[vorbis] 16 KHz clip-off?

Gian-Carlo Pascutto gcp at sjeng.org
Tue Aug 14 05:43:01 PDT 2001



On Tue, 14 Aug 2001, Maik Merten wrote:

> Hello,
>
> congratulations to the Ogg Vorbis team - RC2 sounds good.
> But... RC2 in 128 kbps mode seems to clip off all frequencys
> beyond 16 KHz. On the tracks I tested Beta 4 gave response
> even beyond 18 KHz.

It is better to cut off at 16kHz than to try to
go higher and only introduce artifacts anyway in these
modes.

Even if you can hear above 16kHz, that does _NOT_ mean
you will be able to detect a cutoff at 16kHz. In fact,
this is extremely hard.

It is much better to encode what we can hear good rather
than to encode what we can't hear and artifact what we can.

MP3 was never designed to handle more than 16kHz either.
Trying to go beind that usually only worsens the sound
unless real high (>=256kbps) bitrates are used.

Vorbis will encode fine above 16kHz in 160kbps and higher
modes, but 128kbps is simply not enough (for now).

Judging a codec by its bandwidth (like with frequency sweeps and
spectral plots) is like judging wine by its colour.

> Some testings on a randomly chosen track:
> (other tracks gave similar results)
>
> Artist: Judas Priest
> Album: Jugulator
> Title: Bullet Train
>
> Beta4: 127 kbps, ~ 18 KHz (!)
>   RC2: 132 kbps (!), ~ 16 KHz
>
> This seems even more strange to me as RC2 benefits from channel
> coupling...

Have you done a listening test on both files to see which _sounds_
better?

(Granted, RC2 isn't tuned that well yet so it may be worse.)

>
> BTW: As 128 kbps mode produces average bitrates higher than 130 kbps
> (please keep in mind that I am testing with Heavy Metal music, so
> results for Britney Spears and Back Street Boys may vary ;-) )I would
> be happy if 112 kbps mode is back in RC 3...

I think this is normal. Metal tends to bloat the bitrates of VBR
perceptual codecs.


-- 
GCP

--- >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