[vorbis] 16 KHz clip-off?

Monty xiphmont at xiph.org
Wed Aug 15 13:52:44 PDT 2001



> > In Beta4 it was enough. (At least according to the material
> > I encoded. I could not hear artifacts - the high-freq-area
> > sounded very nice.)
> 
> There were problems in the other end of the spectrum IIRC.

I believe that what people are seeing as different about rc2 is tjust
that... the noise metrics are new, and thus behavior is different.  If
90 of 100 test tracks are considerably better in rc2 (and 10 are
worse), we're going to get complaints about the 10 ;-)

That may or may not be the case here.

> That doesn't have to have to do with the 16kHz cutoff
> but could be because of general encoder problems.
> (which I do believe RC2 has)

It turns out old egcs (which RedHat clung to until 6.1+ and which we
used to make RPMs) miscompiles floor1.c badly.  The result is that the
floor fitting is consistently undershooting the real floor,
occasionally by up to 50dB in places, and this is overflowing all the
codebook ranges in the bad segment.  The artifact sounds like a
flutter or serious harshness in tones.

We're working on a solution (workaround to the compiler bug).

> Quite a bit of people seem to be of the opinion that beta4 had
> better quality, or at least that RC2 is not all that better.

rc2 was shooting for greater consistency.  I find it hard to believe
only I hear the differences as 'improvement'...  Run b4 on
bassrumble.. not only is there bassrumble, the highs are utterly
squashed.  rc2 is bright, clear and consistent at 128.

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.



More information about the Vorbis mailing list