Difference between revisions of "Cx88 Color problem patch"

From LinuxTVWiki
Jump to: navigation, search
 
m (7 revision(s))
 
(6 intermediate revisions by 4 users not shown)
Line 1: Line 1:
On 28 April, Sami Kiiskila wrote to the v4l mailing list about a color problem on cx88 chips:
+
On 28 April 2005, Sami Kiiskila wrote to the v4l mailing list about a color problem on cx88 chips:
  
 
Do you see any difference in these pictures:
 
Do you see any difference in these pictures:
  
 
http://www.cs.tut.fi/~kiiskils/kuvat/TTV-Testikuva-tv-bad-2005-04-23-19.png
 
http://www.cs.tut.fi/~kiiskils/kuvat/TTV-Testikuva-tv-bad-2005-04-23-19.png
 
  
 
http://www.cs.tut.fi/~kiiskils/kuvat/TTV-Testikuva-tv-ok-2005-04-23-19.png
 
http://www.cs.tut.fi/~kiiskils/kuvat/TTV-Testikuva-tv-ok-2005-04-23-19.png
 
  
 
especially when zoomed to 2x size ?
 
especially when zoomed to 2x size ?
Line 16: Line 14:
 
interference in your own recordings then the "patch" is simple.
 
interference in your own recordings then the "patch" is simple.
  
 +
Edit  : /usr/src/linux/drivers/media/video/cx88/cx88-core.c
 
Change : HLNotchFilter135PAL to HLNotchFilter4xFsc
 
Change : HLNotchFilter135PAL to HLNotchFilter4xFsc
 
on line 752 of cx88-core.c and recompile (snapshot video4linux-20050310-100754)
 
on line 752 of cx88-core.c and recompile (snapshot video4linux-20050310-100754)
Line 24: Line 23:
 
- hardware is Hauppauge WinTV GO with cx88xx chip
 
- hardware is Hauppauge WinTV GO with cx88xx chip
  
What happens is probably that the HLNotchFilter135PAL is designed for
+
What happens is probably that the HLNotchFilter135PAL is designed for 13.5MHz sampling rate and the driver tolds the chip to use 27 MHz.
13.5MHz sampling rate and the driver tolds the chip to use 27 MHz.
+
So the notch  frequency is shifted too to 8.866 MHz instead of 4.433 MHz, it doesn't do much good there. So my original guess that the comb filter is not working was wrong, which I couldn't be sure until I tried out both of the notch control registers.
So the notch  frequency is shifted too to 8.866 MHz instead of 4.433 MHz,
+
it doesn't do much good there. So my original guess that the comb filter is
+
not working was wrong, which I couldn't be sure until I tried out both of the  
+
notch control registers.
+
  
 
If the sampling really is done at 27MHz then similar fix would propably help
 
If the sampling really is done at 27MHz then similar fix would propably help
Line 36: Line 31:
 
Added 2005-04-29:
 
Added 2005-04-29:
  
I tested it on American NTSC using DViCO HDTV 3 Gold hardware. It functions properly.  
+
I tested it on American NTSC using DViCO HDTV 3 Gold hardware. It functions properly.  
I don't have a test pattern to be able to determine if there is an immprovement, but it seems better.
+
I don't have a test pattern to be able to determine if there is an immprovement, but it seems better.
 +
 
 +
Added 2005-11-07:
 +
 
 +
It works on British PAL-I too (on a Hauppauge WinTV card) -- the patterning was very annoying, and the picture quality's much better now. -- ats
 +
 
 +
Added 2006-04-01 by Igor M. Liplianin
 +
 
 +
After few days serching I found it ! It works great on my V-STREAM(Kworld) Xpert TV-PVR 883.
 +
My TV system is SECAM, kernel 2.6.16-git9. Change made in /usr/src/linux/drivers/media/video/cx88/cx88-core.c  line 660.

Latest revision as of 20:15, 7 December 2008

On 28 April 2005, Sami Kiiskila wrote to the v4l mailing list about a color problem on cx88 chips:

Do you see any difference in these pictures:

http://www.cs.tut.fi/~kiiskils/kuvat/TTV-Testikuva-tv-bad-2005-04-23-19.png

http://www.cs.tut.fi/~kiiskils/kuvat/TTV-Testikuva-tv-ok-2005-04-23-19.png

especially when zoomed to 2x size ?

If you don't then no need to read further  :)

In case you notice something on the color areas and if you are having similar interference in your own recordings then the "patch" is simple.

Edit  : /usr/src/linux/drivers/media/video/cx88/cx88-core.c Change : HLNotchFilter135PAL to HLNotchFilter4xFsc on line 752 of cx88-core.c and recompile (snapshot video4linux-20050310-100754)

Some warnings: - it might break something - the only known format it works for sure is PAL-BG (used in Finland) - hardware is Hauppauge WinTV GO with cx88xx chip

What happens is probably that the HLNotchFilter135PAL is designed for 13.5MHz sampling rate and the driver tolds the chip to use 27 MHz. So the notch frequency is shifted too to 8.866 MHz instead of 4.433 MHz, it doesn't do much good there. So my original guess that the comb filter is not working was wrong, which I couldn't be sure until I tried out both of the notch control registers.

If the sampling really is done at 27MHz then similar fix would propably help with NTSC too, just change HLNotchFilter135NTSC. But I can't test that here.

Added 2005-04-29:

I tested it on American NTSC using DViCO HDTV 3 Gold hardware. It functions properly. I don't have a test pattern to be able to determine if there is an immprovement, but it seems better.

Added 2005-11-07:

It works on British PAL-I too (on a Hauppauge WinTV card) -- the patterning was very annoying, and the picture quality's much better now. -- ats

Added 2006-04-01 by Igor M. Liplianin

After few days serching I found it ! It works great on my V-STREAM(Kworld) Xpert TV-PVR 883. My TV system is SECAM, kernel 2.6.16-git9. Change made in /usr/src/linux/drivers/media/video/cx88/cx88-core.c line 660.