How light is too light for light frames?

[imakebeer]

N.B. I'm using Siril for stacking etc

I left my gear out last night, taking some pics of M51 (fingers crossed!). I set it to take lots of light frames knowing full well that dawn would soon come and the later ones would end up being scrapped. But what about those last few frames where you can see the day light starting to creep in? How can I decide which frames to stack and which ones to exclude? Or do I just throw them all in and then will Siril automatically decide what it can/can't use? Three examples (unprocessed JPEGS!) below from 0300, 0330 & 0400hrs:

Thanks in advance 👍🔭😉

0300hrs

 

0330hrs

 

0400hrs

 

[vlaiv]

Unless you have very sophisticated algorithms to deal with that - mixing of different SNR subs can be rather damaging. You can end up with worse result compared to not using those subs in the first place.

It behaves pretty much as light pollution, and if possible - minimize it. Most people image only during full night and avoid any sort of twilight. I know it is impossible to do so in summer time in northern latitudes (no true night, only twilight), and for that reason some imagers skip summer months all together - or use them for practice and trying out things - without having high expectations of good results.

[wesdon1]

50 minutes ago, vlaiv said:

Unless you have very sophisticated algorithms to deal with that - mixing of different SNR subs can be rather damaging. You can end up with worse result compared to not using those subs in the first place.

It behaves pretty much as light pollution, and if possible - minimize it. Most people image only during full night and avoid any sort of twilight. I know it is impossible to do so in summer time in northern latitudes (no true night, only twilight), and for that reason some imagers skip summer months all together - or use them for practice and trying out things - without having high expectations of good results.

Thank You for answering this question to @imakebeer. I had a ton of subs that looked almost as bad as his last sub, the "bluey blown out sub" above, and I was only doing 152sec subs, ISO1600. I had taken them around 12midnight to 2am on 15/05/23 and although still "dark" they were blown out, and I was wondering, was it because at the moment in UK, it's not truly "dark" due to such short nights in summer and the fact the Sun never dips very low below horizon. So again thank you for clearing this up! 

Wes, Bortle 7 

Edited May 21, 2023 by wesdon1
missed a bit

[imakebeer]

Thanks @vlaiv 👍 Explaining it in terms of light pollution and SNR it makes perfect sense 😁

[vlaiv]

Here is a handy website that calculates astronomical darkness:

https://www.timeanddate.com/sun/uk/liverpool

(I selected Liverpool as example - but you can easily choose different location).

You can switch the view to Day/Night length and you'll get handy graph showing you period without true darkness:

you can also check the exact times on any given date.

This helps you to plan your sessions as you can easily check if there will be enough true darkness for complete session on a given date.

(by the way - that website has a bunch of useful calculators for astronomy, it's worth checking out other features as well).

[ONIKKINEN]

I would keep only the first image and others with comparable levels.

Have a look at the different graphs siril draws in the Plot tab after registration and deselect the images that are significantly worse than the best of the night. Maybe choose a cut off point of twice the background signal compared to the darkest subs of the night and stack only those. You should still use a weighing method for stacking, such as weighed fwhm in Siril.

What the graph will look like:

This was my last image of the year in late April, you can see that i started imaging before darkness and it never truly settled to proper darkness and then started to rise again. Easy to select only the lights you want to work with here and continue with those. Then in the stacking tab you should probably select either wFWHM or #stars as the weighing method for stacking. Number of stars is a simple but effective method of determining the quality of an image because as the background brightens you lose the dimmest stars, or if guiding had an issue during that sub you also lose the dimmest stars and so on. The weighed FWHM method takes into consideration the number of stars and their size and so will give highest weights for images which have lots of tight round stars and the worst weights for images with a low number of oblong stars.

[imakebeer]

@vlaiv @ONIKKINEN many thanks for the comments, much appreciated 👍

@vlaiv Exactly as you say, it's not the best time of year to be doing it, I knew this at the outset but I couldn't let another opportunity go to waste - if nothing else it's more days to practice with, and at first glance looks not too bad at all by my standards.

@ONIKKINEN At the moment I'm still just using the pre-supplied scripts in Siril to do the stacking etc. , along with one of the basic tutorials for some initial post-processing before moving over to GIMP - so I haven't discovered the features you mention yet. But I'm aware there's a lot more to Siril than what I'm using so far and I'd like to dig deeper - something to keep me out of mischief in the lighter months perhaps!

[ONIKKINEN]

You can open the sequence that the OSC_preprocessing script creates to do manual stacking and do the decisions yourself, along with getting to use the plot tab with some extra work. For the subject of this thread it could be very beneficial to do so.

After running the script that spits out the stacked image (ignore it for now), change your Siril home folder to the "Process" folder that the stacking script creates. Then in the sequence tab search sequence, you should see several and 2 of these are what you want to do something about. The pp_lights sequence is your preprocessed (calibrated) lights, these are not registered yet so nothing in the plot tab to be seen. The r_pp_lights sequence is your registered lights sequence. Open this to gain access to the plot tab.

Although i recommend you start with the pp_lights sequence first so that you can reject the bad stuff before registering them. I'll go over shortly how to do the registering/stacking process manually. The UI is a bit intimidating but in the end its quite simple.

Starting with the pp_lights sequence opened, go to the registration tab and select the "Two-Pass Global Star Alignment (deep-sky)" method for registration. Leave everything else at default. This method is the best because it goes over all of your subs and then selects the best possible frame as the reference frame, unlike other methods that use the first sub of the sequence which is often not ideal (the script does this too). Siril will compute registration data at this point, but not apply it yet so that you can inspect the data before actually applying the star alignment.

At this point, check the Plot tab. You can select from the bottom drop down menus what you want the graphs to represent. Go over the "background" graph at least and deselect the bad stuff by dragging a selection box over the graph and right clicking. You should also go over the FWHM and roundness graphs too so that you can deselect out of focus and trailed images. The wFWHM tab is inverted where top entries in the graph are bad and bottom entries are good, all other tabs are "normal".

After going over the data you have to actually apply the registration before stacking. Go back to the Registration tab and find the "Apply existing Registration" option and choose that. At this point make sure the image selection is chosen as "Register selected images only". Then click Go register and Siril will star align only the subs you left in the Plot tab and reject the rest. After this point you can stack the registered subs in the Stacking tab. Below are parameters that i would recommend, which i believe are default or at least close to.

The awfully named rejection method i have here is excellent for large datasets but it can take a while to process. If your PC is struggling you could choose Winsorized sigma clipping with 3.0 for both low and high sigma (usually ok). If you click the Create rejection maps option you will find a reject high and reject low image in the stacking folder after the process completes. Here you see all the satellite trails and other anomalies that the algorithm has chosen to get rid of.

You dont have to select anything for the "image rejection" tab here, because you already rejected the bad images manually from the Plot tab.

[imakebeer]

Thanks very much for those suggestions @ONIKKINEN .

Hopefully I can find some time one evening this week to give them a go 🤞