[evlatests] Troubles with Durations

[Rick Perley]

    With some left-over 'startup' time, I tried a wide-angle L-band 
holography last evening.  Because the start time was unknown, I used 
durations in the file.  This revealed some significant issues with the 
way scan start/stop times are being generated. 

    The file started with two calibrator scans, the first a five minute 
'dummy', to allow things to set up, and a subsequent short observation 
to get the (hoped-for) calibration.  From there, the file contained 
alternating holography and calibration scans.  Below is a table of the 
durations requested, and the time actually spent. 

Scan       Requested         Got                Comments
-------------------------------------------------------------------------
1               5:00                4:43              All EVLA antenna 
amplitudes low by a factor ~3.  Phases stable.
2               1:20                0:06              EVLA amplitudes 
nearly o.k. (data were flagged). 
3             20:20               21:23             This holography scan 
is o.k. (except for one ten-second record -- see below)
4               1:20                0:03               Just a single 
(flagged) record was recorded
5             20:20               22:40             First 50 seconds 
flagged -- target antennas were moving to source!!!
6               1:20                0:06               Just two 
(flagged) records recorded
7             20:20               22:36             Same as preceding 
holography scan
8               2:40                3:33               Last 48 seconds 
are nonsense
----------------------------------------------------------------------------------

    It is pretty clear what is going wrong here.  The holography scans 
are too long by the length of time that we were supposed to be on the 
following calibrator (which is the target source itself).  At the end of 
each holography scan, the target (moving) antennas are left at their 
extreme positions, and never return to on-axis position -- this time was 
factored into the durations of the calibrator scan.  When the next 
holography scan came, the moving antennas were in the wrong place -- 
hence the beginning of the second and third holography scans were lost 
due to antenna motion, as they hurried to get where they were supposed 
to be.   The first holography scan is o.k., as the antennas were on-axis 
at the preceding observation.  

    None of these issues occur when the file is constructed with stop 
times. 

    In addition to these duration/timing issues, I note the following 
three problems:

    1) The EVLA amplitudes were all low by a factor of about three for 
the entire duration of the first calibration scan.  The next calibration 
scan, which followed directly is (mostly) o.k. -- the amplitudes 
magically sprang to the correct levels.  Only a single (3 second) record 
lies in between these two.  The difference is not a matter of motion. 

    2) The last 48 seconds of the last scan are rubbish data.  There are 
recognizable amplitudes and phases, but they are not correct.  These 48 
seconds in fact lie after the official end of the scan, as defined by 
the duration in the observe file.  The executor seems to be (briefly) 
paused, wondering what to do next ...

    3) In the first holography scan, we see a curious problem which is 
global to *all* holography runs (whether the script is set in durations 
or stop times):  There is an initial 10-second duration of stable data 
-- which is not recognized by the holography programs as valid 
holography (meaning the sub-mode is apparently not correctly 
identified), followed by a 10-second duration where the amplitudes are 
correct, *** but ALL EVLA antennas have a different phase ***, followed 
by a 10-second duration where the phases are back to the correct values, 
and the amplitudes reduced -- this record is in fact the correct 2nd 
holography position. 
       The phase offset on the 2nd record is ** not global ** -- it is 
different for each EVLA antenna.  No VLA antennas show this effect. 
       To the holography programs, the (valid!) first 10-second position 
is ignored.  The second (invalid phase, but correct amplitude) is seen 
as the first valid holography record.  All others following are correct, 
both in amplitude and phase. 
       It makes no difference what the integration time is -- something, 
operating on 10-second ticks -- is invalidating the first (good) 
position, messing the EVLA phases of the second (but without actually 
moving the antennas), after which, everything works as it should. 

    I hope these issues can be readily resolved!