We are using two CVBS cameras (PAL) connected to Ain1 input (camera 'A') and Ain2 input (camera 'B') of the ADV7180WBCP32Z.
See attached "Camera-IF_1A_1B.pdf" for the schematic section around the ADV7180. The Ain1 and Ain2 inputs are driven locally from video buffers, hence the input divider of 1k/1k as per some your application notes; the Ain3 is just a reserved spare input (available on a pin header) and not driven at all.
We used INSEL[3:0] set to a default 0000 for CVBS input on Ain1, then just controlled MUX0 to select Ain2.
And we have noticed that after a power-on cycle, our application failed to obtain the embedded data from the 'B' camera (within preset time-out) if the 'A' camera is also connected.
If camera 'A' is disconnected, camera 'B' will be detected within preset time-out (if connected).
However if both 'A' and 'B' are connected, only 'A' will be detected.
The root cause seems to be a failure of the ADV7180 decoder's clamp-restore circuitry to establish the dc level of the input signal within the application's camera detection timeout.
Currently our application has a camera detection timeout of 840ms. The attached "Tek00007_A-and-B-cameras.tif" image shows that with both cameras connected the decoder seems to require > 1.3 seconds. Compare this to the other image showing the same signals with only camera 'B' connected.
The 'scope channels in the attached images are allocated as follows:
Chan 1: Debugging test point; indicates the moment that the decoder MUX0 is written (marker).
Chan 2: Decoder VS/FIELD output.
Chan 3: Ain1 input to the video decoder, i.e. camera 'A' data (pink).
Chan 4: Ain2 input to the video decoder, i.e. camera 'B' data (green).
But then we have tried to use only the INSEL[3:0] to change between Ain1 and Ain2 inputs, i.e. changing it from 0000 to 0011, without controlling the MUX0. We have also left the decoder to autodetect the input format (PAL or NTSC).
And in this case there wasn't any long delays in switching from Ain1 input to Ain2 input - with both cameras connected - and both cameras were detected well within the time-out.
Why is such a big difference in the operation? Could you provide some advise on dynamic behaviour, related to the muxing and acquiring the input signals?