[evlatests] Effects of the 4-band dipoles

[Rick Perley]

    Although the circumstantial evidence that the 4-band dipoles were 
responsible for the high and variable L-band polarization noted in tests 
made last fall, it was felt that a clear 'on-off' test to prove the case 
was needed. 

    To do this, it was decided to temporarily mount four 4-band dipoles 
on four antennas at the beginning of the two-day maintenance time, and 
to do careful polarimetry observations both before the dipoles were 
installed, and afterwards. 

    The antennas chosen for dipole mounting were #1, 21, 23, and 26.   

    The observations were set up this way:

    A 6-hour block of time, 16:30 through 22:30 LST was reserved for 
both Tuesday and Wednesday mornings.  The dipoles were installed Tuesday 
during the day.
    The observations were identical for each day:  2-minute durations of 
a strong, unresolved calibrator, 1924+334, at the default 50MHz BW 
frequencies in L, C, and X band. 

    EVLA antennas 11, 13, 14, 16, 17, 18, and 19 were unavailable on the 
second day, due to the central antenna rack being down.  This had no 
effect on the results.   EVLA antennas which were present on both days, 
and did not receive dipoles were 4, 5, and 24. 

    Other than minor problems already noted elsewhere, the data were of 
outstanding quality. 

    Results:

    A) Polarization

    The variable polarization noted last fall at L-band is reproduced on 
the four modified antennas at L-band, and on no others.  The source has 
a small (8 mJy, or 0.2%) linear polarization.  For the VLA, and for all 
EVLA antennas without dipoles, (and excluding a short period around 
meridian transit), the calibrated data show a steady ~10 mJy amplitude 
and 0 phase for Stokes Q.  (I rotated the R-L phase so the polarized 
signal lies entirely within the Q visibility). 
    For the four antennas with dipoles, the residual signals show 
time-variable amplitudes of up to 60 mJy in both Q and U and phases of 
arbitrary (non-zero) value.   This is about a 1.5% false and variable 
polarization, far above the system requirements.  The variability 
timescale is an hour or so, exactly as noted in the fall, and there are 
no obvious symmetries with hour angle, elevation, or parallactic angle. 

    No effect of the dipoles on polarization is seen at C or X bands. 

    B) Sensitivity

    Measurement made long ago indicated that the 4-band dipoles cause a 
loss of G/T of about 7% at L-band.  No clear effect was noted at any 
other band, but these old measurements were of mixed quality. 
    Use of correlation coefficients makes G/T measurements dead simple.  
Accuracies better than 1% are straightforward.  I find that the dipoles 
cause of loss in G/T at L-band of 6%.  I find also that there is an 
equal loss at C-band, and a slightly smaller effect, 3 to 4% loss in G/T 
at X-band. 
    In hindsight, it's clear I should have extended the observations to 
K and Q bands ...

    This raises the question of whether the P-band dipoles have a 
noticeable effect on high-frequency performance.   I'd be happy to make 
the observations, if we can arrange to temporarily remove them.