The Earth is an imperfect flat field target because it is too bright for the WFPC2 in the broad-band green and red filters. The rapid motion of the HST also creates streaks across the flat field images. The removal of the streaks requires the combination of multiple Earth observations with the streaks at different angles on the CCDs. An extensive discussion of the generation of Earth flat fields is available in Chapter 6 of the WF/PC-1 IDT OV/SV Report.
The flat field calibration system works by imaging an illuminated diffuser plate onto the WFPC2 exit pupil (relay secondary) by means of an MgF_2 lens. Two lamps provide optical and FUV illumination, providing a flat field which resembles the input beam from the OTA between 1600Å and 10000Å. During SLTV, flat fields were obtained using both the flat field calibration module and the WFPC2 optical stimulus (HST simulator), to generate a database of ratio images which link the internal flats to external flats.
The external stimulus flats have been updated by comparison against on-orbit streak flats, obtained for a small subset of narrow band filters (F375N, F502N and F953N). These were obtained during the initial in-flight calibration of the instrument, to effectively calibrate low frequencies in the internal flat-fielding source. Eventually, sky flats for the broad photometric filters will be made by combining many frames.
Flat fields have been obtained for all filters by combining information from SLTV test flats (which are good for all but the lowest spatial frequencies), and Earth flats in a limited number of filters (which fix the low frequency terms). The calibration channel monitors for time dependent changes in the flat fields, and none have been seen to date in the visible filters. Some redundancy is provided by the internal flat channel, but it does not form a part of the baseline calibration.
During early 1996 flat fields for many filters were updated using an improved illumination pattern derived from large numbers of streak flats. Corrections were typically 1%, though the outermost corner of WF2 showed a 7% correction. The improved flats compared very favorably with sky flats.
Note that the flat fields presently used in the pipeline are based on gain 14 data. The gain ratios are not constant from chip to chip, and therefore a small correction to photometric results derived from gain 7 data should be applied (see Table 4.3 on page 82