[vorbis-dev] Vorbis

[Monty]

> I've been reading through the source, and I can't understand where
> compression is occuring!

It isn't :-)  This is an unfinished codebase and the in-progress back end
coding is not finished or committed.

> It appears that it should be ouputing all of
> the LSP and residule data

It is.  The entropy at that point is about 1.1-1.2 bits/sample, but the full
dynamic range of the output residuals are about 6 bits.  Vorbis is not yet
'compressing' it, so it's outputting the full 6 bits/sample (even though the
majority of the bits wasted).

> The output sounds okay, but has signifantly worse S/N then mp3 at 320.

Yes, the given output should be compared against 112-128 kbps mp3.  One the
backend coder is there, that's the expected output rate.

Note that at the moment, the encoder psychoacoustics are meant to be usable,
but the code is still an in-progress test bed for the format.  For that reason,
the analysis stage is still very naieve.  The first cut release will be
comparable to mp3 due to cleverness in the format, not a brilliant encoder.
We'll be getting the current S/N quality below 112kbps as time goes on.

Monty

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/