As far as I know, 20 bits was the original S/PDIF format, and AFAIK there are a few devices that will not handle bits # 21 to 24 (not even in the digital domain). I would suppose that in those cases one might just as well have fb2k dither the 24-bits files down to 20 than the DAC cropping down to 20.
On a related note, how and when do bit choice and dithering on the output menu really work? I presume that if the file has larger wordlength but the excess bits are all zero, my choice is irrelevant if the volume is at full and no RG is used (right?). But how does it e.g. output a 16-bit file at less than full volume when
- output is 16, not dithered
- output is 16, dithered
- output is 24 (w or w/o)?
On a related note, how and when do bit choice and dithering on the output menu really work?
In Windows 7 (Vista too I guess) the mix format with DS/WASAPI shared is 32-bit IEEE float, so dithering is in the hand of the audio engine, not fb2k.
foo_out_wasapi gives you the choice of dithering with <= 16 bit output.