[vorbis-dev] Impulses

[Monty]

> 
> Do you guys really think window switching is so bad?  It clearly
> works very well and is not just 'mp3' quailty, since it is used
> in AAC which is pretty much the best encoder out there.

AAC, on paper, does not rely only on window switching.  Its 'noise shaping'
(although not done in the time domain) is effectively a nifty time domain hack
to better deal with very wide dynamic range variance across a window.

> The only problem I can see is that the encoding
> is not as efficient - you always need to allocated extra bits for
> short MDCT windows.  But except for extreme cases like castanets.wav,
> the amount of attacks/pulses is usually less than 5%.  Assuming 50%
> more bits for the lossless encoding, a more sophisticated technique
> would save at most 2.5%.

It's not so much a question of efficiency as it is a question of flexibility
and promise. My very hacked preliminary time domain envelope shaping was too
easily befuddled for real use, but with a good number of samples (eg,
castanets), the results were so good that I didn't even *need* short blocks.  I
also understand the failure cases but didn't really have time to chase them
yet.  

So, time domain impulse/envelope encoding is something that at least falls into
the bucket of 'promising'.

> Also, I believe Vorbis is using a 2048 sample MDCT window?  (like AAC,
> but almost twice that of  mp3).   Such a large window results in
> more spreading, making short windows even more important?

Vorbis uses whatever powers of two the encoder feels like.  I've had it set to
2048/2048, 2048/512, 2048/256 and 1024/256 for testing.

> Monty: you've mentioned comparisions between Vorbis and AAC in
> the past - which AAC encoder/decoder were you using?  If you get
> a chance, could you the output from a decoded AAC encoding of 
> castanets.wav?

Sadly, I've only compared Vorbis and AAC on paper.  I hadn't found any AAC
encoders that were substantially better than mp3 yet last time I looked (which
was a little while ago).

MP3, obviously, is easier to compare (and MP3 encoders are still improving as 
well).

Monty

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