[linux-dvb] Re: FF cards alter PTS'?

[Ralph Metzler <rjkm@metzlerbros.de>]

Johannes Stezenbach writes:
 > Ralph Metzler wrote:
 > > Johannes Stezenbach writes:
 > > 
 > >  > The av7110 destroys the adaptation fields if you record and watch
 > >  > at the same time (record with DMX_PES_OTHER to prevent this). However,
 > >  > the DMX_GET_STC should give you correct values. Also, AFAIK the PES
 > >  > header is not altered by the av7110 or firmware, so the PTS should
 > > 
 > > Video PES which are being decoded are taken apart. ES and PTS (lower
 > > 32 Bit only) go into separate ring buffers.
 > > 
 > >  > be correct. Or maybe the 33rd bit of the STC/PTS gets lost?
 > > 
 > > Yes, for decoded video always.
 > 
 > I thought the TS packets are grabbed from the TPP before they
 > reach the decoder? So we should get unaltered PES?

No, AFAIK, you cannot get all packets of the video PID before they reach
the decoder. The TPP does all the writing into the ES ring buffer by
itself. The only exception seem to be packets with PUSI bit set. They are
parsed by the firmware ROM. It also handles the PTS buffer. Here one
can also find the infamous PTS length counting bug which (indirectly) 
caused the "jerking in playback" bug.
But this is all from disassembling the firmware ROM and lots of 
guessing. There is no documentation on this.

Maybe there is a better way to do this. One could do more rewriting in
the EOP FIQ (as I did to get the time shifting working). One
could always handle live viewing like playback, writing to the
decode buffers in software. This would give you the complete video
packets. It would probably also get rid of TPP lockups which seem to
be caused by bad video streams from weak channels or during tuning.
But is it really worth putting more work into this old hardware?


Ralph


-- 
Info:
To unsubscribe send a mail to ecartis@linuxtv.org with "unsubscribe linux-dvb" as subject.