[vdr] Re: VDR files to vob

[Andreas Schultz <aschultz@cs.uni-magdeburg.de>]

On Tuesday 16 July 2002 23:37, Klaus Schmidinger wrote:
> Andreas Schultz wrote:
> > Hi,
> >
> > On Tuesday 16 July 2002 20:04, Rene Bartsch wrote:
> > > I agree to SCR-implementation as I realized MPlayer plays the VDR-files
> > > faster than recorded by vdr which results in EOF after some time of
> > > timeshifting.
> >
> > I looked a bit more into the remuxing and i will not spend any more time
> > on it. The remuxer will have to be rewritten from ground up to get
> > sensible results
>
> What's wrong with it? I thought it would just take an extra data
> record to satisfy dumb players?

The players YES, but not the VOB format. I think the ultimate goal should be 
to be able to record the resulting file without further processing to a DVD 
(at least that would be my goal). For that the PS packets have to be 2048 
bytes, which in turn means that the contained PES packets can be max. 2034 
bytes long (including the PES header).
Now, the TS stream contains an MPEG stream broken down into PES packets of 
2048 bytes size (including the PES header). That are excatly 14 bytes too 
much. The current remuxer does not change that. It can break down large PES 
packets into smaller ones, but it will _NOT_ recombine multiple PES packets. 
Therefore adding the PS header and resizing the contained PES packets will 
result in 2 PS packets, one with 2048 bytes, the other with 28 bytes. Also, 
the audio streams usualy have much smaller packets. The resulting PS stream 
will therefore be filled with a lot of almost empty packets, increasing the 
bitstream rate by at least 1/3 if not more.

The only way to avoid that overhead would be to extract the MPEG stream 
contained in the PES packets, and rebuild the PS stream from scratch. In a 
VOB PS packet an I-frame will always start with an new PS packet. It also 
appears as if B and P frames are never packed into PS packets that already 
contain the tail of an I-frame However, B and P frames can apparently be 
packed into the same PS packet (maybe P frames are not allowed to follwo B 
frames in the same packet - just a guess). Also video and audio packets are 
never packed into the same PS packet.

The current remuxer will IMHO have to be changed significantly to fullfill 
those requirements.

That is the situation as i see it at the moment. Maybe i'm wrong and there is 
an easy way, you certainly have a better insight in the remuxer.

Have fun

   Andreas
-- 
Andreas Schultz <aschultz@cs.uni-magdeburg.de>
Student of computer science

"In accordance with plans for Linux OS world domination
      infiltration of governments is vital (:-))."