[evlatests] Results from First 'Band Switch' test

[Rick Perley]

    We have generated a 2-hour script, whose purpose is to give the band 
switches a decent workout in order to assess the efficacy of the 
proposed replacement switches. 

    The script observes the northern point source 0217+738, changing the 
band every 45 seconds, following the sequence: 
L-> X -> Q -> Ka -> K -> U -> S -> C -> L ...

    The first trial was run Wednesday morning.  A total of about 105 
band changes were done.   At that time, there were no antennas in the 
array with the new switches installed.  (Antenna 21 has the new 
switches, but is not yet returned to service;  Antenna 22 got its new 
switches yesterday).

    Observations were taken in standard OSRO setup -- two spectral 
windows in each band.  Dumps were done at 1 Hz.  'Set and Remember' mode 
was on, so we expect the PDif values, at each band, to remain closely 
constant -- or at the worst should show only slow drifts which would 
reflect slow changes in electronic gain.  In no case should we see a 
notably different  PDif value for any given observation at a band which 
is not clearly part of a trend with the other ~12 observations made at 
that band. 
    This is a good methodology to isolate switch problems -- by not 
changes sources we avoid the problems noted with PDif being sensitive to 
total power.  And by observing a single northern source, we minimize any 
changes associated with changing elevation. 

    To simplify the results, a 'gain change' table was constructed, 
where two kinds of changes in the recorded value of PDif were noted:  
Those which are 1 -- 5% deviant from any trend, and those greater than 
5%.  As the Project Book requires calibration accuracy of ~1%, the level 
of 1% for a note to be made seems a reasonable level to declare a given 
antenna as failing to meet requirements.   

    Some Conclusions:

    1) Not one antenna meets the 1% stability requirements at all bands 
in all IFs (although some are close).  On the other hand, no antenna 
fails the stability requirement at all bands. 
    2) Some bands show much better stability in the PDif values than 
others:  Notably good are C and K bands.  Notably bad are Q, X, and L 
bands.
    3) Some antennas show bad stability at multiple bands, giving a good 
clue as to the origin:
             ea25 has poor stability and K, A, and Q bands in RCP only
             ea20 has very poor (> 5%) stability at Ku, K, A, and Q 
bands in LCP
             ea07 has very poor stability and K, A, and Q bands in RCP
    4) About 80% of the band changes resulted in acceptable (better than 
1%) return to gain. 
    5) Some antennas are notably worse than others -- these are 
candidates for future switch changes:  antennas 16, 23, and 25 are worse 
than others. 

    On the Bright Side:

    Virtually *all* of the gain changes seen in the switched power are 
matched by changes in the visibilities.  In other words, application of 
the switched power calibration (PDif) eliminates the gain change in 
nearly all cases.  I'll report on the important exceptions in a later 
email. 

    We should rerun this test when the two (or more?) antennas with new 
band switches are back in the array. 


             Notably