Both GHRS photocathodes have granularity - irregularities in response The effects of these irregularities could in principle be removed by obtaining a flat field measurement at every position on the photocathode, but that is, in general, impractical. Instead, the observing strategy is to rotate the carrousel slightly between separate exposures and so use different portions of the photocathode. This procedure is called an FP-SPLIT, and with it each exposure is divided into two or four separate-but-equal parts, with the carrousel moving the spectrum about 5.2 diode widths each time in the direction of dispersion. These individual spectra can be combined together during the reduction phase.
COMB
The diodes in the Digicons also have (slight) response irregularities. The biggest effect is a systematic offset of about 1% in response of the odd-numbered diodes relative to the even-numbered ones. This effect can be almost entirely defeated by use of the default COMB addition procedure. COMB addition deflects the spectrum by an integral number of diodes between subexposures and has the additional benefit of working around dead diodes in the instrument that would otherwise leave image defects.
STEP-PATT
The Digicons' diodes are about the same width as the FWHM of the point spread function (PSF) for HST. Thus the true resolution of the spectrum cannot be realized unless it is adequately sampled. That is done by making the magnetic field move the spectrum by fractions of the width of a diode, by either half- or quarter-diode widths, and then storing those as separate spectra in the onboard memory. These are merged into a single spectrum in the data reduction phase. The manner in which this is done is specified by the STEP-PATT parameter, described in more detail HERE. The choice of STEP-PATT also determines how the background around the spectrum is measured.