What subframes do you keep / How many subframes?

Hi Rob, 

Thank you for bringing up this topic. It will be very interesting for me too to learn from more experienced people here. 

In my case, first of all I do the blinking for obvious issues, like trees and heavy clouds etc. Then I evaluate my subs for eccentricity, star quantity and SNR (in this sequence) for each separate channel. Eccentricity lets me discard all the frames, where the tracking was not good, low star quantity will mean poor visibility or thin clouds, which I did not notice in blinking. SNR will let me select the best of what is left. 

CS 

D

Hi Robert,

Attached is an image I did of NGC 6960 12 years ago with equipment I haven’t used in 9 years. As for what I do, I pretty much do just what David Nozadze dose, I don’t do the math or anything that pre-Judges my work. I open everything using CCDStack and Blink all the images; I look at the stars for bloating, noise, and any guiding issues that is very rare. Keep in mind that this image was done with a Tak E-210 f/3 and an SBIG-ST-10XME. But to your question, what to keep and how many? This image is just 4 each x900s Ha-Olll, bin 1x1, and 3 x 240s each RGB  bin 2x2. All pre-processing at the time was with Maxlm-dl, and then moved to Photoshop. If I were to image this today in NB, I would max out at 5 hours each, Ha, Sll, Olll, at 900s or 1200s, I will rarely ever do more then 5 hrs in NB, and maybe some short 120s RGB’s for the stars. For an LRGB, I would probably do maybe 2/3 hours each, maybe 12x600s for L, and 12x900s for RGB, and again, I would just open all subs and toss anything I didn’t like. I get it if the sky condition will not let you do long exposures, but the process would be the same with lots of shorter exposures.        

I have found that unless the frames are severely washed out, then every photon counts.  I am trying to get the SGPro guys to include "Sky quality" as a "Key" in the filename, since I have an Eagle4 Pro with a Sky Quality camera, I could then filter the frames below a certain value very easily.

If after i create a stack I notice any oddities, I will then do what others have done and "Blink" the images in PixInsight to find the offending image(s).  I have lots of storage so that is not an issue, and since I use an 18 Core System, re-stacking the images is not really an issue either.

In the past I used to go off the 80/20 rule, but realised that the 20% of images I was discarding were actually usable.

The other side of things is with modern CMOS cameras like the ASI6200MM Pro which I have are fantastic at shorter exposure in high gain mode, which means you can actually take more image frames and boost that SNR in the final stack.  I make it a rule that with 60S exposures I do no less than 151 frames in each filter, and for 150S exposures, no less than 101 frames in each filter, why so short exposures you might ask?  Well I am imaging at F3.2 in Bortle4/5 and SGPro recommends those exposure times based on frame analysis.  So I generally stick to 60S with LRGB and 150S on NB, unless it's a weird target like M31 where I will also do much lower exposure times for the core.

So unless your frames are really bad, don't discard them, they will get weighted on quality in the stack anyway, so an image with a lower quality score will have less of a bearing on the final stack than an image with a higher quality score.

Just my opinion and my experience.

Simon

Generally, I toss anything that has tree bits, clouds, or trailed stars. Everything else I keep, unless something appears after integration that warrants an investigation. That usually doesn't happen though.

Very interesting responses.  So what I am hearing is that low quality frames are worth keeping and processing.  I guess I assumed that if the image quality was low, it would impact the completed image.

I keep everything, except obviously flawed ones with huuuge startrails due to guiding failing due to clouds, trees.
I very very rarely have subs which have not pinpoint stars, and if it's only slightly eggy, I integrate them, APP algorithm takes good care of that.
Satellite trails, planes -> I integrate it. Slightly lighter background - I integrate it.
My FWHM cutoff in APP is 4", everything bigger gets tossed away because I figured for me personally, stuff like satellites, lighter backgrounds, gradients, and even slightly (we are talking slightly here - only recognizable if you zoom in) - get taken care of either while stacking or in postprocessing quiet easily. Big, out of focus stars, or bloated stars due to bad seeing are imo way harder to fix in postprocessing - and while it's doable, it takes time, needs a good star mask, and you still risk getting artifacts. It's easier if you have an object, that stands isolated in the starfield, like a galaxy, or a planetary nebula, than a nebula, that takes up the whole sensor - you risk getting artifacts in your nebula, and you just don't want it - easier to fix, again, if the stars only stand on black background.
So the more nebulosity, the stricter I am with FWHM as nearly my only criteria to toss out subs, and the lesser nebulosity, the less rigid I get - if it's only a few - let's say 10 out of 200 120s subs that are 4" and over, and I went for a galaxy, I just toss them. If it's half of the 200 subs that are like this, seeing was bad, and I just have to deal with it in post processing, can be painful though.