Depending on the observation mode a certain fraction of the events recorded are due to out-of-time events, i.e. are registered during readout of a CCD and thus will be assigned an incorrect position in RAWY and consequently an incorrect CTI correction. Typical values are 6.3% for full frame mode, and 2.3% for extended full frame mode (with frame time parameter 3).
If withoutoftime=Y the task epevents does not create the ``normal'' events file but uses the output of epframes instead to create an event file where all events are treated as out-of-time events. After the pattern recognition for the same TIME, PHA, and RAWX a new RAWY is simulated by randomly shifting the pattern along the RAWY axis and performing the gain and CTI correction afterwards.
This out-of-time events file has the same temporal variations and pattern distribution as the ``normal'' event file. All events have an energy correction corresponding to the (randomly distributed) RAWY column. You can apply a selection on both the ``normal'' and ``out-of-time'' event file, scale with a mode-dependent factor and subtract the ``out-of-time'' contribution from the other. You can create ``out-of-time''-corrected spectra and images for an arbitrary time and energy range.
The current (experimental) implementation does not automatically detect the instrument and bad pixel setting. The RAWY range to distribute the out-of-time events is instead derived from the RAWY range of the actual events in the data. The mode-dependent conversion factor has thus to be applied manually and is about 95% of the out-of-time event fraction (e.g. 0.060 for full frame mode) due to setting the top 10 rows to ``bad''. This will we improved in later versions.
A different approach is done via the setting withctisrcpos=Y: all event energy corrections do not use the actual RAWY coordinate but the corresponding RAWY location of the source (SRCPOS) as determined by the task epframes (either automatically or via setting withsrccoords=Y). So all events are assumed to originate from the source position. Events that really occur at a RAWY different from SRCPOS get a wrong (usually too high) CTI correction instead. (Note, that this procedure is the correct handling for TIMING and BURST mode - where it is performed automatically). A spatial selection should be applied in the final event analysis to deal only with detector portions dominated by Out-of-Time events. epchain may be used with the setting ccds=4 withctisrcpos=Y to process only the CCD 4 where usually the bright source is located that causes the Out-of-Time events.