[Ffmpeg-devel] Re: [Ffmpeg-user] ffmpeg encoding, reallyslowvideo when recording with audio

[Erik Slagter]

On Fri, 2006-02-03 at 15:39 +0100, Luca Abeni wrote:

> I am stress-testing the v4l2 input and I cannot see big synchronization
> problems. I also tried to disconnect the video input, to reproduce the
> drivers' bug you described in a previous mail, but everything worked as
> expected (no ioctl() failed). When I disconnected the video input, the
> driver just provided blu frames at a low rate (about 1 frame per second,
> I think), but v4l2.c could easily cope with this.

The saa7134 driver (and possibly cx88) is much more sensitive to loss of
sync. AFAIK this is purely a driver issue.

> > These have a bad crystal that make audio and
> > video go out of sync :-( Workaround is in the driver, but has to be
> > enabled and tuned by the user.
> Uhmm... Are you talking about the fact that there is a drift between the
> "card time" (I mean, the one that is supposed to be increased by 40ms at
> every video frame) and the "system time" (the one returned by
> gettimeofday())? If yes, then that cannot be avoided in a real world :)
> But ffmpeg can correctly cope with it (by dropping or duplicating video
> frames... About one frame per hour on my system, if I remember well, but
> I am not sure).
> Or am I misunderstanding the problem?

Yep. See other message. The chip tries to make something right that is
already right and in the process making it wrong ;-) (this all only
counts if you have the broken crystal, otherwise it's actually quite
smart)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 2771 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20060203/1adbd5fc/attachment.bin>