Amplifiers are discussed about which design and topology is the best.
Often people do care only about numbers in specs sheets, but is that enough?
Transistors or tubes all are nonlinear devices, some more or some less. Distortion happens, when signal is amplified by nonlinear device, (it has nonlinear gain).
Square law is closest to perfect amplifying characteristics. (device with straight lines was not made yet and probably never will) Similarity between square law graph and directly heated triode:
Harmonic distortion and intermodulation distortion are related to nonlinear gain in devices. HD is measured by single test tone, IMD is measured with two tones, or more. Negative feedbacks are used to improve bandwidth, stabilize the gain and lower the distortion and decrease output impedance. It looks nicely on paper to use much as possible negative feedback. But really is it that simple? When signal passes a nonlinear device, damage to signal is done forever, not possible to take changes back. Negative feedback does REDUCE distortion, but it does not restore output to original shape.
Harmonic distortion- mainly even order 2nd, 4th are not much disturbing to human ears; It has been measured that it is hard to detect by ears a high amount of 2nd HD. 3rd HD in low amounts has a low deformation effect on sinewaves, therefore it is also tolerated by ears a little. Combination of 2nd,3rd,4th distortion is what makes amplifiers sound warm, raw etc. (sonic character)
Odd order distortion- 5th,7th, 9th,.. ears are very sensitive to these, because they are dissonant -not a even multiples of fundamental frequency.
Some people prefer DHT SE amps with no feedbacks (count me in). DHT´s have very good linearity, compared to other devices. With high quality components, is possible to avoid feedbacks. Other group of people believes in modern-more efficient amplification, with strong feedbacks. Very often spec sheet numbers on such amplifiers are looking much better than Single Ended amps (x% THD).
While negative feedback looks like miracle, it does not behave like that. When feedback is increased, it improves low order distortion. But at the same time, it creates new HIGHER order distortion.(!) (distortion complexity increases, distortion peaks much higher than average value) It was proven by simulations and measurements years ago.
Now when you created new, higher order distortion, its destructive effect gets increased with intermodulation distortion, creating a "noise floor" which sucks music details out.. The end result is boring, or unpleasant to listen. With simple music it will be not so obvious.(less signals to interact with each other)
But when you listen to complex (chamber) music, it will sound bad.
This detail-masking effect increases with each additional amplifier stages.
If you realize, how many stages a typical modern amplifier has, it gets mindblowing.
Precious details from music are washed out.
Big question is, how to measure this effect, measurement setup needs to be very complex and expensive. More than 10 ultra low distortion signal generators feeding together tested amplifier. All generators frequency spreaded across HIFI range, each one´s amplitude adjusted to typical energy present in music genres.
Last generator would be used for injecting a weak test signal, and measuring it´s properties at the amplifier´s output.
With global FB, phase is not always linear with frequency. Therefore it´s strenght or effects will vary, which can cause oscillations or ringing. Therefore avoid global feedback, use instead local feedbacks (if possible).
Semiconductors guys are not entirely doomed, just use audio rated transistors, which have more linear gains. That way, a lower FB is enough. Avoid transistors for switching use, in amplifiers.