Vignetting Over-correction

I have a solution, though it doesn't address the root cause.

Even at 35k ADU, apparently my flats were under-exposed.  Using PixInsight's PixelMath feature, I added a base value of .3 ($T + 0.3) to the Master Flat file created by APP.  Then I calibrated/registered/normalized/integrated using that modified master flat and... it worked!  I'm going to adjust it a bit more (probably to .35 or so) and try again, but .3 is pretty close:

The next time I capture flats, I'll try to go to 40, 45k ADU to see if it helps.  I've never taken flats so bright before; aside from clipping, I worry about pushing the signal into the zone where the sensor isn't linear anymore.  If that doesn't work, PixelMath is an option (though a bit involved; I'd like to see a similar flat adjustment tool in APP - maybe I missed it).

I see the same behavior with my ASI6200MM and a fast telescope: the flats need to be really bright. At F/5 I can use around 15k ADU and at F/4.4 I need to push that to over 30k. I can only imagine that at F/3 and faster the flats need to be even brighter. 

That's interesting about the F speed (this is f/2.22)... going much brighter isn't much of an option on my other full-frame, the ASI6200MC-Pro.  It's much more sensitive in blue, so my flats turn out like this at 32k ADU:

I would love a way in APP to be able to add a fixed value to every pixel, ala Pixel Math in PI.  I will need to do this with every stack I process in APP; I see it in every filter, with both cameras.

Believe it or not, but I also see the same issue with my full frame DSLRs. I had to run the exposure up to get around it.

@imnewhere Thanks for that additional comment. I would have been surprised if it had been an ASI6200 issue only. Full frame cameras easily suffer from a lot of vignetting and more so at low focal ratios.

What F-ratio did you notice this with?

so is there a solution on the way? I constantly have this issue with 6200mm and 2600mc, with f/5 or f/2, no matter what. An adaption subroutine implemented would be nice. I cannot go up with exposure any further

@elgol I have asked Mabula if APP can do anything about this. For now the solution seems to be to make sure to create flats with a high center ADU so the corners can be corrected properly.

@wvreeven 35.000 is quite high I'say ...

@elgol I agree, however if it works it works, doesn't it? Honestly I don't think this is an issue with APP but a general issue with flats where strong vignetting occurs. And like I said, I asked Mabula about this so please give him some time to answer.

I meant I use 35.000 and the issue is still there

@elgol You write that you cannot go up with the exposure time. For both cameras (ASI6200MM and ASI2600MC) 35000 ADU is just over half of what's possible. You should be able to bump this to 40000 and even higher. Why can't you do that?

not if you use gain 100. I could of course try and set it to gain 0 and thus higher, leaving another dark flat round

Hi all,

These new CMOS camera's (even the ones from about 5 years old) are really good in terms of linearity. The old CCD camera's and DSLRs of more than 10 years old, they can suffer a bit with linearity on high ADU values.

If you can only get correct flat-field correction by adding an artificial pedestal to your flats, it means that they are simply still under exposed 😉 So the artificial pedestal can solve that but it is only a workaround and not the best/ideal way to solve the issue.

The best solution is simply to expose the flats longer and use a proper flat-field panel with white light when you are using such fast optics with very strong vignetting.

All the fuss about ideal ADU value for flats is a bit nonsense (if you would ask me) with todays camera's. Simply make sure that the histogram is not clipped on the left or right side. Aiming for higher ADU values of the peak in the histogram will simply give you better flats (better Signal to Noise Ratio) with the camera's that you can buy these days 😉 because the are very linear in response on the entire ADU range.

So I would easily aim for ADU values of 50.000 in 16bits if the histogram does not clip on the right side 😉 (and left side) for all channels (mono or R,G,B).

Oh and personally, I alwasys shoot flats at the camera's highest dynamic range, so at gain 0, the benefit is that the quality again will be higher of your flats and you can still get good exposure times even with such fast focal ratios. Shooting flats with exposure times less than 0,1 seconds can give rise to complications for sure, so try to prevent that...

Mabula

Hi all,

In addition, in APP 1.082 you can easily add an artifical pedestal with the Batch Modify tool to all of your flats.

In APP 1.083 you will be able to do this with a new tool : Batch Add/Multiply Tool 🙂

Mabula

Posted by: @elgol

not if you use gain 100. I could of course try and set it to gain 0 and thus higher, leaving another dark flat round

Can you explain that please? As long as the histogram is not clipped on the right, you can expose longer.

you said it, clipping. anyway, I did a 50.000 ADU test, same thing: over correction it seems ...

@elgol Hmm yes there is some overcorrection there. Is there a difference with when using the 35000 ADU flat? Is the amount of overcorrection the same, or less, or perhaps even more? And did you take the flats at gain 100 in the 50000 ADU case?

wanted to push it so yes with gain 100. and clipped. it got less but not enough it seems. hope the announced pedestrial in 1.083 will come soon hahaha. I am aware of this discussions. still ....

I have the same problem too, since i changed my focal ratio from 5,1 (apex starizona 0,65X reducer) to f7,8 (no reducer or flattener).

Doublet Apo 125 at its full native focal length of 975 mm at f7,8, with no reducer/flattener: small Asi533 sensor.

As you can see in my attached screenshot, where i pumped up stretching and saturarion just for this purpose to show better artifacts, i tried all different ADU flats with 2s exposure, gain 0 and 0,5 second exposure as well: practically it doesn't change anything.

The strange fact is that as soon as i apply the "remove light pollution" tool, almost every sort of vignetting evidence disappears... 🙄 😲

@ippiu Yes, very familiar this. But applying the very good LP tool still should not be necessary especially when thinking about some objects where you just want the same procedure, like different sezzions, mosaics etc. 

what is new to me and it looks like for some here, too, is that there seems to be more behind this than a maloperation of the user

Hello everyone, I also had such problems in the past. What helped me was a post by Adam Block on YouTube. The solution to the problem was that the DARK frames were too old. Sensors can change their behavior within a few months. Then the whole calibration no longer works. One of the effects is then visible as overcorrection of the flats. The exposure times of the flats can also play a role. My ASI 294 no longer behaves linearly when the exposure time is below 2s. The ADU's do not play a role. There can also be an overcorrection. There are articles about this in the Cloudynights forum. Maybe this helps.

Posted by: @minusman

Hello everyone, I also had such problems in the past. What helped me was a post by Adam Block on YouTube. The solution to the problem was that the DARK frames were too old. 

But my Darks are just 6 month older, they have been taken at the beginning of October 2020, so not so old as i would expect for this type of calibration files...

Anyway, I will give it a try this night for sure, due to the fact it's cloudy outside...

Interesting possible solution: thanks @minusman

There is a way to check it. I assume that the flat darks are more current than the DARK frames. Compare the median value in pixinsight of the two images. For the DARK frames it must be a bit higher than the flat darks. If it is the other way around, there is a problem. Adam Block was able to identify the problem in the video.

Posted by: @minusman

There is a way to check it. I assume that the flat darks are more current than the DARK frames. Compare the median value in pixinsight of the two images. For the DARK frames it must be a bit higher than the flat darks. If it is the other way around, there is a problem. Adam Block was able to identify the problem in the video.

I don't have PI so i used APP to see median values.

It seems pretty much the same value for both...

What do you think?

If i use Fits Liberator, i have the same results: median value is 2797 for dark-flat vs 2798 for 300 s darks

I would better compare a DARK and flatdark. Not the master files. You can also display the values in the imaging software. Mostly the mean is used there. It is almost the same as the median. Like in SGP or APT, Nina.

What is also important, do not use bias frames with CMOS cameras! Because of possible non-linearity at very short exposure times. What has also made problems with me true the original USB cable from ZWO. Had a bad shielding. I replaced it with a high quality one.

Hi @elgol and @ippiu,

There is also the possibility that the methods that you use just don't suffice to make good flats. Have you considered this option? Or put differently, which methods have you tried to create flat frames? I know for sure that flats can fail if you shoot them during daylight or if you make sky-flats or t-shirt flats.

You can call it overcorrection, but the correction might be perfect and the problem is in how you make your flats. The corners could simply be under-exposed in the method that you use to create the flats.

I know for a fact that many others shooting at focal ratio's below 3 don't have your issue (the lower the f-ratio, the harder it is to create proper flats), so I think it is likely that the cause of the problem is in your data and not in the APP routines. A very small light leak in your optical train can cause your problem easily as well. I have experienced this myself with a bad Takahashi adapter...

I am not ruling out the problem is in APP as well. If you want me to have a look at your data, please share

• 5x lights
• 5x of any calibration frame type that you are applying

Please upload it here:

https://upload.astropixelprocessor.com/

username: upload5

password: upload5

Please make a folder with your name and let us know once uploaded 😉 Then we will investigate.

Mabula

First of all thanks @mabula-admin for your patience and kindness.

I have been taking flats for 2 years with following method that never created me any sort of problems: i usually use a led Lacerta panel, same gain and temperature as lights, 2 seconds exposure, around 30.000 ADU on Asiair Pro. 50 flats calibrated with 2 s exposure dark-flats (same temperature and gain as flats and lights as well). Then i'm used to use Bad pixel map previously created with darks (in this case, 35 darks, same temp and exposure as lights).

Anyway, i just uploaded everything you need. 

https://upload.astropixelprocessor.com/index.cgi?launchApp=SYNO.SDS.App.FileStation3.Instance&launchParam=openfile%3D%252FAstroPixelProcessorUpload%252Fippiu_flat_overcorrection%252F

Thanks

It looks like the flats are still underexposed.
Especially in the red and blue channel. The green channel has 30000 ADU. See picture.

Maybe aim for 50000 ADU in ASIair so that red and blue also have a better signal.