[MPlayer-DOCS] [PATCH] update -channels to match observed behavior

[Rich Felker]

On Fri, Sep 02, 2005 at 11:44:20AM -0700, Corey Hickey wrote:
> Rich Felker wrote:
> > On Fri, Sep 02, 2005 at 03:07:34AM -0700, Corey Hickey wrote:
> > 
> >>Maybe a reasonable compromise would be to automatically use -af
> >>channels=2 _unless_ -channels (or -af channels) is specified on the
> >>command line or config file. In that case it's up to the user to make
> >>sure the output channels are correct.
> > 
> > 
> > Just fix the af init code to auto-load af_channels (at the END of the
> > chain) if the output device does not support the number of channels
> > requested. HOWEVER: it's probably wrong anyway, since the first 2
> > channels are usually not L/R.
> 
> They aren't? I was under the impression (from testing AC3 5.1) that the
> usual channel order is:
> 
> 0 - left front
> 1 - right front
> 2 - left rear
> 3 - right rear
> 4 - center front
> 5 - subwoofer

Yes, I think so. However for AAC, it's something stupid like:
0 - left front
1 - center front
2 - right front
...?

Actually this has nothing to do with AAC itself, but the broken FAAD
decoder which interleaves the samples in this idiotic order..

> I haven't been able to find many sample files, though. If other files
> typically have a different channel order I'd like to document it. Dang.
> I just made a 6-channel ogg with oggenc -C 6 and it came out:
> 
> 0 - left rear
> 1 - right rear
> 2 - center front
> 3 - subwoofer
> 4 - left front
> 5 - right front
> 
> Is that correct or is that a bug?

No idea..

> I agree that truncating to two channels isn't a great solution, though,
> even if the channel order were standard. It'll still sound really wrong
> unless at the very least the center channel is mixed with left front and
> right front. The best solution would be for all audio codecs to be able
> to downmix to two channels from whatever numbers of channels that format
> supports. liba52 does, and that probably covers 99% of the files. vorbis
> and pcm don't. For lack of being able to find samples I don't know about
> any others.

It's necessary that the filter layer/player know what the channels
mean, not just how many there are. Unfortunately MPlayer is too broken
to do this. Imagine if for video you just knew that you had "3 color
channel" video, but didn't know if it was RGB, YUV, etc. or even what
order they were in..... :( It would be nonsense. Whoever designed the
>2 channel audio support for MPlayer was not thinking straight at
all..

Rich