Two basic criteria are used to determine if a pixel in a particular chip should be declared dead and masked off from use in the data processing pipeline:
The second category includes excessively noisey or erratic pixels, and also pixels with low responsivity and thus excessive noise after responsivity correction. Areas of the sky covered by masked pixels will show increased noise due to a reduction in the number of measurements at that point and will decrease the compleatness if the pattern of masked pixels results in places with zero coverage.
For the protocamera data processed to date, ``bad'' pixel masks have been determined by analyzing sequences of frames taken on a dark slide (``darks''), and analyzing dark sky data during computation of sky flats. The following criteria were used:
The following table gives statistics on dead and anomalous pixels from protocamera data analysis:
The first three rows of this table are for the original chip, which developed several bad columns which progressively worsened with time. This chip eventually shattered when the optics were upgraded in the camera and was then replaced. Statistics are available for the new chip from only one observing run in June of 1994. The bad columns were originally oriented parallel to the scan direction, which was a particularly bad configuration, spliting the scan coverage down the middle. The chip was rotated for the next run so the dead columns were perpendicular to the scan direction and were covered by overlapping frames.