[evlatests] Results from Monday mode-switching tests

[Rick Perley]

    I reported last week on a test performed by Michael, in which the 
correlator was reconfigured once a minute, with a different mode.  
Interesting patterns in the 'perfect zeros' were noted, which the 
pundits (Michael, Ken, and Martin) declared could be avoided by using a 
different set of baseline boards. 

    This 'reconfiguration stress test' was repeated on Monday morning, 
with a slight variation -- instead of changing mode each minute, each of 
the four different modes was observed three times, for one minute each.  
We thus should be able to discriminate between effects caused by a mode 
change from those caused by a simple scan change.  

    The observation cycle was then:

    OSRO1 with BW = 64 MHz (three observations of 1 minute each)
    OSRO 2 with BW = .03125   (ditto)
    OSRO1 with BW = 4             (ditto)
    OSRO2 with BW = 16           (ditto)
   
    This group of four was repeated three times. 

    Averaging/dump times were 1 second, as usual.   The frequencies used 
were 4896 and 5024 MHz for all observations.

    Results:

    1) 24 antennas were in the array -- all fringed at all times, with a 
single exception:  antenna 3 was flagged by the system for one scan 
only.   I don't know why. 

    2) The curious pattern of 'perfect zeros' which initiated this 
repeat test was completely absent!  All four databases reported the same 
fraction of 'perfect zeros' -- 5%, corresponding to the first 10 seconds 
following the change of mode.  No zeros are associated with any change 
of scan not associated with a change in mode.  Within the 10 seconds 
noted above, the data characteristics are always the same (for all IFs 
and all modes):  2 seconds of rubbish, followed by 8 seconds of perfect 
zeros.   This is a Big Improvement over the previous test (and is in 
accord with the report from James circulated last night).

    3) Amplitude stability was fabulous (meaning, no changes in gain 
outside the noise), with the following three exceptions: 

        a)  The five innermost antennas on the west arm showed what 
appeared to be large changes in amplitude (20% or more in amplitude) 
upon change of mode.  But upon further review, it is clear that this is 
due to these five antennas being heavily shadowed:  the gains were in 
fact continuously changing as the antennas became progressively 
unblocked.  The 'jumps' were only apparent, as the array was observing 
in a different mode on each side of the 'jump'. 
       *** It would be helpful to not observe objects at 14 degrees 
elevation while in the D configuration!!!*** 
        b)  Antenna 17, in LCP on both IFpairs (B and D), showed a 1 dB 
gain change at 14:10:27 -- corresponding to one of the changes in mode.  
Did we not include the 'set and remember' command?  Or is the 'remember' 
part only in force for a particular mode? 
       c)  A very small number (two or three) of the 48 individual scans 
had bad data on the last record of the scan, in all cases only a few 
baselines were affected. 

    4) Phase stability was as expected from atmospherics on a good day, 
with the following two exceptions:

       a) Antenna 5, on IFs A and C, is still hopping 60 degrees in 
phase upon change of scan or mode.   The cable length adjustment, done 
last Friday, did not solve this problem. 
       b) Phase wobbles are seen in four baselines, on all four 
parallel-hand combinations (AA, CC, BB, DD).  The amplitude of the 
wobbles is 2 to 4 degrees, the periods are 2 to 4 seconds.  The four 
baselines are:  2 x 14, 4 x 12, 8 x 24, and 9 x 27.  These are (as 
usual) a clearly non-closing effect (not factorable to particular 
antennas).  No discernible amplitude wobble was visible. 

    5) Bandpasses were wonderful on all antennas, in all modes.  No 
attempt was made to check BP stability (too little data over too short a 
time to be useful).