Before going through stage-2, the SF contains all the information from a given observation, irrespective of such things as data quality and Earth occultation. To filter and select the data they want, investigators will run a selector program which operates on the concatenated SF (consulting the concatenated HKF and calibration files) to produce a Filtered Science File (FSF). The FSF has the same structure and format as the SF, except for an additional extension which describes how the multi-dimensional event descriptions (such as the 3 x 3 pixel map in SIS faint mode) are to be ``collapsed" into a single PHA value for each event. This augmentation is not implemented automatically in stage-1 since it depends on the preferences of the investigator, although the GOF will use a ``standard" implementation for pipeline processing.
Generic analysis programs such as IRAF/PROS and XANADU will be able to read the FSF directly. Further transformations, e.g. to produce an image file, will depend on the analysis program used.