[AG-DEV] RAT corrupt sound

[Christoph Willing]

This is very interesting since we're all interested in better quality  
audio. However rat development is done elsewhere - could I suggest you  
join the list at:
     http://oakham.cs.ucl.ac.uk/mailman/listinfo/sumover-tech


chris


On 10/11/2010, at 8:37 PM, Ушаков Андрей wrote:

> Hello all!
>
> I have a problem with RAT on Fedora 11, 12. Received sound has a lot
> of jamming what sounds like ugly electronic noise. I try to solve this
> problem by myself since may of 2010, and I found some solution but
> still not found problem source and cannot fix it but lower sound
> corruption. I interesting only 48kHz Mono mode, btw. How can I fix it?
> First, I checkout latest version of RAT from repository and build a
> debug version of common lib and rat itself. And first thing what I
> find is horrible spamming of message:
>
> RTP packet queue overflow.
>
> So I just enlarge a little this queue. Sound become slightly better.
> Next my target was find why messages
>
> Insert X samples.
> Playout buffer shift back X samples.
>
> occurs and sound has jamming and flipping. I find out what it cause
> because there is some unsynchronization detecte in function
>
> int source_check_buffering(source *src) in file source.c
>
> I have no idea why it does. And I don't know how to debug it. Hardware
> problem? Software? ALSA? No idea. Then I go to the ALSA settings and
> starts an experiments to configure default ALSA device. And I find a
> configuration what almost do a sound well!! So there it is:
>
> 1) I add .asoundrc file in my home directory:
> defaults.ctl.!card Intel
> defaults.pcm.!card Intel
>
> defaults.pcm.dmix.!rate 48000
> defaults.pcm.dmix.!format S16_LE
>
> 2) And, important, I manually setup buffers sizes for default ALSA
> device directly in dsnoop.conf and dmix.conf. And channel count and
> binding also. There it is:
>
> I add to dsnoop.conf:
>
> channels 2
> period_time 10000
> periods 16
> bindings.0 0
> (very  short fragments)
>
> And to dmix.conf:
>
> channels 1
> period_time 300000
> periods 16
> (laaaarge fragments)
>
> And VOILA! sound become clean and very good ... almost. Once-twice in
> 5 min interval randomly occurs same flips and messages:
>
> Insert X samples.
> Playout buffer shift back X samples.
>
> That is my story. If some one know this problem and know how to debug
> and fix it, lets discuss! Also, I can send my hardware and software
> configuration if needed.
>
> And at last, I wrote some CLI interface to RAT what useful when RAT
> work in -noui mode and I can share it if some one interested about it
> :) Its not perfect may be because I have no one friend who understand
> the mbus and RAT :) But it works.

Christoph Willing                       +61 7 3365 8316
QCIF Access Grid Manager
University of Queensland