On Wed, Jun 20, 2018 at 11:31:37PM +0200, Denis Buga wrote: > sorry for bad english
no problem > if simplify, seems like we have three pairs of concepts, while in audio path > from file to card > file (Hz/bit), sndiod (Hz/bit), card (Hz/bit) > first and third is more or less i understand > by second i mean sndiod's "dsp" > i understand, that, when i set up "sndiod_flags" with frequency "N" and > "aucat" file with same "N" > - i can think about it as, simplifying, "passthrough by frequency" yes, true > seems like, the same thing i can think about "bits" in "sndiod_flags" > my question is about "ADATA_BITS" in "dsp.h": sndiod always uses ADATA_BITS internally, no matter the file format, the sound-card format or the sndiod_flags. By default, ADATA_BITS=16, so it uses 16-bit, signed, native byte order (aka "s16"). > to get "bit-perfect" audio, do i need to set this value in all the way with > file, sndiod_flags > if we say that we need "perfect line", is it nessesary to have > ADATA_BITS same as "-e ...N... " in sndiod_flags > ? With ADATA_BITS=16, the only way to get bit-perfect data is to use files with the "s16" format on a sound-card supporting "s16". The volume must be set to the maximum (-v 127 -w off), so that sndiod wont modify the samples, and only one program must be playing. Note that if the file is 8-bit or 16-bit non-native byte order, the resulting data won't have the same binary representation, but I'd qualify the resulting sound as bit-perfect. This is because the original bits representation can be recovered back from the sndiod output. As long as sndiod and the sound-card have better precision than the file, the sound restitution will be perfect (assuming there's no sample frequency conversion, volume changes, etc).
