[Vorbis] multi channel Ogg Vorbis file corruption

[Ulrich Windl]

On 9 Jul 2008 at 8:54, Michael Smith wrote:

> The libvorbis encoder only treats stereo specially. Any other number
> of channels - 1, 3, 6, or 16 - are encoded simply as that number of
> independent channels.

Hi!

I have no special idea, but I can imagine that decoding 16 channels need 
significant CPU power (plus I/O bandwidth). Maybe there's a performance limit?

Regards,
Ulrich


> 
> So, corruption specific to the channel count is fairly unlikely
> (though of course not entirely impossible). It's also not something
> that sounds at all familiar.
> 
> Since you say you've tested this on other platforms, a compiler bug
> seems unlikely.
> 
> Can you make one of these corrupt files available for people to take a look at?
> 
> Mike
> 
> 
> On Wed, Jul 9, 2008 at 7:48 AM, Robert Bailey
> <rob at the-treehouse-studio.com> wrote:
> > Has anyone else created Vorbis files with more than 6 channels?  I'm a
> > game developer trying to use multichannel Ogg Vorbis files for a PS3
> > and XBOX 360 title.  Using the latest release libs and tools, I've
> > created a Mac application that weaves up to 8 stereo AIFF files into
> > one 16 channel Ogg file.  What I'm finding is that if anything more
> > than 3 stereo files (corresponding to a 5.1 audio file) are encoded,
> > then the encoded output becomes unstable.  I hear vocoder like
> > artifacts and garbled output on some of the stereo pairs.  It should
> > be noted that I'm using a linear permutation matrix, so that the
> > output file maintains the original stereo pairs, ie 16 channels in ->
> > 16 channels out.
> > I've built Universal Binary versions of the latest Ogg and Vorbis libs
> > available for download, and meticulously tested my sample_read
> > callback, and everything is doing the right thing.  I've even built
> > raw and AIFF versions of a 16 track file, verified the integrity of
> > the data,  and run them through the readily available command line
> > encoders for PC and Unix.  The output corruption is identical to what
> > I experience on my own encoder.   The Vorbis encoder code itself is
> > extremely dense, and I suspect that any fix will be the domain of the
> > authors
> > So has anyone else had this experience?  As Ogg Vorbis is open source,
> > has anyone put up a patch for this?  Its incredibly frustrating, and
> > seems to be content dependent.  In some file groupings, it performs
> > brilliantly (I love the sound of Vorbis much more than MP3) for 3
> > minutes, and then goes to crap.  In others, the corruption is evident
> > much earlier.   It always appears at the same place for each file
> > group, no matter what the encoding settings.   Any and all advice is
> > greatly appreciated.
> >
> >
> > Robert Bailey
> > rob at the-treehouse-studio.com
> >
> >
> >
> > _______________________________________________
> > Vorbis mailing list
> > Vorbis at xiph.org
> > http://lists.xiph.org/mailman/listinfo/vorbis
> >
> _______________________________________________
> Vorbis mailing list
> Vorbis at xiph.org
> http://lists.xiph.org/mailman/listinfo/vorbis