[evlatests] Time stamps: BDF vs. SDM vs. CASA & AIPS

[Michael Rupen]

I compared the time stamps given in the binary data files, metadata (SDM's
Main, Scan, and Subscan Tables), CASA, and AIPS.  Basically everything
looks good.

* Binary data are aligned with the preceding UT midnight.
   Thus for 1sec integrations all integrations should start and
   stop on UT seconds, and the corresponding midpoint times stored
   in the binary data files should occur have UTs ending in 0.5.
   This is indeed the case.

* SDM Tables take their times from observation documents which give
   the requested scan start times.  The OPT produces scans in LST,
   which unless you work hard are aligned on LST seconds.  Thus
   we expect scans to begin and end at odd UT times (not integer seconds).
   There will be some dead time at the beginning and end of each scan,
   since the requested scan interval will not in general contain an integral
   number of 1sec UT integrations aligned on UT seconds, and we currently
   toss any partial integrations.  We thus expect that:
   (1) the times in the Main, Scan, and Subscan Tables will agree
   (2) the timestamps and intervals for all the integrations in the
     binary data will lie within the time ranges given in those SDM Tables
   Again these expectations are indeed met in practice.

* If one creates a file in UT times with an integral number of UT
   seconds per scan, the SDM times should be identical with those
   in the binary data files, and the number of integrations should
   be as expected to exactly fill the scan.
   Once again, yes! we do see this.

* The time stamps on the visibilities as seen by CASA and AIPS
   should match those given in the binary data files, rather than
   those in the SDM Tables.
   This too is done correctly, using either asdm2MS or BDF2AIPS,
   for scripts with scans aligned either in LST or in UT.

* Scan boundaries should match those requested by the observing script,
   and any changes at scan boundaries should be seen exactly at those
   scan boundaries.

   As described yesterday we set up a script which changed
   the attenuators (a very rapid process) at scan boundaries, and
   put those scan boundaries on 30sec UT boundaries.

   For this script the scan boundaries in both the binary data files and the
   SDM Tables do indeed occur on the expected 30sec UT boundaries,
   for both 1sec and 100msec integrations (dump times out of the CBE).

   The one outstanding issue is the first-integration problem described
   yesterday, wherein the data from the first integration of a scan seems
   actually to belong to the preceding scan.  Martin has a theory and is
   putting a fix in place which we will test (and then describe, if it works)
   this afternoon.


We have a couple more checks to do (e.g., is everything still fine
when we average several LTA integrations in the CBE?) but so far everything
looks quite good.

Cheers,

            Michael