[mythtv] Wrong time in Myth with PVR350 PAL output

[Torsten Schenkel]

Am Mo, den 22.12.2003 schrieb Torsten Schenkel um 20:02:
> Hallo,
> 
> there is some oddity with the pvr350's tv-out feature of mythtv.
> 
> When watching everything is fine, then I want to skip a commercial and
> the video jumps back for some amount with the first skip forward. This
> is with old and new versions of ivtv and mythtv, not a new problem.
> Isaac says it's related to the pvr350's tv-out and yes, I didn't notice
> it when using the vga.
> 
> ATM I have David Engel's patch running, and while this takes some pvr
> problems outof the picture, this one still remains. David also said he
> didn't see this.
> 
> So I understook some tests, and here are the results:
> 
> I watch about 50 minutes of recorded video (according to the osd), then
> I escape out of the playback, saving the bookmark. When I reenter the
> same show, it jumps to the bookmark (50:00 on OSD) but is not a the same
> scene I was at. I then skip forward (it doesn't jump back immedeately
> after a search) to the scene I left and it's 60mins on OSD.
> 
> So the whole thing seems to be a problem with the pvr giving back wrong
> information about frames processed or mythtv doing wrong maths on the
> information it gets from the pvr.
> 
> Summary: The OSD shows correct time when searching and myth doesn't skip
> backwards directly after a search. The displayed time does lag behind
> actual time by 1/6th. When doing a search it therefore skips back to the
> location myth thinks it currently is.
> 
> The 1/6th is interesting, this hints towards a relation to PAL/NTSC
> discrepancy. When myth (or the pvr) takes 30fps as standard, it will
> count 1 second for every 30 frames, but will have displayed 1.2 seconds
> of video. This amounts to the 50mins on the OSD while 60mins have been
> watched. When I skip forward then it will search in the file to
> 50:30mins and starts the video there.
> 
> Might my assumptions be correct? Is this a PAL-only problem? Can it be
> fixed?

Ok, I did try a bit more. I'm not a C++ coder and know near to nothing
about the internals of the ivtv code, so I cannot offer any suggestion
to fix this.

Well, what I dit is:

1. Check the video_frame_rate setting while running, it is indeed 25fps
which is correct.

2. Changed all math that's done to use 25.0 instead of video_frame_rate.
To no avail. (I know step one should have proven this, but since it was
a parallel approach, I didn't mind to give it a try)

3. Changed mythcommflag to have fps=25.0 and run it, still the same
problem with the wrong time in the osd. It did find 5 breaks. On the
PVR's TV-out the breaks were wrong (you guess it 1/6th early).

4. Nuvexport the file to a clean install of myth-cvs on my workstation
(XV playback via Radeon 9500Pro -> CRT).

	Results:
	
	No time skip, this I assumed, since this was the machine I
	tested myth on prior to building the myth-box with the 350.

	The breaks found by mythcommflag were at the correct positions
	now.


This tells me:

The problem is with pvr350 out, not with encoding or the database. It's
cause has therefore to be found somewere in NuppelVideoPlayer,
ivtvdecoder, nuppeldecoder or videoout_ivtv.cpp (which only provides a
class for NuppelVideoPlayer, IIC)

Any ideas were to look for it? My first guess was video_frame_rate, but
this was a red herring, I assume.

Torsten
-- 
Config files for PVR350 TV-Out:
http://www-isl.mach.uni-karlsruhe.de/~hi93/ivtv-pvr-350-conf.tgz
Walkthrough: MythTV on Epia with PVR350 using Debian:
http://www-isl.mach.uni-karlsruhe.de/~hi93/myth/mythtv_debian_epia_pvr350_walkthrough