On Fri, 21 Mar 2008 12:00:02 +0100 Niko Mikkila nm@phnet.fi wrote:

All XvMC implementations that I know of have poor deinterlacing, which makes them pretty useless for decoding interlaced video. Your description suggests that openchrome just drops the other field, so you lose half of both the vertical resolution and temporal definition (50 fields/s -> 25 frames/s). Continuous small skips (at regular or seemingly random intervals) could also be caused by a refresh-rate mismatch, but if Xv is ok, this is not the main problem in your case.

Hmm, it seems that I've not understood something... I assumed that if I don't say anything about deinterlacing, then the (s-video) output would be interlaced. This apparently isn't true, they do something (strange) to the video, because putting deinterlacing on (bob-deinterlacing) produces quite nice results! (well, 40-50% CPU usage, and a little-bit of jerking, 1/s, but the ticker is very sharp now).

But shouldn't the device be able to send completely non-deinterlaced stuff to the tv-out? Maybe the unichrome chip does something, but can that something be turned off?

Anyway, thanks for the info, it led me to a bit more tolerable results :-)