On Thu, 19 Dec 2002, Christian Schmidt-Madsen wrote:
> After quite a few attempts to manually fine tune the dcf files I use to
> set up my meteor2 card, I got hold of a win box with Matrox Intellicam
> (a utility for manipulating dcf files). Using Intellicam I managed to
> get seemingly perfect images, which where grabbed with the Intellicam
> program. Loading the same dcf file on my linux box and grabbing an image
> with the 1.0 version of Cord Seele's meteor2 driver (I have used the
> 'dcf' and 'ppmgrab' programs from the driver's 'utils' dir), I got
> somewhat different results.
>
> The lines in the grabbed image seem to be slightly shifted
> horizontically and each line seems to influence the line before and
> after. To demonstrate this effect I have attached (a small part of) two
> images grabbed with the same dcf file, one from a win box and one from
> my linux box; unfortunately the win image is quite dark, but the
> difference is clear enough.
>
> The artefacts/errors in the linux image seems to indicate a timing/sync
> problem (bad fporch/bporch/hsync/...), but the image from the win box
> seems perfectly ok. In my attempt to locate the source of the problem I
> have tried two different cameras:
> an older JVC KY-F55B (3ccd rgb) with a homebrewed dcf
> and
> a more modern Sony XC-003P (3ccd rgb) with Matrox' dcf (from the
> Mil-Lite library)
Are these interlaced cameras?
There's a known problem with b/w interlaced cameras, that probably also
applies to color cams (not tested). The effect is that the two field are
stored in the wrong order. I havn't found a reliable way to fix this yet :-(
As a workaround, you can force a certain field order by playing with two
configuration bits (VIA_GRABCTRL_GFLDPOL and VIA_GRABCTRL_GREGUDT_SHIFT) in
via_grab_setup() in file via.c. This should help.
Please report any problems.
Cord
