M13 (maybe work in progress?)

[dx_ron]

Thanks for all this.

I loaded your PI-stacked file into ST and treated it the same as I have been doing with mine - crop, uncal1, optidev with IFD at 6px and the same RoI, then straight to color with just turning sat down to about 145%.

I can't say I see much difference.

PI version:

Mike_bilinear.jpg (704.37 KiB) Viewed 5048 times

Siril:

Siril_DCB.jpg (691.32 KiB) Viewed 5048 times

But maybe I'm missing what you refer to as "harsh splotching"?

BUT - last night I was able to gather ~ an hour with the Astronomik L3 filter that cuts out some of the very far blue. I think I need a 2nd post to add the images - brb

[dx_ron]

(before slewing to M13, I tried to add a couple of hours to my NGC 5982 data - but I screwed up and shot 60s subs when all the others are 120s )

Anyway - here's the L3 version:

L3_starting_balance.jpg (498.67 KiB) Viewed 5048 times

I bumped blue up some:

L3_increase-blue.jpg (491.91 KiB) Viewed 5048 times

And here's the overall image:

M13_L3_06-09-23_uncal1_no-deconv.jpg (631.52 KiB) Viewed 5048 times

The saturated blue star is more or less white (though some kinda pinkish, to be honest?). But the red stars are basically red and the small blue stars appear to me to be much more blue and less purple.

I wish someone could explain to me why I decided to remove the L3 filter and go back to the Optolong... So now I need another 8 hours or so, assuming I really want a 2023 AT130 M13 image. Though there's not a whole lot else for broadband at 910mm after midnight right now.

Oh - as for the scopes. The TS version has the same mechanicals (though I think they offer a choice of focusers) but definitely different glass in the lens. So they're not comparable as far as color correction. Would I like to have the TS version? Sure - but it costs over twice what I paid for my AT130.

[Mike in Rancho]

Much of interest here...

I had no idea there were identical OTA's out there - mechanically - but with different lenses. And those must be some lenses if the TS is double the hit to the wallet.

Also, I had assumed this M13 was the L3. Thought it was a while ago you were moving to that versus the L2, or something L2-like. So all 426 minutes here had some expanded blue wavelength? Will be nice if the L3 provides better color as it seems to be. And I too noticed the odd other color apart from the purple. Not quite pink I thought, but really no idea how to describe it.

Were you trying the full OSC PI stack? Yeah to me those still require splitting and star alignment. They all seem to. Otherwise, I felt the PI stack was cleaner than Siril and ASTAP, especially as to background color/noise. In fact I wasn't even needing to go to UnCal1, though of course if one does that'll hide most differences.

The hue problems were still there, as that is likely scope and filter combo, but I felt the centering was much improved. I tried to replicate your MaxRGB, but not knowing your ROI box I just guessed there. Loaded the split and aligned RGB into compose, no bin, crop two mouse wheel clicks on all edges, UnCal1, IFD 6 OptiDev with ROI, and Color. Which still requires tweaking (plus green cap, plus highlight repair) no matter if one field samples or star samples. The cores within stars seem better centered as well.

PI bilinear split aligned MaxRGB.jpg (483.38 KiB) Viewed 5036 times

Anyway the splotching in the Siril stack... Is it just me, or something I am doing wrong maybe? I guess I just can't unsee them. On the left (Siril) of the three-pane comparison I posted on the prior page, the left side of the stars have severe anomalies that the others don't. Again I don't know if that's a bug in Siril itself, or due to including the other 300 subs. That's the red channel, in which they are most prominent, but they are in the other channels too if one scrolls around the image enough. They kind of look like someone took a sharpie and drew an eyebrow on the star.

The result in the processed RGB was that the anomalous area (which is essentially a shadow) lets another color channel poke through it, creating a strange half-arch of the wrong color.

If you don't see them, hopefully someone else can try the Siril stack and let me know if I need to go to the eye doctor.

[dx_ron]

A new TS 130 with shipping and duty is roughly 2x what I paid for my *used* scope.

Yes, I did buy the L3 filter back when I was working on M42. But somewhere along the way I managed to convince myself that it wasn't really necessary and put the Optolong (L2-equivalent) back in. I appear to have been rather badly mistaken in that conclusion.

Now - as long as you are talking about the red stars having a blue splotch on one side then we're talking about the same thing (which I think is also most of the pepsi effect). I just wasn't sure if there might be some *other* weird color splotches (in the background, perhaps) that you were trying to describe that I wasn't seeing.

Then there's the centering issue. I did not realize you had not channel-aligned the PI-produced stack. But it doesn't seem that PI's debayering isn't doing anything terribly different from Siril's (unless I'm still missing the point). It seems to be far less a problem in the L3 stack, though it's still visible. I shall work on trying to do color channel alignment in Siril - or is that something where PI is far superior?

[Mike in Rancho]

Well, it may be that you need to match filters to what a particular scope works best with. Seems the L3 will be a winner with the AT130. Might take a little testing with the other refractor to see if you want the fuller spectrum in there, as it can likely handle it.

Yes splotches in the stars, not background. But also not just red stars. I just wanted to make sure you were also seeing it when only one channel is loaded in as if mono, since that clearly shows it is some sort of error in stacking. On those stars yes it would be the bulk of any glaring Pepsi effect. But Pepsi stars are common enough elsewise, whether from CA or dispersion, so just pointing out this is a totally separate deal. Not that you might not have had both going on.

My bad on the PI stack then. I'm not super proficient at it yet, and thus am not aware of any option button that will automatically do a post-stacking channel split, re-registration of the three channels, and the recombination back into RGB (like DSS has). That's why I included the separated R, G, and B files. Plus that allows use of Compose module and therefore "widefield galaxy" etc.

So, I probably should have found out how to recombine those registered split channels back into one RGB file. Next on my list I guess.

And while I know how to extract RGB channels into three mono in Siril, I don't know how to star align them, really. I mean the Siril RGB combination/compose does have an alignment feature for the channels, but I believe that is just x-y displacement, either manually or by selecting one star. But that would be unlikely to work here, you need full registration at a sub-pixel level plus interpolation. I know how to do that in DSS (albeit awkwardly), ASTAP (easier), and now PI (super easy).

Anyway as long as you get to the root of that stacking bug in the Siril 426 sub stack, you should be able to then readily split channels and then realign if needed. I don't think PI has any improvement over Siril in debayering. The subframe weighting and local normalization is nice though, and I thought made for a cleaner background despite 1/4 the integration. But that takes computer horsepower plus time.

[dx_ron]

So it turns out that Siril can do a full regular alignment on the split RGB channels. It used to involve more manual fiddling with files and sequences, but the latest 1.2 version includes a script. Make a folder with 3 files named R G & B (you can append other bits to the file name, it is only important that the B file be the first name in an alphabetic sort, etc).

With the original Optolong filter stack it results in:

siril_rgb-align.jpg (499.68 KiB) Viewed 4953 times

Not perfect, but better. Still violet, of course, but maybe I'll play with fringe-killer to try to salvage those data.

[Mike in Rancho]

That does look better aligned.

I'd still be concerned about those odd stacking anomalies in the 426m stack though. They were still there when I split and star-aligned that one.

I have yet to download the new Siril beta, still been sitting on 1.06 or something like that. Is it a truly included script or something you have to go hunt for and download? I've never bothered learning scripting for Siril, and consider it kind of an annoyance that's needed in order to effectively use the program. I guess that's one thing I can appreciate about the often otherwise confounding PI - the scripts implement as pop-up windows with all sorts of settings, file loading, and so on. No having to set up folders "just so" and hitting the script button, hoping for the best.

[dx_ron]

Yes, Siril 1.2 includes the RGB Align script. It just automates the process of making a sequence with the 3 split files and registering them to the green channel, but its nice to have. Anyway, I've moved on and will just consider that a 7.7-hour reminder to always use the L3 filter with the AT130.

I'm now up to 222 minutes (usable minutes, anyway) with the L3 filter.

M13_L3_dim_222x60s_fwhm-80_bkg-90_rgb_pre-heal_700k.jpg (633.54 KiB) Viewed 4900 times

There are the obvious ringing artifacts on the brightest stars. I'm already using both Linearity and Deringing Detect at 1%. I've tried Deringing in Shrink, but can't find any combination of mask + parameters that have much of any effect at all. The two brighter red stars in the lower part of the frame have also a nasty artifact that might be related to stacking. RGB aligning this stack slightly reduced that (and didn't change much else about the colors, which is good). Here's a closeup:

artifacts.jpg (275.8 KiB) Viewed 4900 times

That's really obnoxious. I watched those stars closely through the process. The artifact is detectable but not terribly obvious up until deconvolution. At this point it might take some pixel-by-pixel work in Heal to deal with both those weird arcs and the basic ringing problem in all the brightest stars.

[Mike in Rancho]

Well, as the saying goes, the splotch thickens.

The fact that these are completely new subs really makes it more baffling. So I took the prior set and tried out Siril, since I already ran ASTAP and PI on it. First 1.06, just using the downloaded no darks prefab script. I don't know how to otherwise use Siril stacking. Everything else in settings was at default, except changed to 32 bit. Debayer was RCD, seemed the safest of the given options. I used the same 91 selected lights that I used in PI, plus 20 flats 20 bias.

These being from the Optolong L2 data, yes the star shapes amongst the channels were wonky, and misaligned, with just crazy color errors, but no splotches. I then downloaded and installed Siril 1.2rc and ran the same thing, again no splotches.



All I can think of at the moment is that the uploaded data was all from one session, and moreover one side of meridian flip. That's just a guess, and would have to be tested out if multi-session is somehow engendering these splotches.

At least Siril seems pretty darn fast, so such testing would not be a major problem. Some of that may be the files starting out as fits, rather than RAW as when I last tried Siril stacking years ago, but mostly it's probably the computer. My old heap of a desktop just couldn't handle it (DSS was fine), so I was pretty dismissive when so many were claiming how fast Siril was.

Deringing in SVD seems awfully data dependent. My Newt files can be trouble. Your 130 seems to be. But the other day I tried a stack from an AP 130 and it mostly came out just fine, with far less aggressive use of the new deringing controls.

That said, there are still some small stars that don't seem to get any deringing applied to them no matter what. Easy to fix those in Shrink. The medium-large stars can still cause me grief though, and going too low on the fuzz slider starts creating some funny geometric shapes about the core, so I typically have to keep that 15 or higher.

[dx_ron]

Did the data from the AP130 include any fully saturated star cores? There are two separate issues (I think) in my M13 data. One is the hard ringing on all of the stars with saturated pixels at the core, most obvious on the bright blue star. Extreme settings of the deringing sliders minimizes it, but its still quite obvious.

Then there's the red stars where the red channel isn't round. I'm not yet convinced that the issues with the red stars in the L2 vs L3 filtered data are the same (nor am I convinced they are different...). I've started registering and stacking subsets to see where the non-round-red crops up.