A few years ago I bought a 12" Orion XT12G Goto Dobsonian for observation. It’s been great but then I foolishly put a camera in it and lucky imaged the moon, then the planets, then I bought a deep sky camera to shoot deep sky targets, and one thing led to another. It’s been a blast, and I’m happy with my pictures but it made me wonder:
If I stack with SIRIL I use the OSC processing scripts. However, I’m assuming the default settings are “good general settings” tuned for the typical user. That typical user is probably shooting 60-300 second exposures and from a few dozen to a few hundred.
I’m shooting (at F3.7 and F4.9) 4-8 second exposures and typically 1000-3000. “Gosh” you say, “That must be fun to stack”. Well, let’s just say I have a fast computer and still, I’ve learned to postpone immediate gratification. Thank goodness SIRIL is so darn fast, wowzers! But when it’s all said and done, 1-3 hours later I do get a good image.
But this makes me wonder, are there some pretty likely or obvious setting changes I should use in SIRIL based on stacking, say 1500 * 6 second exposures, versus 75 * 120 second exposures? My exposures are certainly noisier, so that probably doesn’t make it easy for SIRIL to make some critical measurements, but I’m much less sensitive to small individual errors that get averaged out. So, my situation is just different, in many ways I expect. One thing I do get is a lot of field rotation over a session, so the lights are certainly rotated heavily against the reference frame and I appreciate SIRIL stacking into a larger output frame with the “-framing=max” option that I updated my script to use.
So, any thoughts on if there are a few settings that would likely improve my results based on tons of short, noisy, exposures?
Thanks!
Steven
PS: Just in case you are wondering if I can get a decent image with such short exposures, here is the Eastern Veil nebula shot with an Optolong L-Ultimate with 6 second exposures for about 3 hours, stacked with SIRIL (yay!):
Thank you. I did not do the post processing in SIRIL. I am a recent adopter of SIRIL so I’m just learning it. Most of my experience is with Affinity Photo and Astro Pixel Processor, and since I wanted to use BlurXterminator and that only works in PixInsight, so my process is currently crazy. So after stacking in SIRIL I did BlurX in PI, then I used APP for the basic stretch, background correction (and normally star based color correction but not on this image), then in Affinity I did everything else including StarX and NoiseX, fine tuning the stretch, etc… It’ll take some time for me to learn SIRILs processing tools… so much to keep in my brain at once! Lol…
Hello, you are asking the correct questions. The more you shorten the exposures, the more you need to filter images out from processing, keeping only the best, if you want to have the upside of this technique.
4-8s is not very short though, and you may still have the disadvantages of long exposures with those of many files to process, although if you have field rotation you cannot really go higher. The limitations will be the sensor noise and seeing enough stars. If you have a low read-out noise camera, you can probably go as low as a second with f/3.7, as long as you see enough stars in images to register them.
You can filter out images to keep the 30% best for example, to see the difference in quality compared to the default script that keeps all of them. It should be significant, but then you get more noise in images because you combine less exposures… So you really need thousands of images in input;
Thanks for thinking about the topic. Yes, I do have a very low read noise camera and in broadband, even with 4-8 second exposures, read noise is below 2% of total stack noise. With narrowband, it’s higher at about 10%, so not fantastic, but not a big issue.
However, in this case, I’m not shooting short exposures for the advantages of lucky imaging, I’m shooting because I have to. But still, you are right, I need to cull a percentage of my lights and that’s typically 5-20% depending on various factors, with the amount of wind being the primary factor. I often cull based on NINA’s reported FWHM which I have added to the beginning of the filename so I can easily short and dispose of the worst images just from the sorted file names. Then I examine the next worst, and draw a cutoff, and I will also cut all those that have a very bright background or low star count, usually from some passing high clouds. Then I simply stack the remaining.
This works well and I’ve tried more sophisticated sorting and culling based on many criteria, but the results are pretty much the same as simple culling.
So what I’m wondering, if once I do that are there settings that would improve the stacking of, say, 1800 * 6 second exposures because each is somewhat noisier than a 60 or 300 second exposures, but there are a ton so small exceptions will simply get averaged out. I mean, one sub-optimal pixel out of 1800 isn’t going to make much difference, so perhaps I need to be less aggressive in rejecting pixels. I’m not sure.
For rejection, the last two algorithms of the list Stacking — Siril 1.2.0 documentation are more adapted to larger datasets.
Make sure you enable fast normalization too if you use normalization.
If you shoot shorter exposures, you could also acquire in 8 bits, because the wells wouldn’t fill anyway, and regain the dynamic by using sum stacking instead of the default mean with rejection stacking. This is much faster too. For 4-8 seconds, I’m not sure it’s the correct way to go.
Funny you found the M63 example as I’m processing that right now… not 18,000 frames though, more like only 2000. Thanks for finding that, I’ll examine his stacking parameters.
Somewhat related, here is my version of lucky imaging, but cheated and did 0.5 sec exposures for the core and longer exposures for the outer nebulosity and merged the two. Also, for the core, I think I used autostakkert, but I can’t 100% recall as it was almost a year ago. But in the future, I think I’ll be using SIRIL, I just didn’t know about SIRIL at the time but I think SIRIL would be more robust as it registers/aligns based on stars which are more visible than the target features.
Interesting about 8 bit. Just like I do with planetary, lunar, solar lucky imaging. As it turns out, I usually increase the gain to where the well has about 1000-4000 levels, so unfortunately (or fortunately depending on your point of view), 8-bits wouldn’t be enough without significant quantification.
If I ever shoot even shorter exposures as super high gain (ex: lucky deep sky imaging) and drop the well down to about 256 levels, I may try 8-bit just as an experiment. certainly when you stack/average many thousands of frames, you regenerate a very level-rich image, much like what is done in planetary.
