I think what takes so long is the time it takes for the vertical
oscillator in the digitizer to sync to the new signal, which could
be up to one or two fields.
This is what I originally thought. Now, though, I get a problem where even
at 2fps per camera, one camera seems to drift in and out of a situation
where the frame is torn in half. I can't think why the chip would take so
long (around 5 or 6 fields) to successfully sync to the new signal.
Another idea would be to use video
delays to match the signals in time, (timebase corrector?). I have
no idea what that would cost.
Another problem, I believe that contributes to the one above, is that
cameras' frame rates seem to drift around a bit.. like one will run at
24.98fps, another will run at 25.05... after a minute or two, one camera has
built up an "extra" frame.
The ultimate solution, I think, would be to have 4 external double buffers,
and have 1 DAC reading from each in turn and outputting the signal to a
single composite input on the BT878, and the source number to GPIO (or from
GPIO!)... or, try to use the BT878 in SPI mode and just read the data
straight, using GPIO. But... we lose the benefits of the 878 being a good
video digitiser - UltraLock(tm) etc..
Bottom line, use genlocked cameras or avoid this problem. I dunno, maybe
someone smarter than me could work it out.
_________________________________________________________________
The new MSN 8: smart spam protection and 2 months FREE*
http://join.msn.com/?page=features/junkmail
