[saxman]

How do OGG files handle bit depths? Like, a WAV file has a byte that tells the software how many bits should be read for each sample, rather it's 8, 16, 24, or whatever. I'm researching OGG, and I can't find anything like that at all. I have looked on Wikipedia and also the official specification -- http://xiph.org/vorb...bis_I_spec.html

Can someone help me understand how software is supposed to know if the PCM conversion should be played back at 16-bit or something else?

[Sik]

Doesn't Vorbis work with floating point numbers?

EDIT: shouldn't you be looking at the Vorbis API reference instead? =/

This post has been edited by Sik: 27 October 2011 - 03:42 AM

[dsrb]

Almost all lossy formats use floating-point internally. There is probably little benefit in going beyond 16-bit, although you may want to dither. A couple of sources from a forum that you may find useful:
* Discussion on decoding and bit-depths: http://www.hydrogena...showtopic=55336
* How to decode to fixed- or floating-point with libvorbis: http://www.hydrogena...showtopic=47135

Quote

Can someone help me understand how software is supposed to know if the PCM conversion should be played back at 16-bit or something else?

If it were a media player worth its salt, it would decode—and perform processing including volume adjustment, DSPs, etc.—in something like 32 bit floating-point, and only then quantise directly before output and to the bit-depth chosen by the user. As to other applications, I don't know; but, again, 16 bit is probably a safe bet in most cases.

This post has been edited by dsrb: 27 October 2011 - 09:32 AM