AstroTortilla Indexs Calculation

[SniffTheGlove]

Hello,

Can anybody here shed some light on how to work out the calculation to find which indexs you need to download and use?

What has got me confussed is the calculation they use in the HELP file.

They say to work it all out you need to know the sensor size and the FL you are working with, so lets just go with THEIR example.

FOV =

57.3 x (Sensor Size 17mm) = 974.1

974.1/ (Focal Length 1200mm) = 0.81175

0.81175 * 60 = 48.705

OK, that was the calculation using their own figures, but they get their final answer as 50 and not as 48.705 (or 48 if you make it an Integer)

OK, let just go along with 50 for now. From the HELP file it then states that with 50 you will use levels 4004 to 4009 which is a range 8 to 60

Where did this 8 and 60 come from?

There is some text that the quads will be 10% to 100%, so 10% of 50 is 5 so this can not be the 8 they are talking about, even from the text 5 releates to Index 4002 (4 to 5.6), so why are they choosing 4004 which is 8 to 11).

Really confused by all this naff maths. Looked elsewhere for a description and no-one knows and just guess what indexs they use.

Does any here understand how to do the calculation and workout which index I might need.

Canon350d sensor size 22.2mm x 14.8mm

Lenses 50mm Prime Lens (EOS 50mm 1.8) and a 300mm 5.6 Soligor Prime

Thanks

[themos]

The only way to know for sure is to grab all the indices and see which ones are being used for your various fov's.

[madatter]

I gave up trying and downloaded 4002 to 4019, i then ran a few photos from my 40d and found it was solving using the 4004 index so i backed up all the index so i have them safe then deleted them apart from 4003, 4004 and 4005. It takes between 20 to 30s to solve on my old laptop.

[IanL]

Upload an image to astrometry.net, which will (hopefully) solve it blind. That will give you some starting information:

- Radius of the image in degrees, double it to get the rough FOV size that you need.

- Image scale in arcseconds per pixel, again can be used to work out which index files to start with, and also useful for setting the upper and lower limits in AT (remember to change the scale parameter to app).

- If you click the "Submission nnnn" link on the solution page, it will show you the log file for that solve. Near the bottom it will tell you which index file solved the plate. These will be the old 200 series index files, but you can use those to find the corresponding index files in the new 4000 series files (or maybe follow the process to request access to the 200 series and use them instead).

Whilst it should, in theory, be possible to work out exactly which indices you need, in practice it doesn't seem to work out like that. The theoretical image scales calculated using formulae don't always correspond to practical use so best to measure them with a plate solve. The 10% to 100% of image scale rule is just a heuristic. Which index will solve a plate depends on lots of variables; the features in the plate (and their scale) for one obvious thing; the quality of your image for another; the parameters used in AT for a third; it doesn't seem to work out of the box for most people and setting sigma values, etc. will affect what is actually detected as a source and what is rejected.

If you have the time, disk space and bandwidth to do so, there is no real harm in downloading all of the available index files. If you don't or can't do that, what I would do is this:

- Take an image of a known target and have astrometry.net solve it. That will prove it is solveable by engine (which is the same online as used by AT). If you don't, you may be tearing your hair out trying to solve a bad image!

- Then use the readme to work out which index files cover the area that you are imaging, at each of the available index scales. Most of the bandwidth sapping downloads are for the numerous and very large small-scale index files.

- Download all of the 'whole sky' large scale index files, and then only the index file that includes your target area for each of the smaller scales.

- Install those files and now try to solve your known plate. Play around with the AT parameters (see the other thread under software for guidance). Once you have a solve you will know which index file worked.

- You can now download the rest of index files of that scale plus start working up and down the scales as needed.

- Remember that it is unlikely you will need all of the files for a given smaller scale; spend a bit of time working out which bits of the sky you can see from your latitude and which you can't, and don't bother with the rest of the index files that you can't see.

However you do it, once you have a working set-up, remove the unwanted/unused index files from the directory so that AT doesn't waste time looking in them. Too many inappropriate index files is one of the causes of slow solving.

[Gina]

Thank you for all that I have just installed AT on my netbook ready to try out when the weather permits.

--- snip ---

However you do it, once you have a working set-up, remove the unwanted/unused index files from the directory so that AT doesn't waste time looking in them. Too many inappropriate index files is one of the causes of slow solving.

May do that later but I'll probably leave all the index files I've downloaded (4005-4019) in-situ for a while. I can cover most of the sky at one time or another and use focal lengths from 55mm to 600mm ATM. I hardly need AT for the shorter FLs as CdC gets me close enough but at 200, 510 and 600mm it should be very useful. Currently I can spend half an hour getting my target into the frame of my 314L+ with the ED80 at 510 or 600mm FL. I've been using my DSLR finder (1100D with 200mm lens) after slewing with CdC to move the target into the right place in the finder image using 20-30s exposures and EQASCOM to move a bit at a time. Once the target is in the CCD camera frame it's easy to frame it using Artemis and binning with short exposure looping.

[IanL]

Thank you for all that I have just installed AT on my netbook ready to try out when the weather permits.

May do that later but I'll probably leave all the index files I've downloaded (4005-4019) in-situ for a while. I can cover most of the sky at one time or another and use focal lengths from 55mm to 600mm ATM. I hardly need AT for the shorter FLs as CdC gets me close enough but at 200, 510 and 600mm it should be very useful. Currently I can spend half an hour getting my target into the frame of my 314L+ with the ED80 at 510 or 600mm FL. I've been using my DSLR finder (1100D with 200mm lens) after slewing with CdC to move the target into the right place in the finder image using 20-30s exposures and EQASCOM to move a bit at a time. Once the target is in the CCD camera frame it's easy to frame it using Artemis and binning with short exposure looping.

- You may already have some images on your computer from each of the focal lengths you are using. use a single unprocessed sub, as that is what you will be using in the field. You don't have to be connected to the scope or camera to test/sort out the parameters, AT will solve and show you the RA/Dec just fine by using the 'file open' camera.

- You will need a different options set for each focal length you are using. It is easy enough to to save and load a set of named options for each camera/lens. It is quicker to specify the no-plot option for production use, but you may find it useful to turn on plotting when you are testing so you can look at the solution produced.

- If you have only gone down to scale 4005 then you definitely need all the index files for UK latitudes. If you find end up needing any of the files from 4004-* downwards, then consult the healpix map here:

http://trac.astrometry.net/browser/trunk/src/astrometry/util/hp2.png

In that case you will not need any of the files ending in numbers that are below the lowest declination that you plan to image, which saves a huge amount of bandwidth and avoids AT trawling redundant indices when it does a blind solve (i.e when it is not connected to a scope and doesn't know the rough location of the image).

- Do a blind solve on an image, I find uploading to astrometry.net is the best bet as it usually solves, where AT may fail if you haven't got it sorted out yet. Look at the reported image scale (arcseconds per pixel), and that is what you need to determine the minimum and maximum image scale settings in AT. Go a bit under and a bit over for the min and max settings. Start with a wide margin, maybe half and double the reported image scale, you can refine this to a narrower range once you have it working. Remember to set the units for the image scale to "app".

- Set your search radius to 45 degrees. When you are doing a blind solve this is ignored anyway, but when you are connected to a mount and do 'capture and solve', AT will not search for a solution more than 45 degrees away from the mount's reported position. You may be able to reduce the search radius a lot more than this, but I haven't tried yet.

- Remember to use the -r and -c <number> options in the 'additional parameters' box, as recommended on the AT project homepage. This has made a huge difference to solve times and success for me. The -r option sorts the sources (stars) in order of brightness and solves on the brightest ones first. The -c option limits the number of sources used to a hard number, which is more useful than simply using the sigma option to reduce the number of sources detected. I'd try maybe 200, 100 and 50 as experimental values, as if you can't get a solve on the 200 brightest stars in the image, you are probably going to be waiting a long time only to have it fail anyway.

- Then it is a matter of experimenting with the sigma value. Start with 100 and see how you go. If you are getting very few sources detected (low tens) then decrease the sigma value. If you are getting too many (high hundreds or thousands) increase the sigma value.

- The downscaling parameter may also be useful if you have really big images, as it speeds up solving, but equally it may shrink the image too much to have stars successfully detected in some cases.

- If you are still having major problems after all this, check the edges of the image for coma. A good test is to crop out a central portion of the image and solve that, since I did have issues with non-round stars when my LP filter wasn't properly screwed in to the holder. I can't remember the parameter, but there is one on the AT page to relax the star shape algorithm to cope with less well-focussed images that may help.

[Gina]

Thank you very much Ian - very useful info

[Gina]

It works! I just set it going on one of my subs without setting any parameters (except camera to "Open file dialog") and it solved it Took nearly 400s but it got there Now to tweak things... I've also fed two subs into astrometry.net and got them solved so I know some good starting settings.

[Gina]

I tried adding the -r and -c switches as suggested ie. --sigma 200 -r -c 200 and it stopped working. Removed the -c 200 and it's working again. Anyway it's solving in about a minute, with whatever I've thrown at it so far (on the desktop). Single sub TIFF files. astronometry.net told me the 314L+ with ED80 at 600mm gives just over 2 arcsec/pixel and at the other extreme with 55mm lens it's 23 arcsec/pixel. Setting the range to 2 and 40 and it solves everything I throw at it.

I did have a little play out in the obsy (no stars though). AT will recognise and take images from my QHY5 guide camera but inspite of the ASCOM camera showing Atik and Atik 2 I was unable to get it to recognise the Atik camera - it just hung up. I guess I need to install a driver or something to use the Atik camera directly for image grabbing. This would be the ideal setup and I think it should work. I could use the QHY5 before I set up guiding I guess.

[IanL]

Sorry my bad, the -c flag is the one that controls star roundness defaults to -c 0.01 but the authors recommend -c 0.02 (maybe a bit more) if you are a 4000 index series user. The number of objects flag is --objs N ( note is is a double -- not a single - for this parameter, typically you should be aiming for 50 to 500 objects, but I found more than 200 took a long time). Definitely use the -r flag.

I'd definitely use different set-ups for different focal lengths to speed things up, 2 to 40 app is a big range and it is trivial to load a different parameter set to speed this up.

Don't know about your camera driver woes; are you trying to call the ASCOM cameras from within AT? If you have APT installed, maybe try the CCD mode with the Atik and see whether it works. If so you could just use the APT setting in AT to take the alignment images. Worst comes to the worst, use the file open, it's still quicker than anything else. You should be aiming to get solves below 30s, anything over that is a bonus but I have got to 6.5s on some images.

[Gina]

Sorry my bad, the -c flag is the one that controls star roundness defaults to -c 0.01 but the authors recommend -c 0.02 (maybe a bit more) if you are a 4000 index series user. The number of objects flag is --objs N ( note is is a double -- not a single - for this parameter, typically you should be aiming for 50 to 500 objects, but I found more than 200 took a long time). Definitely use the -r flag.

Thank you Ian - that explains it Yes, I'm using the -r flag but was surprised to see it actually solve on the second or third set of 10 objects (at least I think that's what it was telling me). I'll try the --objs N, that ties in with the info on the web site

I'd definitely use different set-ups for different focal lengths to speed things up, 2 to 40 app is a big range and it is trivial to load a different parameter set to speed this up.

Yes, I will do that, particularly as I expect the netbook to be a lot slower.

Don't know about your camera driver woes; are you trying to call the ASCOM cameras from within AT?

Yes, I found the ASCOM camera option and I feel sure I've read that the Atik cameras can be accessed that way (could be wrong - it's not unusual )

If you have APT installed, maybe try the CCD mode with the Atik and see whether it works. If so you could just use the APT setting in AT to take the alignment images.

Yes, I can certainly try that.

Worst comes to the worst, use the file open, it's still quicker than anything else. You should be aiming to get solves below 30s, anything over that is a bonus but I have got to 6.5s on some images.

Yes, there is that option, of course Still much quicker than the faffing about I have been doing.