[mythtv] 8902 HD transcode => MPEG4 --> a/v out of sync

[Geoffrey Hausheer]

On 2/8/06, Jesse Crews wrote:
> Hi,
>
> I've been having an issue with transcoding anything recorded from a
> particular ATSC transponder (NBC affiiliate, OTA). The output always
> shows the video somewhere around .5 seconds earlier than the audio.
> Maybe even an entire second.
I have seen this on our local PBS-HD affiliate occasionally, though in
my case it is actuallly something withe the stream from the card, as
even mplayer plays the stream off, and I've looked at the PTS values,
and the a/v really is off in the original.  Are you sure that it is
the transcoder that is doing it?  Have you tried playing it with
mplayer?

> I did run mythtranscode with mpeg2fix process messages enabled, and
> nothing came out.
Are you trying mpeg2->mpeg4 or mpeg2->mpeg2 (mpeg2fix can only handle
the latter) or was this just an experiment?
What do you mean by 'nothing came out'?  You'd need to tell me what
the failure is, and perhaps run with '-v all' (this will be huge, so
zip it up and send it to me off-list)

> When I play the file back via mythfrontend, it also throws the guess
> message, but the A/V sync is correct.
Odd...can mplayer handle it correctly?
> What could be different in the
> way that NVP calls are being utiized by mythfrontend vs. mpeg2fix in
> mythtranscode?
Well, they really do decode completely differently, so that isn't an
easy question to answer.

> There are some comments in there about how some
> operations being done "might not be a good idea."
There are a lot of corner cases that are handled in mpeg2fix, and I
need to use some heuristics to make them work right.  They may not
work in every situation though.

> Should I add some messages to print aPTS, vPTS, PCR (really, whatever
> the PCR is equivalent to in a TS), and try to see where drift is being
> introduced?
using '-v all' with mythtranscode will print out enough PTS info to be
useful (we don't pay attention to the SCR in mythtranscode (though
maybe we should) and there isn't currently any code to print it out.

>
> There does appear to be some logic to figure out if the above values
> are wrong, and I recieved no such messages.
>
It depends on how you run mythtranscode as to whether those messages
will print.  by default, only 'MPF_IMPORTANT (equivalent to
VB_IMPORTANT) will print out).  We do handle all sorts of corruption
cases, and I can probably handle yours too with a bit of patience and
some debugging.

.Geoff