[vdr] Re: [linux-dvb] Re: vbv delay/buffer overflows on VBR statmux

[Oliver Endriss <o.endriss@gmx.de>]

Klaus Schmidinger wrote:
> Gregor Lawatscheck wrote:
> > At 21:30 20/05/2003, you wrote:
> > > > It simply patches the bit_rate in the sequence header to 15
> > > > MBit. Works fine here.
> > >
> > >I just tried this with a recording of today's "Quarks & Co" and
> > > I'm afraid this patch somehow causes video glitches. There is at
> > > least one place where I have picture distortions with the patch
> > > that are not there without it.
> >
> > What if you set it to 0x00 0x00 or 0xff 0xff? Maybe we need to find
> > a value where the RTSL can't rely on the data passed to it and has
> > to fall back on other data to demux.
>
> I'm afraid I can't do that today, because my VDR is recording...

Will not help, as the patch triggered on a fake sequence header.
I fixed it.

The patch value 0x249f (15 Mbit) is perfect for this stream,
as the patch will not modify the stream at all. ;-)

> The file is at ftp://ftp.cadsoft.de/pub/etc/test.tar (ca. 8MB).
> The recording contains two marks. The first one indicates
> a glitch that was either in the original material or happened
> during recording - it is visible with both patched and unpatched
> driver (just mentioned it to not put you on the wrong track).

Sometimes I also get this kind of glitch. In this case I see something like

Mai 18 13:12:10 orion vdr[3789]: ERROR: not sync'ed to TS packet on device 1
Mai 18 13:12:10 orion vdr[3789]: ERROR: skipped 188 bytes to sync on TS packet on device 1

or 

Mai 18 13:13:17 orion vdr[3789]: ERROR: not sync'ed to TS packet on device 1
Mai 18 13:13:17 orion vdr[3789]: ERROR: skipped 116 bytes to sync on TS packet on device 1
Mai 18 13:13:17 orion vdr[3789]: ERROR: not sync'ed to TS packet on device 1
Mai 18 13:13:17 orion vdr[3789]: ERROR: skipped 72 bytes to sync on TS packet on device 1

in the syslog. Both types skip 188 bytes (one TS packet).

If I disable
  #define DO_REC_AND_PLAY_ON_PRIMARY_DEVICE 1
  #define DO_MULTIPLE_RECORDINGS 1
I never see these glitches.

> The second mark is around the position where the picture is shortly
> distorted. Maybe I didn't mark the exact point, so please just let
> the replay run and be ready to see the glitch around the second mark.

Please try the new patch.

> Also note that if you try moving the marks with the patched driver
> you don't get clean still frames, while with the unpatched driver
> this works.

Yes, sometimes this happens, but there is no difference whether I use 
the patch or not.

Oliver


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