Artifacts when using flat frames


I usually don’t use flat frames, but this time there is visible vignetting ruining the photo. If I don’t use flat, the artifact/noise does not appear. It shows only after stacking, regardless of the method. Camera is Nikon Z6, tracked, but unguided with bad alignment (I don’t see north pole from my balcony). Flat is taken by pointing the camera to LCD screen (brightness reduced) with same settings, t-shirt over the lens and only 1/400s exposure (surprisingly, longer exposures cause overcompensation and bright corners, took me several days and >100 flats to figure that out). It happens regardless of whether I use bias frame or not. Noise is present on all 3 channels, especially blue. Siril is configured to use 16 bit FITs and compression.

I’ve read in the past on this forum and elsewhere that it’s the camera induced correlated noise when alignment is near-perfect and to use dithering, but in my case alignment is far from perfect and it shows only when subtracting flat. Flat frames itself look ok, min value is 1072, max 5497, mean and median ~1260 (I can upload if someone wants to look at it) which is somewhat low, however no such pattern is visible. Hmm, now when I think, could it be possible that, when stacking pre-processed light frames, we’re also stacking (now negative) flats which are always at the same position and thus correlated and that dithering would need to be applied while subtracting flats? Perhaps I need to stack many more flats, currently I have 5-10 of them.



Hi Vedran,

took the liberty to plate solve your image to see the orientation of the pattern ( So it seems the main direction is along RA axis, which would suggest that what we see is mainly walking noise generated by the periodic error of your mount unguided. There is also some small Dec component, probably due to your PA being off by a little. This kind of pattern is only visible after stacking, never in individual frames.
Now, to get rid of this noise, you should indeed be dithering. There is no need for the alignement to be near perfect to have walking noise and to have to dither. Unlike slowly-varying drift induced by PA being off or mount PE which causes a reproducible back-and-forth movement, dithering is totally random and will spread around the pixels causing this.
Now why would this appear only when you apply flats does not make much sense to me. I would guess that the same pattern is present when you do not apply flats. But because there is important vignetting without flats, the stretching of the stacked image being adapted to cover the range from the dark corners to the bright center does not really make the pattern visible. And it only appears when your image has been made flat.
If you want, you can share the stack with and without flats applied, as well as one flat and one bias frame so I can have a closer look.



Thanks for reply. Right, I was able to find similar examples of noise elsewhere online, however I use Star Adventurer mini mount and don’t use computer guiding so unfortunately dithering is not possible as far as I know (unless somehow done manually). That being sad this is first time I saw it in years, perhaps because signal this time is very low and image needs significant stretching while fighting with haze and skyglow.
Interesting thing is that I was eventually able to mostly get rid of it just by stacking many more flats, when I stacked 33 flats the noise was pretty much gone. I had no idea that stacking many flats has any benefit, but I guess I was wrong.
However in the end, using flats didn’t improve the image significantly. I don’t have to significantly crop it anymore as corners look much better, but after background extraction there still remain faint colored “ring-like blotches” in the image, mostly in green and red channels, less pronounced than before without flats, but still visible.

Sure I can share, do you need just a single original flat/bias frames or ones obtained after stacking (given that stacking them apparently improves situation a lot). Actually, if you have enough bandwidth, time and disk space, I could upload somewhere all original raw .nef files (~15GB: ~460 lights, ~30 flats, couple bias).

Final image without using flat frames:

Using flat frames (postprocessing done was similar in both cases):