Thanks, I never even thought of the issue of timing of transients for amplifiers before. That makes a lot of sense. But then for conventional amplifier designs, people also talk a lot about feedback. So even in the scenario where feedback is "properly done", is it just a matter of trade-offs currently where if you have more feedback, you're going to get better low-level linearity but it may cause more noise floor modulation and worsen the timing of transients? Whereas if you have less feedback, you're going to have more distortion and worse low-level linearity but then you're going to get slightly better timing of transients?
OK, you are talking about a number of different problems together, so lets try to un-tease your Gordian knot, without recourse to swords!
In class D we do indeed have a problem of timing of transients, and it sounds just like the problems I have with digital in converting sampled data back to the original continuous analogue signal - you can't hear the starting and stopping of notes correctly, and the music does not flow properly. In the case of Class D, the signal is being sampled but the timing is signal related, and the OP LC filter reconstructs the timing of transients - but it can not do an effective enough job of doing that. Another massive problem with Class D is noise floor modulation.
With linear amps, of whatever type, here are some notes:
1. Noise floor modulation - this is feedback independent, as the first stage is normally responsible for this - so subsequent gain and feedback can't fix an input stage problem. To do this you need lots of decoupling, RF filtering, and careful grounding and other techniques.
2. Low level non linearity - this isn't solved by feedback either, as most of the problems comes from the direct signal path - the coupling components and the feedback components themselves. You can only solve this by reducing the number of metal/metal interfaces and reducing the oxides and impurities at the interface..
3. There is nothing wrong with feedback per se. What the problem is is poor high frequency performance - and that can be caused by poor open loop distortion, or not enough feedback for high frequencies - its way more complex than simple "no global feedback" or "feedback is bad" argument.
4. Timing in linear amplifiers - this is not the same as the digital timing problem, as its more of a continuous non linearity problem. It is where the delay changes with signal level - this can occur with cored inductors or with OP stages that have signal related propagation delay. This distortion is known as phase intermodulation distortion (PID), and it sounds different to the timing of transients problem - reducing PID makes it sound smoother and darker, with better instrument separation and focus.
The problems of 3 and 4 can be solved by using an analogue noise shaper approach and this is something I have used well before with Dave. It pretty much eliminates HF distortion, and solves the PID problem of output stages too.
I hope this clarifies a very complex subject,
Rob