VGA to PAL using AD725

Hi,

I'm currently developping a system to output PAL video from a camera.
I worked on a former project where the ouput video was in VGA. The video structure was like this : FPGA => DAC => VGA.

In order to add PAL format to this project, I add an AD725 to convert VGA to PAL. I proceed as shown page 12 on the datasheet : http://www.analog.com/static/imported-files/data_sheets/AD725.pdf

I code the new timings signals (hsync and vsync) according to PAL standard as shown on http://martin.hinner.info/vga/pal.html
I used two state machines to do it with a 13.5Mhz clock

constant v_long_sync : integer := 371; -- 27.5us
constant v_short_sync : integer := 32 ; -- 2.4us
constant v_long_delay : integer := 400; -- 29.6us
constant v_short_delay : integer := 61; -- 4.5us

constant h_front_porch : integer := 0;
constant h_sync : integer := 54; -- 4 us
constant h_back_porch : integer := 108; -- 8us
constant h_left_border : integer := 0;
constant h_video : integer := 702; -- 52us
constant h_right_border : integer := 0;

I tried my code on ModelSim successfully.


However, when i displayed a pattern on a TV, the image quality is not what i expected (some kind of diagonal lines moving but quite perceptible) . I try to test the signal directly at the output of the AD725 and I get this signal for the composite output :



I think that the noise is responsible for the quality of the picture and i'd like to solve this noise problem. Any idea ?
Furthermore, I measured the frequence of the pal quartz I used and this one is quite different from the PAL clock needed (17.734475MHz) (about 600-1000Hz difference). Do this clock need to be exact ?

I really want to improve the quality of the picture.

Kind Regards.

Joeka

Here is an example of what i see on a TV :

Yes, most probably it's due to noise generated in ground loops. Sheilding and earthing of board, signal cables will solve the problem. Grounding metal cans of crystal. Try tv and vga signal source with removed from mains earth. This is some kind of high frequency, see if switching supply is not the cause.