[evlatests] EVLA Longer Term Gain/Phase Stability

[Rick Perley]

    Convinced (sort of) that our short-term instability issues are 
resolved, I used 1.5 hours
of test time early this morning to track a strong point source at a 
single frequency, in order to watch how things
change on this timescale.

    Frequencies chosen were 4885 and 4835 (AC and BD).  BW = 12.5 MHz, 
the correlator
mode was '4', giving four parallel-hand correlations with 16 channels in 
each.  Thus, there were
no cross-hand (linear polarization) data.    Averaging was 3.3 seconds. 

    Sources observed were 3C286, to get a semi-absolute gain, and 
1407+284, a strong point
source.    90% of the time was spent on the latter. 

    Observations:

    1) All antennas gave good stable signals. 

    2)  Short-term amplitude and phase stability were as good as VLA 
antennas.  No drops,
spins, dead zones, or any notable bad behavior was seen. 

    3) Bandpasses were as always.  The 'phase hook' (40 degrees) in the 
lowest frequency
channel, and its associated low amplitude (10 - 20 %) are notable.  
Beyond that, there is a
phase slope of about 40 degrees across the rest of the 12.5MHz 
bandpass.  This is likely
the source of the 'closure errors' seen in 50 MHz wide continuum. 

    4)  Closure errors on the bandpass-normalized data are negligible.  
At the level of
1% and 1 degree, the only notable errors are for a VLA antenna -- and 
this in one
IF only.  (The cause of this is evident in the bandpasses -- 27B has a 
HUGE Gibbs'-
looking ring in it). 

    5)  There are interesting small effects seen in the longer term 
behavior of amplitude and
phase:

       5A) Amplitude.  The required behavior -- shown by VLA antennas -- 
is that the
apparent flux density of a strong calibrator not change by more than a 
few tenths of
one percent over the 1.5 hours.   As the correlation coefficient 
produced by the
correlator is multiplied by the geometric mean of the SEFDs for the two 
antennas
concerned, this stability requirement means the system temperatures and 
efficiencies
must be monitored (or known) with a relative accuracy of half (more or 
less) the
desired stability. 
                Antenna 13:  Channels A and B (RCP)  are beautifully 
stable to the required
accuracy.  Channels C and D (LCP) are not, with the apparent
amplitude rising by ~3 percent over the initial 20 minutes.   The 
observed variation in
C and D are identical -- they are seeing, or being influenced by, 
identical effects. 
Looking at the recorded
system temperature (back-end), we note the values changing by remarkably 
large
values -- by over 2K (larger than any other antenna) over the 
observation.  These values
are used to correct the correlation coefficient, so the fact that the 
observed amplitude
of antenna 13 in RCP is very stable indicates the observed Tsys changes 
are real, or
or at least reflecting and correcting real changes in correlation 
coefficient.  On the
LCP, the same Tsys changes are seen ***except during the initial 20 
minutes***,
during which the recorded temperatures differ from the trends shown on 
the RCP
side by about 1K -- about the right amount to explain the change in 
source amplitude. 

             Antennas 14 and 16:  Both antennas show slow long-term 
drifts in apparent
amplitude (gain), which for both antennas are the same for the two 
channels in each
polarization.  Thus, for example, antenna 14C and 14D decline in 
amplitude by about
2% over the period, while 14A and B rise and fall by about 1%.  Antenna 
16 is
qualitatively similar, with 16A and B declining by 2%, and 16C and D 
rising by
about the same amount.  However, these trends are not seen in the 
recorded Tsys --
which for all four IFs is very stable, to better than 1%, over the 
period.  Hence, we
are seeing small and slow changes in correlation coefficient which are 
not reflected
in the Tsys monitoring.  

          Because I was observing two different sources, with flux 
density different by
about 5 Jy, I expect to see about 0.5K difference in Tsys.  This is seen 
on all
antennas *except* antenna 13, where the temporal changes noted above are of
such magnitude that the small source-dependent changes are barely visible. 
          I note in passing that the Tsys values recorded for antennas 
13, 14 and 16
are all reasonable (20s) except for 16C and 16D, where they are about 
17K.  This
is clearly an error in the Tcal value used to convert the sync detector 
voltage to
a temperature, as the on-off source differential for those two IFs 
(only) is about
0.4K. 

       5B) Phase

       Long-term effects are also seen here. 

    Antenna 13:  The well-known baseline error (really big -- phase 
changes by one
turn in 80 minutes) dominates all.  We should solve for the x-y error 
from this, to
at least get this out. 

    Antenna 14:  We see effects which are IF-dependent:  14A and 14C show a
slow phase change of 15 degrees over the 80 minutes which is absent in 
14B and D.
All IFs show identical shorter-term (minutes) fluctuations which are 
undoubtedly due to
atmosphere. 

    Antenna 16:  The same picture as for 14:  16A and 16C show a similar 
-- but larger--
trend in phase (20 degrees in 80 minutes) which is absent in 16B and 
16D.  (16 is at
E24, 14 is on N16).