[evlatests] Adventures with Switched Power -- (continued ...)

[Rick Perley]

    This is an update for the summary report sent last week. 

    A four-hour long observation at C-band of three calibrators was 
taken last Wednesday morning (Nov 3, 08:20 through 12:20 IAT -- 2:20 
through 6:20 local time).  The AC IF pair was tuned to 4.5 GHz, the BD 
pair to 6.5 GHz.  The observation was in two parts:

    a) A two hour block alternatively observing two sources, one rising, 
the other setting. 

    b) A two-hour block simply tracking 3C147 immediately after transit. 

    Of these two, the more immediately useful is the latter, whose 
results are given here. 

    3C147 was observed starting 15 minutes after transit, and an 
elevation of 74 degrees, tracking it continuously until elevation 60 
degrees.  Given the time of day (before dawn), weather (calm, clear and 
dry), and elevation (high), no major changes were expected to be seen in 
any monitored characteristic:  Pdif, Psum, Tsys, or fringe power.  This 
expectation was met for most -- but not all -- antennas. 

    A reminder here that the EVLA correlator does not provide normalized 
correlation values (as the VLA correlator did) -- we record 'straight 
cross power', which will reflect changes in the entire signal path, 
including atmospheric absorption, antenna pointing, and electronic gains 
(including digital gains).  All changes following the introduction of 
the switched power should be monitored (and correctable) by the 
demodulated switched power -- 'PDif'.  For this correction, the 'total 
power' 'PSum' is not needed.  This value is needed for computing the 
system temperature:  Tsys = PDif/PSum*Tcal. 

    For nearly all antennas, the changes in the observed parameters:  
PDif, PSum, Tsys, and fringe power, were small and easily explained by 
minor changes in system gain and changes in spillover (which are in good 
agreement with on-antenna measurements made by Bob Hayward and me).  
There is every reason to believe that application of the switched power 
'PDif' will correct for antenna electronics gain changes.  (See below 
for why I cannot prove this statement). 

    For the following antennas, non-proper behavior is observed:

    1) Antenna 10:  For all four IFs, a periodic (but not sinusoidal) 
variation in Psum, Pdif, and fringe power is seen.  The period is about 
1 hour, and the amplitude of the effect is about 3%, equal in all IFs.  
This variation is *not* seen in Tsys.  It is clear that we are seeing a 
real change in electronics gain.  It cannot be due to pointing 
variations, as these will not change the Psum or Pdif.   This variation 
will be nicely corrected by application of the switched power. 

    2) Antenna 14 has a variation of the above -- a significant secular 
change was seen, similar on all four IFs, but of very different 
amplitude:  2% in A, 6% in C, 1.5% in B, 3% in D.  It is present in 
Pdif, Psum, and fringe power, but not in Tsys.  The variation is slow, 
and non-periodic.  It should be perfectly correctable by switched power. 

    3) Antenna 18 shows a large, triangular periodicity in switched 
power (PDif) *only*.  There is no effect in Psum or fringe power.  The 
characteristics of the periodicity are quite odd:  In A and C, the 
amplitude is about 6%, with a 70 minute period.  But the A and C 
switched powers change with opposite sense -- when A is going up, C is 
going down!  Meanwhile, and B and D, a different pattern is seen:  20% 
in B, 5% in D, with an 80 minute period, and in phase -- they both rise 
and fall in synchronism.  This is clearly a problem with the switched 
power mechanism itself. 

    4) Antenna 25, in IF 'B' shows Psum and Pdif switching between two 
different values, with the period shortening from ~10 minutes at the 
beginning to 5 minutes at the end.  The most interesting feature is that 
the Psum and Pdif values are in anti-phase -- when one goes up, the 
other goes down, so the effect on Tsys is the sum of the two (8% in 
Psum, 3% in Pdif, so 11% in apparent Tsys).  As there is no such effect 
seen in the actual fringe power, I conclude this indicates a problem in 
the switched power mechanism. 

    5) Antenna 17, IF 'B' only shows two large changes in Pdif, Psum, 
and fringe power -- clearly indicating changes in attenuation, by 3dB in 
the first step, and 6 dB in the second.  All indications are that the 
change in observed fringe power will be corrected by application of the 
switched power.  A curious effect is that the first step (down in power) 
showed to effect in Tsys, while the second (also down) showed a small 
(0.5K) reduction in Tsys.  Perhaps this is due to starving the sampler? 
-- the 9 dB in power is nearly 3 bits, rather severe for the 4-bit 
requantizer...

    6) Antenna 12, IF 'B' shows a rapid oscillation in PSum, of period ~ 
15 seconds, and amplitude ~ 10%.  This is very weakly seen in PDif, so 
the effect is also seen in Tsys.  There is no visible effect in the 
fringe power, so this again appears to be an artifact of the switched 
power mechanism. 

    I've noted that most of the observed fringe power variations should 
be corrected by application of the PDif values.  Unfortunately, for this 
2-hour segment of the test, I cannot successfully apply this:  The AIPS 
program 'TYAPL', when asked to do this operation, applies no 
corrections, and flags all the data (by negating the weights).  There 
are no input options in the program which allow variations.  Apparently, 
there is some internal test for data quality which is failing, and 
results in no corrections and (unsurprisingly) flagged output values.  
Curiously, the program correctly modifies the first two hours of the 
test (that part with the two sources on opposite sides of the sky). 
    However, visual inspection of the various quantities gives strong 
support to the conclusion that application of switched power -- in most 
cases -- is resulting in proper correction for system gain changes.