[Ffmpeg-devel] Real-time mpegts encoding and streaming

[Brian Brice]

Luca Abeni wrote:
> Hi Brian,
> 
> On Fri, 2007-03-02 at 16:16 -0600, Brian Brice wrote:
> [...]
>>> Now, we only need someone implementing correct audio/video
>>> interleaving ;-)
>> It also needs correct PCR calculations.  HDV cameras use transport
>> stream, and they are *really* strict when it comes to PCR.  Encoding
>> a transport stream from ffmpeg and sending that to cam doesn't work at
>> all.
> Are you really sure about this? I mean: are you sure that the problem is
> the PCR? I do not doubt that a TS produced by ffmpeg is not good for HDV
> cameras, but I suspect the real problem is audio-video interleaving. Do
> you have some of such hardware for performing tests? If yes, can I send
> you a dirty patch for testing?
> 
> I've been told many times about how important the PCR is... But when I
> performed some real experiments on real hardware, I noticed that almost
> all the DVB-S and DVB-T set-top-boxes do not use the PCR at all... They
> just look at the audio PTS, and use it as a sort of reference clock.
> 
> But I never tried HD hardware... Maybe the high bitrate needed by HD
> forces to use the PCR instead of the audio PTS?

I'm sure the video and audio need to be interleaved properly, but any
problems I've had in the past were due to the PCR.  Some of these 
cameras expected PCR info every 25 ms and would mess up if it was too
high.

Standard transport stream isn't compatible with HDV 1080i.  There are
some nice, undocumented private streams that the camera absolutely
depends on.  HDV 720p is much more flexible in allowing only audio,
video, and PCR packets.

I have cameras at work.  If I can get access to them, I can test.


-- 
Brian Brice