[Openchrome-users] opechrome tv-output shuffles fields

Christiaan van Dijk dvbmail
Tue Apr 17 10:51:00 PDT 2007


Hi Nico,

I'm familiar with the swapped field problem on the via boards and yes, 
this is very annoying.
The problem seems to come from the TV encoder which has no (known) 
method of indicating which field it is displaying. The video driver 
therefore has to guess if it's displaying an odd or even frame.
When playing MPEG2 with XvMC this can be solved by using the BOB 
deinterlace method. This method doubles the frame rate and somehow the 
TV encoder can pick the right field, I don't have any details on how 
this works but the result looks good. So use deinterlacing to display 
interlaced material on an interlaced screen :-\ 
When using standard video playback also deinterlace, BOB is giving me 
troubles with MythTV so I use kernel for playback of non MPEG2 material 
(myth always uses BOB in case of XvMC).
Frame rates have to match so use the right modeline, can be tricky, mine 
is still shifted a bit too much to the left.

I wanted to see if there are tricks with the TV encoder to get the field 
information but via is not very willing to share datasheets :-(

Well that's my experience with this stuff. I think there have to be ways 
to really solve the issue but I don't have the information to figure it 
out. Would for example be very nice to add a component out to the board 
which the TV encoder seems to have.

Regds,
Christiaan.

nico berndt wrote:
> Hello everybody!
>
> I happily found the TV-Out of my VIA Epia EN12000EG to be one of the 
> rare cases to provide me with an unscaled and very good looking 
> PAL-picture. I was looking for such thing on a sbc for ages, since I 
> depend on proper TV-Outputs for my work.
> Sadly the TV-output is not all the way proper: It shuffles fields. Did 
> anybody notice this? Terry Barnaby wrote about the field order possibly 
> being changed, but it actually is worse.. I suppose you would not really 
> see this on a regular tv set, but when captured again you do.
>
> To explain a little deeper what I mean by shuffling fields:
> Proper video consists of frames consisting of even and odd fields: 
> [o1/e1] [o2/e2] [o3/e3] [o4/4] [o5/e5] [o6/e6]
> For progessive content, such as most DVD movies, both fields within a 
> frame sum up to a progressive frame again and the captured output should 
> again look line described above. But it looks like this [o1/e1] [o2/e3] 
> [o3/e3] [o4/e4] [o5/e6] [o6/e6]. I analyzed thousands of frames and 
> there is no pattern behind it, totally not! I don't quite understand 
> what is going on. I've seen video outputs mess up with the geometry and 
> badly shuffling fields, even starting the next field right in the middle 
> of another field, all torn, so this is not really a surprise. The 
> surprise is that the odd and even fields _always_ come in the right 
> order but from wrong frames! And that really bugs me, because I can not 
> image how this is happening. It looks like the encoder chips were surely 
> capable of doing everything right. One point that leads me to this 
> conclusion ist the stubburn order of the fields. Since it never mixes 
> up, I suppose the modelines to be very accurate. An often seen problem 
> is that the vga chip uses a refresh rate that is not exactly 50.00 Hz 
> (for PAL) and the encoder chip chokes up on trying to capture 50.10 Hz 
> into 50 fields.
>
> If tere is any interest I could privide my testvideo wich consists of 50 
> different frames displaying photos and large numbers from 01 to 25. The 
> position of the numbers changes all the time and there is also a stamp 
> "odd" or "even". "odd" is stamped into odd lines only, and "even" 
> accordingly. The whole thing comes as a dv type avi file, so the video 
> is I-frame only and the fields are encoded seperately. Having this and a 
> capture card one can spot errors immediately just by reading "number" 
> and o/e.
>
> So if there is any interest i'd provide this one and offer my help in 
> thinking and testing. I am no coder, sorry, no...
>
> Best regards and thanks to everyone involved building this driver!
>
> ../nico berndt
>
> _______________________________________________
> openchrome-users mailing list
> openchrome-users at openchrome.org
> http://wiki.openchrome.org/mailman/listinfo/openchrome-users
> Main page:
> http://www.openchrome.org
> Wiki:
> http://wiki.openchrome.org
> User Forum:
> http://wiki.openchrome.org/tikiwiki/tiki-view_forum.php?forumId=1
>
>
>
>
>   




More information about the Openchrome-users mailing list