Flat-field accuracy can be as good as 5
when
the signal is stabilized. The flat-field was found to be very stable
all along the calibration; this allows the use of a library flat-field,
instead of loosing observing time by making flat-field calibration
with the ICD linked with each observation. However, a limitation
comes from the differential stabilization times between the
pixels. It would be prohibitive in flight to wait for a full
stabilization of the array, and the observer will have to make use
of the images obtained in still unstable configurations. Usually,
stabilization is quantified by comparing the mean pixel signal to
the stabilized value. In our case, it is better to determine the
stabilization by measuring the pattern noise on flat-fielded
images. The differential stabilization generates a pattern across
the images which increases the spatial noise. Low level effects
can be observed for several minutes after a flux step. The main
problems are located on the pixels at the edges of the array, and on
a line of pixels in the lower left hand side quadrant of the array which
has a slightly lower responsivity than the average. They appear in
dark on Figure 
.
A good way to remove these effects, as well as the low frequency
noise mentioned earlier (which in fact may be related to these
effects), is to use a beam-switching procedure switching the
pointing from the source field to an adjacent reference field, as is
done with ground-based telescopes. Alternatively, a micro-
scanning procedure can be used, displacing the field of view by a
few arcsec on the sky every few exposures; this can be realised
using the raster pointing mode of ISO (see also section ).