Inspirations from Thierry

 

7/04/2025. After a few clear mornings and dozens of trials, here's my best current workflow:

• Capture in 8-bit, gain 80-231 (100-150 is fine), set gamma to 22-32, bin 2x2. Quark set one notch above dead-center. Captures with the TMB92SS uncompressed will need 7-12ms for chromospheric detail. Disable gamma correction and use 25-50ms for limb pyrotechnics. Flats are crucial and should be taken whenever you change from mid-disk to the solar periphery or, of course, when changing gamma. Avoid extreme gamma since it will emphasize the uneven bandpass the Quark can exhibit which is distinct from uneven illumination which can be handled with a flat.
  
  
• In Autostakkart!4, use 50-100 frames from whatever length clip is in hand (500-4,000). Use about 1000 stabilization points and drizzle 3x. I know we're already oversampled, but drizzling seems to help a good bit anyway.
  
  
• Square the frame by layering two copies in multiply mode (to get most of the benefit while preserving shadow details, use curves to bring the top layer's minimum up to ~30%, and bring the white threshhold back to ~75%). It's OK and generally beneficial to sharpen the hell out of this layer. Opacity can be reduced to keep things in bounds. Flatten and proceed. [The quantitative portions of this item are almost certainly wrong; see next page.]
  
  
• In ImPPG, let L-R do most of the work: 70 passes, anti-ringing enabled, set L-R as needed, but err on the side of minimal sharpening.
  
  
• Curves, levels, high-pass, noise-reduction, highlights and shadows, USM, etc as needed.
  
  

Presumably, there's an optimal number of frames. Look at the 101-frame batch-generated stacks from 7/3/2025 clips, and see which work best. After working up the best results, consult the .TXT files to see which capture settings worked best. Essentially there are enough clips and enough variables on file now that they amount to blind trials; the .TXT files document their collection.

Today I tried a Harris shutter effect on a complex prominence. I also practiced on some chromospheric details -- no killer spot complexes today, but some good detail here and there just the same. Rather than eat up bandwidth with intermediate results, I'll wait for another session or two.

Also today I finally cleaned the TMB92SS's optics after using it for 8 years. I honestly don't think it was necessary, but it bugged me seeing dust on the inner surface of the objective light up when solar energy is reflected back up the tube. Let's pretend it could help with contrast, even if only a tiny bit. I used a microfiber cloth and Baader Optical Wonder fluid. I also cleaned the IR/UV filter, the diagonal mirror, both sides of the Quark's optics, and the cover window on the ASI178MM. Along the way, I discovered that the collimation screws on the TMB were very loose (unscrew the dewshield from its collar to access them). I rotated the collimation bolts to finger contact, not to finger tightness.

7/05/2025. I did a 12-panel mosaic today (just to keep my hand in), and the usual survey of active and/or interesting regions. The workflow up above has been amended in light of today's experience.

Later that same day: holy shit! I've found a whole new set of tools! SolarToolbox for PixInsight: www.youtube.com/watch?v=3yzfH5x5Smg

I'm running these tools against the multiplied, un-deconvolved data and it looks really promising. Best tests so far:

Contrast Enhancement :: Histogram EQ, kernel 115, contrast limit 1.8, amount 0.56. Sharpening: kernel 5, boost 15, denoise 0, denoise kernel 2. Many more parms to play with and tweak.

or, Contrast Enhancement :: local contrast, contrast 7. Sharpening: kernel 6, boost 15, denoise 0.7, denoise kernel 5.

There are many parms to play with and tweak.

 

7/07/2025: A long day later, it seems clear that you're going to earn a good image one way or the other. Either you go around the barn with my previous workflow, or you tweak and adjust SolarToolbox and then work on the best output to remove faults and improve that result. The result can be spectacular, but it still feels hit or miss (the assessment that there are a lot of parms to tweak remains spot on). The STB will get you 80% there, the next 20% will still take some work. After a full day of experiments, the best way I've found to use it is: (1) Use NoiseXterminator on the raw file (or to the multiplied raw file) to aggressively remove high frequency noise and leave low frequencies alone. The Toolbox is really good at sniffing out noise and emphasizing it, and NXT is far more effective than the noise reduction facilities in the toolbox. (2) Run the SolarToolBox. (For now, I just check my PI saved icons for a good starting point. I will tell more when I know more.) When the image is close, save it and then (3) run ImPPG with very low USM settings and modest L-R settings. Polish that result as needed.

This is not really a lot simpler than the original workflow, but it can be made more routine, less extemporaneous. At long last (1:55 AM after two days' experimenting), it can produce an undeniably improved result. Sometimes.

7/08/2025: The Sun has gone quiet, but I have new software to learn and a new workflow to smooth out. So I unpacked a dataset from an active afternoon in the summer of 2024 with excellent seeing and started with that. This feels more routine as I get a grip on what does what in Solar Toolbox:

Before & After from August 8, 2024. Yes, you can damn well click it.
Best 200 of 4,000 frames.
AutoStakkart!3 output at top.
Refined with NoiseXterminator, Solar Toolbox, ImPPG, and Photoshop at bottom.

 

07/13/2025. Mornings have been clear and steady, afternoons loud and wet. That works. I've done tons of new images and mostly learned an old lesson: capture with gamma off and you can easily recover in post all the tones you see on the screen at low gamma settings. Into the bargain, you get better shadows and highlights, a few more tones to work with, and shorter exposures. So take a flat at zero gamma; focus at low gamma (high contrast) where the image really snaps in and out of focus; return to no gamma for capture. Keep gain low(ish). 150 or less if possible. Exposures of 5-25ms are fine, so are longer ones on steady days. In 4,000 frames, 50 will be sharp even with generously exposed subframes (keep the histogram plausibly centered), and at low(ish) gain, 50 frames will be enough to control noise. 200 is sometimes better, so build both in post; it's just another several seconds and storage is cheap. 2x2 binning with my current kit is fine; with 3x drizzling, ROIs that produce up to 50fps are useful. The internally mounted 0.5x compressor should probably be the go-to optical train. Add 17-20mm for widefields and full-disk mosaics; use a snout and no compressor when image scale is everything and the air allows. Examples to come; things are happening on the Sun.

 

07/16/2025. My best workflow so far: stack the best 50 of however many you have (4,000 or 500 or whatever). Run NoiseExterminator in Pixinsight (60% reduction at high frequency, 10% at low, crossover at 2.7 pixels). Open in Photoshop making good use of highlight and shadow sliders (avoid extremes and take it easy with other sliders if you use them at all). Crop out the border intrusions. Use levels to move the peak levels up to 70% or so (that is, make it very light without clipping highlights). Stack two copies of the image and blend them via multiplication (no need to fiddle with curves if the source files are of the right brightness). Run ImPPG ( 70 iterations of L-R deconvolution with ringing suppressed, and take it easy -- in a 3x drizzled file derived from "snout-compressed" data, 5-ish is about right). That'll get you 90% of the way there. Continue with a bit of this and that (USM, sharpen, highlight and shadow... whatever works). FWIW, the file I'm basing this on was 8-bit, 2x2, gamma 22, gain 150, exposure 7.1 ms; best 50 of 2,078 frames, ROI 1280x800, about 35fps. Filename 1715-6 from 7/15.

One other key: focus for all you are worth!

 

 

 

 

 

:: top ::