M13 (maybe work in progress?)

[dx_ron]

Thanks for climbing down into the rabbit hole with me, Mike! ("when you find yourself in a hole, stop digging"? - nah, dig faster and deeper!)

I pulled out a subset of about an hour's worth of images taken just before and another from just after the meridian flip, so as to minimize atmospheric diffraction effects. The I calibrated with a couple of different choices of de-Bayer algorithm (I normally use the RCD default). To ST, where I kept all the same settings for Crop, Wipe (just uncal1) and Optidev (6.0 px IFD, and kept the same RoI). Then straight to Color, where the MaxRGB really shows the issue, which appears to mostly affect red stars, but maybe something wonky is going on with blue stars that is just more hidden.

They always look like this - none of the de-Bayer variants appeared to make any difference:

pre-flip_uncal1.jpg (161.19 KiB) Viewed 4998 times

So, uhh? It doesn't look like a red circle that is simply embedded within or displaced to the right from the blue circle.

I should try a completely different program for calibration and stacking. I did have a go in ASTAP, but failed. Apparently just throwing some lights, a master bias and a master flat in with all default settings is not the right approach (it's probably trying to calibrate the already calibrated master flat, I guess). Will take new flats (do not have the original flat subs any more) and try again.

[Mike in Rancho]

Yes I get pretty much the same in MaxRGB. I did try loading just B and G, which ended up more normal.

Some kind of stacking or debayering artifact happened, primarily in the red channel. I do see some flaws in the others, mostly on the left side, but red is everywhere, and not just what will ultimately be red stars.

I'm at a loss to figure out what would cause such results. I think this is separate from the purple issue, but hey who knows? Do you have any way to readily blink through subs to look for duds that could have resulted in bad references or averaging? Maybe an owl landed on the scope or something.

This is just the red channel. Not even from the split channels I made, but just composing the OSC stack only into the R slot.

red channel only.jpg (218.9 KiB) Viewed 4972 times

Yeah always a mystery if calibration masters will work cross-platform. They are often best considered proprietary to the program that created them. That said, what tabs did you load into? If the Siril master flat is already bias-subtracted, to test ASTAP you would just throw that into flat, and the master bias into darks.

[dx_ron]

So many variables and complications. As immortalized in the movie Brazil: "My complications had complications"

Took all new calibration frames and fed them into ASTAP. Got pretty much the exact same outcome as Siril. Went back to other data sets, with varying results. Some with similar half-red/half-blue, some not.

Finally opened a single calibrated M13 sub in Siril.

Red:

1sub-red.jpg (15.72 KiB) Viewed 4947 times

Blue:

1sub-blue.jpg (14.62 KiB) Viewed 4947 times

Now *that's* interesting. It's the blue that is oblong while red is round.

The M13 was shot through an Optolong uv-ir filter, which has about the same cutoffs as Astronomik L2. I had previously shot some things with an L3, which more stringently cuts off the extreme blue.

Looks like I can only add 3 images, so the M51 star goes in a new post

[dx_ron]

M51, L3 filter, red:

m51_1sub-red.jpg (16.76 KiB) Viewed 4947 times

blue:

m51_1sub-blue.jpg (14.87 KiB) Viewed 4947 times

I think it is entirely plausible that the odd shape of the stars in the blue channel is the culprit (though I have been equally sure with plenty other snap judgements that later proved wrong, so...).

Will swap the L3 filter back in. Might test some tonight, though i am in an area with getting the worst of the high-altitude smoke, according to the maps. You could even see the smoke plumes in the visible satellite images yesterday.

[Mike in Rancho]

Well, I suppose that's a possibility (lots of things are at this point! )

But the red isn't exactly round, I can see a hint of the same oval, in the same direction, as the blue below it. It just seems the blue is tighter and more focused, perhaps? Also is it a red star with perhaps less signal in the blue band?

Still, some Pepsi-ism, while annoying and requiring some work in post, isn't necessarily a problem in stacking. If some CA or atmospheric dispersion or both, that's pretty normal.

But does the ASTAP re-stack have those same side-of-star partial-arch artifacts? You saw those in my posted crop right? Or just load the stacks (original Siril for sure) into Siril and mosey around the image zoomed in whilst altering the RGB channel tabs. Again, primarily in the red channel, but while fewer there were odd artifacts in the other channels too.

If that's not in the ASTAP file, I have to imagine the result will be cleaner, even if there's still CA or dispersion.

Of course none of that necessarily explains the purple cores. I've had plenty of Pepsi stars in my data, but they've just stayed Pepsi, never grape.

[dx_ron]

Yes, the ASTAP stack has the same look to the same stars:

M13_ASTAP-stack.jpg (255.29 KiB) Viewed 4904 times

Weird finding (though aren't they all?) - back in Siril, *exactly* the same pre-processed and registered sequence, but re-stacked this time with "No normalization" instead of "Additive with scaling" and the red/blue balance on stars is shifted what looks like a fair bit. When fully processed they are still pepsified, but a bit less so:

M13_no-norm.jpg (511.83 KiB) Viewed 4904 times

Meanwhile, I tried to collect some new M13 data but with the narrower L3 uv/ir filter (narrower only on the blue end). The wildfire smoke was so bad that there's way less overall signal and everything has a very ruddy reddish cast, so not going to bother with that. The L3 filter test will have to wait.

I would think the lum filter change will slightly change focus. I have Ekos focus on the un-debayered image, so it should be dominated by green (I would think). Bah - maybe its time to bite the mono bullet (if only I hadn't just splurged on a couple of quality eyepieces to use with the AT130...).

Meanwhile, I'd appreciate a quick primer on how you deal with the pepsi stars. Is it mostly a matter of desaturation, or are you actually able to reconstruct a single-hued star? On a different front - when looking super-up-close at stars there does appear to be about a 1-pixel shift in star centroid positions between the blue vs red channels. If I have time I will try CFA separation and/or some kind of channel realignment (dss? though in the past I never seemed to get that to successfully shift anything).

[Mike in Rancho]

Ok so if you look at only one channel you can see those patches of weirdness?

If ASTAP does the same, I think we have to start looking at the subs. There must be a "some of these are not like the others" situation.

Those patches are so obviously improper that I don't think we can necessarily trust any centroid locating or split channel star realignment at this point. That also makes other fix-it techniques like highlight repair premature I would say.

I think there's a huge amount of subs here, possibly too much to zip and place on Drive or Dropbox? But if you can, or if you want to upload some smaller representative chunks of lights plus calibration files, I'll run them through whatever PI can throw at it for analytics, plus WBPP, to see if any source anomalies can be found.

[dx_ron]

OK - I don't use g-drive much, hopefully I did this right. The folder should have 20 bias, 20 flats and 100 lights. The lights are the 100 minutes right before the meridian on 5/30

Let me know if you'd rather just have the Siril-stacked master bias (100 frames) and master flat (50x). The flats were taken indoors a couple of days later (because I noticed that a dust mote had moved since the prior set of flats I was trying to re-use for that night). But I assume slight differences in the flat are not be the root of the nonsense.

I appreciate you taking all this time!

Here's the link: https://drive.google.com/drive/folders/ ... sp=sharing

[Mike in Rancho]

Yeah you're probably right Ron, but what is exactly at the root of the nonsense?

I still scoped out the flats and the masters that were created. Indeed it seems a big dust spot moved a little, but it just corrected a bit funny and otherwise was lost in the Wipe and O-Dev. Pretty unique contour map on the master flat, but I guess they match the lights so good enough. Still interesting to debayer the master, split the channels, and separately stretch them out.

I started off with ASTAP just to have a control, figuring it would turn out like the original Siril stack, yet it didn't. Noisier with only 1/4 the integration, but while the general chromatic issue still persisted, the harsh splotching regions were gone. So query whether that is something being done by Siril, or is due to whatever is in the other 300 light subs.

I used Blink to run through all the lights, and ended up tossing maybe 10 of them due to what seemed like passing clouds, or perhaps a neighbor's porch lights, then just ran WBPP defaults. First time using it for OSC, so other than my usual no cosmetic correction, all the bells and whistles were on. Granted I probably should have dug into the settings for more of the ST recommended options and maybe tightened the sigma.

I think it was a bit better yet again, though perhaps just incrementally cleaner than ASTAP. I then split the channels and star-aligned to the green channel, which made minor improvements again. Mostly it was in the smaller background stars, which went from multicolored to more solid like we would expect.

The purple mid-cores and some Pepsi aberration to the large and medium stars still persisted, though maybe not as badly, and thus more amenable to some processing tweaks for cleaning up.

Here's a side-by-side-by-side of loading the file only into the red channel in Compose, followed by a stock Wipe and then a stretch. The 436 or so subs of the Siril stack are cleaner overall, except for those splotches. Then 102 subs for ASTAP, and 91 subs in WBPP.

Red Channel crops.jpg (210.43 KiB) Viewed 4804 times

In case you want to compare directly or just inspect the stacking results and evaluate the (hefty) price of admission, here's a link with the full OSC stack out of WBPP, and also the split and star-aligned separate channels. https://drive.google.com/drive/folders/ ... sp=sharing

The culled light subs were 280, 316, 319, 321, 329-333, 339, and 340, if you feel like running exact dups in other stackers.

To try to ground myself I also dug out a pretty bad M13 I did back in 2021. I'm not even sure I was guiding yet. Semi-comparable as it was OSC with a stock D5300, 4-inch doublet refractor at 900mm, also DSS. So in all aspects inferior to your equipment, yet I don't see anything like the color aberrations here.

Is the AT130 anything like the TS130? Maybe we can bring Freddy into this. I think some of his 130 images have had similar split star issues, though I can't remember if also the purple/magenta stuff.

So other than maybe that's just what it does, or perhaps a lens cell collimation matter, all else I can think of is perhaps focus position. I vaguely remember that I could get a white star leaning towards blue, or red, just by wobbling the focus knob to either side. I always tried to just land it in the middle, but of course stuff like that changes through the night.

IIRC Telescope Greg has a 5-inch ~900mm refractor also (plus a 2600mc), I wonder how his does for star colors.

[Mike in Rancho]

I added more files to the linked Drive folder.

Found the WBPP setting for the more ST-suggested bilinear debayering and restacked. After that, again split the channels and star-aligned them to the green.

I think that composing the split and registered bilinear RGB has the best color star alignment of any of them so far. Not perfect, but better than the others including the previous PI stack (must have been VNG). The hue issue, mostly purple, still take some work with the RGB bias sliders, green cap, and probably some highlight repair. Full field or star sampling for Color seems to only make for a starting point. I can't quite figure what's going on here, scope lenses or filtering, but at least a lot of the split-colored star issue is tamed down.