here a few runs with different ADU/exposure times. Looks like the exposure time is divided by 10, than the number eventually becomes numericaly instable , sometimes even negative.
3x ADU20 5 sec exposure flats:
hold on ... are you using Ekos ADU % to calculate optimal exposure time for a particular ADU %? If you see the ADU % to anything but 0, it will try to automatically figure out the exposure time required to get to that ADU. I think this could be causing it, let me investigate Ekos algorithm.
Ok, so it is the ADU flat algorithm, it was only tested on the simulator before. Can you send me a serious of 10 flats with different ADUs? That would help me a bit in the testing, thanks!
So I updated INDI::CCD to check for exposure limits, it was accepting negative value before!! Furthermore, I updated Ekos to fail if the calculation does not yield useful results. So don't send me anything yet, tomorrow both kstars-bleeding and INDI will be updated in the PPA. Please update and then start KStars from the console/terminal, and take flats again with a particular ADU % in Ekos, then send me the KStars log from the terminal, it should display the ADU & Exposure settings along with the calculated slope in each go. Send me the log and the captured fits so I can investigate further.