SIGNAL-TO-NOISE RATIOS IN IUE LOW-DISPERSION SPECTRA .2. PHOTOMETRICALLY CORRECTED IMAGES

The character of detector noise is explored in photometrically correct ed images from the short-wavelength and long-wavelength prime intensif ied vidicon cameras of the International Ultraviolet Explorer. A proto col is proposed for deriving realistic ''noise models''-crucial to the application of Optimal extraction algorithms like that of Kinney, Boh lin, and Neill (1991, PASP, 103, 694)-from the available collections o f UV-Flood calibration images. The protocol includes evaluation of the ''noise-filtering'' properties of the SWP and LWP cameras through 2-D spatial power spectrum analysis. The two vidicon cameras behave nearl y identically. For both, the incomplete removal of the pixel-to-pixel sensitivity pattern can lead to a factor of up to two enhancement in t he apparent noise, depending on position in the image. Even with good suppression of the pixel granularity, however, the remaining random no ise can exhibit saturation behavior that causes the S/N to cease impro ving with increasing exposure. The random noise itself exhibits a two- component character: a normal white-noise field superimposed on a filt ered (Gaussian-like smoothing) background. The influence of the smooth component varies strongly with position. Nevertheless, when all of th e relevant effects are considered, the underlying ''pristine'' noise m odels show essentially no dependence on spatial position, except for a n unusually noisy patch on the LWP camera. Two additional sources of n oise, beyond the largely photometric contributions documented above, a re microphonics and cosmic particle radiation. Microphonics are import ant in only a few exceptional circumstances, but cosmic ray ''bright s pots'' set an effective limit of almost-equal-to 4 hr on useful SWP-LO exposures of (unresolved) emission-line objects, even those conducted during low-radiation time.