[evlatests] A strange apparent non-linearity

[Rick Perley]

    The recent work by Bob and me on X-band sensitivity and efficiency 
has led to a remarkable apparent non-linearity.  To follow up on the 
possibility that Cygnus A has 'too much flux' at X-band (thus leading to 
the anomalous high efficiency we measured), I observed Cyg A and some 
standard calibrators in an 'on-off' mode at L, S, C, and X bands.  This 
was done in 'set and remember' mode, so the gains are set fixed after 
the initial observation on a cold-sky object.  The resulting change in 
system temperature should directly translate into a flux density, after 
utilizing the known efficiency. 

    The results were strange, to say the least.  The X-band change in 
Tsys was what we expected to see (indicating that Cyg A is indeed 
stronger than Baars says it is).  But the C-band observation, using 
antenna 24 in RCP,  indicated that Cyg A has two times the flux 
indicated by Baars -- this is completely impossible.  The S and L band 
observations indicated slight excesses in Cyg A -- nothing that couldn't 
be explained by small errors in efficiency, or Tcal. 

    The cause of the C-band excess was easy to find -- the 'PDif' value 
(difference between the detected powers between 'on Cygnus' and 'off 
Cygnus' declined by about 30% -- this should not happen, given that we 
had utilized a fixed gain.  The observation indicates that the gain is 
flattened by about 30% when the input system power is doubled (the 
increment provided by Cyg A is about equal to the cold-sky noise power). 

    This observation sparked a review of *all* the changes in switched 
power, for all antennas, IFs, and bands, when on and off Cygnus A.  The 
result of this is:

    A) Most X-band antennas retain good linearity.  Exceptions are  
antenna 7 (compression by 4%), and antenna 2 (expansion (!) by ~ 5%). 

    B) About half the c-band antennas show small (typically 10%) 
compression when on Cygnus A.  The worst is antenna 24 in RCP (note -- 
this is our trial antenna).  Antenna 1 is also particuarly bad 
(typically compression by 15%).  Some antenna show expansion, (28C, 4A, 
17B). 

    C) Nearly all S-band antennas show compression of 10 to 20 % (note 
Cyg A nearly triples the system power).  Antenna 15 is particularly bad, 
with compression by nearly 50% in LCP, and 25% in RCP.

    D) Nearly all L-band antennas show compression of 10 to 20% (Cyg A 
more than quadruples system power).  antennas 9, 10, 12, 16 are 
particularly bad.  Antenna 28, in LCP only, compresses by a factor of 
three, but hardly at all in RCP. 

    There is some evidence that this effect is not due to electronics 
upstream from the T304 -- the exhaustive measurements made by Bob 
Hayward and me on antenna 24 at L, S, and C bands (also Ku, K, and Q) 
have never shown any hint of compression -- our power measurements are 
made before the T304.  In particular, antenna 24A, which showed 30% 
compression at C-band in the observations noted above, shows no hint of 
any compression in the data we took 10 days ago on the antenna.  Our 
on-antenna measurements were made at the output of the T302, while the 
problems noted above are made on the correlator station board.  The 
problem is then likely to be between these two points (T304, and the 
digital system).