[evlatests] Hour-long X-band observation results

[Rick Perley]

    Most of this is old news, but just for the record, I write this 
down.   

 The last hour of Software Time yesterday was utilized for tracking a 
single strong calibrator at a single band -- Xband.   The CMP timing 
issues were resolved prior to this observation.  A number of interesting 
characteristics were found from inspection of the data. 

    1) Flagging. 

    a) Unflagged data appeared in the archive at 22:50:20, but the 
antennas did not reach the source until about 22:50:38.  Some occasional 
flagging of EVLA antennas prior to this time was noted.  No flagging was 
seen on any VLA antenna. 

    b) The operator's log noted antenna 13 having a focus error for 
about two minutes during the observation.  This was caught by the 
flagger.   A similar event (as judged by the amplitude loss) on the same 
antenna at a later time -- but not noted in the log -- was also flagged 
correctly. 

    c) Antenna 19 was noted by the operator as having a large rotation 
error.  The bad data were correctly flagged. 

    d) However, a loss of good data from antenna 9 (a VLA antenna) in 
the middle of the observation was not flagged. 

    2) Correlator Controller Reorganizations

    Two events which are the signature of the correlator controller 
getting itself reorganized occurred.   Neither was flagged.  The 
signature is this:  There are a few (<10) seconds of bad data (low and 
ragged amplitudes), followed by a 10-second gap (which starts on a 
ten-second tick), followed by good data. 

    3) CMP rebooting effects. 

    Just prior to the beginning of this run, the CMP was rebooted by 
Hichem.  The subsequent data are quite remarkable -- all the following 
effects can be ascribed safely to the rebooting process:
       a) The amplitudes on all antennas suddenly changed to a new 
value, typically 10% different.  The cause of this was easily found to 
be due to the Tsys values -- prior to the change, the Tsys values were 
fixed at a constant (and incorrect) value.  When the CMP got its act 
together, correct data, and correct visibility amplitudes began. 
    b)  The phases on all VLA antennas were spinning at a remarkable 
rate.  At a particular moment (different for each antenna), which was 
always the first record following a ten-second tick, the phases began to 
track properly.    The resynchronization clearly followed a DCS ordering. 
          The moment of phase lock-in lagged that of correct amplitudes 
by 9 to 18 seconds. 

    As noted, all the phenomena here are explained by the reboot 
process, and should not be a concern.  What is notable is how long this 
takes -- the last antenna to get synchronized (#27, at N56) was nearly 3 
minutes following the beginning of the data stream.  We don't know 
exactly when the CMP was rebooted.  It might be prudent to allow for 5 
minutes, following the initialization of a CMP reboot, before data 
should be taken. 

    4) And odd pointing situation? 

    Four antennas -- 1, 7, 9, and 27, showed rhythmic amplitude drops, 
with a ten-second period, equal on all IFs, and synchronized with each 
other, for about one minute.  These look like pointing offsets.  The 
amplitudes drop by about 10%.  The situation corrected itself, and no 
other such instances were seen. 

    5) Baselines. 

    The hour-long observation showed that the VLA antenna phases are 
changing at a rate compatible with Vivek's estimation of about a 1-meter 
error (on the longest spacings) in the Bx, By coordinates which, if due 
to rotation of the coordinate frame, corresonds to a time error of 
something less than 1 second.   This problem affects VLA antennas only.  
EVLA antennas look good.