I am new to plate solving and only have used it once successfully, two nights ago. Of course convenience of 'solve and sync' is tremendous, so I naturally tried it again last night, but it was a compete fail - despite having the exact same optical, hardware and software setup as the night before. I have attached the log file, but I think the relevant bit is here:
Warning 21:38:13.331561 #1 Plate Solving Failed with 255 : Read file /tmp/tmp.ppm.k196p4: 1920 x 1080 pixels x 1 color(s); maxval 65535 Using 16-bit output
fitsbin.c:467:read_chunk: Couldn't find table "kdtree_data_stars" in file "/usr/share/astrometry/data/index-4203-22.fits"
Could not find that error when doing a Google search, except one case where a person deleted the offending index file, so I tried that and still got the same error. Also, it seems like the search was taking a long time (minutes). Previously, it took only seconds. I am using ASCOM 6.5 with POTH so SC4 should have been getting the coordinates from the mount.
Note that I successfully ran the Platesolve test in the SC settings tab and also that Polar align plate solving was working just fine.
Any ideas what went wrong? I am thinking of reinstalling the indexes, but even if that fixes the problem, I am at a loss why it would break for no reason.
Info 21:51:28.265828 #1 Plate solving : simplexy: found 907 sources.
I have found that plate solving tends to work better with smaller numbers of detected stars (100-200). You can reduce the numbers by increasing the 'sigma' settting in the SharpCap plate solve settings - a little trial and error should find a good value. You may have had less stars found on a previous attempt if your camera exposure or gain settings were lower.
Another reason that the process could fail is if the telescope mount co-ordinates are wronge - SharpCap limits the search to within 15 degrees of the current co-ordinates, which should be fine, but if your telescope alignment is not set then plate solving will fail. You can always save a frame and upload to nova.astrometry.net for online plate solving (which is very reliable) to check.
The error you are getting certainly looks like a corrupt index file - it looks like you renamed the file to '.old', but the plate solving engine still used it. Maybe delete all the 4203* files completely and then reinstall that index?
>I have found that plate solving tends to work better with smaller numbers of detected stars (100-200). You can reduce the numbers by increasing the 'sigma' setting in the SharpCap plate solve settings
Ah, I was not sure what 'sources' were but assumed the more the better!
Is there a way to test plate solving on previous images I captured? It would be nice to be able to get this dialed in during the day. Darkness comes late here and I often have to setup so want to hit the ground running so to speak.
It is very possible that both the problems you mentioned occurred. I felt like the RA coordinates reported by POTH were wrong (way off) and my images looked strange as well. To top it off, my laptop mouse was misbehaving and jumping all over the place causing many settings to get inadvertently changed. All in all, a very bad night!
You can certainly test your previously captured images - easiest to use 'Test Camera 1 (Deep Sky)' - that shows a picture of M42 by default, but there is a control on the right to allow you to browse for your own image (FITS, PNG, etc) to use.
Thanks for the reply. I went out last night (before I saw your reply) and tried again and although other things worked better, the plate solve was still spotty. I tried solving in the home position first thing and was pleased to see that it did, so I thought I was set. Then slewed near Alioth in the Big Dipper and as usual, my mount was off, so I tried the solve and sync, but no luck. Long story short, ONE HOUR LATER, I finally got my second image to solve. This was near the zenith in the Cynga, so lots of stars. I used this procedure: take snapshot, if it looked good (lots of stars), I would try a solve and watch the number of sources. If the number was above 200, I would cancel, change camera settings and try again.
Along the way I found the following:
- the sigma control in the PlateSolve settings had NO effect on the number of sources
- the number of sources ranged from around 800 to almost 1400 at one point, despite the fact that my sensor is only 2MP!
- I tried using 2x binning, lower res, different picture formats
- the ONLY thing that lowered the number of sources was to lower the exposure and/or gain
Finally, I got down to two sources, kept my target the same and gradually increased the exposure and gain until I got up to about 50-60 sources by trial and error. In the end, I was back to the settings that were close to what I had used 40 minutes earlier! I was able to stay in this 'goldilocks zone' for quite a while allowing me to slew to several targets. Eventually, I tried moving to a different area and needed to do another 'solve and sync' and once again, the sources jumped up to over 900.
I think there is something wrong with the solving algorithm and/or my setup. I am using ansvr and SC4. I looked at my SC log file from last night and unfortunately, there is no info regarding the number of sources found. Maybe there is an ansvr log? I do have several of the images, along with the cameras settings, that I took while trying to get a decent number of sources. I might be able to try to solve them later to see if I can recreate what I was seeing.
UPDATE- After much tedious testing, I found the following:
- Most of my sample images loaded through the test camera do solve, but it typically take over 5 minutes. This appears to be due to my small FOV. Typical solves use index 4203 or 4204. Some images solved in less than a minute. These were all blind since I don't know a way to feed this info in this mode.
- I had too many indexes loaded, so got rid of some. This helped a little
- The images that my laptop takes 5-6 minutes to solve, Astrometry will solve in less than 1 minute.
- Changing the sigma value for these images, which typically had less than 100 sources DID have an affect and I could see a direct correlation between the sigma value and the number of sources.
- I did not run into any images that had more than 200 sources, so still mystified where the 900+ came from
- In my review of last night, I forgot that I also quit after about 5 minutes without a solve. Once I let it run for 20 and it was still going.
- The algorithm runs through the larger FOV indexes quite quickly, but when it gets to the smaller indexes (4204 and below), it can get REAL SLOW, taking several seconds per sub-index. The fewer sources, the faster it goes, but too few and it won't solve.
- I changed the ansvr setup to use binning since I saw that the online version does this.
SOOO
- I 'guess' the reason Astrometry is so fast is that it has some serious processing power, but I thought I read people thought the local process went FASTER.
- Still don't know where the 900+ sources comes from and hope I don't see it again, but suspect I will
- I hope once I get POTH connected and feeding approx coordinates to the solver, it will go more quickly.
Ran into the same problem last night - SC shows an impossibly high number of sources when trying to solve and sync. This time, I captured the images for later analysis and as I suspected, they solve just fine with a reasonable number of found sources. I am going to file a bug report on this issue.
so do you have copies of the images that generated the very large number of sources? I guess they must differ in some way from the more normal images - if you can share one or two images then we can investigate further.
