Hi, colleagues!
I’ve improved the input stage of the amplifier by adding a tracking power supply to the comparison node. The results are shown in the images below. I see no reason to modify the second stage— even if the actual distortions are 100 times worse, the harmonics at 20 kHz with 116 W (4 Ω) output will still be below 0.001%. At other power levels and frequencies, the distortions will be even lower.
