[ogg-dev] OggPCM : Need more justification for chunked data

[Rene Herman]

Silvia.Pfeiffer at csiro.au wrote:

> we have discussed the issue of different formats per channel, e.g.
> different sampling rates. It was not clear whether with PCM sampling
> of devices this is actually a common (or even used) case. Do you know
> how multi-channel sound is sampled? Is it created with different
> widths/rates on different channels?

I must say I'm not at all aware of how often it is done, but for example 
the DVD-A specification allows for independent sampling rate and bit 
depth for the 6 channels. It allows for 24, 20 and 16-bit widths and 
192000,96000,48000 and 176400/88200/44100 sampling rates for the front 
and surround channels (and adds 24000 and 22050 for the LFE channel) and 
allows those to be freely combined -- the rates only within one family; 
N * 24000 or N * 22050.

Now, whether or not there would be many DVD-A's using multiple formats 
for the channels when/if the thing catches on as a medium is for anyone 
to say but it does make sense to me personally. Say, 24/96 for the front 
channels, 16/48 for the surround channels and, maybe, 16/24 for the LFE 
channel. _If_ it catches on it will probably become the main source of 
multiple channel PCM audio in the future (SACD is DSD, not PCM, which is 
probably a better idea anyway but the market decides -- possibly on 
neither).

> If it is not a common case, we can leave the solution to different
> streams and keep it outside OggPCM. If it's common and the data comes
> interleaved in one stream, we should discuss including it into
> OggPCM.

But having said all that, if it would be a pain to support, and multiple 
logical streams would work just as well, that might be good enough. I 
don't have much insight into what the added complexity would amount to. 
Perhaps if the rates are limited to integer multiples that it wouldn't 
be too bad?

Rene.