StarTools

Hello, having recently acquired a new PC I am finally processing various pictures from this winter.

This one in particular has been giving me some headaches, it's a stack of about 70 120" lights, 20 flats and biases shot at ISO 400 with a modified Canon 100D and a Samyang 135 at f2 on a star adventurer from a bortle 4 location. Images were captured directly in .fits with astroberry and then stacked in siril. As can be seen from the severe walking noise I did not dither, I am looking into a DIY motorization of the dec axis to deal with this in the future.

Here is the stacked file if anyone would like to give it a spin.

And this is my definitive attempt at processing with the related log.
Removing the gradient caused by a streetlight proved tough, I had to push aggressiveness to 98% and there still appears to be some light pollution around the upper right corner. Is it bad to bring that parameter almost to the max? Would it be better to leave it at a lower value and crop out remaining gradients? I masked some dark portions of the image (the horsehead and the dark clouds of M78) as I had the impression that they were messing with the result, I am unsure whether it's actually something I should be doing, or perhaps the dark anomaly filter is sufficient?

Usually to colour balance I sample from a star mask with very good results, but this time in several attemts it has always resulted in a strong green cast which I tried to deal with in "Max RGB" mode. I imagine it has do with chromatic aberration and fringes around the stars, but balance was still off after shrinking the mask to exclude them.

There is also some banding, how should I remove it?

Hi,

There is definitely a lot of fast undulating unevenness going on that is hard to discern by people (and algorithms) with 100% certainty.
That gradient is indeed pretty bad, but I cannot be completely sure your flats are perfectly working either (you will know from your other datasets - do they perfectly work there?). The "hole" at 8 o'clock is somewhat suspicious, with a uneven lit border around the edges.
Vertical banding (I would have though would be caught by your bias frames) conspires with the above to make very fast undulating uneven lighting differences, which cannot be removed by modelling a slowly changing gradient.

Dithering would help a lot to mitigate all these issues. Any location-based unevenness, will be spread out, making it easier to reject by the stacker.

With regards to the gradient, I think you did a decent job in Wipe (you could also try the Uncalibrated 1 and Ucalibrtaed 2 presets). Gradient removal will only get you so far; once the low frequency gradients are removed, any remaining unevenness and uncertainty in the signal will remain visible if stretched a lot - Wipe's Temporary diagnostics stretch does precisely that (e.g stretching too far on purpose).

If your dataset exhibits strange behaviour around the edges, you can try a different Gradient Edge Behavior mode (I tried "Grow opposite axis", which makes some assumptions about the nature of the gradient near the edge being of artificial origin rather than celestial).

The vertical banding can be quite effectively mitigated by selecting Adaptive Vertical Bias for the Synthetic Dark/Bias setting.
If I cannot be 100% sure all faint detail in my entire image is real, then, out of an abundance of caution, I would just specify a Region of Interest over the object exhibiting the greatest dynamic range (e.g the object that runs from darkest to lightest the most), which in this case is M42; Jumping to the Color module, I found it came up with a decent default color balance for the nebulosity. The only exception, as you point out, being some of the stars. This could be due to alignment issues or stacker misconfiguration.

The quick and dirty way to address this is to use the Legacy preset, which simply desaturates the highlights like other software does, conveniently "hiding" the issue. You can add Cap Green and Highlight Repair to that for good measure if you wish.

You will end up with something like this; Hope this helps!

That gradient is indeed pretty bad, but I cannot be completely sure your flats are perfectly working either (you will know from your other datasets - do they perfectly work there?). The "hole" at 8 o'clock is somewhat suspicious, with a uneven lit border around the edges.
Vertical banding (I would have though would be caught by your bias frames) conspires with the above to make very fast undulating uneven lighting differences, which cannot be removed by modelling a slowly changing gradient.

I am not sure what's wrong with this data. I have never noticed banding in calibrated images with this camera, could it have to do with the flats? I tried stacking without them and although less evident it looks like it's still there. Here is the flatless stack for reference.

As usual thank you for your help.

EmanueleDM wrote: ↑Sun Apr 11, 2021 9:26 pm I am not sure what's wrong with this data. I have never noticed banding in calibrated images with this camera, could it have to do with the flats? I tried stacking without them and although less evident it looks like it's still there. Here is the flatless stack for reference.

As you get deeper into the data, which proper calibration and dithering will let you do, then minor signal aberrations will start to stand out more.
The banding is visible in the stack without flats as well. Dithering would probably take care of a lot of this.

I tried stacking again using older bias files - from before I began using astroberry - and the banding problem seems to be gone, I must be doing something wrong in EKOS.

However, I am struggling to get rid of the light pollution gradient. Even at 98% aggressiveness it's very strong in the upper part of the image, of course I made sure to crop out stacking artifacts. It's better at 99% and 100%, but I have the impression that it's wiping away some neublosity too. Are my eyes deceiving me or is there something strange going on?

Here is the latest stack: https://drive.google.com/file/d/1VR_MhK ... sp=sharing

This is my best attempt, the upper right corner doesn't quite convince me, maybe I am expecting too much from this data?

Hi,

When dealing with widefields like these, gradients can be expected to undulate (rise, fall off) quite a bit quicker (as you're capturing a lot more of the sky in your image).

Therefore, if they are severe (and they seem fairly severe in this image), you will indeed want to use a higher aggressiveness; normally, the trick is to pick a value that removes faster undulating stuff in your image, but never detail (which undulates very fast from pixel-to-pixel). Generally it very rare for a true light pollution dome/gradient to undulate faster than real celestial detail.

Besides gradients, however, uneven lighting is a different story and it requires a different approach. It is obviously ideally solved by using good flats.

Which brings us to your dataset. In this dataset, there is conspicuous dark patch in the lower left corner, which seems to indicate your flats are not describing the uneven lighting signature in your optical train properly. Meanwhile, the "gradient" is somewhat suspect as well, and seems to be only partially of celestial origin.

When Wipe needs to be really pushed even out an image, there is usually a latent issue with your calibration frames. Indeed, looking at the dataset you posted a couple of weeks back without flat frames, the gradient is entirely absent and only some vignetting remains, leading me to believe the "gradient" (and dark patch, and a faint localised gradient at the top border) is largely introduced by your flat frames and is not of celestial origin.

Your signal is good and deep, and you're really starting to get into a territory where good calibration frames make all the difference in bringing out the faintest detail.

Hope this helps!

EDIT: I just re-read my post a couple of weeks back, mentioning the same flats issues and observations. We're not dealing with a light pollution gradient here. In that case, you will get best results by tweaking one of the "Uncalibrated" presets.

I did not mean to sound obtuse, I should have mentioned that I did take notice of your observation about flat frames from the previous reply of course, I tried stacking with a couple of different sets of flats from different sessions without much success, but I see now indeed that there is no such gradient in the stack without flats, so there must be something wrong with those. I can't figure out why my results would be inconsistent using always the same setup (tablet displaying a white screen in front of the lens). This master flat looks appropriate, yes? https://drive.google.com/file/d/1tQwexr ... sp=sharing

About the dark patch in the lower left, aren't those the dust clouds around M78 that are very dark?

Again, thank you for your useful pointers.

EmanueleDM wrote: ↑Thu Apr 29, 2021 8:08 am I did not mean to sound obtuse, I should have mentioned that I did take notice of your observation about flat frames from the previous reply of course, I tried stacking with a couple of different sets of flats from different sessions without much success, but I see now indeed that there is no such gradient in the stack without flats, so there must be something wrong with those. I can't figure out why my results would be inconsistent using always the same setup (tablet displaying a white screen in front of the lens). This master flat looks appropriate, yes? https://drive.google.com/file/d/1tQwexr ... sp=sharing

About the dark patch in the lower left, aren't those the dust clouds around M78 that are very dark?

Again, thank you for your useful pointers.

Thank you for uploading that. It's always very hard to tell whether a master flat describes the uneven lighting well. It definitely appears to correct for the vignetting well. However, if the gradient disappears when not using flats, then there is of course definitely something not quite right with them.