Jump to content

sgl_imaging_challenge_2021_annual.thumb.jpg.3fc34f695a81b16210333189a3162ac7.jpg

AstroArt5 and Flats


samtheeagle
 Share

Recommended Posts

Me again, with yet another AA5 question... I've had this software since my birthday in April, and I'm still not having much luck with it. It looks awesome, with so many features that I need, but I am really struggling to get good ANY images from it. I'm determined to figure it out, it's far too expensive to put to the side and go back to DeppSkyStacker. Mind you, I've given up on getting good guiding from AA5, I've reverted to PHD because it actually works. Anyway, I digress...

So I got about an hours worth of data on NGC6960 last night, and I've been attempting to process it with AA5. I've been following the advice of Olly given in one of my earlier posts about my woes with this software: http://stargazerslou...p/#entry1441245 and I'm getting a number of issues, but the main one for this post is the application of flats. They just don't seem to work for me, look it these pics...

post-3645-0-22391700-1349640690_thumb.jp post-3645-0-75660200-1349640705_thumb.jp

The first is a hyper-stretched image showing the vignetting with no flat applied. The second shows the same image, hyper-stretched again, with a flat. All it seems to have done is invert the vignetting, not remove it. What the hell?

I'm sure it must be me, AA5 is very well regarded in the community, but me and it just can't get on right now. Can somebody offer us some relationship counselling?

Link to comment
Share on other sites

  • Replies 33
  • Created
  • Last Reply

Top Posters In This Topic

I have similar results unless I nail the exposure time for the flats. Seems to differ depending on the sky quality I had for the lights, so normally end up doing 3 sets with different exposure times. Then it's pretty quick to do a test stack with each master flat to see which one work the best. You would think that since it's a simple division that it doesn't matter - it's all relative - but I find that it does. That could be due to skyglow and internal reflections in my RC though for the lights.

I have my Geoptic flat panel set quite low, so that a flat takes 4-8 seconds.

And I debayer, calibrate and stack in one go. The little status window flickers so quickly that I can't determine if AA5 debayers after or before calibration, so that remains a mystery to me. But it does work.

Give that a go?

Edited by Jessun
Link to comment
Share on other sites

Cheers Jesper, I'll try some fixed exposure lengths then. I realized today that I'd made a bit of a faux pas in that I had the camera dial set to A-Dep instead of AV, so perhaps that's upset things? But looking at the histogram of the flats the peak is bang in the centre, which is where I think it should be... I'll take a series of flats and see what sort of exposure does the trick, I'm using an EL panel, so hopefully whatever value I hit upon should be good for a while until the panel starts to fade. I guess the point at which the histogram peaks is the real value I need to record...

Link to comment
Share on other sites

Right then, so I tried to get some more flats last night, one of the VERY handy features of having an obsy where you leave your kit setup :) I set the dial to AV mode this time and it shifted the histogram peak from the centre to about 2/3 of the way along. As a quick test I divided one of my subs by this new flat and it looked like it had worked well. But when I came to process the whole lot a bit later on I once again found myself with the vignetting inverted as before :( So I can only assume that it's something to do with how I'm using the Pre-Processing tool in AA5. I know that the flats work ok because they do the business when thrown at DSS.

I'm trying to follow the steps Olly described in the post linked above, but I wonder if anyone can flesh out those steps for me? For example when creating the master bias which combine method should I use?

Link to comment
Share on other sites

Going back to the first thread, the step you missed was image-shift by 0.5 on both axes to debayer the flats. That's sorted.

Right, your flats are now over correcting. On two of my setups (TEC140 and FSQ85) with Atik 4000s mono and OSC, I do not get this. The flats work perfectly. However, now for the strange bit; on Yves' instrument (ODK14, SX H36 mono, Nebulosity) I get exactly the same over correction as you. We are working on it and will post as soon as we have anything useful to say. Obviously I use exactly the same routine and even the same panel on occasion. I have tried exposing the flats to below 1/3 instead of 1/2 and it made no difference at all. Maddening.

Olly

Link to comment
Share on other sites

I assumed it was something to do with my QHY8L camera at the time.

I came up with a work-around in the end. I found that if I added an offset to all the flats I could get them too work. If the offset was too small the flats over-corrected and if the offset was too large they under-corrected. But with a bit of trial and error it was possible to get the flats to work. As I said I put this down to my camera at the time and not AA5. In fact I think a friend ran my subs through Maxim and had the same issue.

Cheers,

Chris

Link to comment
Share on other sites

Just to run through what I have already tried, none of which made a gnat's crotchet of difference;

-exposing the flats to just 8000 counts.

-exposing them to 20,000 and 40,000 as well.

-applying bias as darks for flats. (my usual routine.)

-shooting dedicated darks-for-flats at the same exposures and temperature.

All to no avail. Yves did something in PixelMath in Pixinsight and I'm preparing new flats for a further exploration of this lead.

Olly

PS Chris's offset post came in while I was on the phone. Could you give a fuller explanation, Chris?

Edited by ollypenrice
Link to comment
Share on other sites

PS Chris's offset post came in while I was on the phone. Could you give a fuller explanation, Chris?

It was an idea I came up with while studying the maths involved with applying flats. I simply used AA5 to add a constant value to every pixel in each flat (say 10000). The idea is that the more offset you add, the more insignificant the gradient variation in your flats is and the less effect it has in correcting the final image. If you get the right offset value then the flats seem to work.

Not a solution as such, but it did allow me to rescue some data.

Link to comment
Share on other sites

Here's Fabio's reply to my post on the AA5 support forums:

Hello, that is a classical problem that happens when the flat field is not dark-corrected (or at least biasvalue-corrected), the division gives a wrong result:

instead of: corrected image = image / flat

you get: corrected image = image / (flat + offset)

or: (image + offset1) / (image + offset2)

Additional info can be found in the chapter "image calibration".

obviously also the image must be dark corrected. If you don't want to correct the dark frames (recommended for low noise cameras) then at least you must subtract a constant equal to the average value of the dark frame.

P.S.

For non-scientific images "remove gradient" is usually easier than using flat fields.

He is correct that in my case I'm not using darks, but I believe Olly's process here is bias correcting the flats, which he implies should also do the trick... So I'm not sure that I'm any the wiser.

Link to comment
Share on other sites

Most of us use a master bias to calibrate our flats. It is easier than making dedicated darks for flats and should be insignificantly different. Anyway I have tried uncalibrated flats, bias calibrated flats and dark calibrated flats and with the Starlight H36/ODK14 they ALL give over correction. On my other setups they work. I'm shooting yet another set of flats right now to see if I can get anywhere.

I'm afraid that using 'remove gradient' or DBE in Pixinsight simply is not as effective as applying good flats and that's that. How does a software programme distinguish a dust bunny from a patch of nebulosity. Obviously it can't.

I too am still none the wiser.

Olly

Edited by ollypenrice
Link to comment
Share on other sites

I've just been playing with the same data, and decided to throw caution to the wind and used the pre-processing tool to do everything in one pass. I tell you what, it looks like it's applied the flats correctly doing it that way! Here's what I did...

I put the flat frames and the lights frames into their respective boxes

I put the master bias frame as the flat dark frame and the dark frame

I did a sigma combine, translate + rotate alignment, normalize background, and reject bad images with star elongation 1.2

I applied hot pixels filters single and group after dark/flat correction

Here's the hyper-stretched result, which has a boat load of sensor noise in the lower right corner, but otherwise seem to be much better corrected.

post-3645-0-97570100-1349805241_thumb.jp

Link to comment
Share on other sites

Hi

If the vinetting with your set up is quite severe , the difficulty of flats become even harder and make the maths very fussy--something that seems to a problem with reflectors and thier illumination profile.

I use a master bias as a dark which works very well most of the time, but a dark has some dark current over a bias i.e my bias of my h35 is around 1200 adu but a dark might be 1250 ( I ignore all other noise )

So if I add 50adu to my bias with pixel maths ( making a fake dark ) I can get better correction.

Your figures may be different and this is similar to the ofset mentioned above but it is the dark that is causing the problem , it only takes 20 or 30 adu to cause these effects.

Now my secret is out , so don't tell everyone :p

Kind regards Harry

Link to comment
Share on other sites

This sounds interesting Harry. Not my field so I'll put Yves onto the case. Is it the illumination profile of the reflector, though, or is it to do with the H36? The ODK14 has an astonishingly even illumination, more even than my Tak FSQ85.

Edit, would a good test of a flat be to apply a master flat to a set of flats which are then stacked to see if the result were a truly flat image? I'd apply a master bias as a dark in this test.

Olly

Edited by ollypenrice
Link to comment
Share on other sites

Hi

I only get the same problem as you with my old newt , the refractors do not cause a problem :Envy: .

As for your flat idea -- divide a rubbish flat by a equally rubbish flat and you get a even illuminated image so I do not think this would help

Harry

Link to comment
Share on other sites

Having played with the image I got from the one pass pre-processing last night I have found that I've lost a load of colour information :( The flats seem to have removed the vignetting, and the dust bunnies, which is nice, but the colour seems to be gone too... One step forward, another back :S

Here are 3 pictures of my processing efforts thus far, the first with all the data thrown into DSS and just letting it do it's thing, the second from AA5 that's just a stack of lights, and the third one is the other AA5 one pass process which includes the flats. It's interesting to see how DSS gives me a red/brown output, but AA5 is much more blue... But the third image is just a wash out next to the others.

post-3645-0-59249600-1349854646_thumb.jppost-3645-0-81394800-1349854654_thumb.jppost-3645-0-76572200-1349854661_thumb.jp

Link to comment
Share on other sites

What you have found here is that the colour information has been damaged because of the one-pass approach. I find you need to calibrate (with darks, flats and 'bias as darks for flats' first and only then can you stack and debayer in a second pass. I hope Harry will come back on this because I think I remember him saying AA5 could do it in one pass (or did I dream it?). But when I tried it I found I got the same problems as you did.

I'm still on the 2 pass approach.

It looks like throwing an offset onto the bias when using them to calibrate flats is the best line to persue. A little project for today...

Olly

Link to comment
Share on other sites

Update. It gets worse, and better.

I shot all new flats, darks and bias last night. The flats I exposed to 18000 counts. Lowish. The bias were at the same temp.

I then ran my usual routine which never works with this setup.

Flats in the flats box, new master bias as darks for flats. New darks in the darks box. Lights in the Lights box.

The result was... perfect.

I am totally stuck for an explanation.

Olly

Link to comment
Share on other sites

Riiiiiight then. Hmmm.

So I think the key thing I've discovered so far is that AA5 doesn't like me skipping darks. Which to be fair I should be capturing to do the job properly, but always seem like a chore. I'll have a play with some old darks of the same exposure and see what happens, as well as the bias offset work around...

Link to comment
Share on other sites

This thread, interesting as it is is getting a little bit confusing :wacko: ... What works and why? I suppose that's some of the dark arts that makes this hobby interesting...

I may just be lucky to be able to throw all in the mix in one go. I capture in Nebulosity - and do the master calibration frames in Nebulosity too to make them match the lights. I don't use darks, but rather a bias for both lights and flats - and it works in one click. I only adjust exposure time to match the lights. Why this works is a mystery.

When your flats don't work does that mean that dust bunnies are 'inverted' too, ie overcorrected? In theory they should be as a function of something going wrong altogether...

Olly,

Going back to the first thread, the step you missed was image-shift by 0.5 on both axes to debayer the flats.

What does this step do?

And I can't remember if you use AA4 or 5, but in recommending the two-step approach, did you have time to test the same data with the two-step vs a one-step run? A massive stretch should reveal color loss straight away. I do however tick the 'Normalize background under 'Options', I wonder if that is how I cheat my way around this problem?

Link to comment
Share on other sites

Hi

AA5 does not require the shift 0.5 trick as long as you select debayer in preprocessing it will do it for you :)

As for the 3rd version of the image I cant see any colour did you select debayer :kiss: , also most packages create slightley different colour and they attempt to correct for any colour inballance using the neutralize background trick

with varying amount of success , hence why I use pixinsight as it gives me more control on colour

Kind regards Harry

Link to comment
Share on other sites

@harry - Thanks for the clarification :)

@jessun - Sorry it's getting a bit confusing, but that's basically the crux of the problem. We're finding that AA5 is behaving quite strangely, with different results being given when following a "standard" process. It seems that it's a bit sensitive to certain data, sometimes flats will work as expected, and others, for whatever reason, will cause an over-correction.

Link to comment
Share on other sites

Yeah I know, it is confusing. Do your dust bunnies - if you have any - come out inverted too? Everytime I had flats problems that looked like your over corrected result the image also had the tell tale sign of dust bunnies looking like craters. And changing the flats exposure time cured it, but the works behind that I can only speculate about... I'm not quite the mathematician....

I use bias as flat darks btw.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue. By using this site, you agree to our Terms of Use.