The CFHT + WFC + UH8K Optical System

In the UH8K camera the chips are laid out in two rows each of four 2K by 4K chips laid side by side. There are gaps of about 60 pixels between the sides of the chips and about 20 pixels between the two rows. However, the chips are not precisely laid out on the grid but are slightly rotated and shifted with respect to an ideal tiling grid. In addition to the somewhat irregular chip layout, CFHT suffers from a field distortion introduced by the telescope wide-field corrector; a pin-cushion type distortion with shift amplitude of about 40 pixels at the corner of our roughly 1/2 degree square field. We have solved for the projection from chip coordinates onto the sky under the assumption that for each 2K×4K image there is some smooth mapping from pixel coordinates to the rectilinear sky coordinates, but we do not assume that there is any static relationship between the mapping from exposure to exposure — i.e we are assuming that the telescope/detector system remains rigid throughout an exposure, but can deform in a smooth but otherwise arbitrary manner between exposures. As we shall see, this gives a solution of much greater accuracy. With this model one can also combine data taken with different instruments. We have solved for the parameters of this model — the ‘astrometric solution’ — by minimising the residuals in the predicted sky positions of a set of reference stars. In principle, given certain conditions on the geometry of the telescope pointings, one can perform an ‘internal’ solution using only the mosaic data in hand and without reference to any external astrometric system (one should then be able to tie this into any other chosen astrometric system by inspection of the coordinates of any pair of stars). However, with the rather small offsets between exposures employed here, this turns