Our problem is still present, but I've continued my research, so I can give you more information.
In our system, we have a quite long PCB trace coming from a crosspoint switch, which is connected to the ADV7619's input through AC coupling. The ADV7619's pixel bus is connected to an FPGA, which has a video measurement block, so I can read out the parameters of the sent video.
So I wrote the script, which switches video to the ADV7619's input, waits a little, then reads out the measured data.
Using this tool, I've determined, that there are two cases.
The first:
- ADV7619 locks on the incoming video
- The ADV7619 knows the correct resolution
- The data from the FPGA shows, that the output is correct
In this case, the chip will be continously locked to the video without any glitches or dropped frames. The FPGA will measure the same parameters all the time.
The second:
- ADV7619 lock on the incoming video, but V_LOCKED_ST becomes 1 very often (see previous posts)
- The ADV7619 mistakes with the vertical resolution (usually with +-4 lines). Sometimes it say the correct resolution, but the value is changing.
- The FPGA shows changes in the VSYNC parameters indicating that there are glitches on the VSYNC line. This can be measured with a scope too.
The most strange thing in this problem, is the connection between these two states. I've seen, that the bad state can somehow fix itself (going from (2) to (1)), but that's very rare.
I think, that which case we will see is determined during the transition when I switch the video on the IC. The fact, that we receive V_LOCKED_ST interrupts indicates, that the problem is located at the early stage of receiving.
So I think, that the problem comes from the automatic equalizer. I think that sometimes this automatism fails to find the correct EQ parameters. I've seen that there's a manual EQ recommendation for low speed signals, but I could not find any information on the manual EQ settings except the given register values.
I think that the best way to find out if this theory is valid is to read out the EQ values determined by the automatic procedure for case 1, and apply it as a manual EQ setting. If this eliminates the problem, then we've found it's cause.
Can you help me with information regarding the equalizer? Or do you have any other idea what should I check?
Thank you,
Arpad Maroti