The topic of controlling scan timing is discussed in detail in the Scan Times section.
PRESTART allows the recording to be started before good data is expected. It was used to give time for earlier correlators to synchronize playback. All correlators now in use are able to sync quickly so the need to adjust start times to allow for synchronization is basically gone. In practice, the the Mark5A units take few seconds to get going on record which is why the default is 5 seconds for the legacy systems. For VLBA/RDBE and VLA/WIDAR systems, the default is zero. The following discussion is kept to describe the function of PRESTART while the parameter is available.
PRESTART gives a time by which to start the recording at a site before the nominal start time as established by START, GAP, DWELL, PRESCAN etc. Like PRESCAN, if the requested PRESTART moves the start of motion to before the stop time of the previous scan, the recording will just be kept moving. But, unlike PRESCAN, this time shift is done on a per-station basis -- if a station was not participating in the previous scan, it's recording start will be shifted the full PRESTART time, even if that means starting while other stations are still recording the previous scan. Also unlike PRESCAN, the shift due to PRESTART is not reflected in the scan times reported in the summary or sch files.
The amount of time by which the recording start is shifted before the nominal start time can be displayed using the TPSTART option in the SUMITEM list.
See the section on scan times for more information on controlling scan times.
For PCFS (VEX) controled stations, the user should bear in mind that all start times currently (Oct. 2001) must be equal. SCHED will issue a warning if this condition is violated.
Starting the recording early allows time for the correlator to synchronize the inputs from different stations before the start of good data. Anytime the recording stops, the correlator takes time to resync. On the old hardware VLBA correlator, that time is empirically about 8, 13, and 25 seconds for speedup factors 1, 2 and 4 respectively. Therefore it was best to keep recording through short gaps. On other correlators playing Mark5 media, sync is rapid -- on the order 1 second or less at the JIVE and MarkIV correlators and the DiFX software correlator (current VLBA), so these concerns are not so great.
One exception when it may be a good idea to stop recording is when there will be a formatter reconfigure. These happen when the internal switching in the formatter, or the pcal detection setup, change (number of channels, BBC assignments, sidebands, pulse cal detector frequencies etc.). The media does not receive good data while this is happening and the correlator can have problems recovering sync afterward. At least a normal resync must happen and occasionally (maybe 10-20% of the time) the correlator will get in a bad state and take a long time, maybe over a minute, to resync. If the media are stopped during a reconfigure, a resync will be done normally.