Mark Adams a écrit :
>>* the image on the VGA monitor lacks a few pixel on the right (approx
>>6-7 pixel columns missing): CRTC registers to be adjusted ?
>>* the image on the TV lacks even more pixels on the right. I had to
>>cheat with GH_ACTIVE, with 773 instead of 720 to be able to get enough
>>of the image on the visible area
> 
>  On my machine, the output is perfect -- I've checked the timing on a scope. 
> There could be more differences between the VT1622 and the VT1622A or there 
> may be something else wrong. Note that you should _not_ be able to see all 
> 720 pixels: the analogue active line of a PAL TV picture is 702 clocks of 27 
> MHz so 9 pixels should be hidden at either end. If you can see all 720 
> pixels, the active line is either extending into the line blanking period or 
> there is a horizontal squeeze.

I also read http://www.tvforum.co.uk/forum/viewtopic.php?p=193549 (last
post), with similar information, detailled differently.
I just expected to see the whole 720 lines at a time, with a little
overscan.
Now I know that in fact there is no such 720, but better 702 pixels per
line. My 720x576 main test image should then not show the same part on
top/bottom and left/right : the corner white circles are hidden on the
left/right edges. I just expected them to just fit at the left/right
visible part of the image.
I should then tune with 702x576 (not 720) test images.
This also explains why some DVB channels have some kind of little
right/left black bars : these are the 9 non-fitting pixels at each side.

>>* HSCALE_FAC is not 0, ie. there is an horizontal (down-)scaling clearly
>>visible when viewing the moire_2x2.png test image. Could this be the
>>27MHz vs. 28Mhz dotclock effect ? 27MHz should give a wider image than
>>28MHz, that I have to compensate using HSCALE_FAC, which is not perfect.
> 
>  I don't understand this. I have HSCALE_FAC = 0 and there is no pattening on 
> your moire test image. If I set HSCALE_FAC to your value, I see extensive 
> banding.
>  Perhaps now you've got it working, you might want to try the original code 
> I sent with just the #define VT1622A changed and see how that turns out?

I'll retest and tune more, after understanding that.

>>Here is the result of the "D" command in the vt1622 program:
> 
>  Here's mine if it's any help:

I'll compare.

Thanks for all.

-- 
NH

_______________________________________________
directfb-users mailing list
[email protected]
http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-users

Reply via email to