[vlc-devel] Re: VLC skips on "reduced" MPEG2-TS streams

[Samuel Rivas]

Christophe Massiot wrote:
> Indeed the delay between PCR and PTS is too low, and there is not enough
> time from the moment when the packet enters the decoder until the moment
> we're supposed to display it. Theoritically the minimum is 120 ms (one I
> + 2 B's), to which we add 300 ms, so even at the minimum we should have
> time to decode the frames. I suppose the PTS's are in the past then.
> 
> AFAIK there is no specification for this value, at least not in MPEG.
> Conventional broadcaster use a value of 500 ms between the PCR and the
> next PTS.

  MPEG defines the maximum length of the buffers needed to decode an
input transport stream. It is responsibility of the muxers to create a
stream that does not overflow, nor underflow, those buffers. Thus, the
delay should be enough to have the whole pictures and audio frames in
the buffer at the time they should be decoded and present.

  However, the theoretical decoder that MPEG defines assumes synchronous
transmissions (e.g. satellite). For IP networks transport stream is far
from being the ideal streaming format, requiring larger buffers in the
decoder size than the specified in the standard.

  Regards
-- 
	Samuel

-- 
This is the vlc-devel mailing-list, see http://www.videolan.org/vlc/
To unsubscribe, please read http://developers.videolan.org/lists.html