Intel GMA 900 (915G) - interlaced modes?

Will . nodenet at
Mon Jan 8 14:57:48 PST 2007

>I just want to move from old EPIA-M which has problems with TV field
>parity and can't do VGA->SCART RGB due to lack of interlaced modes.
>I'm thinking about buying something cheap like MSI Hetis with GMA900
>but it would have to do interlaced modes (I assume with 2x pixel
>clock, something like PAL 1440x576i with hardware scaling from
>720x576 YUV etc).

Hi, I have an EPIA (think its an SP1300) and an AOpen i915GMmHFS, I've been 
trying to get high quality tv out for a while thought I'd share my 
experiences. I've used the DVE PAL disc for setup, and an assortment of 
movies as source material. I run them through the computer and usually again 
through another DVD player for comparison both connected to the same plasma 
screen. I avoid doing any testing on my computer LCD display as I've found 
the results to be missleading (even using the same resolution and refresh 
rates - the nature of the displays is different anyway).

>I know there are specialized TV (S-VIDEO) encoders but SCART RGB
>would be much better in terms of quality (tested with Radeon+RGB
>vs. EPIA with "noscale") and stability.

S-Video on the EPIA and from another PC with a radeon haven't given me great 
results but I know the S-Video input on my tv isnt particularly good anyway.

I've been running on the VGA ouput at 720x576 @49.986.. (the closest I could 
get to 50Hz without the TV encoder). Full frame straight into the VGA input 
on the plasma. Results are generally very good, although it took alot of 
tweaking and messing to get there. I encountered some problems sending video 
at 50frames per second (using software de-interlacing) when I'm only 
displaying circa 49.99, cpu useage spiked up to 100% and it would become 
very difficult to get into the menus or quit Xine. This could of been due to 
the hacked driver I was using, I think the back buffer would eventually be 
in use for a whole frame requiring most cpu time to be spent in X waiting 
for the buffer to become free, preventing other tasks from running properly 
(and presumably X from doing other things too). I've compared the software 
deinterlacers built into Xine both against each other and the deinterlacer 
in the plasma using a variety of DVDs. I got a mixture of results, there 
were very noticeable differences in performance when viewing Video /mixed 
material  e.g. Star Trek TNG most of the deinterlacers introduced a lot of 
artifacts that shouldnt of been there. IMHO vertical was closest to the 
inbuilt deinterlacer on the plasma and didnt give me any strange flickering, 
noise, stair steps, excessive ghosting etc, it seems to be a good 

I've never tried to bring up an interlaced mode on the aopen board although 
I did consider running VGA->scart. Scart is still my prefered choice as it 
integrates with the rest of my system properly, VGA out doesnt. I'm 
currently using the chrontel tv out, I got some good advice here (xorg) 
about how to trace the BIOS so I've started testing using the tv out 
recently. I soldered up the Scart out from the chip (not advised) I believe 
most chrontel chips have scart out but I dont think any mobo manufacturers 
provide the pin headers or output sockets / cabling required. I've done the 
basic tuning to be as close as possible to a regular DVD player again using 
DVE PAL. Although results are good I think I'm getting some slight noise 
from my wiring, I may have to get some better cable and try again. I get a 
measured field rate of 50.001.. which with a margin for error probably means 
it's as close to the PAL spec as possible (I think this may give the VGA out 
a refresh rate of 50hz). Xine seems happy - no dropped frames and no CPU 
burn with the standard video out loop. According to the blurb the tv out 
uses 10 bit DACs, although I'm not sure either how this compares to the VGA 
DACs or whether the full range is being utilised via XV (I havent noticed 
any obvious banding during testing which is not already present elsewhere).

I've not had any luck with the field parity, I have to run with the software 
deinterlacer to get interlaced video working smoothly. Presumably using the 
software interlacer it's just luck if the output field lines up with the 
computer generated lines or the original ones. Progressive video does seem 
to work very nicely with the deinterlacer off, of course if everything was 
progressive things would be much easier however thats not the case. IMHO 
where the source is 50 fields per second or circa 60 fields per second 
trying to generate a 25 frames per second or circa 30 frames per second 
progressive output results in a lack of smoothness and sometimes a loss of 
resolution too depending on how the deinterlacer works (commenting on the 
difference of a typically transcoded video file vs realtime deinterlacing of 
the source stream at full frame rate). I think there may be a fix for the 
field parity on the EPIA although I'm not sure if it works under X or not, 
and I think scart out from the TV chip on the EPIA requires a soldering iron 
too. If you get the interlace modes working it would probably make the VGA 
out a good way to go to get scart, althought I think you'll still need to 
run with a software deinterlacer to get things working.

Type your favorite song.  Get a customized station.  Try MSN Radio powered 
by Pandora.

More information about the xorg mailing list