Well, admittedly this is a minor issue... indeed now i think i finished
Basically this is a cleaned up version from my previous post including
some simple logic to avoid overshoots by waiting the feedback to take
effect. Fortunately it's easy to do so, since we know the number of
buffers in audio fifo.
I tested this logic against several streams, simulating a sample rate
error at sound card drivers. I was able to consistently recover from
+-1% sample rate error without a single dropped frame or audio buffer
during dvd playback.
For low quality streams and broken ones the limit as the corrector don't
drop frames/audio will be less the 1%, but indeed much better than
So, i'd like to commit this patch to cvs. The new code is as simple as
the current one, the main difference is where gap error is applied. If
someone doesn't agree with this, or could provide any example where it
will not work, please fell free to say "don't do that"! ;)
It's nice to see that there's a good comment in there describing what
the code does!
IMO, some more of that stuff would definitely make a Xine developers
life much easier and avoid the constant reimplementations that seems to
go on in some of the "darker" parts of the code where nobody, except
perhaps the original author, really seems to understand exactly how the
code is supposed to work. :-D
Get latest updates about Open Source Projects, Conferences and News.