On Sat, May 21, 2005 at 11:45:18PM +0100, nyk wrote: > On Sat, May 21, 2005 at 11:32:38PM +0100, Dave Chapman wrote: > > Nico Sabbi wrote: > > > Dave Chapman wrote: > > > > [Ch4 /DEC-2000T/AMD64 weirdness] > > > > The audio stream seems perfect, but the PES headers in the video stream > > are very, err.., minimal: > > > > 00 00 01 e0 03 fb 80 00 00 > > > > This indicates a PES_packet_length of 0x03fb (it's normally zero for > > video packets) and no PTS or DTS values. When I wrote the driver I had no previous experience of mpeg packets of any type, so I just stuck as much information into the packets as the PVA provided. So, yes, they probably look a little funky. > > Is the problem in the part of the DEC2000T driver that recreates the > > vidoe PES headers? (I'm assuming that the DEC2000T doesn't give access > > to the original transport stream). > > According to the driver, the data is sent along the bus as PVA and > translated on arrival, but unfortunately the PVA stream hasn't given > much better results. > > Perhaps this is a usb problem? I don't think so. PVA is quite odd, in that you get ES video packets, and PES audio packets. The PES packets the driver just uses verbatim, but for video packets it has to try to create a PES packet for them. There is a flag in the PVA packet header that says whether the PTS immediately follows the PVA header or not. If it does, the driver uses it. In PVA mode, the driver spits out the PVA verbatim, so if that is just as bad as the TS, then the problem lies elsewhere. Something that might be worth looking at is the PCR pid. This seems to be ignored by scan (unless in vdr mode) and tzap, and I imagine for a lot of other apps channel configs, since it appears to not be needed. However, if it's not sent with the video and audio pids to the dec, the output from the scart will be out of sync A/V-wise. As a hack, the driver sends the video pid for the PCR, since for most channels this will work - but Channel 4 is one of the ones that has a different PCR pid. I never checked the resulting streams from the driver, since they always use to play in sync... -- Alex
