In ACCUM mode, a raw image is constructed by STIS during the exposure. In TIME-TAG mode, STIS instead records the position and approximate arrival time of each photon. Processing on the ground converts spacecraft data to an event list in FITS format (filename suffix "_tag.fits"). Ground processing of TIME-TAG data then builds a raw image ("_raw.fits") from the event list.

We have identified four errors in the current handling of TIME-TAG data. Roughly once every 10 million events, a time reported by STIS is corrupted. Only one event is affected in the data sent to the ground, but all events until the next BUFFER-DUMP are affected in the event list ("_tag" file) produced by the current ground processing software. Events recorded during the affected time interval will all be missing, shifted by a large time offset, or artificially compressed into a shorter time interval.

Roughly once every 100,000 events, a time reported by STIS is too small by 0.000125 seconds, which is the finest time step reported by STIS. This error leads to a small negative step in time in the event list, but the times for subsequent events are correct. Given that this error is at the timing resolution limit of STIS, this error will affect very few programs.

In raw spacecraft data, roughly once every 10,000 events, a time reported by STIS is too small by about 0.032 seconds. Errors of this type are generally trapped and corrected by existing ground software, but the current algorithm is imperfect. For long exposures at global count rates less than about 500 counts/sec, roughly once every million events, a time in the processed event list ("_tag" file) will be too small by about 0.032 seconds. In some datasets where the STIS buffer fills before the scheduled BUFFER-DUMP time, processed event lists contain so many -0.032 second jumps in time that all the events are artificially compressed into a few seconds.

Finally, event list ("_tag") files contain in the last FITS extension a table of good time intervals, which indicate time intervals when events were actually recorded. The exposure time in the FITS header currently contains the requested exposure time, rather than the total duration of the good time intervals. This leads to incorrect net count rates and fluxes in extracted spectra ("_x1d" files), if the actual exposure time is less than the requested exposure time. Fluxes will be incorrect when a corrupted time creates a gap in the event list (see above) or when the STIS buffer filled in less than the requested BUFFER-DUMP time, which effectively truncates the exposure time.

Currently, good time intervals do not indicate gaps in an event list due to corrupted times. Since good time intervals currently are not used to calculate an effective exposure time, this behavior does not introduce additional errors. New ground processing software will recover events from existing gaps and correct the good time intervals.

We are testing changes to the ground software that will fix all of the problems described above. We anticipate that the new TIME-TAG software will be installed by the end of June. On-the-fly-reprocessing (OTFR) will (and must) be installed before the new TIME-TAG software can be effective. Once we announce that the new TIME-TAG software has been installed, people with affected datasets are strongly encouraged to retrieve their data again from the archive.

To help assess whether existing event lists ("_tag" files) are affected by any of these problems, we have created the program "checktag.e". Checktag prints all time intervals containing events and then reports the requested and apparent exposure times. Apparent exposure time is incremented by 0.032 seconds for each (spurious) negative jump in time of -0.032 seconds.

If the requested and apparent times agree, then count rates, fluxes, and uncertainties in existing extracted spectrum files ("_x1d" and "_x2d" files) are probably correct. If checktag.e reports negative time steps, gaps, or large time offsets, the event list ("_tag" file) should be retrieved again from the archive, once the new TIME_TAG software is installed.

If the requested and apparent times reported by checktag.e do not agree, then the count rates, fluxes, and uncertainties in existing "_x1d" and "_x2d" files are incorrect. To be consistent with the apparent exposure time, extracted quantities in these files should be scaled by the ratio of requested to apparent exposure time. We provide IRAF scripts (scalex1d.cl and scalex2d.cl) to facilitate this scaling, however we still recommend checking against new "_x1d" and "_x2d" retrieved from the archive, when the new TIME-TAG software is installed. Simple scaling should not be trusted if checktag.e prints the summary message, "Warning: This file appears to be bad."

Detailed checktag and scalex instructions are available. Download the software at ftp://ftp.stsci.edu/pub/instruments/stis/timetag.tar.gz