22 Jul
2010
22 Jul
'10
12:58 a.m.
On Wed, Jul 21, 2010 at 11:10:07AM +0200, rom wrote:
On 21/07/10 09:43, Arnold Krille wrote:
libsamplerate doesnt like a constantly changing resample coefficient.
that gives quite some noise in the high freqs.
the algorithm in alsa_in obtains a pretty constant coeff after a few
seconds.
the phase offset continues to drift around a bit, but not so much.
So please, if you create homepage for the app,
state the problem and the fact
that your (nice!) app is not a solution but only a patch to the problem...
:-D eheh, yes yes i will... i can even provide a sample recording that
clearly shows the problems of this approach: basically, wiring all the
inputs to the same white noise signal and recording it, shows some
random phase shifts in the mixdown, and the hi freqs just sound bad. But, when i record from independent mics, actually
crappy mics (cheap
electret cartridges), in a room with lots of reverb, and lots of other
problems, that problem seems less evident.
Anyway, now i'll try to use that "alsa_in" tool that Adrian mentioned
before; if it has the same functionality, i'll use that instead. It's
strange, there doesn't seem to be very much information about it...
yeah... alsa_in and alsa_out are pretty much ignored.
thanks for pointing out.
theoretically your tool could achieve better sound quality than alsa_in,
because it doesnt need to be realtime. (it could also do perfect phase
alignment)
so the basic question is, if you are still motivated to continue working
on it, when alsa_in provides you with a realtime capable solution, which
seems to be better than your current implementation.
Thank you all for the replies
Alberto
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev(a)lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-dev
--
torben Hohn