[FFmpeg-devel] [PATCH] jack: fix timefilter feedback factor computation.

[Nicolas George]

Le septidi 7 nivôse, an CCXX, Nicolas George a écrit :
> This was not tested beyond build on Debian stable, but the PDF file in the
> comments of timefilter.c say b=\sqrt{2}\omega rathar than b=\sqrt{2\omega},
> and experimentally (on software-generated noised signal), it makes a huge
> difference on the efficiency of the filter.

Ok, I got most of the maths behind the \sqrt{2}\omega; I needed to look at
it as a differential equation, not as a sequence. I can confirm that it is
definitely not \sqrt{2\omega}.

I will include my findings in the comment at the top of timefilter.c I
submitted a few days ago and push it; after all, a comment can not break
anything.

As for the change itself, someone actually testing it would be nice. The
idea would be something like that:

./ffprobe_g -f jack jack | grep pts_time | head -n 1000 | sed 's/.*=//' > pts1
<apply patch, build>
./ffprobe_g -f jack jack | grep pts_time | head -n 1000 | sed 's/.*=//' > pts2
gnuplot <<<'plot "pts1" using 0:1, "pts2" using 0:1'

Regards,

-- 
  Nicolas George
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20111228/631c9e8c/attachment.asc>