Two recent visitor runs with the 4-m Prime Focus CCD Camera have impressed us with the need to carefully exorcise ghosts prior to flat-fielding in some PFCCD imaging applications. As described in the June 1997 issue of the NOAO Newsletter, the new prime focus corrector produces a faint ghost image of the telescope pupil. Since that time, the "Sol-Gel" antireflection coating has deteriorated; tests conducted by George Jacoby indicate that the ghost amplitude is about a factor of 2 worse than when the corrector coating was fresh last spring. The amplitude of the ghost is currently 1% or less at B, V, R, and under 2% at broad-band U, and 4% or less at I. The amplitude is larger for narrow-band interference filters (14% at 5007), and one can expect it to be even worse at the red or UV.
What to do about the ghost? Team Mosaic is investigating means of recoating the offending corrector element with a longer-lasting coating; we are also considering offsetting the PFCCD camera to remove the ghost from the FOV, or possibly reinstalling the old corrector for use between Mosaic runs. (Mosaic requires the larger corrector.) We do emphasize that one can adequately remove the ghost from the data by care in processing. In the direct imaging manual we rather cavalierly suggest that this ghost flat-fields out, but as both Howard Bond and Bill Harris recently pointed out to us, ghost reflections are additive rather than multiplicative, and really should be subtracted first from the dome flats, and secondly from the program frame. A prescription for removing the ghosts was suggested by Bill Harris and Doug Geisler: (a) using a narrow-band interference filter, obtain a high-contrast image of the telescope pupil ghost. Measure the background in four corners, subtract a mean background value, and thus produce a zero-background version of just the ghost. (b) on your flat-field images scale the ghost until you find an empirical match to the intensity above local background, and subtract, producing a ghost-less version of the flat-field. (c) use this image to flat-field your program field. (d) if necessary, scale the ghost to the residual in your program field, and subtract. This step should only be necessarily in cases where you are not measuring local sky.
We have conducted some on-the-sky tests which users of the PFCCD may find comforting. Using broad-band U, we did photometry of multiple exposures of a star both on and off the ghost image. We processed the frames in the normal way; i.e., without any effort to remove this ghost in either the flat-field or the program data, and compared the result to what we got by removing the ghost in the flat-field image. (After all, this is the real question: how well can you do photometry?) The six measurements showed an rms of 0.007 mag when reduced in the standard way (no ghost subtraction). Removing the ghost in the flat field alone reduced this scatter to 0.004 mag. The maximum difference we measured from "on ghost" to "off ghost" was 0.015 mag for the standard reductions; for the corrected image it was 0.008 mag. We thus believe that many people interested in photometry at U, B, V, and R can simply ignore these ghosts, although we encourage everyone to carefully examine their data.
Users of Mosaic should consult the Mosaic manual and/or the local gurus (tarmandroff@noao.edu; gjacoby@noao.edu) for recommendations concerning Mosaic. We do not know at present what solution we will be offering to PFCCD users in the spring, but observers should be prepared to perform exorcism as needed.
Phil Massey