Since I am doing my PhD on Class D amplifiers, I thought I would weigh in on the feedback discussion of this topology. In general, there are two feedback loops - take that! haha. It depends on the complexity of the control system but often there is a current feedback and a voltage feedback. The tracking error is controlled via both methods since the modulation frequency varies with the input (usually the case for audio class D amps) and thus makes regulating the gain tougher (done with a comparator). Having two feedback loops allows better tracking of the signal since the switches are more-or-less decoupled from the signal - unlike a linear amp. You can even go more complex and have feedforward loops that will compensate for changes in input line voltages thereby creating built-in power conditioning.
I am leaving out a lot of details because there are zillions of topologies and I am not familiar with all of them. I mainly focus on Pentium power supplies which are have even higher switching frequencies than audio amp topologies and even then, timing error is really not that bad. What I dislike about Class D audio amps is the fact that the signal is quantized...and there isn't any real way around that - and has a greater impact than feedback delays on the signal integrity. The CD doing it is enough for me. I have however read the CI white papers and have attend talks by the Philips engineers that created the basis of many of today's audio amp Class D circuits and they undoubtably work.
The effect of feedback in audio will continue to be a debate until we have better equipment to measure the differences (or discover new variables) that, apparently, only our ears know about today. One thing for sure is that less feedback is used in linear amplifiers today than was used in the 80s mainly because transistors have gotten much better - which is also part of the reason the sound is much better too...zero global negative feedback is getting more praise than it deserves IMO.
I am not sure we know everything about their circuits but if Ayre uses no local feedback, I won't be buying one of their amps. The bias would have to be reset every month - unless you listen to it everyday in which case you probably wouldn't even notice...
I am leaving out a lot of details because there are zillions of topologies and I am not familiar with all of them. I mainly focus on Pentium power supplies which are have even higher switching frequencies than audio amp topologies and even then, timing error is really not that bad. What I dislike about Class D audio amps is the fact that the signal is quantized...and there isn't any real way around that - and has a greater impact than feedback delays on the signal integrity. The CD doing it is enough for me. I have however read the CI white papers and have attend talks by the Philips engineers that created the basis of many of today's audio amp Class D circuits and they undoubtably work.
The effect of feedback in audio will continue to be a debate until we have better equipment to measure the differences (or discover new variables) that, apparently, only our ears know about today. One thing for sure is that less feedback is used in linear amplifiers today than was used in the 80s mainly because transistors have gotten much better - which is also part of the reason the sound is much better too...zero global negative feedback is getting more praise than it deserves IMO.
I am not sure we know everything about their circuits but if Ayre uses no local feedback, I won't be buying one of their amps. The bias would have to be reset every month - unless you listen to it everyday in which case you probably wouldn't even notice...