On Sun, Aug 12, 2012 at 07:38:58PM +0200, "Jérôme M. Berger" wrote:
Actually, that's one point where PA is right (even though it's wrong on a lot of other points): doing it like (2) avoids amplifying the quantification noise if the sound card applies the master gain in analog (or uses higher bit depths internally before the DACs as some do).
True, if the master is after the DAC, but even then irrelevant. Quantisation noise for a typical 20-bit (or even 16-bit) DA is low enough so it doesn't matter. See previous post.
When cascading amplifiers, it is always better to put the highest possible gain on the first stages (always leaving enough headroom to avoid clipping/distortion) so that later stages will not amplify the noise from the first stages (or so that they will reduce it along with the signal). The only case when this rule does not hold is when doing digital processing in floating point (because then the quantification noise is defined as a proportion of the actual signal instead of its potential maximum).
Correct again. But there's not reason why a software mixer shouldn't use floating point, or a fixed point format (e.g. 32 bit integers) that provides enough room above and below. Please don't tell me that PA is using 16-bit for its internal operations - that would really prove it's complete crap. Ciao, -- FA A world of exhaustive, reliable metadata would be an utopia. It's also a pipe-dream, founded on self-delusion, nerd hubris and hysterically inflated market opportunities. (Cory Doctorow)