Aleksandar Dovnikovic wrote:>
> Frank Klemm made the clip to test short block switching -
> it's made of series of short periodical 'pulses', and this
period
> gets smaller as time passes.
Another simple testcase, not only for blocklength switching but related:
use mostly zeroes, put put some one sample spikes in it, at different
offsets in the MDCT window. For example, 16384 times 0, 1 time 1000,
16384 times zero, 1 time 1000, etc. You will notice the spikes will not
all sound the same after encoding. This is because the psycha model
will not think all of these are equal. Not only vorbis suffers from this,
but most transform based encoders do. Vorbis does reasonably well,
actually.
> I couldn't use mode -b 350 because oggenc always hangs at 98% with
> this clip and the file produced has annoying distorsions...
>
> Obviosly (like Monty said) Vorbis is still learning and needs more
> tweaking in the short-block detection department (and must also
> avoid some patents)...
I actually managed to reliably crash the Vorbis encoder today.
Investigating...
Dagdag,
Segher
--- >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@xiph.org'
containing only the word 'unsubscribe' in the body. No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.