[ExtractStream] Sequence number bogus errors (FAQ?)

[Josh Harding]

--- "Makhratchev, Konstantin" <konstantin.makhratchev@l...> wrote:
> netweork. It appears that some audio packets are lost in transition
> at higher rate than video packets (say, 1 in a 100 for audio vs. 1 in
> a 1000 for video)--could be some poor TCP/IP stack implementation.

At that error rate, if it were a network problem, you'd get errors in
every file you tried to rdist, nfs or ftp to the TiVo. We can rule out
the TCP/IP stack as one of the possible problems.

> ExtractStream does not provide a reliable connection so the error
will
> be cumulative and the recording will turn from synced in the
beginning
> into the mess at the end of the show.

ExtractStream isn't providing a connection... it only does local file
access.

It is interesting to note that audio loses time faster than video. 
This could be a small clue. I don't know off hand all the numbers for
the bit rates and such, but here's an idea: if you miss blocks of a
fixed size each from the audio and video streams, time-wise that's a
bigger chunk for audio. It could be 10ms of audio but less than one
frame of video. Over time, that would make the audio stream be ahead
(end sooner) compared to the video (which you'd expect to have glitches
in some frames if not dropped frames).

It's times like this that I wished I knew more about MPEG encoding...
do the audio frames have time code? or atleast sequence numbering? 
Can they be directly correlated to video frames? How does the TiVo get
them to line up consistently?



=====
--The Amigo

__________________________________________________
Terrorist Attacks on U.S. - How can you help?
Donate cash, emergency relief information
http://dailynews.yahoo.com/fc/US/Emergency_Information/