Dear Guenter,
Happy new year, and many thanks for your comprehensive answer.
To answer your question, the noise does not get worst with heating, but a very small amount of coolant on the top of the ADV instantly cures it. I have to apply a lot more cooling to get the same effect anywhere else on the board.
The ADV is running on two isolated PSUs - please see our design below.
Now that I am analysing our original design, I can see room for improvement. E.g. we do not have a 400 ms delay between 3v3 and 1v8 start up, and our EMC filters do not have a ground tap (as used in the reference design). But we have made over 1000 units without ever seeing this issue.
It has only started to happen on our newest batch..
I tried your idea for replacing the on board PSU with an external PSU (I selected the 1v8 rail for this trial). It made no difference, but I did find that by gradually reducing the voltage (1v8 -> 1v6) the noise stopped shortly before I completely lost the image. It was like something within the ADV "switched off" earlier than the A/D converters which causes the noise to stop...
On board, the 1v8 and 3v3 buck converters run at a fixed frequency (~1.25 MHz). However, I tried to cool the PSU chips as their curves suggest this would slightly lower their frequency but it made no difference to the noise.
Our board does not have a uP - but we do use FPGAs to process the digital video and new video clock. However, from what I can see the noise is being picked up at the point of the analogue sampling. I say this because the noise is always white, despite my red and green inputs both being disconnected..
Thank you for your input this issue. Where would you suggest I go from here?
Regards,
Adam