[evlatests] Switched Power Compression not seen in visibilities

[Rick Perley]

    Joe has asked (for some time) that I determine whether the peculiar 
switched power compression problem seen on many antennas at L, S, and C 
bands also affects the data.  I had delayed answering this question 
mostly because I was sure I knew the answer -- it had to be! 

    But this is wrong. 

    To recap:  The 'Switched Power Compression' problem is a reduction 
in the monitored PDif value when the system observed a source strong 
enough to elevate the system temperature.  As we are observing with a 
fixed gain, no such reduction should occur.  The phenomenon is 
repeatable, and is roughly proportional to the added system 
temperature.  It is band-specific, which argues against the IF chain 
being responsible, and polarization independent -- the RCP and LCP can 
change in opposite ways.  There is a correlation between IFs A and B, 
and between C and D, but it is not very strong.  Finally, the effect is 
not always one of compression -- some antenna-IFs give PDif Expansion!  
Very weird. 

    I utilized the '4M' project data to determine if the effect is 
present in the visibilities.  The target source 3C273 adds about 2K to 
the system temperature at 6 GHz.  The calibrators 3C286 and J1150-0023 
are weaker.  Plots of PDif clearly demonstrate the phenomenon, which in 
all cases is in the same sense as the compression (or expansion) seen 
when observing Cygnus A.  The only difference is in scale.  The table 
below gives a sense of the size of the phenomenon:

    Antenna-IF           Cygnus A           3C273
-----------------------------------------------------------
            1A                      -17%                 -4%
            3A                      -13%                 -3%
            4A                      +10%                +1.5%
           24A                     -25%                  -5%
            28A                     -13%                -2%
            28C                     +10%                +1.5%
---------------------------------------------------------------

    At C-band, Cyg A adds ~30K to the system temperature, while 3C273 
adds about 2K.  

    This effect is simply not seen in the visibility data.  Even for 
those antennas where it is the strongest, the gain solutions demonstrate 
no convincing correlation with the (apparently false!) indications of 
system gain change provided by the PDif values. 
    Yet the PDif values certainly *are* monitoring gain changes due to 
amplifier drifts and malfunctioning switches.  That this is so is very 
clear from these same data! 

    Somehow, the synchronous system is being tricked into reporting a 
gain change -- in some antennas at some bands -- when the input power 
level changes...