My experience with the Fosi Audio BT30D that i got last week:
* 36V 5A power supply
* compared to a Rotel Pre + Rotel Class AB Amp from the 90's with pimped power supply (bigger, better caps).
* output comes from a very good pimped Rotel DAC
my listening experience:
* the biggest difference was the noise in higher frequencies (like the hihat on a drum set or "S"-vocals in voices)
for the normal listener that does not make a huge difference, but for an experienced hifi/high-end listener this difference is the first thing that you can notice
it is not nice.
* the rest of the listenening experience is good / very good.
* the channel separation is not perfect but good.
* the bass punch is not perfect but good
* no other imperfections where noticable
my interpretation what's going on:
i think that the noise of the switching power supply is going through the class-d amp. the result is a kind of amplitude modulation of the power supply noise with the class-d switching. if you are familiar with amplitude modulation you will understand that there are two side bands. below and over the main AM frequency. that means the noise from the power supply can get mirrored below the switching frequency of the class-d amp. and because you don't have a single switching frequency in the class-d amp but more adaptive to the signal (if my understanding is right), then the modulation of the power supply noise and the switching of the class-d amp result in a crazy modulation mess that is audible in the high frequency range of the audio ouput signal.
this is my interpretation. i did not do any measurements. with a high-res measurement of a digi-scope on the output of the power supply while playing music or a 1khz sine wave test signal, it would be easy to measure.
if i am right - which i don't know - what could be the solution?
==> A signle or double L-C lowpass between the power supply and the amp.
A capacitor alone would not do it, because usually switching power supplies do not like high capacity loads.
That means an L-C lowpass filter that can handle currents for at least 5 amps.
maybe two stages in series, first with a corner frequency of 1kHz, second with a corner frequency of 100Hz (is this too much regarding part size and price???)
another benefit with such an L-C lowpass filter (with a low-ESR cap at the end) could be a better channel separation.