Consequences of high-feedback amplifier designs?

The headphone amplifier market has been dominated by opamps, thanks to their ability to acheive extremely low distortion with relatively simple designs and cheap parts.

As I understand it, this thanks to the ability of an opamp to easily feed tons of closed loop feedback current from the output into the input for error correction. (Someone please correct me here if I'm misrepresenting something here)

Making the input of an amplifier dependent on the output intuitively leads me to think there would be some sort of consequences in the time-domain, like a sort of "smearing", but I'm not aware of any technical data or measurements that would support this, if it's even true at all. However, the general lack of time-domain measurements in audio equipment leads me to the possibility that maybe something is being overlooked.

I do like the sound of a good opamp-based amp, but when I listen to other amps that don't have opamps, they seem to have better transient response. Of course, that's not necessarily because of the high levels of feedback or the opamps at all; even discrete designs will usually use feedback loops. Still, I'd like to know if anyone has experieced the same differences.

 

I got a computer science friend interested in these type of questions and the burst attack/decay measurement research @purr1n did but Marv won't respond to my messages regarding it so I assume he feels it is a black hole pursuit or simply has lost interest in favor of other projects.

 

The genesis starts here in 2017. Basically Marv measures impulses at different frequencies from different headphones and creates attack and decay charts. He tries to find the best way to display that information.

Comes up again here in regard to Yggdrasil MIL and "splattery" treble. Marv brings up transient distortion measurement but ejects from further involvement.

 

This is absolutely true. I would link you to some papers on the subject but they are all written in Russian. Actually it is not uncommon in Russian-language audio forums to refer to negative feedback as 'OOS' (out-of-sync), as that is exactly what it is, a feedback loop creating layered signals which are out of sync with the original. This is subjectively perceived as a weakening of sound dynamics and liveliness. Local loops are considered less destructive to music than global ones, precisely because a shorter loop will cause less OOS delay and therefore less destruction.

You can look up the following paper: https://ieeexplore.ieee.org/document/1162117
The author blames the reduction of sound quality on transient intermodulation distortion, and this is a popularly accepted opinion, but his measurements and theory only consider steady-state bursts. This completely ignores the reality of the effects of negative feedback on a real loudspeaker and a variable music signal, where the back EMF from the loudspeaker causes a 'memory' effect within the feedback loop, resulting in the phase shift of the delayed loop signal (smearing) and a loss of clarity.

It is this sort of misleading engineering thinking that has limited the investigation of the effects of negative feedback on unexpected amplitude bursts in a variable signal (i.e. music). It does not help that these effects are largely invisible to measurement equipment. It is for this reason the sound quality of audio frequency amplifiers has been in steady decline since the popularized use of negative feedback in the 1940s.

 

Thanks for the reading material @roshambo123 and @k4rstar ! Those graphs were more or less what I would have expected, I'll need to give some thought to how to keep the work going or possibly improve the measurements, this topic should be very important to the audio community but seems to be overlooked.

Aside from my listening experiences, this talk by Rob Watts from Chord electronics helped kick off my thoughts on exploring time domain measurements: