On Friday 21 March 2003 10:50, Gerd Knorr wrote: > Gabor Kerenyi <wom@xxxxxxxxxxx> writes: > > > fetch latest bits from http://bytesex.org/snapshot/ > > > > Thanks, it works. Though it doesn't solve my problem > > as I thought. > > I have to decide the next field (top or bottom) in the IRQ > > handler depending whether the bt chip previously > > captured a wrong field or not. > > Point being? Why not simply capture at full field rate and throw > away the ones where the bt878 chip didn't manage to pick the > correct field? (due to resync after input switch I assume?). Yes you're are right. It is needed to achieve higher frame rate when multiplexing inputs. The problem is that after switching inputs the BT chip sometimes captures a wrong field. No problem, on the Euresys card there is a chip which outputs the correct field and I can get the signal through GPIO. I tried to capture at full framerate but the result was not good enough. The frame rate was max. 2.5 fps/channel (4 channels). Previously I used FIELD_SEQ_TB and at that time I could manage to get 2.5 to 3.8 fps/channel. (for very short time 4.4) That time I used an algorithm something like this in the IRQ handler: CapturePicture IF COUNTER>0 OR FIRST_FRAME=TOP switch channel IF (FIRST_FRAME=TOP) buffer->reserved[1]=TOP ELSE buffer->reserved[1]=BOTTOM COUNTER=0; ELSE COUNTER=COUNTER+1 FI So it's very simple. After the switch the first field in the buffer will probably wrong so we have to drop it. But the second field in the buffer will be good, so if the second one matches the requested FIELD type we can use it, otherwise we capture another 2 fields. So at worst case 2 frames (4 fields) are needed to be captured to get the correct field, but sometimes 2 are enough. The disadvantage is when the second capture is made and we use the first field not the second one in the buffer we would not need to capture the second field. At FIELD_SEQ_TB the 2 fields in the buffers are never the same, but the order can change. (TOP-BOTTOM, BOTTOM-TOP). with an extra chip on board it is possible to determine the first or the second image matches the requested field. With FIELD_LATERNATE I thought it's possible decrease the number of frames needed to capture at the worst case from 4 to 3. Gabor
