Subject: Re: [ecasound] -erc effect causes delay
From: Kai Vehmanen (k_AT_eca.cx)
Date: Fri May 30 2003 - 04:47:39 EEST
On Thu, 29 May 2003, Jeremy Hall wrote:
> ecasound -f:f32_le,1,48000 -i Track_3-1.wav -erc:1,2 -f:s32_le,2,48000 -o
> alsa,out -a:jhall -i null -o null
[...]
> it didn't matter whether -o was null or rtnull it still got XRUNs.
Xruns from ALSA? What kind of messages? Nuoll objects really can't produce
xruns, but of course they might slow down processing and cause increased
xrun risk for ALSA objects...
> It should be pointed out the delay was not a simple delay, rather it was a
> sample-synched copy on both channels, followed by delays that back off in
> the right channel. It made for a nice phasing effect at small buffer
> sizes.
[...]
> You might be able to reproduce by splitting a file into separate files,
> one for each channel, then trying to rtcombine them with the example I
Ok, this turned out to be a real bug, and a very nasty one (audio is
leaked from mixdown round n-1 to unused source channels in mixdown round
n). This explains why the result varies depending on buffersize setting.
I've just put the fix to CVS. Can you verify it solves the problem?
In any case, as this is a bug in the core audio mixing engine, we are
talking about a serious bug. I'll probably have to release 2.3.0 because
of this, and hopefully soon.
Jeremy, thanks once again for the bug report!
-- http://www.eca.cx Audio software for Linux!
This archive was generated by hypermail 2b28 : Fri May 30 2003 - 04:48:32 EEST