[vorbis-dev] Vorbis bitstream specification...

[Monty]

On Fri, Feb 08, 2002 at 12:38:57PM -0800, Brendan Dowling wrote:
> Hi,
> 
> I'm looking for more documentation on the Vorbis bitstream format. 
> The goal for me is to write an optimized decoder using only integer
> or fixed point math for use on the Phatnoise Car Audio System
> (see http://www.phatnoise.com).  
> 
> I've already found the info on the Ogg framing system and I've already
> written my own thing for parsing through Ogg frames (easy).  

'Tis one of my priorities for the next month or so (unfortunately,
along with a bunch of other stuff).

But.... I can start you off.  You can take the mainline encoder and do
a simple but good translation to fixed point, using the following tips:

1) you want 8.24 precision for the floor vector.  An integerized
version of the floor0 LSP algorithm is on mainline, and floor1 is
already integer only.  For floor1, you just need to replace a lookup
table.

2) the residue vector assembly has to be in 24.8. 

The implication of the above two is that the codebooks can't be
pre-unpacked to a specific fixed point position.  Codebooks are
general and may need to be anywhere from 1.31 to 24.8.  You can either
unpack upon first use or do on-the fly unpacking from the unnormalized
vorbis codebook packed float 32.  The former is easier, the latter
slower but more flexible.  I can guarantee either will work for Vorbis
I, Vorbis II may break a simple version of the former.

3) the MDCT on mainline has a stable dynamic range and is easy to
integerize.  The mainline integer version assumes a 14 bit fraction on
the trig lookups, which frankly is far too little.  This is done only
because C doesn't have a 32x32->64 bit two result register multiply.
If you're on ARM (for example) and have SMULL, use 2.30 for the
lookups, only save the RdHi register and you'll have very good S/N.

Sure, that's glossing over alot, but it turns out that integerization
is mostly busywork.  There's nothing tricky in an integerization; it's
just that no one has taken the time.

(As a note, I was contracted to take the time, and have written and am
currently maintaining a commercial version for iObjects/Fullplay.  For
this reason I can't assist in coding, but I can answer any questions
you have about the mainline code or specs.)

Monty

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.