The third processing stage filters the combined event list to produce the filtered event list, which contains the collection of node-specific exposure maps. It also creates the separate combined exposure map. The filtering performed here is not general; it is based only on event flags and frame timestamps, and the purpose is simply to remove data that are not usable for any purpose of analysis. Combining the event filtering with exposure map construction in one step assures mutual consistency. The combined event list is not used beyond this point, and can be deleted (parameter expunge) with little risk that anything of value will be lost.
Typically the interactive user will wish to adjust the time-filtering of the data, which is done by manually constructing a supplemental GTI table (see task gtibuild) and restarting rgsproc here. If the original combined event list is missing (as, for example, in the PPS) the original filtered event list is acceptable as a substitute if no changes have been made to the original attitude and housekeeping GTI. In fact, rgsproc will automatically replace a missing combined event list by copying the filtered event list when it is started at this or the previous stage. However, this behavior can lead to unexpected results. rgsproc cannot determine whether the filtering already applied in the existing filtered event list is compatible with the filtering about to be applied. The result will always be a fully self-consistent and valid filtered event list, but it may have more events and exposure removed from it than the user is expecting. The ExtraFiltering warning message is issued whenever an automatic copy is performed, to remind the user that the result may not be the same as would be obtained using the original combined event list.
Note that the time filtering automatically includes the attitude-drift and housekeeping GTI computed in the first stage, if these files are present in the current working directory. Deleting these files is therefore one way to disable them when it appears that something may be wrong with the housekeeping or attitude data. Although these files do not exist in a PPS distribution, this is not really a problem because the combined event list does not exist either and applying them to the filtered event list would be redundant.