[evlatests] Results from WIDAR 'stress' tests, Saturday afternoon.

[Rick Perley]

    Michael took WIDAR data with the standard 'test' setup Saturday 
evening.  These were run at C and K bands, with the standard setup.  
Following the C and K band observations, Michael ran the same script 
with the VLA's correlator, providing some very useful comparisons. 
    The two sources, 0217+738 and 0228+673, were separated in elevation 
by 3 degrees.  The observations were 40 minute long. 
    Calibration followed a now-familiar path:

    1) Use QUACK to retain only the last 1.0 minutes of every scan.  The 
scans are 1minute 23 seconds long.
    2) Use FRING to find a global (over all 8 contiguous sub-bands) 
delay and phase solution for a single 1 minute long piece of data, *not* 
selected from the first scan.  Use CLCOR to implement that solution for 
all sources.  (6 hour boxcar).
    3) Use CALIB to solve for amplitude and phase for each sub-band and 
polarization, using 50 channels (25 MHz) from the middle of each 
sub-band.  Use CLCOR to interpolate this (using 0217+738 as the 
calibrator and 0228+673 as the target), with a 1-minute smooth. 
    4) Use BPASS to solve for the bandpass, for each source and each 
scan (to check for bandpass stability, etc.). 

    Results are generally very, very good, with a few important 
exceptions.  Key points:

    A)  C-band. 
   
    1) The first scan has distinctly different phase than all others.  
There is a remarkable pattern:  the phase change between the first and 
second scans has one of 5 different values:  0, 60, 120, 180, and 300 
degrees.   Both polarization and all subbands for a given antenna jump 
by the same amount.  The antennas group the following way (antenna 4 was 
the reference:
    0:  antennas 4 and 27
    60:  3 and 25
    120:  5, 9, and 24
    180:  19
    300:  2, 8, 15, and 28. 

    2) Amplitude stability was outstanding, with no identifiable drops 
in amplitude, other than antenna 9, which seems to take a couple of 
extra seconds to get back on source after a change. 

    3) 'Zero' records are very rare:  typically six for a given antenna 
throughout the run (out of about 50,000 total).  'Zero' characteristics are:
          - baseline based:  if a particular baseline shows a 'zero' 
amplitude, there is no tendency for either antenna to have a zero with 
any other antenna at that time.
          -  different between polarizations.  The number of zeros is 
about the same in R as in L, but the occurences are completely different.
          - the same for all channels for a given baseline, subband and 
polarization.  If one channel is zero, all are. 
          - different between subbands.  Each subband had about the same 
number of 'zeros' in the run, but the occurences are different for each 
subband. 

    4) Phase connection between the sources is excellent in general.  
Antenna 24 may have a repeatable 5 degree offset betwen sources.  Nearly 
all antennas connect to with 1 or 2 degrees every time.  *** In 
particularly, any offset between sources for antenna 8 is limited to 
less than 1 degree***.  Near the beginning of the run (at 1:18 IAT), a 
plausibly atmospheric event caused phase jumps of 10 to 15 degrees 
between the two sources, on the longer baselines only.  The size of the 
jump increases with baseline length. 

    5) Bandpass stability is outstanding.  All subband bandpasses 
connect in amplitude and phase.  Subband #1, as usual, is useless on its 
lower edge.   There is something quite wrong with antenna 8, LCP -- the 
bandpass shows amazing (but stable) structure  of about 3.2 MHz period... 

    B)  K-Band. 

    1) The initial scan phase jump is only weakly seen.  When present, 
it is 30 degrees or less. 

    2) Amplitude stability is similarly outstanding.   The same 'late to 
get on' behavior with antenna 9 is seen.   Antenna 27 has very low 
amplitudes in both polarizations and on all subbands, representing a 
real SNR loss -- as it did last Friday. 

    3) There were *no* zero records that I could find. 

    4) Phase connection is generally good, except on antenna 8, which 
has the same 15 degree offset between sources that I found last Friday 
(with a different pair of sources, but the same elevation separation).  
Antenna 24 also has a clear offset at most times, typically 5 to 10 
degrees.   Complicating interpretation is evidence of an unusual 
atmospheric-like event which passed by during the first 20 minutes or so 
of the run -- slowly oscillating phases (period of 8 minutes) which is 
clearly a coherent pattern over the entire array.  The pk-pk of these 
waves (from end of West arm to end of East) is well over 130 degrees!  
The phase amplitude clearly scales with distance down the arm.  If not 
due to the atmosphere, the cause certainly knows the location of the 
antennas ...  The phenomenon died away by the end of the run. 

    5)  The bandpass characteristics are the same as at C-band. 

    C)  VLA K-band comparison. 

    Immediately following the WIDAR K-band observation, the same script 
was run with the VLA's correlator.  Key results are:

    1) The 15 degree phase oscillation between the pair of sources seen 
at K-band on antenna 8 is also seen, with identical characteristics, in 
the VLA correlator data. 
    2) The smaller phase step seen in antenna 24 is also seen in VLA 
correlator data. 
    3) Phase offsets between the sources for non-WIDAR antennas are very 
common in the VLA data:  antenna 1 has a 15 degree step.  Antennas 10 
and 12 have a 10 degree step.  Some non-WIDAR antennas have quite poor 
phase stability-- or steps -- these are hard to discriminate.  The 
central point is that I see no evidence that these issues have anything 
to do with WIDAR.   They occur in equal measure with the VLA correlator. 
    4) The low sensitivity seen with antenna 27 at K-band with WIDAR is 
***not*** seen in the VLA correlator data.   This might be interpreted 
as something caused by the formatter -- but this antenna gives perfectly 
good data at C-band through WIDAR.  So the only plausible solution (that 
comes to my mind) is in pointing.  But this would require that the VLA 
K-band observation have referenced pointing done.   Looking at the logs 
indicates this is not possible -- only one minute passed between the end 
of the WIDAR K-band test and the beginning of the VLA corrrelator test...