Two class D Cherry amplifiers that i use now (a small integrated
Maraschino and a larger Megamaraschino) do reproduce with little
distortion, clear and detailed as far as i can see.
Class D might be the one way out of this. The issue is something called 'gain bandwidth product' and has to do with how much gain you have combined with bandwidth. That's easy enough to understand; where it gets tricky is when you apply feedback. If the bandwidth is limited, you can't apply much feedback because the amplifier will have phase shift at its bandwidth limits and this phase shift can result in the negative feedback becoming positive if something isn't done to limit it.
But the problem is that the application of feedback introduces distortions of its own. An amplifier with good linearity prior to feedback, and having a fairly benign distortion signature (meaning a predominant 2nd or 3rd with not much above that) will have many more harmonics and intermodulations (which form at the feedback point), while doing a pretty good job of suppressing the lower orders.
Since the ear converts all distortions into tonality, the addition of the higher orders and the IMD makes the amp sound brighter and harder. To get rid of this issue, you have to have enough feedback so that the feedback is able to compensate for the distortions it adds. Generally speaking this value is in excess of 35dB; if the amp has less the feedback has detrimental effects although it might look 'good' on paper.
(This is why feedback has a bad reputation and why there are a good number of products that use none.)
So to get 30dB of gain out of the amp, you need 75dB as a minimum if the amp is operated open loop (no feedback). That's a lot! That's why gain bandwidth product becomes a useful phrase because getting the gain and the bandwidth at the same time gets tricky to say the least.
Now a class D amp can get around this problem through a fairly simple mechanism called self-oscillation. The idea here is to add so much feedback that the phase margin of the amp (where negative feedback becomes positive due to phase shift as I mentioned earlier) causes the amp to go into oscillation. You simply use that oscillation as the switching frequency. Class D amps make gain through the comparison of the switching waveform (usually a triangle wave; its fairly easy to convert a squarewave to triangle, so with a small bit of circuitry we use the switching of the amp itself for this) to the input waveform. So all you have to do is set the ratio between the two to set the gain, and then run as much feedback as you need. This can result in extremely low distortion and no phase shift, even if the output filter frequency is set fairly low. Time will tell if this proves out to be a viable path.
