The Current State of Things
7/17/2025. How about a case study?
TMB92SS, Quark Chromosphere, ASI178MM. Captured in 8-bit mode, gain 150, gamma 22, bin 2x2. Quark set one notch clockwise of dead-center. 7.1ms exposures using 0.5x Antares compressor inside snout at minimum distance from sensor. This is a stack made from the best 50 of 2,071 frames capturing a 1280x800 pixel ROI using 1,750 alignment points (45 pixels each) in AutoStakkart!4. Drizzled 3x. NoiseXterminator in Pixinsight (60% reduction at high frequency, 10% at low, crossover at 2.7 pixels). At this stage you can't see the benefit of NX, but it really matters later.
Click to enlarge any of these images.
(1)
Square the frame by layering two copies in multiply mode. Before duplicating the layer, use curves to lighten the entire histogram -- bring the peak up to around 70%. Make sure the leftside tail is at least up around 10-20%. (After multiplying the two layers, a pixel with brightness 0.70 will be around 0.49; one with brightness 0.10 will be around 0.01, so it's very easy to lose the shadow end if you start with too-dark data. It's much harder to lose highlights.) After multiplication, stretch a bit with levels or curves, work on highlights and shadows with Photoshop's RAW filter, etc. Whatever artifacts you introduce now, whatever noise you enhance, whatever harshness you invite, will be made worse in the next step, so take it easy here.
(2)
In ImPPG, let Lucy-Richardson deconvolution do most of the work: 70 passes, anti-ringing enabled, but err on the side of minimal sharpening (4.0 to 6.0... whatever the image scale demands or allows without degrading into noise). Unsharp masking may help, but Photoshop and PixInsight each offer better space-domain image enhancement algorithms than the similar tools ImPPG provides. Use ImPPG for what it does best, the L-R toolset:
(3)
Curves, levels, high-pass, noise-reduction, offset emboss blends, highlights and shadows, USM... try whatever you like to bring out the details without bringing out the mess, which is all too easy to do. See? Is the image below better than the one above? I sometimes think it is, and I sometimes think it's just harsher. I'm still working on this stuff. I expect never to be done with it.
And as I said last time: in the beginning, focus for all you are worth. You'll always get better results from well-focussed data. It doesn't take long to fiddle with that, and there are better things to spend post-processing time and effort on than fixing sloppy focus.
Goddamn! On the same day I was taking the data above, Thierry Legault was hard at work with his new Soleye 300 telescope and a pair of Lunt 40 H-a etalons, something like 9,000 feet up in the Alps where he waited three days and finally got steady skies. So, with a 12-inch solar telescope, excellent seeing, and expert technique, this is what you get. I mean, damn.
7/19/2025. What if I tried two or three sub-aperture masks to control total energy flux while preserving (nearly) full-aperture wavefront sampling? Why wouldn't that work? It's not like there's any shortage of light. At first I thought dual apertures would do, but that would, at best, produce much better resolution in one dimension than in the other. So maybe three apertures? Anyway, I tried the simplest possible plausibility check this morning: two holes cut in an aluminum foil mask over the 92mm TMB objective. Results were inconclusive, which is to say, "not a roaring success." Not an abject failure, either, though, so let's keep thinking that just maybe the 10-inch R-C could have a future under the Sun as well as under the stars. The way light behaves when treated this way is not particularly easy to anticipate.
What did work today was controlling my urge to push L-C deconvolution too damned hard. I batched all the videos, then batched NoiseXterminator against all the stacks, then squared the ones that seemed interesting, and then with those took pains to not over-do sharpening early, in the ImPPG stage -- or anywhere else. Several modest steps work better than one or two sledgehammer approaches. And take it easy all. the. time. Push L-R to see what's there, then back it off until you can barely see the buried details. Then pull those out using the rest of the bag of tricks (USM, high-pass filters, edge-emphasis, etc etc). Noise and "ropiness" is the alternative. Also, while shutting down for the day, I saw that the Quark was set a few notches clockwise from its usual position. Finally, today marked the debut of an actual shaded enclosure for the computer (replacing a ricketty cardboard box). Small improvements count.
7/20/2025. Another small improvement: the ASI178MM / Firecapture combination offers a high speed capture mode useful for up to 10-bit data transfer. (How did I not know or remember that?) Since I'm using 8-bits these days, it seemed worth checking out. Yep. Works a treat. 1280x800 regions of interest run at close to 70 frames per second rather than 45. It could matter when shooting through breaks in the clouds or when trying to get a generous sample in 30 seconds or so because of plasma motion. In fact, I snagged a decent photo of a bright flare in that departing sunspot group in about 1.5 seconds of clear air. Here's the best 50 frames during the flare compared to the usual 200 of 2,000 view in clearer air about 90 minutes later. The air was none-too-steady for either take, but you image with the air you've got.
Clickable? Of course. There's also a video (not yet linked).
7/23/2025. I've been using this filter for a couple of years, and it's crossed my mind that I might not be using it at its best. The internal heater adjusts the filter's peak wavelength to either side of 6562.8 Angstroms, more or less. So under today's good seeing, I spent an hour taking video clips at three positions: long, short, and what I've long thought was on-band. There were about a dozen files in all, enough that I had no idea which were which (except by referring to my notes). I produced the best images I could from all of the clips without knowing which were expected to be on-band, which long, which short. I only consulted my notes after deciding which clips had produced the best images. Spot on! The ones made with the filter set as I have been using it are the best of the bunch (none are garbage, but there are pretty clear degrees). Sometimes things work out.
Best 50 of 4,000, filter tuned to one click clockwise beyond top center
8-bits, 2x2 binned, high speed mode
14.7ms, gamma 27, gain 160, ROI 1280x800, no optical compression (check this), drizzled 3x
Autostakkart!4, NoiseXterminator, Photoshop Multiply, ImPPG.
PhotoShop for USM, HighPass, Downsample for display, Shadows & Highlights
Now let's work on time-lapse techniques. Solar prominences in motion would complete my quest for video of the sort that Mr. Bill used to air on Asheville's WLOS-TV 13 in amongst the morning cartoons (at least that's how I remember seeing it; I could have conflated a couple of pre-school broadcasts -- it's been 60+ years).
:: top ::