We present two subtle charge transport problems revealed by the statistics of flat fields. Mark Downing has presented photon transfer curves showing variance dips of order 25% at signal levels around 80% of blooming. These dips appear when substrate voltage is raised above zero, for - 0V to 8V parallel clock swing. We present a modified parallel transfer sequence that eliminates the dip, based on the hypothesis that it is caused by charge spillage from last line to the 2nd last line. We discuss an experiment to test whether the electrode map is incorrectly reported in the data sheet. A more subtle dip in the variance occurs at signals around 6000 e-. This is eliminated by increasing serial clock high by a few volts, suggesting the existence of a small structural trap at the parallel-serial interface. Tails above blooming stars are suppressed using an inverted clocking during readout and a positive clocking during exposure to maintain sharpness of the PTC. We show that integrating under three parallel phases, instead of the two recommended, reduces pixel area variations from 0.39% to 0.28%, while also eliminating striations observed along central columns in pixel area maps. We show that systematic line and column width errors at stitching boundaries (~15 nm) are now an order of magnitude less than the random pixel area variations.
Download