We have a design with 2 DS90UB914 parts on it and 2 DS90UB913 parts.
The DS90U914 parts are set up for power over coax as in the demo board design.
We have had the most trouble getting a stable connection to the DS90UB913A_CXEVM demo board. I have been able to get a more stable connection by making changes and cutting some traces in our board to clean up the layout.
One thing I have noticed is that longer cables seem to work better than shorter cables. For example, one of the two channels will not lock consistently with a 12in cable, but will achieve stable lock with a 24 in and 60 in cable. I see that your recommendation for the emo board is a 60 in length cable. I also have one channel that does not consistently send I2C communication over the link to the remote camera sensor. I can get it to program the sensor after multiple tries, but it seems to work better over a longer cable length. I'm wondering if this would point to anything in the routing or if there is some other reason this would be the case.
I'm working on a second revision of the board and I want to make sure I get the routing as clean as possible. I'm using the board in single ended power over coax mode. I see on the demo board that the RIN- signal is routed about the same length as the RIN+ signal. Is it necessary to keep the trace lengths the same for single ended operation, or is this just so that the demo board can accommodate differential mode as well as single ended?
Thanks,
Ben