[Comm2011] astrometric solution

[Konrad Kuijken]

Now that the rotator issue has been traced to the software that combined the information from the encoders of the rotator, we have a workaround (only use one encoder instead of the four existing ones) that seems to work fine. No more backlash, and the angular preset is now correct and predictable.

We're close to having an astrometric solution that is good enough for automatic guide star acquisition - that will be a nice milestone.
The WCS that is in the image headers at the moment is good to about 25 pix all over the field, including the aux.CCDs.

If you overplot USNO stars over a full science array in skycat you still see a radial pattern of residuals, up to 25 pix in the outskirts. This implies that the plate scale is wrong by about 50/16000, ~ 0.3%.
[NB there was a bug in the header writing up to now - the CD values only had 3 significant figures. Not quite enough to explain this effect though].

However, on individual chips the residuals change by about 5 pixels over the length of the CCD, implying the plate scale is wrong by 5/4000, ~ 0.12% .
So apparently there is an issue both with the crpix values and with the CD matrix.

Could someone look into this and provide updated values? I think we can assume here that the plate scale is constant over the full array, but it needs to be determined together with the proper crpix's for each CCD. Once this is set we are in business. (science frame and AG stars).

I put a frame (OBS217_0011) and a simulataneous full-frame readout of the AG CCDs in

ftp://ftp.strw.leidenuniv.nl/pub/kuijken/ocmwcs/

Thanks

Koen