Solve-field ignoring flags

[Nate]

I've been running astrometry.net, version 0.44 on my local machine, and I've been trying to solve about 200 fields that all come from the same telescope/detector.  So far, I've had a "success" rate (insofar as ".solved" files are produced) of only 64%.  I've noticed that when I pass certain flags to solve-field, it doesn't seem to notice them.  This is my current command:

'solve-field --no-verify --scale-units degwidth --scale-low 0.15 --scale-high 0.16 --no-plots --cpulimit 60 '

First, some fields will solve quickly while others will hang up, and when they do hang up, I've noticed that the solver will try a lot of index files that correspond to fields outside of the scales that I have set (my field is about 9.3' x 9.3').  For example, it will check indices 4202-* 4203-*, and 4205-*.  My understanding is that it should be checking indices between 0.1-1 x the width of the field, so why would it check 4205-*?  Doesn't that correspond to 11'-16'?  

Second, it seems to completely ignore the cpulimit I've set for it.  It just goes and goes with no indication of when it will give up.  In fact, it seems to ignore the default 600 second limit, as I've left it to solve images overnight and come back the next day to find it still working on an image!

I'm calling astrometry.net using a Python script, as an os.system call, if that helps.

Thanks,

Nate

[Dustin Lang]

Hi,

4205 contains features 11 to 16 arcmin, and your field is 13.2 arcmin across the diagonal... so there is some overlap, and it will load that index.  It also does some checks so that it drops features larger than 13.2 arcmin early in the processing.

It's possible to reduce that 0.1 to 1 range, via --quad-size-min, --quad-size-max.

The CPU time limits are messed up.  It *may* work correctly if you have the "inparallel" option set in your config file.  Otherwise I think it applies the limit to each phase (10 objects in a single index), rather than cumulative.  I forget, I may have improved this since 0.44... (someone should force me to write a CHANGELOG).

Another thing you can do is set --objs 100 or something, which cuts the object list down to that may of the brightest objects.  It'll still take a long time to churn through that many though!  (And it only uses those stars for verification, so don't set it too low.)  And there is also --depth, to set the number of objects it will try to build quads from.  You could have a look at the ones that are succeeding to see what reasonable values will be for those.

cheers,
--dustin