[evlatests] EVLA Phase Stability

[Rick Perley]

    I utilized about 3 hours of 'dynamic' time last evening to test the 
phase stability of the array. 

    Short Answer: 

Phase Stability is Truly Fabulous!  The 'Global' phase jumps are 
completely absent.  Only one EVLA antenna shows notable individual phase 
issues of significant size. 

    Longer Answer:

    The observations were taken in the 'holography' mode:  Rocking the 
antennas up and down, between the half-power points of the beam.  The 
system moved the antennas by about 3 arcminutes every 10 seconds until 
the half-power point is reached, then reverses direction.  The movement 
was in elevation.  After ten such cuts, a short calibration observation 
was made.  (This mode is used because it resets the system every ten 
seconds, so if phase or other problems are associated with such resets, 
we are more likely to see them than in the more standard tracking modes, 
where resets are typically minutes apart). 

    All 8 EVLA antennas were in the array, and all fringed beautifully, 
with the exception of:

    Antenna 26, IFs A and C were dead. 
    Antenna 23, which has a significant number of odd phase changes (see 
below for description).
    Antenna 17, which had a 10-second square-wave phase pattern on IFs B 
and C only.  Ken immediately knew what caused this, and has fixed it.  
No more need be written.  (He has also 'pre-fixed' antenna 21 for the 
effect). 
********************************************************************************************************
    Result #1:  (This for holography types only). 

The VLA and EVLA are driving the antennas -- while in the scanning modes 
-- in utterly identical fashion.  Each holography cut had 11 'stopping' 
points, each 10 seconds long.  I was conservative, and assigned 140 
seconds to each cut -- 30 seconds longer than the absolute minimum 
required.  For the VLA and EVLA, what actually happened was:
       The first 10 stopping points got exactly the 10 seconds requested.
       The last stopping point got 20 seconds of duration. 
       There was a 10 second gap. 
       The system then began the next holography cut.  The actual time 
spent on a cut was then 10*10 + 20 + 10 = 130 seconds, while the script 
requested 140 seconds.  In other words, both the VLA and EVLA holography 
modes are effectively ignoring the stop times given in the scripts!  
This is not a bad thing (since both arrays are doing the same thing, and 
the data are correctly labelled with submode and antenna positions) -- 
but has an interesting consequence when we get to the calibrator 
scan.    Read on ...

       There were 10 holography cuts taken in a block, followed by a 
calibration.  Because of the 10-second discrepancy between requested and 
actual times (as described above), the antennas reached the end of the 
10-scan block 100 seconds earlier than planned.
The VLA and EVLA now do different things with this 100 second excess!  
The VLA recognizes that the next scan is a calibrator scan, and moves 
the antennas to the on-axis position.  Hence, we get an extra 100 
seconds of good calibration.  But the EVLA apparently does not look 
ahead like this, and uses the 100 extra seconds to remain at the 
half-power position.  After this 100 seconds, it 'wakes up', and moves 
to the calibrator position.  So in this case, the EVLA is paying 
attention to the given stop time. 
    This is a minor issue -- it effect is easily seen in the 
calibration, and easily flagged.  (But it would be useful to get rid of 
this). 

    A Conclusion for Holographers:  The appropriate DWELL time to use 
when constructing holography files is 10*(N+1) seconds for each 
holography scan, where N is the number of stopping points along the scan. 

******************************************************************************************************

    Result #2. 

    There were zero (0) global EVLA phase jumps in this test.

********************************************************************************************************

    Result #3.

    Antenna 23 is the only EVLA antenna with significant phase issues.  
All four IFs showed a curious pattern wherein the antenna phases changed 
by 20, 50, or 100 degrees (more or less).  The antenna took ~7 seconds 
to get to one of those offsets, dwelled about 7 seconds in that state, 
then took ~7 seconds to return to the correct phase.  This occured 
perhaps 20 or 30 times in the 3 hours (so only ~5 minutes out of 180 had 
a bad phase).  There is no signature of this effect visible in the 
amplitudes.

*********************************************************************************************************

    Result #4.

    Antenna 13, IFs B and D has a modest number of short-duration 
amplitude dropouts which were accompanied by incorrect phase.  The 
durations were 1 or 2 seconds.  Ken says there were communication 
problems with this antenna yesterday, and these dropouts are likely a 
result. 

********************************************************************************************************

    Result #5.

The phase stability of antenna 16 is notably worse than the others -- 
with slow changes of 10 or 20 degrees. 

********************************************************************************************************

    Result #6.

All EVLA antennas showed a strong phase drift with time, identical 
between polarizations, but very different between IFs (which were at 
1465 and 1385 MHz for IFs 1 and 2, respectively).  The observed slopes 
are, w.r.t VLA antenna 2 (at the center of the array):

    Antenna            AC               BD
-----------------------------------------------
    13                   -10 deg/hr       +16
    14                   -27                     0
    16                   -33                    +3
    17                      0                   +27
    18                    -7                    +20
    23                    -10                  +20
    24                   -67                   -60
    26                   dead                -16
----------------------------------------------------

No VLA antennas showed anything like this -- which is clearly not a 
baseline error.