Overcorrection of Oiii flats


I am having trouble with calibration of my mono Oiii data, in the form of my flats over-correcting. The issue does not arise when using my mono Ha filter.

All of my lights and calibration frames are perfectly time and temperature matched, but try as I might, I cannot fix this issue.

I have heard that there is a Pixelmath workaround in Pi, is there any way I can fix this in Siril?

Apparently, this appears to be a common issue with Oiii filters.


Wow. No pixelmath could solve this.
Are you sure it is only the flat that make it? Did you tried a background extraction?

If your flat is the only responsible then … Redo your flats.

Hi Ciril.

Yes I have tried a background extraction but to no avail. I know it is related to the flats as when I calibrate using flats and matching flatdarks from an earlier Ha integration, most of the overcorrection is gone. Obviously using Ha flats going forward is not a fix.

The original flats were sky flats (which I never usually do, as I have an artificial light box), but when I reshot them with my light box but the result was the same. Both were shot with a mean ADU of 20400.

Should I lower the ADU of my flats?


My concern is about the shape. If it is just a flat issue the shape is absolutely not normal.
Please show us your flat in histogram equalization mode.

As requested, histogram equalized master flat.

OK. So your flat is really not good.
Either your light source is bad, really bad and not homogeneous, or … Actually I don’t see another option.
Why it is visible only in OIII? Because OIII needs longer exposure.

Right - I have finally pinned this problem down.

I tried reshooting my flats+darkflats with decreasing ADU down to 1000, with no success.

Tried blackening a surface of my step-down ring, on the objective end of my optics, to minimise internal reflections - didn’t work.

Last throw of the dice was to swap out the halogen bulb in my lightbox to a high power LED - success! Flat frames are much more evenly illuminated and mirror the optics.

Final stack is now how it should look. Here is the new masterflat.

Thanks for your assistance, as I was close to conceding defeat ! :+1:


I don’t recall having ever seen a flat that has bright corners and dark sides like that, for me it’s still not alright.