[vorbis] APPLAUD.WAV problems

[Monty]

> It isn't at all a bad idea. Is it possible to add a switch to oggenc that
> eliminates the lossy stereo modes?

It will be in rc4.

> I agree, but many songs contain applauds and a premium encoder such as OGG
> shouldn't produce such CLEARLY EVIDENT artifacts.

At mid-to-low bitrate, it's not fair to blame *any* encoder.  Some do
better than others, but if audiophile quality is your concern, -q4 in
Ogg or 128 in LAME is just to low to take seriously (I think we both
agree here).

> You are right. But if a person hears such an evident artefact on a common
> sound like an applaud, he would think again before re-using the encoder
> below -q 5.

Artifact != difference.  I hear 'serious problems' in the LAME .mp3
that aren't in the Ogg.  I'll take noise in the center of the stereo
image over flanging any day, although neither, in the ideal world, is
acceptible.

However... that's not to say that there's a loooong road of possible
stereo analysis improvements possible in the future, nor am I saying
that there's no room for improvement in applaud.wav that won't also
benefit other samples.

> I'm sorry to have created confusion. What I actually ment was that GT2 -b
> 128 compresses APPLAUD.WAV better than RC3 oggenc -q 4.9 at the same
> bitrate.

Yes, lossless coupling is a big help to this sample.  The encoder does
not have any analysis code currently to look at noise resolution
taking into account an out-of-phase stereo field.  Right now it
believes that moving that noise energy to the center is acceptible
(which it obviously isn't), and it's performing correctly as per the
current algorithm.  Analysis needs to be strengthened to take this
into account.

LAME maintains the stereo image because it isn't capable of
automatically moving noise localization around to save bits.  It
maintains the stereo field by accident ;-)

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-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.