[vorbis-dev] the reason for the gain of high frequencies

[Sebastian Gesemann]

Hi, there !

I'm pretty sure that the following text (and attached picture)
explains the reason for the gain of high frequencies.

when quantizing a vector scalar by scalar by simply rounding
each scalar to the nearest level, the quantization-error-vector
and the original signal-vector can be assumed to be orthogonal
(average case)
This is a problem when we want to preserve the energy level
while using a low SNR, because a low SNR will lead to a large
error-vector which increases energy because of the
orthogonality property.

    |signal+error|
==> -------------- ~= sqrt(1+exp(-SNR*c+d) > 1
       |signal|

(again, this is the average case. 'c' should be around 8.6 and
'd' around 0.0 - i don't know for sure)

in the picture (see attachment) we can see that the gain is
correlated with the SNR within a frequency region just the way
i explained above...
(except for the freq-cutoff, of course ;) )

the current encoder only seems to care about the SNR
and not about the resulting energy level as well.

any suggestions ?
(Monty?)

<p>bye,
Sebastian

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

-------------- next part --------------
A non-text attachment was scrubbed...
Name: vtest.png
Type: image/png
Size: 20249 bytes
Desc: vtest.png
Url : http://lists.xiph.org/pipermail/vorbis-dev/attachments/20020905/2473df82/vtest.png