[evlatests] Large Bz error?

[Rick Perley]

    Further observations based on the Thursday night data ...

    The observations on Thursday evening were of four sources, very 
closely spaced -- maximum extent in RA = 10 degrees, maximum extent in 
Dec = 4.0 degrees.  Following removal of the bad data (previously 
described), a CALIB was run.   The solutions revealed the following:

    1) For most antennas, the temporal phase drift is fairly small -- 
typically 10 to 20 degrees over the 1.5 hours.  This indicates an (x,y) 
baseline error of a couple of centimeters.   One antenna, however, is 
very different:  Antenna #1 (at W9) shows a clearly sinusoidal change of 
250 degrees over the 90 minutes -- this indicates an error of at least 
25 centimeters in (x,y). 

    2) There is a very clear, and very large, time independent offset in 
phase between the four sources.  The offset is clearly a function of the 
source declination.  Such an offset is most likely due to a error in the 
antenna position z-coordinate.    Not all antennas show this effect, but 
those that show it strongly are all at the ends of the arms.  Here is a 
short table showing the phase offset between two sources at 32.3 and 
28.4 degrees declination, for each antenna.  Antenna 25 (at N2) was 
taken as reference.

    1   W9    -80 degrees
    2   E2       0
    3   E9      +50
    4   W1     0
    8   N12    -90
    9   E6       0
    18   N18  -120
    19   W4   0
    23   E7   +60
    24   N10   ~0
    25   N2    Ref
    28   N8   -30
----------------------------------------
       Roughly, a difference of 100 degrees corresponds to an error 
about 30 cm in the z-coordinate.   Phase offsets less than ~20 degrees 
are essentially buried in the atmospheric noise. 
    Note that the large negative offset are on the W and N arms, while 
the E arm shows positive offsets.  However, the offsets are not linear 
with distance -- antenna 24 (N10) shows no offset at all, while the 
antennas on each side do, while antenna 9, at E6 shows no offset, but is 
immediately adjacent to 23 on E7, which has a large offset ... 

    These position errors should explain why phase calibration of an 
object which is a few degrees away from a calibrator returns a very poor 
image.