[evlatests] Switched Power -- Present even when Absent!!!

[Rick Perley]

    A remarkable discovery has been made concerning our synchronous 
switched power system which should help considerably the diagnosis of 
the problem.

    To recap the problem:
   
    *  Observations of Cygnus A, at L, S, and C bands, showed that the 
PDif values (which monitor system gain) are typically 10% less when on 
Cygnus A than when observing cold sky.  As we are observing in a 'fixed 
gain' mode, this apparent reduction should not occur.  Some antennas as 
much worse than others, and some even show the reverse effect. 

    * This effect is seen on sources of lower flux density, with the 
effect roughly proportional to the increment in total power contributed 
by the target object. 

    * The effect is not caused by a change of gain -- the visibilities 
are not affected. 

    * The effect is different for different bands, but is the same for 
subbands within the same polarization.  (In other words, IFs A and B are 
affected similarly, as are C and D, but there is no correlation between 
A and C). 

    * Numerous tests have shown that the IF stages (T302 and T304) 
remain linear within the power range between cold sky and Cygnus A. 

    New Observations. 

    Ken and Michael arranged for the array to observe on and off Cygnus 
A, with the noise diodes alternating between on and off.  We thus have 
measurements of PDif on and off Cygnus A with the diodes switching, and 
with them not switching.   These observations were made at 5.3 GHz, with 
the AC and BD IFs tuned to the same frequency, in 'OSRO' mode (just two 
subbands).  The SY table produced by this was then examined within AIPS. 

    New, and Amazing Results:

    The 'compression' of PDif is seen as always, with the usual antennas 
being far worse than the others:  ea24 in RCP, and ea01 in all 
polarizations. 
    But  (hold onto your hats here):  There is a residual, usually  
negative,  'PDif' for those antennas which show 'PDif Compression' when 
the diodes are off!  This means that, despite having now input power, 
there is a coherent modulation of receiver output power with the 
expected period, but in anti-phase.  In other words, the receiver power 
drops when the noise diode would have been adding power (had it been 
turned on).
    Furthermore (and this is the key point) -- the 'negative PDif' is 
larger -- much larger -- when the antenna is pointed onto Cygnus A then 
when on cold sky. 
    There is a clear relationship between the size of the negative 
offset (with switched power off) and the size of the 'compression' when 
observing Cygnus A.  But the absolute values of the no-switched-power 
offset, when added to the value when switched power is on, do not 
restore the on-source switched power values to correct values.  This 
will be made clearer by the following table, for antenna ea24A:  (recall 
that Tsys when on Cygnus is about 2X the cold-sky value):

       Setup                            PDif
-----------------------------------------------
    On Cyg, switched        125 millicount
   Off Cyg, switched         165 millicount
    On Cyg, not switched    -30 millicounts
   Off Cyg, not switched     -15 millicounts
------------------------------------------------------

    Thus -- the PDif values when the diodes are on are not simply offset 
by the values when the noise diodes are turned off. 

    Not all antennas have this problem -- quite a good number show no 
effect at all, and one antenna is 'inverted' -- it shows PDif 
'expansion' when observing Cygnus A.  All antennas that do not show the 
'PDif compression' all have zero PDif when the noise diodes are turned 
off. 

A table of the good, bad, and ugly is given below. 

    Antenna                   Behavior
--------------------------------------------------
    ea01                     Large PDif compression on both polarizations
    ea04                     Small PDif compression on both polarizations
    ea05                     Negligible compression in RCP, small 
compression in LCP
    ea06                     Small compression in both.
    ea07                     Small compression in both.
    ea08                     Small compression in RCP.  LCP is hugely 
noisy (as reported before).
    ea09                     Negligible compression in both. 
    ea10                     Moderate compression in RCP, small 
compression in LCP
    ea11                     Negligible compression in both.
    ea12                     Small compression in RCP, negligible in LCP
    ea13                     Negligible compression in both.
    ea14                     Negligible compression in both.
    ea15                     Negligible compression in both.
    ea16                     Small compression in both.
    ea21                     Very large *expansion* in RCP, but the 
signal is very weak.  Large compression in LCP.
    ea22                     Negligible compression in both.
    ea23                     Moderate *expansion* in RCP, small 
compression in LCP.
    ea24                     Very large compression in RCP, small 
compression in LCP.
    ea25                     Small compression on both polarizations.
    ea27                     Small compression in RCP, no compression in LCP
    ea28                     Moderate compression in RCP, small 
*expansion* in LCP.
--------------------------------------------------------------------------------------------------------

    It is optimistically hoped that this evidence is sufficient to allow 
quick diagnosis of this important problem! 
    I have plots to aid understanding...