Test tones are used to show how equipment performs below audible levels but how music performs at listening levels is the deciding criteria. In that regard science fails miserably.Actually it doesn't.
But science is rarely applied when doing measurements. If you want the measurements to show what you need to know in order to know how the equipment will sound, don't hold your breath because spec sheets are generally created as a sales tool.
Now it is a simple fact that we can measure and correlate what we measure to what we hear and that is entirely due to studies of how the ear works, mostly done in the last 50 years.
If you want to know what to look for, take a look at a post I made on another thread (sorry, its long):
Distortion is in all forms of amplification. Of course we want it to be as low as possible but the ear poses some real challenges.
It uses the higher ordered harmonics (5th and above) to sense sound pressure, and to do that it has to be keenly sensitive to them! The ear is more sensitive to the higher ordered harmonics than almost anything else. For this reason, a THD of 0.01% can be very audible if that’s mostly higher orders.
The ear assigns tonality to all forms of distortion! The higher orders get ’harsh and bright’. We’ve all been hearing this in most solid state amps made in the last 70 years. Its why tubes are still around!!
The lower orders (2nd, 3rd and 4th) are nearly inaudible and mostly contribute to ’bloom’ and ’warmth’ using audiophile terms. The ear has a masking principle where louder sounds mask the presence of quieter sounds; if the lower orders are in sufficient quantity, they will mask the presence of the higher orders. When this happens, the amp will sound smooth and because the lower orders are mostly inaudible it will appear to be relatively neutral. If the amp has such a distortion signature this will be the case whether tube or solid state. But for technical reasons (its very hard to build a zero feedback solid state amp), until recently this has mostly been describing tube amplifiers.
The problem has been up until fairly recently that the the devices (whether tube or semiconductor) didn’t exist to allow the amplifier design to have a sufficient amount of feedback (on an engineering basis, the devices didn’t exist to allow for sufficient gain bandwidth product). You have several hurdles to cross; first when adding a lot of feedback you can’t exceed the phase margin of the amp otherwise it becomes unstable and can oscillate. Phase margin is an engineering way of saying that there’s a certain high frequency above which the negative feedback applied is no longer negative due to phase shift in the circuit.
The second problem is you have to have an enormous amount of gain- and with gain you get phase shift- because you really need north of 35dB of feedback in order to allow the amplifier to clean up the distortion caused by the application of feedback itself (which tends to be almost entirely higher ordered harmonics, caused by the process of bifurcation occurring at the point where the feedback is combined with the input signal). These higher ordered harmonics are of course audible which is why feedback has gotten a bad rap in high end audio over the last 40 years.
So you have to blow off 35 db of gain with feedback and still have a good 25dB of gain left over- so this means that at a minimum any amplifier that uses feedback properly will have a total gain (called ’loop gain’) of at least 60dB! Most amps made have far less than that which is why solid state has garnered a reputation for harsh and bright. Between 12 and 20dB is the area where feedback generates the most distortion: its on a bell curve. Yes, it does suppress distortion but my point here is that its makes some of its own too.
Because this is such a tall order, most amps simply didn’t do it. To deal with this problem, the industry (sweeping this under the carpet) only tests harmonic distortion of amplifiers at 100 Hz. At this frequency almost any solid state amp has enough feedback which is why they can play bass so well. But if you measure the same amp at 1KHz or 10KHz you’ll find the distortion is much higher- and of course that is why the amp sounds bright and harsh (its not a frequency response error). This increase of distortion with frequency is a sign that the amp lacks Gain Bandwidth Product. GBP is to feedback what gas is to car. When you use it up by increasing frequency, at some point there’s no more feedback. At any rate distortion is increased!
Tubes avoid this for the most part by having a greater amount of the lower ordered harmonics. So they lack the harshness and brightness not because they are lower distortion but **because the higher orders are masked**.
There are a number of solutions. One way to get tubes to be much lower distortion is to design the circuit to be fully differential and balanced from input to output. In this way, even orders are cancelled not just at the output but throughout the circuit. This results in a 3rd harmonic as the primary distortion and since the 3rd is quite close to the fundamental is treated by the ear the same as the 2nd. But it can easily be at a level 1/10th that of an amp that does not employ this technique, and succeeding harmonics will fall off at a faster rate according to a cubic progression because distortion isn’t compounded from stage to stage. For this reason such an amp is said to have a ’cubic non-linearity’ and is considerably more neutral and transparent than amps that express the 2nd order as dominant (a ’quadratic non-linearity’), yet just as smooth. This is true whether the circuit is tube or solid state.
Feedback can be avoided altogether, thus avoiding the brightness that occurs with its application. SETs are an example of this as well as our OTLs (which are fully balanced and differential) and there are solid state examples as well, such as the Ayre.
Another solution is to simply have enough gain and bandwidth using newer semiconductors so enough feedback can be applied so that the amp has consistent distortion at 1KHz and 10KHz as it does at 100Hz, and won’t oscillate with +35dB of feedback. This is a bit of a trick but it is doable and there are a few solid state amps of traditional design that do this- the Benchmark and Soulution come to mind.
Finally, class D amps can be built that have so much feedback that their phase margin is grossly exceeded and they go into oscillation as soon as they are turned on. The oscillation is then used as the switching frequency. This type of class D amp is known as ’self oscillating’ and can have very low distortion. Because of non-linearities in the encoding scheme and also due to dead time, lower ordered harmonics might be generated. If this is the case, such an amp will sound every bit as smooth and transparent as the best tube amps (due to masking) but with greater neutrality and transparency due to vastly lower distortion overall (in case its not clear, distortion masks detail).
So the bottom line is the distortion signature is more important than how much distortion is actually present. That is what the spec sheets aren't showing and why there's often a disconnect between what you hear and what is measured. Its not that we *can't* measure it, its simple because most of the time we simple *don't* measure it.