I think I found a problem. I was using 2.6.0.6 and took bias, dark and light frames (FITS). The dark and light frames were 600 seconds long but both strongly resembled the bias frames. The light frames looked like a bias frame with a few bright stars on it. All were taken with my obsy computer but I don’t think the EOS 60Da is actually exposing for that time at all. I should have realised from the histogram. The light pollution should have moved the peak to the middle for that kind of time.
My subsequent tests were done with a different computer using the same SGP version and both CR2 and FITS formats. Saving CR2 files to the memory card seems reliable (though there is definitely some form of dark frame compensation going on in the RAW file) but at least the SD and max values of the bias frame are considerably smaller than that from a 5-min dark frame. I’ll load up 2.6.0.10 and try again.
Just found this:
I think I fell into the trap. Drat and double drat, I just wasted two clear nights