Hi Christian,

> as I had to do some testing on my data anyway, I gave it a try comparing
> charon an astronmetry.net on the same machine, charon in a dosbox... The
> results are not compareable. charon took more than 1 hour and was not
> able to get me any astrometry data from the same field as astrometry.net
> did within 12 seconds.

I'd expect astrometry.net to be mostly limited by connection speed and
server demand. At least as described in their paper, the algorithm ought
to be very fast. If astrometry.net can accept a list of point sources
and return the alignment (i.e., you don't have to send the image back
and forth, just the much smaller data about what sources are where and
the astrometric reduction results), it should be _really_ fast... it
sounds as if that's what you're doing?

Charon can be insanely slow, as you found, until you have it set up
with the correct focal length and a suitable tolerance for focal length
error (that is, you may tell it you have a 1.23-meter focal length, but
will still have to allow a percent or two of error on this). The speed
is also very sensitive to the number of stars used for pattern matching.
Of course, if the pixel sizes are incorrectly set, you're almost
assuredly out of luck. But once you've got Charon set up, it shouldn't
take more than a couple of seconds per image.

Getting it set up, though, can be a pain. The focal length given
by the manufacturer is often not very close; don't just enter that and
tell Charon the scale tolerance is 1%. Charon is also picky about
inverted vs. uninverted. When people send me images and ask about how
to get them to work in Charon, I usually try the example image both
ways (inverted and uninverted) with a large scale tolerance.

You may want to try sending me an example FITS image, and I'll see
what I can do. I'd need to know what the target object was. I can then
send you the 'charon.dat' file, and you'd be able to run images without
trouble. (Keeping in mind that Charon still won't update the FITS header
with WCS data and will require an approximate RA/dec, and will still
be running in DOS. Astrometry.net will probably remain your preferred
solution.)

-- Bill