The ISOCAM flat-field comprises two components: (1) the pixel-to-pixel variations in flux sensitivity - the detector flat, and (2) optical vignetting - the optical flat. The latter is most important for data taken with the 6.0 arc-seconds pixel-field-of-view, and can generally be ignored for the other magnifications. The flat-field is split into detector and vignetting components because the vignetted, or purely optical, component can move with respect to the detector due to ISOCAM wheel positioning jitter. The wheel jitter is the mis-alignment of either the filter wheel or the lens wheel with the optical axis of the telescope resulting in similar ``jitter'' in the optical flat-field.
As a general rule, the best CAM flat-fields are those which can be derived from the data. For example, from median combining individual images in a high-background, deep, raster by using 'auto' option in the CIA routine corr_flat. The SLICE package (also available with CIA), offers another approach for constructing the so-called "time-dependent" flat-field. This type of flat-fielding not only accounts for the pixel-to-pixel variations in sensitivity of the detector, but additionally attempts to account for the pixel-to-pixel variation is the transient stabilization times of the detector. The time-dependent flat-field approach has been very successfull for rasters, especially those containing low-contrast extenden structures.
Users should note that incorrect flat-fields can intensify any residual transient artifacts and significantly affect the flux calibration, detection, and morphology of sources. It is a good idea to compare non-flat fielded images with those that have been flat-fielded. In particularly look for features that only pop out after flat-fielding.
If libarary (or other flats not derived from the data) are used, you should also look for the effects of the ISOCAM wheel jitter. These can be recognized in your data by carefully examining the images for residuals which follow the contours of the circular vignetted pattern, or which leave one row or column or corner too bright or too dark. Some images illustrating these effects will be placed here shortly. If you find such artifacts in your data, one solution is go through the entire flat-field library to see if a different flat provides a better reduction.
Recommended reading: