Created attachment 108227 [details] Test log have Kstars/Ekos running on a Raspberry Pi 3 under Ubuntu Mate and controlling a QSI 583 camera. After imaging last night I set up a run for flats using the camera control module, setting 1 second (as recommended in the tutorial) as a starting point for the LRGB filters with a target ADU of 40,000. It started doing test shots, gradually decreasing the exposure time (for the R filter if I recall but it could have been L) but after a number of attempts it jumped the time up above 1 second and starting looping doing this dropping the time and then jumping back up. I decided to redo it, starting the exposure times at 0.05 and that worked for the RGB but failed on the L, where it kept on trying the same exposure time repeatedly. I ended up estimating what the exposure should have been and it worked fine. Additional testing about a week later and things improved after bringing the software up-to-date, however, I still have some issues. I set up a sequence queue for LRGB starting each exposure at 1 second with a target of 40,000 and a tolerance of 1,000. This time it processed the LRG but got into a repeating loop for the B(lue), which appears to be infinite, as seen in the attached text file. This is visible in the first attachment. Today, I tried another test and found a worse result for the Luminance. 2017-10-08T10:32:35 Capture failed. Check INDI Control Panel for details. 2017-10-08T10:32:34 Capturing image... 2017-10-08T10:32:34 Current ADU is 57337 Next exposure is 0.030 seconds. 2017-10-08T10:31:53 Capturing image... 2017-10-08T10:31:53 Current ADU is 64000 Next exposure is 0.094 seconds. 2017-10-08T10:31:12 Capturing image... 2017-10-08T10:31:12 Current ADU is 64000 Next exposure is 0.125 seconds. 2017-10-08T10:30:32 Capturing image... 2017-10-08T10:30:31 Current ADU is 64000 Next exposure is 0.250 seconds. 2017-10-08T10:29:51 Capturing image... 2017-10-08T10:29:51 Current ADU is 64000 Next exposure is 0.500 seconds. 2017-10-08T10:29:09 Capturing image... By my calculation where 0.094 seconds produced a value of 57,337 ADU then it should have tried 0.065 seconds (40000/57337*0.094) but instead tried 0.03 which failed (but this appears to be another bug as 0.03 should be valid and has worked in the past).
After more investigation I have come to the conclusion that the problem is significant inconsistency in the brightness for a short shutter speed. Either this is due to the light source being inconsistent at that time scale or the shutter. It would appear that the shutter speed is quantised and can only make larger adjustments than the 0.01 seconds required. My solution has been to decrease the flat panel brightness and this seems to have overcome the issue. My suggestion is to start with a shorter exposure than the recommended one second and let the algorithm increase exposure as starting bright introduces residual brightness issues. I suggest this can be closed, but I don't know how to do this.
I can we can add this to the documentation so other folks would know. Thanks!