New user struggling with Photometric Color Calibration

[thumdugger (Scott Barnett)]

Hi everyone,

Been doing astrophotography stuff for about 3 weeks, now. I’ve been using Siril pretty much exclusively and enjoy using it … except for plate solving and photometric color calibration parts which are hurting my soul right now.

For the life of me I cannot figure out how to use this tool effectively. It has worked for unknown reasons a few times allowing me to use the WCS overlays and/or get calibrated color. However, well over 90% of the time I just get failure messages that make no sense to me.

The message I least understand is “not enough stars selected” as the dynamic PSF can usually easily find hundreds of stars in the image and certainly more than the six that keeps being asked for.

I have tried an enormous range of images, have kept things as clean as possible, and have also blown the stretching all the way up to, and including, histogram stretch levels and it just never seems to see what it needs or it suggests, generically, that my focal length, etc., values may not be correct.

I have manually selected stars, I have constrained search areas, and I have down sampled until I’m blue in the face but nothing works often or consistently. I have even had the calibration or plate solving work, reloaded the image, repeated the steps that led to success, and then had it just fail to find anything.

This evening I grabbed an image I took of ZTF c/2022 E3 this morning and tried for hours but could not get it to be recognized. I then put the image up on nova dot astrometry dot net to see if it could find anything and it found things perfectly.

I realize astrometry.net is a more powerful and purpose built tool; I’m not comparing the two.

The important take away for me was grabbing the updated fits file with the WCS data AND all of the focal length, pixel size, resolution, etc. information and then trying a photometric calibration on that file (that already had all the resolving done to it). It still failed even with all of the correct information supplied by astrometry dot net.

I even went back to the original file I had uploaded, oriented it exactly like astrometry dot net had it, fed the Image Plate Solver (and the calibrator) the exact same numbers astrometry dot net supplied and it still complained about not enough stars. And dynamic PSF was still finding over 170 stars without issue. And tweaking the file to ridiculous level of contrasts and bloated stars only resulted in “probably wrong focal length, etc.” messages.

I am brand new to all of this and completely willing to believe that I am the problem here. It is just that the messages Siril provides haven’t led me to any better understanding of what is going on, I haven’t found any documentation that goes into detail about issues and solutions to try, and I have no idea why the Image Plate Solver (and the calibrator) can’t find stars even when the dynamic PSF can find hundreds.

Any useful suggestions would be warmly received and greatly appreciated. Sorry for all the “dot net” parts but new users are only allowed four links and even though I’m not linking anything the server thinks I am so … dots it be.

-thumdugger

[lock042]

Hello and welcome,

@thumdugger

The message I least understand is “not enough stars selected” as the dynamic PSF can usually easily find hundreds of stars in the image and certainly more than the six that keeps being asked for.

This could occur if you got an active selection on the image. Then the search is only done in the selection.

@thumdugger

I have tried an enormous range of images, have kept things as clean as possible, and have also blown the stretching all the way up to, and including, histogram stretch levels and it just never seems to see what it needs or it suggests, generically, that my focal length, etc., values may not be correct.

This tool MUST never been applied on stretched image. Always on RAW and linear data.

@thumdugger

The important take away for me was grabbing the updated fits file with the WCS data AND all of the focal length, pixel size, resolution, etc. information and then trying a photometric calibration on that file (that already had all the resolving done to it). It still failed even with all of the correct information supplied by astrometry dot net.

This is perfectly normal with 1.0.6 because It does not use the WCS data and need to redo the astrometry by itself. This behavior will change with 1.2.0.

It is difficult to help you without any images and/or screenshots. We need more information about your data.

[thumdugger (Scott Barnett)]

Hi lock042,

Thanks for getting back with me on this!

@lock042

This could occur if you got an active selection on the image. Then the search is only done in the selection.

Occurs with or without active selections and I am able to verify, I think, at what it is looking at via the console. Here is one from my failures from last night:

Blockquote
21:40:05: Found 515 stars in image, channel #1
21:40:26: Solving on selected area: 0 0 1230 1845
21:45:05: Findstar: processing…
21:45:06: Catalog NOMAD size: 200000 objects
21:45:06: Plate Solving failed. The image could not be aligned with the reference stars.

In this particular case the width of the cropped image is 1230 and the height is 1845. This series of messages would be an example of the “focal length, etc.” not the “need six stars” message.

@lock042

This tool MUST never been applied on stretched image. Always on RAW and linear data.

This one intrigues me.

I only use it in linear view but I’m not sure what you mean by “RAW”.

What I meant in my description was that I’ve done everything from leaving the stacked image untouched (and very, very, dark) all the way up to stretching the image, even to extreme levels, hoping that making the stars “more human visible” would help it (although I can’t see why it would but sometimes wildly flinging oatmeal against the wall gets something to stick that you weren’t expecting).

But what do you mean by “RAW”, in this context?

Should I be doing the plate solving on a sequence (after registering but before stacking)? Or only on uncropped images whether they are stacked, or not? Or should they be done before any adjustments have been made (no “extract background”, stretching, etc.)? Should the images not be rotated even if I’m rotating them to what a person would see if viewing the image in real life?

I think I’ve tried every variation of this I can think of but this might be where I’m screwing things up the most.

Another area of concern is orientation and focal length. These most recent images were shot wide on a Viltrox 13mm f/1.4 lens and because I was shooting basically straight at Polaris to get ZTF c/2022 E3 the camera was mostly on its side and swung upwards at probably 65-ish degrees so my tripod wouldn’t fall over.

The images that are absolutely always failing are all from my Viltrox 13mm f/1.4 lens. Is 13mm to wide? Is there a range of focal lengths that should be avoided or need special care to make them work?

There isn’t any guidance in the GUI or documentation on this point so I’ve assumed it isn’t an issue but as I think about it maybe there is a “too wide” that will pretty much guarantee failure? I have failures on all my lenses include my Nikon 80-200mm f/2.8 and my Nikon 70-300mm f/4.5-5.6 lenses but these could all be failing for different reasons and maybe my 13mm is mostly failing because it is too wide? Maybe?

The only successful calibration I had came from a Sigma 18-50mm f/2.8 lens shot at 23mm. Even then that image only worked because for whatever reason the calibration briefly showed me the area of the image it found stars in (it usually doesn’t show me anything but the failure messages) and I noticed that all of the circled stars were in the very middle of the image. I then did an active selection of that area of the image and it suddenly worked great.

In general neither Image Plate Solver or the calibrator (which I think uses the Image Plate Solver) shows me what it found so that I can get a hint on what it is seeing. I find it helpful when it does because at least I can gauge better why it might be failing even if I don’t know what to do to help it. Even the astrometry dot net files miss tons of stars in the image so it is totally reasonable to me that while it can see a star that doesn’t imply it knows what the star is.

Is noise an issue? Most of the time I’m shooting at iso2000 to iso6400 depending on the target. Sometimes I can get it down to iso800-iso1250 but those are usually my “longer” exposures of 15s to 30s when basically pointing straight up.

Also, how “pure” should the image be? In my case these most recent images have probably at least 40-50% not stars. There are trees, wires, an occasional telephone pole, or even just my (or my neighbors’) houses and roofs. I’ve been cropping and rotating everything out so there is only stars but maybe this is contributing to my failings? I remove them because I noticed that dynamic PSF loves to interpret parts of the trees as stars (birch bark is very white) and all the snow on the roofs also produces lots of false stars, too … so I remove most all of the non-stars before trying to plate solve.

I can totally share images I’m just not sure how to. Maybe I’ll finish the tutorial on this site to see if it shows me how to do that here. Or should I be posting them elsewhere and sharing a link?

Again, thank you for the help. I’m loving this stuff and Siril is my main tool for learning about these things

[vinvin (Vincent Hourdin)]

@thumdugger

Catalog NOMAD size: 200000 objects

I haven’t real everything, this line here is already a bad sign. You should never get so many objects from the catalogue, I’d say the limit magnitude is very wrong and the field too big.

@thumdugger

13mm f/1.4 lens

This tool will probably fail for most images made with a focal length of less than 100mm, there are just too many stars and distorsion, and having an object near the centre will help too.

[lock042]

@thumdugger

The images that are absolutely always failing are all from my Viltrox 13mm f/1.4 lens. Is 13mm to wide? Is there a range of focal lengths that should be avoided or need special care to make them work?

Oh ok. Not a lot of time for a big answers, will do it tonight. But siril will always fail to solve wide field because of distorsion. So in fact this is normal here.

[thumdugger (Scott Barnett)]

Hi vinvin,

@vinvin

I haven’t real everything, this line here is already a bad sign. You should never get so many objects from the catalogue, I’d say the limit magnitude is very wrong and the field too big.

This isn’t the normal return. I, probably out of frustration, appear to have turned of the auto magnitude off an that run and cranked it to 18 to see if it could “find more than six stars, now”. Looking at the console history it looks like the previous 20 attempts were returning 700-800 per attempt. Maybe I should try to go the other way and manually limit things to some brighter magnitude to reduce the star count way down.

@vinvin

This tool will probably fail for most images made with a focal length of less than 100mm, there are just too many stars and distorsion, and having an object near the centre will help too.

This is really good to know. It fails on my 100mm+ focal length images, as well, but this could easily be because of other things I’m doing wrong on top of any “too wide” issues.

The only image that has worked was 23.3mm on my 18-50mm but that one fails most of the time, too, and the only time it didn’t fail was when Siril showed me a bunch of circled stars near the center … I then actively selected that center area and the next attempt worked.

Siril usually doesn’t show me what the plate solver was using and I don’t know how to feed it information it needs. The dynamic PSF often finds tons of “stars” (which I’m thinking only have to pretend to be stars to get picked) which is part of my confusion; I don’t know why dynamic PSF can see so many but plate solver seems to either see less than six or when it sees more than six it still fails and almost never shows me what it did see.

Even manually picking a stars dozen bright stars can still trigger a “need at least six stars” message so I’m not even sure it is paying attention to my manual selections.

Would running a lens distortion correction of the image help at all (like from Light Room or some other similar thing)?

Maybe Siril clarifying in the “focal length wrong, etc.” message it could include a bit about reasonable minimum focal lengths or even disabling the functionality unless the image is above a certain level that has a chance to succeed?

I don’t know just thinking out loud as I have spent many, many, hours trying to figure what I’m doing wrong when the answer for most of this lens’s images may just have been “probably won’t ever work no matter what you do”. It is a rectilinear 13mm lens and even cropping to just the center where there should be no distortion to speak of doesn’t seem to work so I’m really unclear on what the actual issue is.

I’ll try testing this part out more to see if I can verify this as the issue … it might give me some insight on why my longer focal lengths are failing, too.

Thanks for this info … could be a key insight to something that just dawned on me this morning as a reasonable cause of the problems I’m having!

[thumdugger (Scott Barnett)]

Hi lock042,

@lock042

Oh ok. Not a lot of time for a big answers, will do it tonight. But siril will always fail to solve wide field because of distorsion. So in fact this is normal here.

Yeah, vinvin was calling out the same thing so that is what it is looking more and more like.

That Viltrox lens is a rectilinear lens and the area I was cropping the latest image out of is very near, or directly in, the center so I would have thought that distortions would be minimal or even non-existent. I also cropped the area small (hoping to further reduce any issues) and threw adjusted focal length numbers at the plate solver that seemed to work for astrometry dot net but those numbers might be wrong or not particularly helpful.

Being that I’m horribly new to all of this astro photography (less than a month and only four or five so-so bortle 4-5 shooting opportunities or open areas between the clouds) stuff and new to photography in general (maybe six months total) there are certainly going to be massive gaps in my understanding of things. You know, things like what “rectilinear” actually means in a practical sense.

I have been starting to shoot with higher f/stops (f/6.3+) because I have noticed that the stars get crazy bad distortions when I am shooting in the f/1.4 to f/2.8 range). Shooting at f/1.8, which is a full stop below f/1.4 yielded my many squares, some triangles, and more than a few “brushstrokes”, for lack of a better description, when it came to the stars and even f/2.8 returned some questionable star fields.

Shooting at f/6.3 and f/8 seems to give me much better round stars (FWHM of 3.84-4.29 on all my frames) but I don’t actually know what counts as “good” here. The little bit of info I have come across suggests FWHM numbers in this range should certainly be good enough for decent stuff but I haven’t yet fully vetted that assumption.

Anything additional thoughts you have on where I might be messing up and what to do about that I’d be extremely appreciative of. Should I rotate the images to a certain presentation for better success? Should I only crop after doing plate solving (still fails when I do that)? Should the image have all foreground removed? Should I always actively select an area to plate solve and should it be certain dimensions, aspect ratios, etc?

I’m having tons of other learning curve issues that I’m happily working my way through but this one I’ve made almost no headway on in the last couple weeks and can’t see how to improve things. This one forum post has already clarified quite a bit regarding most likely issues and I’m very grateful for the assistance I’ve been given so far.

Thanks, everyone!

[vinvin (Vincent Hourdin)]

The stars you can see in the Dynamic PSF window will be the same as those used by the plate solver. You may try increasing the sigma value, which is the detection threshold, to keep only brightest stars. Then when it does its thing, it tries to identify those stars with the stars of the catalogue, and the number of 6 you are referring to is how many were matched between image and catalogue. That’s the part that is not working well for this kind of image. In general it works better if there are about as many stars in each and a hundred should be enough, so don’t go too high in magnitude on both sides.

I’m not sure running a distortion correction will help.

New version of Siril has the ability to use WCS information from other solvers and perform PCC on all stars from the catalogue, that will improve things a lot.

Good luck

[lock042]

Keep in mind that doing photometry on stars no bigger than one or two pixels makes little sense. So PCC is not really dedicated for wide field images.

[thumdugger (Scott Barnett)]

I didn’t realize there was a minimum-ish size so that is good to know.

Looking at the image a great many of the stars show up in the range 5x5 to 12x12 for the lighter grays of the star body with the darker grays being more like 11x11-20x20 (images from a slightly stretched stack in linear mode):

Are these too small for typical success? What would be more representative of sizes that I should be going for?

My hope for this image was that I could use the photometry to adjust the overall color of the image so that the comet colors, if any were actually caught, could be preserved as much as possible. The “Remove Green Noise” function from the Image Processing menu did exactly what its name implies and left everything, “not green”.

This might be a hopeless pursuit, I don’t know, but while processing the imagery I couldn’t think of any other way other than “eyeballing” it to remove the “green noise” from the imagery without removing the actual green of the comet coma (if the camera even caught any of it which it possibly couldn’t shooting this wide).

Thanks again for the help.