The finer details of FOS exposure timing extend far beyond the scope of this document (refer to Appendix A of FOS ISR 154 as the best available low-level description of FOS data acquisition procedures and timings). In summary, at the end of an observing segment (i.e., what will become an output group) the transfer of detected counts from FOS memory to storage on the spacecraft proceeded only after some communication between the FOS microprocessor and the HST Science Data Formatter (SDF). The normal data-taking commanding of the FOS did not require FOS exposures to start at precise pre-selected start times, and there was no direct record in any telemetry of when an exposure actually started. The engineering telemetry did record the so-called LINESTART time; that is, the precise time when the SDF told the FOS to commence sending data from the FOS memory to the SDF. A truncated version of LINESTART time was recorded in the FOS headers as FPKTTIME (the so-called first-packet-time). It was also possible to calculate exactly the length of time that the FOS had spent actually recording signal by using fundamental FOS data-taking quantities which were recorded accurately in the headers.
Then, in principle, one can start with the LINESTART (or less accurately, FPKTTIME) and subtract both the time required for post data-taking electronic "handshaking" between the FOS and the SDF and the time actually spent taking data to determine the start time of an observation (see Section 33.1).
Unfortunately, the length of time required for the handshake between the FOS and the SDF could vary from 0.001 to 0.255 seconds depending upon the load on the SDF. As a result, FOS observation start times calculated even with the highly precise LINESTART from the engineering telemetry give results 0.001 to 0.255 seconds later than the actual start time. It is not possible to determine the absolute start times more precisely than with this level of uncertainty. Due to the fact that FPKTTIME is a truncation, not a rounding, of LINESTART, an observation start time calculated from FPKTTIME may range from as much as 0.25 seconds earlier to 0.125 seconds later than the actual value.
Further, under normal loading of the SDF and for most FOS exposure time regimes the duration of the handshake interaction could vary by up to approximately 0.02 sec between readouts of the FOS memory. The dominant component of this variation was internal to the SDF. Under normal circumstances the actual interval between start times of individual groups (e.g, in a RAPID exposure) should be nearly the same. Relative group timings can be accurately determined from relative LINESTART timings with the understanding that variations of order 0.02 sec or less commonly will be seen in the length of these intervals due to the influence of SDF loading. Since the interval between successive FOS group starts was fixed by FOS commanding to be an integral number of FOS data-taking cycles, called INTS, it may be possible to improve the accuracy of relative group timings further with post-observation analysis which can only be facilitated with the engineering telemetry and STScI site-specific software. Unfortunately, the very low-level details of FOS commanding cause essentially random variations in the group-to-group intervals of certain RAPID mode observations when the readout time of the FOS memory is comparable to the internal FOS data-taking cycle time (the INT). A description of this phenomenon is beyond the scope of this document. If your LINESTART intervals seem to vary randomly or have substantial excursions outside of the typical 0.02 sec range, you should contact email@example.com. Similarly, if accuracies of better than 0.02 sec in relative timing and/or 0.255 sec in absolute timing are required, please contact STScI.
In any event, always use timings derived from FPKTTIME or the engineering telemetry; do NOT attempt to form times based upon start time of observation and anticipated or requested group integration times.