[evlatests] Tests on dropouts, phase stability, and sensitivity

[Rick Perley]

    Ken and I ran another short experiment with the 'fast continuum' mode,
observing a number of calibrators of varying strengths.  The first twenty
minutes' data were taken with phase switching on, the last twenty minutes
with it turned off antennas 14 and 16. 

    Delays and pointing were believed to be correct (save 14B -- see 
below). 

    a) Dropouts.   Dozens of dropouts, with the usual characteristics, were
seen in the phase-switched data.  No dropouts were seen in the non
phase-switched data.  The only (slight) change in characteristics in the
dropouts were occasional 'double dropouts' -- two consecutive 400 ms
records with the same level of attenuation.  As usual, the dropouts were 
always
at the 25, 50, or 75% level. 

    b) Phase behavior.  This was completely normal for the unphase-switched
data.  We saw abnormal phase behavior in antennas 14 and 16 for the
phase-switched data -- rapid phase winds, as high as 1 radian/second, for
some of the scans.   This (new) phenomenon is very odd:  Antennas 14B and
14D did not show it at all.  All other antenna-IFs did.  But these did 
not show
it on all scans.  For one source, no antenna-IFs 'spun'.  For another, 
14 A and C
did, 16 (all IFs) did not.  For another source, the situation was 
reversed. 
No explanation is offered -- but this is unlikely to be connected to the 
phase switching. 

    c)  Antenna 14B. Very poor sensitivity (which Ken explains is due to
a large remaining delay error).  The multiple amplitude states are still
present -- unchanged from last week's test.   They are present in both
the phase switched and unphase switched data.  There are two families of
amplitude states:  A coarse step, corresponding to drop in amplitude of a
factor 1.4, or about -3dB in power.  Each coarse step is seen to comprise
two sub-steps, separated by ~0.5 dB in power.  The timescale of the
responsible cause is faster than 400 ms -- there is no visible tendency for
the state of any one point to influence that of adjacent points.  OTOH,
it would seem that the responsible cause can't be very much faster than
a few Hz, or else we wouldn't see these separated amplitude states at all. 

    d) Amplitude stability.  Other than 14B, and ignoring the dropouts,
the amplitude stability of 14 and 16 look as good as or better than the VLA
antennas. 

    e) Sensitivity.  After removing discrepant points, I carefully 
checked the
amplitude rms of the observations of a weak (0.35 Jy) source, where the
noise will dominate.  Antennas 14A, 14C, and 16A are definitely slightly
noisier than good VLA antenna-IFs.  The degradation is small -- 10 to
15% at most.  However, antenna 16C is as good as the median good VLA
antenna.  I did not check the B/D IFs as closely, as 14B is quite useless. 
A cursory check indicates the sensitivity conclusions will be similar. 

    f) Back-end System Temperature.  Seven of the eight reported system
temperatures for antennas 14 and 16 are in the normal range -- mid 30s 
to low 40s. 
However,  antenna 16A reported a 64K system temperature -- a value which, if
true, should easily be seen in rms noises about 40% above the others.
This was not seen in the data.  On the other hand, there is a clear 
correlation
amongst VLA antennas between high rms values and high backend system
temperatures.  All VLA antennas whose rms noise levels are as high as the
EVLA antennas (against VLA antennas) have backend Tsys values over
40K.  There are no examples of (VLA) antennas with high Tsys that are 
not matched
by high rms noise.