Pure luck.

7/14/2023. I rushed out to try some reduced aperture full-disk imagery. I'm still disappointed, but it's coming along at least a little. I revisted the big sunspot while I was set up and then decided to go looking around, since I was there and all. I happened into a prominence just detaching from the Sun and being hurled out into space at hundreds of miles per second. For the record, prominence detail is easy to see at numerically high gamma; it's easy to focus at low gamma, but 14-bit files work fine captured with gamma off. Don't be wasting your time, patience, and storage capacity capturing multiple versions. Also, it appears that 2,000 frames is overkill under typical seeing. Half that worked very well.

 

 

 

 

 

 

These are all 250 frame extractions from 1,000 frame clips, unity gain, somewhere around 6-8 ms exposures, one notch "under" noon on the Quark's heater. While not all of these were captured with gamma off, they could've been. Live and learn. Most are 2,000 x 2,000 regions of interest, which interestingly, does not result in faster frame rates but does speed up post processing considerably. Most of these were drizzled 150% in post just to have plenty of pixels to play with. Except for the "Harris Shutter" capture of the eruption (3 clips, 60-90 seconds apart mapped into R, G, and B channels), and the very wide view of both the mighty sunspot and the surprising prominence, all these all single-clip efforts and none of that one for the surface and one for the limb stuff. The single clips were processed twice in some cases though, to allow IMPPG to emphasize one aspect or the other. The wide view of the big spot and the prominence together is a two-frame mosaic using the full chip for each panel using identical exposures but different transfer functions in post.

 

7/23/2023. I've had some decent sessions that I have not posted here. Maybe by and by. Today, the camera throttled for unknown reasons; I applied ice; it remained throttled even though I restarted the camera and restarted the software. Something as yet unconsidered is happening.

In a semi-serious effort to get serious, I set up the computer and camera in the kitchen and tried out all sort of settings. I kept "16-bit" checked and used "high speed mode" which limits the ADC conversion to 2-fewer bits ("high speed mode" is very much faster). This is the way I have been doing all the best stuff lately. Frame sizes below are width x height. Herewith, some of what I found out:

Full-frame: 26 FPS
Full-frame with flat-field: 22 FPS

2000 x 2000: 25 FPS
2000 x 2000 with flat-field: 22 FPS

1400 x 2000: 27 FPS
1400 x 2000 with flat-field: 27 FPS

2000 x 1300: 44 FPS
2000 x 1300 with flat-field: 44 FPS

2200 x 1505: 38 FPS
2200 x 1505 with flat-field: 38 FPS

1280 x 1006: 58 FPS
1280 x 1006 with flat-field: 55 FPS

3088 x 1932: 30 FPS
3088 x 1932 with flat-field: 25 FPS

I saw no appreciable effect on FPS from binning 2x2. Applying gamma adjustments imposed a slight (~10%) degradation on full-frame performance and less on the faster ROIs shown above. As expected, 16-bit mode is consistently half as fast as 8-bit mode, but the results are usually well worth the speed sacrifice. I turned the USB priority down from 90% to 60% to see if it had much effect on frame rates. Yes, it does! Leave it as high as possible.

Flat-fielding imposes a real speed cost on very large fields. I much prefer to flatfield during capture, but keep this in mind in case it ever matters.

Note that a wide frame is faster than a tall one. That's to be expected given that the readout has to be by rows or by columns first. One of those orientations will limit the number of values to be read by simply cutting off reading sooner while the other will retain the overhead of reading part of every row (or column) on the chip. The faster approach is an ROI that is wider rather than taller. Compare the 1400x2000 speed with the 2000x1300 and 2200x1505 spreeds.

Note also that you get a noticeable improvement (30/26 and 25/22) by very slightly cropping the full field into a very large ROI.

In the kitchen, the ambient temperature was about 23C. Over the course of 45 minutes or so, the camera warmed to 38C. It was still rising when I turned everything off. If the heat of operation raises the camera ~15C over ambient, then my use of it in 90F weather would produce sensor temperatures of ~48C. I routinely see 52-54C. The difference is incident solar radiation, either striking the outside of the camera or striking the sensor and its enclosure via focused light. In either case, the surplus heating isn't much and suggests that the 92mm refractor with its dielectric diagonal and 7nm Baader cut filter is not letting a lot of IR energy through. But keep these numbers in mind when using the 152mm scope (and remember that you've already made a 127mm stopdown ring intended to produce an optimal f-ratio but perhaps useful for energy control also).

 

7/24/2023. Putting that to work, I did a couple of mosaics this morning. The first is four frames: two of the prominences, two of the surface.

 

 

Same exposure for all, but two different gamma mappings from sensor to image, best 400 of 1,000 frames in each clip. Custom flat-fields for each gamma setting. Honestly, I'm not sure the gamma settings mattered that much. Deep bit depth might come close to giving me all this in just two frames for breadth. Full-frame, 16-bit captures were running only about 17 frames per second this morning (why? it wasn't even hot), but setting a Region Of Interest a very few pixels smaller than the full size of the chip as shown above gave me ~30fps every time. I have no idea why the difference from kitchen counter experiments. Two more clips trying to capture a wider expanse of active Sun (no gamma, best 200 of 1,000):

 

 

And finally, for those who prefer an "all-natural" capture, here's a single "16-bit" no-gamma clip processed to try to retain both rim and surface details:

 

 

I thought at first that the blurring of the plasma in the middle of that bridge was due to some careless smoothing or noise reduction on my part during the workflow, but no -- when I went back to the original data it's there in the pristine stack. It has to be motion blur from flowing plasma. The thousand-frame clip only took about 35 seconds to capture. That's some fast motion.

 

7/26/2023. The same prominence (the "bridged" one on the left) has rotated fully onto the face of the Sun. Note arcade loops over the small sunspot at right center. Best 400 from 1,000 again.

 

 

 

My solar images beginning in June 2023 are made using a Daystar Quark Chromosphere filter on a TMB92SS refractor. A 2" Baader 7nm Ha filter and an A-P dielectric diagonal provide extra IR protection. The camera (~2021 et seq) is a ZWO ASI178MM chosen for its tiny pixels and fast capture rate. It is notorious for a nasty pattern noise which can be avoided with careful exposure or removed using FFT-based processing. I capture data using FireCapture, stack using AS!3, deconvolve and perform initial histogram adjustments using IMPPG, and polish in Photoshop. The solar kit is mounted on a Skywatcher Star Adventurer "tracking platform."

 

:: top ::