear is one of the most finely tuned and sensitive measuring instruments known. I am not going to dispute this - not so that I will not offend anyone, but because in some respects it is true. Having said that, I must also point out that although extremely sensitive, the ear (or to be more correct, the brain) is also easily fooled. We can imagine that we can hear things that absolutely do not exist, and can just as easily imagine that one amplifier sounds better than another, only to discover that the reverse is true under different circumstances. Listeners have even declared one amp to be clearly superior to another when the amp hasn't been changed at all.

Could it be the influence of speaker cables, or even loudspeakers themselves? This is quite possible, since when amps are reviewed it is generally with the reviewer's favorite speaker and lead combination. This might suit one amplifier perfectly, while the capacitance and inductance of the cable might cause minute instabilities in other otherwise perfectly good amplifiers. Although it’s a fine theory to suggest that a speaker lead should not affect the performance of a well-designed amplifier, there are likely to be some combinations of cable characteristics that simply freak out some amps. Likewise, some amps just might not like the impedance presented by some loudspeakers - this is an area that has been the subject of many studies, and entire amplifiers have been designed specifically to combat these very problems [1].

Many published designs never get the chance of a review, at least not in the same sense as a manufactured amplifier, so it can be difficult (if not impossible) to make worthwhile comparisons. In addition, we sometimes have different reviewers making contradictory remarks about the same amp. Some might think it is wonderful, while others are less enthusiastic. Is this because of different speakers, cables, or some other influence? The answer (of course) is that we have no idea.

We come back to a common problem, which is that the standard amplifier tests are not necessarily appropriate. A frequency response graph showing that an amp is ruler flat from DC to daylight is of absolutely no use if everyone says that the highs are "veiled", or that imaging is poor. Compare this with another amp that is also ruler flat, and (nearly) everyone agrees that the highs are detailed, transparent, and that imaging is superb.

We need to employ different testing methodologies to see if there is a way to determine from bench (i.e. objective) testing, what a listening (i.e. subjective) test might reveal. This is a daunting task, but is one that must be sought vigorously if we are to learn the secrets of amplifier sound. It is there - we just don't know where to look, or what to look for ... yet. Until we have correlation between the two testing methods, we are at the mercy of the purveyors of amplifier snake oil and other magic potions.

 

[Audioholics]

Veteran reviewer Julian Hirsch of Stereo Review tackled the subject of amplifier sound in an article in the 1980s. Hirsch was a meticulous reviewer, a formally-trained engineer who put his faith in rigorous, repeatable, scientifically-valid tests and procedures. He completely eschewed imprecise fads like green Magic Markers, Shun Mook M’pingo discs, and $1000/ft interconnects that had no discernable or measurable differences compared to standard cables.

After exhaustive testing of several amplifiers with several different speakers (some known to be “easy” loads and some known to be “difficult” to drive), Hirsch made some interesting observations that simultaneously explained why there were potential sound differences between amplifiers while at the same time remaining utterly faithful to the quantifiable, repeatable scientific process. It remains, in my view, the best explanation for this entire topic that I have yet seen.

Starting with the huge and all-important assumption that the amplifier under test is not being driven into distortion or being operated in a way that could elicit bad behavior, overheating, or engage any protection circuitry, Hirsch found the following:

• Some pre-amp/power amplifier/speaker systems combined in such a way as to produce a very slightly rising or drooping frequency response across the entire 20-20 kHz audible spectrum. It could be a combination of the way the pre-amp and power amp combined on an input/output impedance basis, it could be because of the way a particular amp behaved with the specific load presented by that specific make/model of speaker, but there were system combinations where the frequency response showed a variation across the band.

These were not big variations. Perhaps on the order of + 0.5 dB in the bass to - 0.5 dB in the treble. A 1 dB total tilt, +/-. But over a wide range of frequencies—many octaves—that’s definitely audible. As a matter of fact, this phenomenon is formally known in audio engineering circles as “spectral tilt.” If system “A” is tilted up a dB across the band and system “B” is titled down a dB across the band, experienced listeners will definitely hear that, without question. Godehard Guenther, head of ADS Speakers in the 1970s, talked at length about this effect when voicing his then-top-of-the-line ADS 910 speaker. Extremely slight changes to the speaker’s “tilt” made a dramatic difference in its perceived bass-treble balance and “weight.” Therefore, depending on the program material being used, the interaction of the different components, etc, such a spectral tilt variation could unquestionably lead a listener or reviewer to characterize the amplifier’s sound as “harsh” or “sweet” or “musical,” etc.

• Hirsch also found that small—minute, even—differences in amps’ comparative signal-to-noise ratios often had a tangible impact on their perceived sound.

The SNR difference was not audible as hiss per se. But he did find that a few extra dB of noise could subliminally mix with the program material and lend a very slightly “brighter” or “harder” character to the sound. Granted if amplifier “A” had a SNR ratio of 80 dB and amplifier “B” was 75 dB, both would be considered more than “quiet” enough for high fidelity purposes. But the difference between their SNRs could show up as a perceived difference in tonal character.

This was a great study, perhaps the best ever done on amplifier sound, for several reasons:

• First, like the consummate scientific professional that he was, Hirsch did not enter his study with any preconceived notion as to how the study should come out. He did not try to fit the data to suit his preferred conclusion, as so many lesser “experts” do.
• Second, his results were based on quantifiable, measurable tests. They were not “feelings” or “impressions” or anything else of an inexact nature.

• Lastly, even though he didn’t subscribe to the “subjectivist” approach to reviewing, he had no hesitation in saying that there were indeed differences, and then he proceeded to explain why, in precise, concrete terms. Too many reviewers have very strong feelings about their approach or personal listening skills and these feelings too often prevent them from making truly objective observations, especially if the observation is contrary to their previously-stated position.

Hirsch suffered from no such ego-based limitations, and therefore his findings were more defensible from a strictly scientific standpoint. In my experience, this study presents the clearest tangible, measurable, repeatable evidence yet that there are definitive audible differences between amplifiers.

So, sometimes (not all the time) there are differences in the sound of amplifiers, but according to Hirsch, probably not for the reasons we’d like to think.  Are there other reasons for audible differences between amplifiers, such as the “gut feeling” one gets when viewing beautifully laid-out amplifier circuitry, massive heatsinks, storage capacitors with impressive uf ratings, and a large heavy torridal transformer, all accompanied by convincing technical literature text and an impressive array of reviewers’ quotes lauding the design’s sonic excellence?

It’s tough for any of us to say we’re unswayed by that.

 

Pass Labs X350.5 Two-Channel Amplifier

Measurable Performance Characteristics

[Elliott]

We need a way to correlate subjective versus objective testing. Both are important, the problem is that one is purely concerned with the way an amplifier behaves on the test bench, and a whole series of more or less identical results can be expected. The other is veiled in "reviewer speak", and although it might be useful if the reviewer is known and trusted, is not measurable or repeatable.

The whole object is to try to determine what physical factors cause amplifiers to sound different, despite that fact that conventional testing indicates that they should sound the same.

A detailed description of the more important (from a sound perspective) of the various amplifier parameters is given below.

• Input Sensitivity: The signal level required to obtain full power at the amplifier's output. This is determined by the gain and power rating of the amp. A 10W amplifier requires far less gain than a 200W amplifier to obtain full power for the same input voltage. It would be useful if all amplifiers had the same gain regardless of power, but this is not the case. Sensitivities vary widely, ranging from about 500mV up to 1.5V or more.

• Total Harmonic Distortion (THD): This is a measure of the amount of distortion (modification) of the input signal, which adds additional signal frequencies to the output that are not present in the input signal. THD is commonly measured as a percentage, and can range from 0.001% to 0.5% for typical hi-fi amplifiers. A theoretically perfect amplifier contributes no distortion.

• Intermodulation Distortion: A form of audio distortion where the distortion products occur at frequencies that are sums and differences of the input signal. For example, if the input is 500 Hz and 2200 Hz, then the IM distortion products will occur at 1700 Hz and 2700 Hz. IM distortion is particularly objectionable, since the distortion is not related harmonically in any way to the original signal, unlike THD.

• Frequency Response: The measure comparing the input signal to the output in terms of frequency versus amplitude. A perfect amplifier will amplify all signals equally, regardless of frequency. Realistically, an amplifier needs a response of about 5Hz to 50kHz to ensure that all audible signals are reproduced with minimal modification.

• Phase Response: This indicates the amount of time that the input signal is delayed before reaching the output, based on the signal frequency. Variations in absolute phase are not audible in an amplifier system, but are generally considered undesirable by the hi-fi press. Since it is not difficult to ensure phase linearity, this is not generally a design issue except with valve amplifiers.

• Output Power: This is most commonly measured into a non-inductive resistive load. This is not done to improve the figures or disguise any possible shortcomings, but to ensure that measurements are accurate and repeatable. Power should only be quoted as "watts RMS" over a specified frequency range at a specified level of distortion. Although “RMS” is not strictly technically correct, is accepted in the industry, and may be measured into 8 ohms, or other impedances that the amplifier is capable of driving.

• Output Current: Not often measured, but sometimes quoted by manufacturers, this represents the maximum current the amplifier can supply into any load. It is rare that any amplifier will be called upon to deliver any current greater than about 3 to 5 times the maximum that the nominal speaker impedance would allow for the amplifier's supply voltage. Greater variations may be possible with some speaker designs, but (IMO) this represents a flaw in the design of the loudspeaker.

• Power Bandwidth: This is usually taken as the maximum frequency at which the amplifier can produce 1/2 of its rated output power (this is the -3dB frequency). A 100W amplifier that can produce 50W at 10 Hz and 50kHz will be deemed as having a 10-50kHz power bandwidth.

• Output Impedance : This is the actual output impedance of the amplifier, and has no bearing on the amount of current that can be supplied by the output stage. Valve amplifiers usually have a relatively high output impedance (typically 1 to 6 Ohms), while solid state amps will normally have an output impedance of a fraction of an Ohm. By use of feedback, it is possible to increase output impedance (> 200 Ohms is quite easy), or it can be made negative. Negative impedance has been tried by many designers (including the author), but has never gained popularity - possibly because most speakers react very poorly to negative impedances and tend to sound awful.

The difficulty is determining just how much of any of the problem items is tolerable, and under what conditions. For example, there are many single ended triode valve designs that have very high distortion figures (comparatively speaking), high output impedance and low output current capability. There are many audio enthusiasts who claim that these sound superior to all other amplifiers, so does this mean that the parameters where they perform badly (or at least not as well as other amps) can be considered unimportant? Not at all!

If a conventional (i.e. not Class-A) solid-state amplifier gave similar figures, it would be considered terrible, and would undoubtedly sound dreadful.

Frequency & Phase Distortion in Amplifiers

Distortion of the frequency response should not be an issue with modern amplifiers, but with some (such as single ended triode valve designs), it does pose some problems. The effect is that not all frequencies are amplified equally, and the first to go are the extremes at both ends of the spectrum. It is uncommon for solid-state amps to have a frequency response at low powers that extends to anything less than the full bandwidth from 20Hz to 20kHz. This is not the case with some of the simple designs, and single ended triode (SET) Class-A - as well as inductance loaded solid state Class-A amps - will often have a less than ideal response.

I would expect any amplifier today should be no more than 0.5dB down at 20Hz and 20kHz, referred to the mid-band frequency (usually taken as 1kHz, but is actually about 905Hz). (My preferred test frequency is 440Hz (concert pitch A, below middle C), but none of this is of great consequence.) 0.5dB loss is acceptable in that it is basically inaudible, but most amps will do much better than this, with virtually no droop in the response from 10Hz to over 50kHz.

Most amplifiers will manage well beyond the range necessary for accurate reproduction, at all power levels required to cater for the requirements of music. So why are some amps described as having poor rendition of the high frequencies? They may be described a "veiled" or something similar, but there is no measurement that can be applied to reveal this when an amplifier is tested. Interestingly, some of the simpler amplifiers (again, such as the single ended triode amps) have poorer response than most of the solid-state designs, yet will regularly be described as having highs that "sparkle", and are "transparent".

These terms are not immediately translatable, since they are subjective, and there is no known measurement that reveals this quality. We must try to determine what measurable effect might cause such a phenomenon. There are few real clues, since amplifiers that should not be classified as exceptional in this area are often described as such. Other amps may be similarly described, and these will not have the distortion of a single ended triode and will have a far better response.

We can (almost) rule out distortion as a factor in this equation, since amps with comparatively high distortion can be comparable to others with negligible distortion. One major difference is that typical single ended triode amplifiers have quite high levels of low order even harmonics. Although these will give the sound a unique character, I doubt that this is the sole reason for the perceived high frequency performance - I could also be wrong.

There must be some mechanism that causes multiple reviewers to describe an amplifier as having a poor high frequency performance, such as (for example) a lack of transparency. There are few real clues that allow us to determine exactly what is happening to cause these reviewers to describe the sound of the amp in such terms, and one may be tempted to put it all down to imagination or "experimenter expectancy". This is likely to be a mistake, and regardless what we might think about reviewers as a species, they do get to listen to many more amplifiers than most of us.

Frequency Response of Axiom Audio Class D Amplifier would change radically depending on the load impedance it was driving.

[Audioholics]

As a general observation, it’s amazing how often regular ol' frequency response turns out to be the causal factor in definitive, repeatable sound differences (assuming no gross distortion). The FR variations may have been caused by many different factors, such as interaction between components or connectors or other issues. The underlying equipment itself may not have any inherent FR errors or problems, but when used in combination with specific associated equipment, FR variations suddenly crop up.

This is one of the potential problems of assembling high fidelity component systems for which the average consumer has no solution or remedy. Pre-amp “A” from this company + power amp “B” from that company + speakers “C,” source unit “D,” interconnects “E,” and speaker wire “F” all seem on paper to add up to one terrific system.

But due to circumstances completely unforeseen and unavoidable, they combine for an ever-so-slight FR variance that leads the user to conclude that this or that component is to “blame,” especially when the variance occurs after the latest component change.

This is either a strength of so-called “closed” all-in-one systems, where every link in the system’s chain is controlled by the system designer and one can reasonably expect every part of the system to work correctly with every other part. Or…the all-in-one customer (shelf system, docking station, soundbar, etc.) is so uncritical that even if there are slight FR errors along the way, that customer never notices or cares.

 

Impedance

[Elliott]

The output impedance of any amplifier is finite. There is no such thing as an amplifier with zero output impedance, so all amps are influenced to some degree by the load. An ideal load is perfectly resistive, and has no reactive elements (inductance or capacitance) at all. Just as there is no such thing as a perfect amplifier, there is also no such thing as a perfect load. Speakers are especially gruesome in this respect, having significant reactance, which varies with frequency.  A genuine zero impedance source is completely unaffected by the load, and it does not matter if it is reactive or not. If such a source were to be connected to a loudspeaker load, the influence of the load will be zero, regardless of frequency, load impedance variations, or anything else.  Since this is not the case in the real world, the goal (or at least one of them) is generally to make the amplifier have the lowest output impedance possible, in the somewhat futile hope that the amp will not be adversely affected by the variable load impedance. In essence, this is futile, since there will always be some output impedance, and therefore the load will always have some influence on the behaviour of the amp.

 

Denon 10CH Amplifier Output Impedance is commendably low

Another approach might be to make the output impedance infinite, and again, the load will have zero effect on the amplifier itself. Alas, this too is impossible. Given that the conventional approaches obviously cannot work, we are faced with the problem that all amplifiers are affected by the load, and therefore all amplifiers must show some degree of sensitivity to the speaker lead and speaker.

The biggest problem is that no-one really knows what an amplifier will do when a reactive load reflects some of the power back into the amp's output. We can hope (without success) that the effects will be negligible, or we can try to make speakers appear as pure resistance (again, without success).

 

Skeptical Audioholics Reviewers Becoming Believers?

[Audioholics]

We had a few skeptical reviewers in house at the Audioholics Showcase Theater Room one day when we were testing the RBH T30-LSE speaker system.  Clint DeBoer was among the skeptics that believed all amplifiers sound the same if not overdriven.  With that I connected and level matched our Denon POA-A1HDCI (MSRP: $7500) 10CH amplifier and a Panasonic SA-XR50 Class D  receiver (MSRP: $299).  I purchased the Panasonic receiver after measuring it because I’ve never measured such lousy performance in an amplifier before. Also this is the very same receiver that some audiophiles over at Audioasylum and AVSforum were raving about.  They made comments such as “this is the closest sound to tubes you can get from solid state” and “incredible sound for the money”.  I wanted to use it as a reference going forward to really demonstrate how amps can sound different sonically.  I didn’t tell the listening panel how bad the Panasonic was.  I connected both amps to Zone 2 and Zone 3, respectively, of my Denon AVP-A1HDCI A/V processor and level matched them.  All I had to do was switch the speaker leads each time so the listeners could instantly compare the difference.  At no time were they aware of which amp they were listening to. 

We ran this test at low listening levels (70dB at the listening area) and at higher listening levels (80dB).  Each time I switched between the amps, the listeners all recognized quite a dramatic difference in sound.  The Panasonic actually conveyed more bass than the Denon but it was very boomy and lacked articulation.  At the higher listening level, things really got ugly for the Panasonic.  It was clearly running out of gas and it just sounded nasty.  The reason the Panasonic exhibited boomy bass was likely because of its high output impedance (measured over 1 ohm!) compared to the Denon that measured in milliohms.  The RBH speakers dip down to about 3 ohms so the system damping factor was really compromised with the Panasonic receiver. 

I asked them to now pay attention to high frequency detail and clarity.  I heard comments like “That amp sounds very harsh” or “The sound is very bright but lacks life and ‘air.’” Can you guess which amp these comments were attributed to?  It was again they Panasonic Class Dl.  Even at low power levels where both amps were operating unclipped, there was a clear preference by our listening panel for the better-designed amplifier.  Clint was no longer a skeptic and closed minds were opened.  As an engineer myself, I still recognize that not everything we can measure matters and not everything that matters is necessarily always properly measured. 

This example goes a long way towards proving that amps that measure definitively different (including the very tangible output impedance difference between the two) will sound different—not exactly a big surprise.

However, the audible differences between amplifiers that “golden ears” speak of occur between equipment acknowledged to be quite good by any standards, with excellent measured performance in all the so-called “important” areas. This Bryston may have 0.05% THD and IM and a SNR of 90 dB, and that Adcom may have THD and IM of 0.05% and a SNR of 90 dB, but some people will swear that one or the other is “sweeter” or “harsher” or “veiled,” etc.  With the Denon vs. the Panasonic, we can see the measured differences and accept intellectually why the sonic differences exist.

That’s not what this article is exploring. We’re trying to figure out why Amp A and Amp B sound different despite the fact that all their measured parameters are not only very close, but supposedly past the point of audible relevance. No one can hear the difference on a complex musical waveform between 0.073% and 0.055% THD or between 94 and 91 dB SNR. But you’ll hear the difference blindfolded at 50 paces between 3.2% upper-order THD of the Panasonic when it’s clipping and 0.076% lower-order THD of the Denon. No wonder the Panasonic sounded lousy when it was “running out of gas", but its poor audio performance at unclipped listening levels was also quite interesting.

 

Conclusion

[Elliott]

Will we ever be able to finally perform an objective test, and be able to predict with a degree of confidence how a given amp will sound?

Any tests that might be devised to do so must satisfy both the subjectivists’ and the objectivists’ camps. We are all looking for the same thing - the flawless reproduction of sound - but the two camps have drifted further and further apart over the years.

These are my musings, and I am open to suggestions for testing methods that may reveal the subtle differences that undeniably exist between amplifiers. At the moment we have a chasm between those who can (or think they can) hear the difference between a valve and an op-amp, a bipolar junction transistor and a MOSFET, or Brand "A" versus Brand "B", and those who claim that there is no difference at all.

The fact that there are differences is obvious. The degree of difference and why there are differences is not. It would be nice for all lovers of music (and the accurate reproduction of same) if we can arrive at a mutually agreeable explanation for these differences that is accurate, repeatable, and measurable.

If these criteria are not met, then the assessment is not useful to either camp, and the chasm will simply widen. This is bad news for all - it is high time we all get together and stop arguing amongst ourselves whether (for example) it is better to use one brand of capacitor in the signal path or another.

New testing methods can also be applied to the measurement of individual components, speaker cables, interconnects and preamps.

 

[Audioholics]

The sound variation between amplifiers is as controversial and inexact a subject as any in audio. It is one of the only topics where a large, well-qualified, well-educated segment of the audio community doesn’t even think the subject exists!  Does it exist? Do amplifiers—when used within their undistorted performance envelope—exhibit real sonic differences from each other?  If so, are these differences due to design approach/circuitry differences and component quality? Or are the sound differences caused by their interaction with other equipment?

We’d love to hear your experiences and your impressions as to why.

 

Acknowledgements

Many thanks to Rod Elliot of ESP for his contributions to this article.