9.1 Noise Models

Two portions of the low-dispersion extraction procedure rely upon an estimate of the detector noise in order to perform their functions. First, the cross-dispersion profile fitting routine utilizes the estimated S/N of each wavelength sample in order to calculate appropriately weighted spline fits to the data. Second, the extraction procedure uses the noise model information to derive an error estimate for each point in the extracted spectrum. The noise models are derived empirically for each camera by measuring the scatter in the flux numbers (FN) around the mean FN in the background regions of several hundred science and flat-field images taken at a variety of exposure levels. Since the sigma as a function of FN is wavelength-dependent, the analysis is performed within 20 equal-sized wavelength bins ($\sim$54Å wide for the SWP and $\sim$85Å wide for the long-wavelength cameras) in the low-dispersion SI. For each wavelength bin the standard deviation in FN versus mean FN is represented by a third-order polynomial. The wavelength-dependence of the four coefficients of this polynomial are then each represented with a third-order polynomial to allow a determination of the expected standard deviation of any pixel given its wavelength and observed FN.