I'm trying to understand the various audio format and resampling options in MPD.
I've read that ALSA's resampling is quite bad, and it's the default in MPD. So it's best to set auto_resample to "no".
samplerate_converter: chooses from 6 different options from best to worst. "best" won't play at all on my Pi; "medium" plays with about 78% CPU usage, "fastest" uses about 40%, etc.
According to the man page, when auto_resample is set to "yes", ALSA takes care of resampling, so the MPD samplerate_converter parameter should have no effect. But that's not what I observe here. When I play 96KHz tracks, samplerate_converter affects the CPU usage regardless of the auto_resample setting. Why is that?
auto_format: This disable's ALSA's format conversion, but it's unclear to me whether this is a good idea for best quality.
I think the Pi supports maximum 16 bit, 48KHz with the built-in analog output. So 44.1KHz and 48KHz tracks will all play at their native sample rate, and anything higher will be resampled to 48KHz, is that right?