UFO Operation
Timeouts
- The DMA timeout is at least 50ms
- The bellow is explained how long it takes to read frame by hardware. We also need to add DMA timeout + PCIe transfer time on top of that before we can expect to acquire frame
ð Yes, I wish to fix this definitively. We can assume for today and for the future that we want to use only microsecond time (not nanosecond) in the follows way: total time = integration time + readout time + back pressure from PC/DMA
ð Integration time à set by Libuca so that you know well (microsecond)
ð Readout time = number of rows * time per row. Now, number of rows à set by Libuca (you know) and the time per row is:
o = 13us for 4MPixel @ 11 or 12 bit
o = 3.3us for 4MPixel @ 10 bit
o = 2.7us for 2MPixel @ clock 48MHz (in this case only 10bit is available)
ð back pressure from PC/DMA à this is very variable and depends from the PC status not from the electronics.
ð The DMA timeout is 67 ms that is too much. What we can do easily is to bring this timeout programmable by software, we will create a new register in the DMA bank register (address 0x64-0x68), so that you can define your timeout.
Frame Size (CMOSIS sensor)
4*129*256*(number_lines -1) + 3*129*256 + 128*256 + 512 bits, or
4*129*number_lines + 32 BYTES