What went wrong?

[Penguin]

Here's alink to a flickr album showing a problem that I have had with my last 2 attempts at imaging. This one was of M37.

https://flic.kr/s/aHskwn1Cqf

What you have here are

• jpg versions of a selection of the 47 30-second lights and 20 darks (stretched by the Canon 450D)
• the result of stacking the raw images (looks already stretched: the lightening is distinctly odd)
• screenshots of the PixInsight DBE processing stages
• finally the autostretched result of the DBE which looks horrible!

I suspect the stacking but I am happy to be corrected. Any advice is gratefully received!

[wimvb]

I've seen similar patterns, and I suspected something being off in the optical path. I'm not sure though. Have you tried processing just one raw frame with DBE, colour calibration and stretching? It will be noisy, but it might answer the question if it's hardware (= optics) related or software (= processing) related.

BTW, I assume you did the stacking etc in PixInsight as well. You could compare stacking result in PI with stacking result using another software (DSS maybe).

Good luck,

[Waldemar]

can you give us some more details? Scope, camera, reducer/flattener? etc. 

[ShrimpBoatSteve]

Maybe try stacking without the darks to see what you get, or stf stretch a dark to see if something is weird there. The lights look OK. I had this happen to me a few months ago but I can't remember what caused it.

[Penguin]

Thanks for the replies.

This was prime focus through a Skywatcher Explorer 150p on an NEQ5-Pro. Camera was an unmodified Canon EOS 450D. No reducers, flatteners or anything else between camera and scope.

I stacked in DSS (I don't know how to do it in PI) 

To be honest, having spent shedloads on PI I only use it do Crop, DBE and Histogram Transform and I don't even understand how those work! Likewise with DSS, I click the recommended options but have no idea what they mean. What is Sigma Clipping anyway?

I will try the various suggestions (processing a single raw, stacking just the lights etc)

[ShrimpBoatSteve]

I fumbled around in PI at the start too, then I followed these tutorials:

http://www.harrysastroshed.com/pixinsight/pixinsight%20video%20html/Pixinsighthome.html

http://www.lightvortexastronomy.com/tutorials.html

[pyrasanth]

I took a look at the raw images & the stack you have posted & it might be condensation on the CCD window. I had a similar problem with my Atik 11000 camera as the dew heater failed & it produced a similar artefact. If you look at the stacked image you can see an irregular ragged pattern. This can be caused by the edges of the condensation drop beginning to evaporate. I would examine that possibility or look for condensation in the imaging train. 

[Waldemar]

It could be condens on your secondary mirror, too...

Did you take flats?

[Penguin]

No I didn't take flats. I haven't built a light-box or any other way to take flats. Would condensation leave any tell-tale signs that I could check for. Also, this has happened on 2 consecutive observing sessions, 7 days apart. Would condensation do that or is it a transitory problem?

[Waldemar]

Condensation would leave some little spots on your mirrors, or on the filter in front of your LSDR camera if you use any.
condensation is the big spoil of a night's imaging. Check your secondary through the focuser opening, If you see tiny round spots, you found the problem.

i.m.h.o. you need to make flats...

[wimvb]

7 hours ago, Penguin said:

Thanks for the replies.

This was prime focus through a Skywatcher Explorer 150p on an NEQ5-Pro. Camera was an unmodified Canon EOS 450D. No reducers, flatteners or anything else between camera and scope.

I stacked in DSS (I don't know how to do it in PI) 

To be honest, having spent shedloads on PI I only use it do Crop, DBE and Histogram Transform and I don't even understand how those work! Likewise with DSS, I click the recommended options but have no idea what they mean. What is Sigma Clipping anyway?

I will try the various suggestions (processing a single raw, stacking just the lights etc)

I too can recommend Harrys videos (see ShrimpBoatSteves link). The tutorials on LightVortex are excellent too, but maybe not for the absolute beginner.

Sigma clipping, median, average, etc. are concepts in the way individual pixels in light frames are selected or discarded when creating the stacked image. Depending on the number of light sub frames, and their quality, one method of selecting and discarding pixels may work better than another. Usually the default settings in any software used for stacking are good as a starting point. Don't bother too much about it until you've learned the basics. Follow the basic tutorials with your own images, and concentrate on the order of steps (crop, DBE, histogram transform), rather than the details of each step. When you've done this and start to understand what the various steps do, you can then try to understand the inner workings and optimum settings for each step. As they say on Mythbusters: failure is always an option. Luckiliy it is also always an option to scrap results and redo processing.

When you start to understand the workings of PI (or Photoshop and DSS for that matter), you'll discover just how marvelous a tool it is.

And don't forget to ask around in this forum.

 

Good luck

[MattJenko]

Did you check that those auto placed sample points did not include part of whole stars? Also, try using the Normalize button under subtraction as well, as this can help. If subtraction doesn't work (as in this case, try Division).

[Penguin]

I re-did the stacking, telling DSS to only use the lights with a score above 6000 (which was about 5 frames out of the 47 I took) and all the darks. This resulted in a far smoother stacked image and the processing from there was much better. I will have a look through the lights one-by-one later and see if there are any strange ones. I will post the latest result when I get a chance.

I also had the scope outside last night taking M81 and M82 and found a simple way of taking flats: point the scope towards a bright source and put a sheet of A4 printer paper over the end. Seemed to work quite well and when I stacked with those, the initial stacked image was much better and gave PI far less cleaning up to do, so thanks for the prompting and I will always do this in future.

I had used Harry's videos before, at least the ones on cropping, DBE and histogram stretching and those are the workflows I tent to blindly follow. The rest seemed to be more relevant to monochrome images so I tend to skip them. The lightvortex walkthroughs look excellent though and I will definitely read through them.

Thanks for the summary of sigma clipping etc. Why does Crop come first? is it mainly to get rid of mismatches around the edges of the frames where they are misaligned? I get DSS to do this for me so I very rarely crop any further until the end of the PixInsight work when I can see what I have. Is this wrong?

Many thanks for all your suggestions.

[MattJenko]

If you don't crop, the jagged edges of the images after stacking get included in the calculations for background removal, which introduces errors.

[happy-kat]

+1 for cropping it is the first thing I do in StarTools because if you don't remove the edge garbage the image goes very strange so I use a generous crop.

[Penguin]

3 hours ago, MattJenko said:

If you don't crop, the jagged edges of the images after stacking get included in the calculations for background removal, which introduces errors.

Ah, I think that's what I get DSS to do for me. I suppose if I stop it doing that then I can do it myself in PixInsight afterwards and that will give me more control.

[Penguin]

Here's my new stacked and processed result using just the lights with a score over 6000. Can't see what is wrong with the raw subs though.

[Uranium235]

Seems to me that you have not taken any flats - DBE (while being quite handy) is absolutely no replacement for flats.

Take your flats, apply them with Bias and Darks in DSS - stack, then move the master stack over to PI. You should only be using DBE if you have a gradient, but while on the subject - your sampling areas are a bit wrong as you have far too many. All you need is about 5x5, then a few additional sampling points in the area that has the most gradient - try to avoid sampling on stars (find areas of blackness)  and avoid sampling the object you are imaging.... thats why you had a massive back halo around it becuase is mistook your cluster as a background level.

[ollypenrice]

Rob's dead right. You absolutely do not need dozens of sampling points. DBE needs to find an underlying gradient. It is not a tool for correcting bunnies, it works at a large and very general scale. Always look at the model it has created. Stretch it fully (CtrlA) to see if it has created anything which resembles your object. If it has, scrap it. If it has dark patches on it they have probably been caused by a sample point too close to a star.

Whether you process in Ps or PI (or both, as I think is best) it is vital to crop the edges clean as step one, every time. If you don't, it will confuse DBE and the histogram will be incoherent on the left hand side.

Olly

[Uranium235]

Ahh yes I forgot to mention that! Crop the edges or it goes a bit wrong

Its aslo useful to place the samples to a stretched preview, store them as a process and then transfer the samples on to the unstretched version... sampling on unstretched data means you might have selected a background fuzzie in error. But once the selection is stored as a process you can tinker around with it without having to start from scratch.

[Penguin]

Hmm, interesting, thanks. I have been blindly following Harry's tutorials and he put the samples per row up to 15 so I just did the same, moving them away from the object of interest and any bright stars (on an auto-stretched view). I will try with fewer in future. Should I turn off the auto-stretch before applying the DBE?

Flats of course are 'harder' to take but the last time I was out I tried placing a sheet of A4 over the end of the OTA and pointing at a bright light. Seemed to work quite well so I will always do this in future.

[Uranium235]

As long as the auto stretch hasnt made any permanent changes to the image, you can probably leave it on (I use a preview thats been copied into a new document and set my samples on that).

But as far as flats go, if you have a large flatscreen monitor - just use that with notepad open (that will give you a white screen) and point your telescope at that (about 8-12 inches away will do), but you may need a couple of A4 sheets to dim and diffuse the light as it  might be a bit too bright.