On Thu, 2008-11-20 at 08:31 +0100, Hans Verkuil wrote:
> On Wednesday 19 November 2008 11:27:05 Alexander Pipelka wrote:
> > Hans,
> >
> > I reinstalled the old system.
> > It was Ubuntu Feisty with the ivtv driver version 0.10.0
> >
> > Here the dmesg snippet:
> >
> > [ 19.404000] ivtv: ==================== START INIT IVTV
> > ====================
> > [   19.404000] ivtv:  version 0.10.1 (tagged release) loading
> > [ 19.404000] ivtv: Linux version: 2.6.20-15-generic SMP mod_unload
> > 586 [ 19.404000] ivtv: In case of problems please include the debug
> > info between
> > [ 19.404000] ivtv: the START INIT IVTV and END INIT IVTV lines, along
> > with
> > [ 19.404000] ivtv: any module options, when mailing the ivtv-users
> > mailinglist.
> > [   19.404000] ivtv0: Autodetected Yuan MPC622 card (cx23416 based)
> > [ 19.404000] ivtv0: Unreasonably low latency timer, setting to 64
> > (was 32)
> > [   20.876000] ivtv0: loaded v4l-cx2341x-enc.fw firmware (262144
> > bytes) [   21.104000] ivtv0: Encoder revision: 0x02050032
> > [   21.104000] ivtv0: Recommended firmware version is 0x02060039.
> > [ 21.188000] cx25840 0-0044: cx25843-23 found @ 0x88 (ivtv i2c driver
> > #0)
> > [   26.188000] ivtv0: Registered device video0 for encoder MPEG (4
> > MB) [   26.188000] ivtv0: Registered device video32 for encoder YUV
> > (2 MB) [   26.188000] ivtv0: Registered device vbi0 for encoder VBI
> > (1 MB) [ 26.188000] ivtv0: Registered device video24 for encoder PCM
> > audio (1 MB)
> > [   26.568000] ivtv0: Initialized Yuan MPC622, card #0
> > [ 26.568000] ivtv: ==================== END INIT IVTV
> > ====================
> >
> >
> > You may believe it or not this driver produces a correct stream
> > without any B-Frames.
> >
> > To prove my words here are two links to captured streams. One with 2
> > B-Frames one with 0 B-Frames.
> >
> > http://www.bms-austria.com/~pipelka/2bframes.mpeg
> > http://www.bms-austria.com/~pipelka/0bframes.mpeg
> >
> > Open them with e.g. avidemux and check the frame types.
> >
> > The following commands where used to create the streams:
> >
> > v4l2-ctl -i 1
> > v4l2-ctl -s pal
> >
> > v4l2-ctl -c video_b_frames=2
> > cat /dev/video0 > 2bframes.mpeg
> >
> > v4l2-ctl -c video_b_frames=0
> > cat /dev/video0 > 0bframes.mpeg
> 
> I'm going to take a look at this during the weekend. Sorry for having 
> you go through all this trouble, but it really surprises me that it 
> ever worked at all.

Might I suggest that perhaps changes to the cx25840 module may be the
cause, and not the ivtv driver nor firmware, especially since you noted
a dependency with the AV decoder chip used.  Perhaps a change modified
the way SAV and EAV codes get emitted in the stream by the CX2584x
(especially when the V & F  bit change) which induced the change.

Just a guess.  Feel free to ignore. :)

Regards,
Andy




> Regards,
> 
>       Hans
> 
> >
> >
> > Regards,
> >
> > Alex
> >
> > Am Di 18.11.2008 20:57 schrieb Hans Verkuil <[EMAIL PROTECTED]>:
> > > On Tuesday 18 November 2008 18:13:04 Alexander Pipelka wrote:
> > > >  Hans,
> > > >
> > > >  The older version of the driver supported 0 B-Frames (I'm 100%
> > > > sure).
> > > >
> > > >  I need this for very fast frame navigation in a MPEG stream.
> > > >  It worked with the driver included in Ubuntu Edgy (I think it
> > > > was version 0.7.4 of the driver).
> > > >
> > > > Unfortunately I updated that machine to Ubuntu Hardy and the
> > > > problem appeared.
> > > >
> > > >  Well, I could reinstall an other machine with Ubuntu Edgy to
> > > > *verify* that the 0 B-Frame thing works, but I already know this,
> > > > ...
> > > >
> > > >  Please help !
> > >
> > > I am not convinced. The only difference I've found between 0.7.4
> > > and the
> > > latest driver when it comes to B-frames was corrected by my patch.
> > > As far as I can tell 0.7.4 used the same firmware as well.
> > >
> > > I vaguely remember that someone tried it with an old ivtv version
> > > as well in the past and got a similar results. However, I can't
> > > find the specifics so I might be wrong.
> > >
> > > The upshot is that I *really* need to see proof that this ever
> > > worked (and worked with the same firmware as well!) before I go and
> > > spend a lot of time figuring out what changed. The fact that you
> > > could set B-frames to 0 using ivtvctl doesn't mean at all that it
> > > actually worked. Actually the value 0 that was passed to the
> > > firmware in 0.7.4 was not according to the fw docs since they
> > > specific that you have to pass '#bframes + 1' to the firmware, so
> > > '0' should be an illegal value.
> > >
> > > Regards,
> > >
> > >   Hans
> > >
> > > >  Alex
> > > >
> > > >  Hans Verkuil schrieb:
> > > > On Tuesday 18 November 2008 08:26:16 Levente Novák wrote:
> > > >
> > > > Hi,
> > > >
> > > > Is there any advance on the topic of B-frames problem on PVR-150s
> > > > and
> > > > PVR-250s? A few days ago I tested Hans' patch and while my
> > > > PVR-150 MCE does now save a non-zero length stream if th number
> > > > of B-frames is set to 0, this stream has not 0 but 1 B-frames per
> > > > P-frames. Beside this, even if it seems to work OK with 1 or 2
> > > > B-frames, the stream is choppy and unusable with 3 or 4.
> > > >
> > > >
> > > > My patch fixes the zero-length stream issue, but that's all. I
> > > > don't believe it ever gave you a stream with 0 B-frames. I
> > > > suspect that the
> > > > firmware forces a minimum of 1 B-frame.
> > > >
> > > > If you can prove that the older driver actually gave you 0
> > > > B-frames, then I'll take another look. But I believe that the
> > > > conclusion simply
> > > > is that you need a minimum of 1 B-frame.
> > > >
> > > > Regards,
> > > >
> > > >         Hans
> > > >
> > > > _______________________________________________
> > > > ivtv-users mailing list
> > > > [email protected]
> > > > http://ivtvdriver.org/mailman/listinfo/ivtv-users
> 
> 
> 
> _______________________________________________
> ivtv-users mailing list
> [email protected]
> http://ivtvdriver.org/mailman/listinfo/ivtv-users
> 


_______________________________________________
ivtv-users mailing list
[email protected]
http://ivtvdriver.org/mailman/listinfo/ivtv-users

Reply via email to