[Tremor] Tremor Result vs Vorbis Floatpoint decoder results

[xiphmont at xiph.org]

> I am not sure whether input test vectors, which i used, are correct one. I
> am not also sure about the behaviour of tremor decoder against Vorbis float
> decoder.
> Please suggest if the results are acceptable.
>
> The test vectors you're using are good.

The files with small PCM differences (one or two linear steps peak) are to
be expected.  Vorbis does not spec bit-exact decode, and Tremor is using
some minor approximations.  The differences will be well below an audible
threshold.

The larger differences from other files are likely a problem, depending on
how they're being measured.  These files all use the older pre-1.0 floor 0,
so this suggests an error in floor0 decode.  However, it's difficult to tell
only by looking at average/RMS differences.

Have you listened to these files, or plotted them as a spectrogram?  These
tests are likely to be more useful after a difference test has pointed out a
possible problem.

Monty
Xiph.Org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.xiph.org/pipermail/tremor/attachments/20110202/2f297530/attachment.htm