Win a Pentax Limited Lens: Raffle Now Live!

I got motivated by DrawsACircle's posts and went back and gave another try at processing my M31 stack of shots from this summer. The total exposure that was included in the final stack was a bit over an hour all done with 20 second shots keeping only the best 60%. All were taken with astrotracer on my k-3 using my 300mm Sigma f/4 APO Macro lens at f/4.5 and ISO 1600. The last few times I ended up with a very blue image and I did this time as well. I am getting a fair amount of detail. The gradient is coming from not adjusting the framing as often as I should so I know about that issue. My question is about the blue cast and if that is a result of my processing (I don't play with individual channels but wondering if it is coming from DSS and having it process the DNGs) or if it is just an artifact of the fact that the vast majority of the shots I have taken of M31 this summer were before 10:00 PM after which I get kicked out of the parks?

During that time of year it is barely civil dark where I am and there is a lot of light pollution in general as I am in a bright bortle class 8 spot.



There were a couple of airplanes the flew by as well as a handful of dimmer satellites that were captured.

I have had similar results with my processing of M51 which was also shot this summer before 10:00 PM.

Processing DNGs is not the cause - I always did that without problems. Is suspect the settings of DSS to be the culprit. But with the given information it is hard to tell. Maybe you can share the autosave.tif and one or two original lights for a quick inspection. And what software do you use for the processing?

You could try using Sequator, it’s also freeware and a lot faster than DSS.
It seems as if you have a picture with a lot of details.

I did have align RGB channels selected when I stacked them in DSS. I am wondering if that is the problem and attempting a restack with it off to see if I get different results.

Below is a link to the zip file that contains a similar crop as well as one of the original DNGs I will remove it after a couple of days.
https://drive.google.com/open?id=1G72P2l7FJ9uibTyel61kgv5CgGdY5vA_

I think I will give that a try as well. It is good to know my options.

Hi Andrew,

I looked at you data and here some first findings and conclusions (I presume that the autosave file is untouched besides cropping?).

There seem to be some systematic problems in your workflow. I do not find this blue tint in the original sub. It is present in the stacking result. So something goes wrong in between. I suspect that the color calibration settings in DSS are not appropiate. You may want to play with the calibration settings in DSS. I do not know the engish versions names but there is an option to adjust the RGB channels in the final image in the first menu tab "(results"). Also there are three options for the background calibration in the "light" menu tab. After you have found the right settings you can save them as a default for the next start. I would not recommend to use Sequator for this kind of images. It brings fast results but the data do not remain linear. This makes problems in the further processing and limits the quality of the final result.

I have loaded the stacked image into the Astro PIxel Processor. After a quick background neutralisation and a simple automatic stretch the "raw" image looks like this:



The tint is gone and the original M31 honey colors are there. The same happens when the white balance is set in PS with the eyedropper. So the blue tint is only the result of unlevelled color channels.

What more? The image has a strong gradient from lightpolution. This is no problem with a propper processing. But the dark right down corner is something else. There seems to be something obscuring the optical path. In the result the corner gets so dark that a strong banding artefact occurs while stretching. No idea what that obstruction comes comes from.

Last but not least the stacking settings themselves should be optimized. With appropiate settings there may be no residuals from any planes visual in a stack of 60+ images. So the outlier rejection settings in DSS stacking options can not be ok. Maybe you did a median or average stack? I would recommend to use a kappa-sigma stacking or median-kappa-sigma stacking with a sigma of three and 5 iterations instead to eliminate outliers.

There is much potential in your M31 data - many details and a good contrast. The core is not blown out. The colors are nice and natural. I think it would be worth it to invest into another processing session.

So I hope you you find this useful. Have a nice weekend!

Pete

That banding, from what I can tell, is a result of DSS and how it stacks when one doesn't really have much for tracking (I use astrotracer) and I don't recompose as often as I should. I have gotten that effect frequently when I let things drift too close to the edge so that the crop includes them.

The light pollution gradient I expect given where I shoot at most of the time and how low it was in the sky when I could be out. Even the darkest place close by (still a 40 minute drive) is only bortle 5.

I have been using the straight average stacking so far.

I went back and processed one of the partials I saved off and got better results. I will probably do a restack trying both of your suggested methods and process them again. Out of curiosity when did you set the channel levels? Was it before or after the first levels stretch?

Thanks for the advise. I am learning and posted the image looking to get pointed in the right direction again hoping I wasn't making too big of a mess again.

I think that "making big mess" is a normal part of the game and a necessary step. To understand what went wrong brings it to another level next time. If you would see my early M31 images you would be proud of yours.

I do not believe that the banding together with the heavy darkening in the down right an up left corner is a result of the improper tracking/recomposition. That would more likely lead to a wider field due to the stacking and/or weak and noisy data in the corners with a lot of noise and stacking artefacts - but not to such a loss of data. Your original sub only shows normal vignetting so the theory that someting was in the way does not prove. What again remains as a conclusion is that something goes wrong with the stacking in DSS. To be honest it looks like the data were edited in DSS and data were clipped with the result saved in the Autosave file. If you like you can save and disclose or send me your DSS settings file and wilI have a look at it.

What comes into my mind is that there may be very bad subs in the stack that ruin the result - especially regarding that you did an averaging stack. Did you inspect each light frame and sort out the bad ones? You can also/additionally have a look at the sorted registration scores in DSS. It makes a lot of sense to throw out subs that show a significant drop of in the score.

You really want to avoid the average/median stacking method because it always delivers worse result than methods with outlier elimination (sigma).

When do I set the channel levels? In Astro Pixel Processor (which I use mostly now) and PI the background neutralization is done on the linear data before stretching which is best.With the right settings in DSS set (background calibration and/or calibrate on each chanel) the channels are already levelled quite nicely before the stacking is performed. The linear stacking result then is quite neutral in most of the times. During the further processing it is essential to do another background calibration as early in the process as possible to avoid stretching the false colors. A click with the grey eyedropper in PS levels is always good for that. The same rule applies for color noise that should be fought as early as possible. The camera raw filter can be used to kill thew color noise without applying a luminance noise reduction that should be applied later when it gets more due to stretching. Also the gradient reduction should be performed as soon as possible so it is not amplified by stretching the data.

The banding in the lower right disappeared when I restacked them using kappa-sigma stacking and median-kappa-sigma. I think they are artifacts of how DSS does average stacking against a pure black background. I have seen some tutorials where the person suggests cropping the edge off of a DSS stack because artifacts like that but those were with stacks taken from a proper equatorial mount so were not as pronounced. As I don't have one it makes sense I ended up with more and stranger ones as there were a few shots where M31 had almost completely migrated out of the frame.

I had 414 lights that I started with each 20 seconds at ISO 1600 (2 hours 18 minutes of exposure) and of that told DSS to use the best 60% so the final image was 248x20s at ISO 1600. For a few of the sessions I have dark and bias frames but I haven't included them yet.

I did some quick processing with your suggestions and got much better results last night but didn't have the time to invest in a proper session. Tonight will be right out for that as well but I will probably do a restack with dark and bias frames included. Hopefully Wednesday night I can spend some time doing the processing again and see what I can get out of it.

Thanks for all the advise. Once I get a handle on this I will probably revisit my M51 shots from this summer to see what I can get them to cough up. Now if the skies here would just clear I could point the camera at some other things too