SVD Items

General discussion about StarTools.
Post Reply
Mike in Rancho
Posts: 1166
Joined: Sun Jun 20, 2021 10:05 pm
Location: Alta Loma, CA

SVD Items

Post by Mike in Rancho »

A couple of thoughts I had while playing with some data last night...

For quick processing and to check out the SNR boost I binned the data to the lowest preset, 25%. SVD did not seem to take kindly to that, and results were poor despite trying several combinations of sample stars. No worries, for this, synthetic should be fine. I turned sampled iterations to OFF, and started working on my synthetic iterations and radius. However, it would not come out of Circle of Confusion mode, which remained grayed-out as if I was doing SVD sampling. Until, that is, I went on a hunt for all of my sample stars and clicked off the blue boxes one by one. After the last one was turned off, my synthetic modeling options suddenly appeared.

This may have been discussed before in theory, but this was the first time I ran into the issue in practice - there does not seem to be a way to clear the samples. Re-running apod mask did not work...those little blue boxes have staying power! I also tried to trick it by using a blank txt file for Load PSF's, but ST just laughed at me. I suppose the quickest way to switch from sampling to synthetic is to just cancel out of SVD and enter again? At least if one wants to choose from the array of modeling options.

Anyway, my other thought was more of a description matter, though it's possible I'm just not thinking about it correctly. SVD with a single sample is described to be like 1.7 deconvolution. Probably setting aside that the masking works very differently, of course. But anyway, what particular base "model" is being used?, as of course there are no choices (Moffat and so on) as there are in 1.7, which can in fact lead to quite differing results. And as noted above, if even one sample is chosen, the synthetic side locks to Circle of Confusion. If one were to then engage that by turning on synthetic iterations, I am guessing that would be akin to 1.7 with a secondary PSF selected?

Not that it really matters as it works great, just curiosity really. :D
hixx
Posts: 254
Joined: Mon Sep 02, 2019 3:36 pm

Re: SVD Items

Post by hixx »

Hi Mike,
this seems to be expected behaviour to me:
1) Deconvolution only works if your image is oversampled. as soon as You bin strongly, oversampling would be removed and no latent resolution will be left to be restored by Decon.
2) As soon as one star is selected for sampling, SVDecon will work in sampling mode, disabling Synthetic options other than COC. Hence all options would appear not before unselecting the last sample.
3) I guess the quickest switch into synthetic Mode would be Clicking the Apod Mask button and clearing the mask
4) Decon module's goal is to compile an optimized PSF for each pixel which works better if star samples are available All those models of the atmosphere are justproviding a global compromise between multiple non-optimal solutions. The idea is that the samples would "construct" the "real" PSF so no further model would be needed unless star are unavailable like in planetary mode which relies on synthetic mode completely
Clear Skies,
jochen
Mike in Rancho
Posts: 1166
Joined: Sun Jun 20, 2021 10:05 pm
Location: Alta Loma, CA

Re: SVD Items

Post by Mike in Rancho »

hixx wrote: Thu Dec 30, 2021 12:56 pm Hi Mike,
this seems to be expected behaviour to me:
1) Deconvolution only works if your image is oversampled. as soon as You bin strongly, oversampling would be removed and no latent resolution will be left to be restored by Decon.
2) As soon as one star is selected for sampling, SVDecon will work in sampling mode, disabling Synthetic options other than COC. Hence all options would appear not before unselecting the last sample.
3) I guess the quickest switch into synthetic Mode would be Clicking the Apod Mask button and clearing the mask
4) Decon module's goal is to compile an optimized PSF for each pixel which works better if star samples are available All those models of the atmosphere are justproviding a global compromise between multiple non-optimal solutions. The idea is that the samples would "construct" the "real" PSF so no further model would be needed unless star are unavailable like in planetary mode which relies on synthetic mode completely
Clear Skies,
jochen
Hi Jochen. Thank you for your help! :D

I will use your same numbers,

1) Yes I am aware of this. I was simply explaining how I got to where I was. A story, so to speak. Also note that the tipping point of where bin has degraded the ability of deconvolution to help is likely to be data dependent?
2) Right, that's what I was already saying. I am pointing out a procedural usage issue. It may or may not be common, depending on how much people play around in the module trying out different things.
3) Thank you for the suggestion; however, it does not work. Note that the apod button does not have a clear option (though you can pick manual generation to switch to the main Mask screen). From either way of getting into Mask and clearing, if you then keep, SVD notices the blank mask and presents you with the choice options again. At that point, theoretically you can select "as is" and go back into SVD with a blank mask. However, remember that the mask in SVD has a secondary usage for deringing, and this applies in synthetic as well. So, you have now lost some of your deringing capability. This is not good, eh? And if you then try to regenerate a mask, even to stay in synthetic and to use just for deringing, the samples are still there. As I said, they have amazing staying power! :P As a side note, I also tried the planetary button for the heck of it, as I know that clears the mask. Again, samples survive that too. They are like the undead! :shock: Planetary also makes other changes, which can only be reset manually - there is no undo or reset back to non-planetary. So once again, really the only way I have found so far to escape from either dilemma semi-expediently is to leave SVD completely back to the main menu and then go back in fully anew to start from scratch.
4) Ah, interesting. I knew that SVD sampling was of course reality-based, but didn't know it was 100% and figured there might still be some kind of atmospheric formula utilized, whether known or new and custom. If the reversal is pure reality, I suppose ST can't become famous for the model itself? Was kind of pulling for it to become "Ivo-Beta 3.14" or something cool like that. Oh well. ;)

Thank you again!

Mike
hixx
Posts: 254
Joined: Mon Sep 02, 2019 3:36 pm

Re: SVD Items

Post by hixx »

Hi Mike,
thanks for Your reply.
1) Yes, it depends on real resolution. There is a formula provided in the inofficial manual, background chapter: Image scale (arcseconds per pixel) = 206 * camera pixel size (microns) / Focal length (mm)
3)
However, remember that the mask in SVD has a secondary usage for deringing,
Yes this seems to be a gap. Even though ringing is less of a problem in 1.8
4)
there might still be some kind of atmospheric formula
Yes there is an overall optical /atmospheric formula called PSF. it will be derived from the star sample(s) directly, hence will not be constant compromise as the synth models are, but would differ from case to case, which is the really cool thing about the "Ivo et. al." - model :)
Undo planetary probably might work using DSO preset, but I am sure bouncing modes will open some cans of worms. Probably Ivo will consider a 100% 1.7 process, I think he stayed away from this so far because 1.7 results will always be inferior to SVDecon
As a side note, I also tried the planetary button for the heck of it, as I know that clears the mask. Again, samples survive that too. They are like the undead!
I guess this is clearly a bug. Lets hope for the final release...
cheers,
jochen
Mike in Rancho
Posts: 1166
Joined: Sun Jun 20, 2021 10:05 pm
Location: Alta Loma, CA

Re: SVD Items

Post by Mike in Rancho »

Hi Jochen, thanks again (and Happy New Year! :obscene-drinkingcheers:)

I have seen that scale formula before. The follow-up question though, I guess, would be is there a way to apply and use it in a useful way? As we were talking about where binning transitions into deconvolution failure (or degraded results). Is it the oversampling/proper sampling/undersampling phasing? Still seems to me it will require trial and error, due to variable seeing. And hence, it therefore might help to more easily swap between the usually superior SVD sampling, to the "lite" synthetic version for mild stellar tightening and detail enhancement.

Sorry if I am not making much sense. :lol: I am full of stuff like this. For example my current quandary is just what does the "%" value for Airy Disk in SS actually mean? (I know that I generally decrease the parameter for wider field).

I think I might be in that zone of trying to move from beginner to intermediate, which Ivo says is difficult for ST users. ;)

A full copy of 1.7 deconv probably isn't necessary - it was more of me misreading the documentation which actually said "similar" and so allowed for some wiggle room. Overall I like 1.8 far better, even synthetic. Although there is no masking out, the deringing and other controls in 1.8 make up for that in all but rare circumstances.

I am not sure I would call this a bug either. More like extra feature(s) not implemented, and perhaps not really needed or worth the trouble. Hitting cancel and entering SVD again is a viable solution. Heck, I used to do that all the time with 1.7 HDR, since if you tried other presets there was no button to return to the exact default entry state. But new and improved 1.8 HDR took care of that and made it moot.

Besides, the button I really want is pre/post tweak in color. :D
hixx
Posts: 254
Joined: Mon Sep 02, 2019 3:36 pm

Re: SVD Items

Post by hixx »

Happy new year, Mike
The binning formula helps calculating if details are larger than 1 pixel. Let´s assume the seeing limit is 2 arcsec. the cam would have 5 µm pixels and your scope would be 1000 mm focal length. Hence image scale would be: 206 x 5 / 1000 = 1,03 arcsec / pixel: Because the smallest detail limit is 2 arcsec (seeing), you woud burn the second pixel, as it will not add any detail. On paper, you could Bin 50% in such case, but as Decon might be able to restore details, I woul give it a shot at 70% or so.
Sorry if I am not making much sense. :lol: I am full of stuff like this. For example my current quandary is just what does the "%" value for Airy Disk in SS actually mean? (I know that I generally decrease the parameter for wider field).
It makes perfectly sense to get to the bottom of ST. Understanding how it works and why is the key to yielding optimal results.
Here`s Guy's description of Airy Disk radius
Airy Disk Radius
Sets the radius of the Airy disk point spread function (PSF) that is used to diffract the light.
Default is 50% ('DimSmall', 'Brighten', 'Isolate', 'Saturate', 'AiryOnly' presets)
Range is 0% to 100%:
I think it governs the pixel radius of the Airy disk, which indeed would be lower in wide fields and / or larger apertures.
cheers,
jochen
Mike in Rancho
Posts: 1166
Joined: Sun Jun 20, 2021 10:05 pm
Location: Alta Loma, CA

Re: SVD Items

Post by Mike in Rancho »

Thanks Jochen,

Yeah I know all my image scales and in fact have all my stuff plugged into Stellarium, which gives me the calculations. But other than general concepts I don't quite know how to apply the numbers and I have no idea how to figure what my seeing is. Bad, probably. :lol:

Similar with the Airy Disk as applied in SS. I've read the documents and notes, but...percentage of what? It is just percentage strength of the slider, or is there an actual unit to the variable?

This is the kind of stuff I have to start doing my learning about. I think that I can ballpark the size of an Airy disk with something like 1.33 x F/#? That would be in microns. And then I can use my camera's pixel size to see how much I am sampling an "Airy disk." (at original scale before binning or anything). But that's just sort of theory. When I take an image, the stars are a huge range of sizes from tiny dots to, let's say, Alnitak lol. So even if I applied the pixel scale how would I know which of these many star sizes to choose to evaluate the Airy disk? And of course back to the matter at hand (which itself strayed some from the original post), is there an objective or formulaic way for me to precisely select that radius percentage in SS?

Anyway, not super important so don't worry about it Jochen! I'm just starting to get more curious about what exactly I am setting when I choose values for certain ST sliders. :think:
hixx
Posts: 254
Joined: Mon Sep 02, 2019 3:36 pm

Re: SVD Items

Post by hixx »

For Seeing, you con use meteoblue. com which has an astronomical seeing section. On a clear night it is usually around 2 arcsec
regaring the Airy disk pertcentage you'll probably have to ask Ivo for the equation. I dont really care. I just set it to whatever looks good
jochen
Mike in Rancho
Posts: 1166
Joined: Sun Jun 20, 2021 10:05 pm
Location: Alta Loma, CA

Re: SVD Items

Post by Mike in Rancho »

hixx wrote: Sun Jan 02, 2022 6:02 am For Seeing, you con use meteoblue. com which has an astronomical seeing section. On a clear night it is usually around 2 arcsec
regaring the Airy disk pertcentage you'll probably have to ask Ivo for the equation. I dont really care. I just set it to whatever looks good
jochen
Thanks, interesting site. I added it to my bookmarks. The seeing it was giving me for last night, about 1", seemed pretty optimistic? But I had to go set up and take advantage of it!

And no worries, I just like to know how all this stuff works and to consider choices that might be based on some criteria apart from just appearance. And from what I've seen the SS Airy radius is pretty difficult to tell, though I did finally run a couple with varying settings so I could blink between them and see that there were in fact differences.
Post Reply