[vorbis] c't listening test: Ogg problem at 128kbps

[Sebastian Gesemann]

Hi there !

On Tue, 3 Sep 2002, Giuliano Pochini wrote:

> > At 128kbps, however, most of the frequency response problems were gone
> > -- except this time, I picked Ogg as the worst, based on an overly
> > exacerbated treble. Others were still noisy and lacked impulse detail,
> > but Ogg/Vorbis was the one I decided first was definitely not close to
> > the original. Listening to it again, I can confirm the difference I
> > heard, although maybe I wouldn't put it in last place this time.
>
> It's caused by the lossless coupling at hi frqs and the effect is

lossless-coupling will be used for -q6 .. -q10 AFAIK.
you probably meant the point-stereo-thingie.

i think it's a quantizing problem.
when quantizing a set of small values like the mdct-coeffs for high
frequencies, the energy difference (between unquantized and quantized
coeffs) may be much graeter than by quantizing large values.

But the point-stereo mechanism seems to be a problem for me, too.
the angle-value of the square-polar-mapped vector will always be set to
zero which leads to a 3dB gain in worst case (worst case = phase
difference of 90 degrees)

To compensate these tings, the encoder should "carefully" encode the floor
and "carefully" quantize the coeffs according to the resulting
energy-level.

i just read the vorbis-spec, but i'm not familar with the internals of the
vorbis-codec, wether/how it takes care of...

<p>bye,
Sebastian

--- Crosspoint v3.0
 * Origin: Point 15 der DCG Main BBS (2:2432/207.15)

--- >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.