Hello,
thank you all for your answers! There are some great ideas that I’ve explored / I’m going to explore!
I continued my research and I think that the pictures are affected by an inconsistency in color reproduction in Sony cameras:
This sucks because there is no easy or deterministic fix.
In addition I played with the RAWs again trying to find a way to mitigate the problem. First I converted them to tif files with dcraw -6 -j -t 0 -o raw -T -b 3 -a. Then I did some trickery with ImageMagick to equalize the RGB values, separate the channels, etc. Finally I subtracted one image from the other (per channel) to kind of create a diff.
I found, that while the BG channels produce an almost pure black image (apart from where the images differ, e.g. moving stars), indicating almost no difference, there’s a distinct red cast for the R channel. Unfortunately the cast is uniformly distributed, but take a look yourself:
That explains (at least to me) why no matter what color correction I tried the images never really looked alike. If I corrected for the cast in the top left and lower right the center would be off and vice versa.
@ggbutcher Are you sure you got the images to match with your software? I too tried to adjust them with the raw black/white point tool in darktable but I was unable to match certain images.
I’ll give rawproc a try to see if there’s inconsistencies with darktable’s rendering.
I also took a bunch of black frames, unfortunately not in the same conditions though. I might put my camera on the balcony over though once it gets to freezing here, but that’ll be January or February. Maybe I’ll get an earlier opportunity. For now I have this:
(all photos in darktable, same WB applied, exposure +10EV)
That’s a series of pictures from ISO 100 to 1250 in 1/3 stop intervals. I only adjusted ISO in between and hit the shutter right away. I really can’t explain the first two (ISO 100 & 125) but the rest looks like expected, a color cast that gets worse with higher ISO)
A series at ISO 1600 with an Intervalometer set to 30s / 1s delay.
Same as above, ISO 320
There’s a notable and arbitrary color cast here, but at least in my eyes uniform.
I’m rather lost here and tend to accept this just at it is. I think next time I’ll just set up my A7RII again for milky way timelapses, together with the big 1635/f2.8 I should get some decent imagery.
Here’s a clip of a part of the series with rather minimal processing, I’d say the color flicker is barely noticeable (Nothing I’d give away in a professional setting though):
(video is HVEC / x265, you might need to download it to view it, at least my browser won’t)
If you’re interested in the black frames, let me know and I’ll upload them (maybe not all, to not overburden the forum…?)
Indeed I think I might’ve left on the camera noise reduction. I’ll turn it off and do another series!
@Peter Thank you for the hint on hraw. I’ll try that too and report back!