[evlatests] P-Band performance, and a polarization mystery ...

[Rick Perley]

     I had time to better review the test data taken 31 December.


     The test was taken on 3C147, probably the best source for P-band 
testing.   The flux density of 3C147 is about 55 Jy (more at lower 
frequencies, less at higher).  And it is completely unpolarized.  Also 
observed was blank sky, about 10 degrees north of 3C147.

     All antennas except ea23 were in the array.

     Basic Results:


     1) *ea14* was 'dead' -- meaning such weak fringes that the 
calibration programs dropped this antenna.

     2) *Bandpasses*:  Nothing different than previous reports have 
stated.  All antennas show the 3 MHz ripple to some (small) degree.  
ea04X has a large scale phase ripple (20 MHz period -- corresponding to 
7.5 meters in free space) visible over the entire bandpass.  ea11X has a 
remarkable resonance ('suck-out'), at 440 MHz, with 10 dB power loss.  
The width of the resonance is about 5 MHz.

     3) *Sensitivity*.  I used the histogram-plotting program UVHGM to 
generate noise histograms of the calibrated blank sky data -- real and 
imaginary amplitudes separately.  A 5 km short-spacing cutoff was 
employed to prevent large-scale structure from biasing the results.  All 
histograms were beautifully gaussian ('text-book quality').

     a) Only one antennas was clearly discrepant:  ea18Y has a 
sensitivity about 1/2 (that is, worse) that of all other functioning 
antennas.

     b) There is a significant difference in sensitivity between the two 
polarizations -- the 'XX' (vertical dipole) sensitivity is 8% better 
than the 'YY' polarization.  The *median* histogram width for the XX 
correlations is 2.8 Jy (1 second averaging, 125 kHz channelwidth -- 
Hanning smoothed).  For the YY correlations, the value is 3.05 Jy.

     Histograms of the antenna sensitivities were generated -- the 
dispersion is quite small (about 0.2 Jy).  Antennas which are clearly 
less sensitive than the median are:

      XX side:  ea01, ea11, ea04, ea09, ea12 (from worst to less-worse).

      YY side:  ea18 (as noted), ea04, ea12, ea19, ea24.

     4) *Polarization*.  A most remarkable characteristic has been 
found.  When reviewing the antennas which have low cross-polarization, 
it is easily seen that the entire array can be cleanly divided into 4 or 
5 subarrays.  The members within each group have low cross polarization 
(less than 5%) among themselves, and have high polarization with the 
members of all other groups. The groups are:

     A) ea01, ea03, ea07, ea12, ea16, ea22, ea25 and ea26

     B) ea02, ea06, ea08, ea18, ea19, ea24, and ea27

     C) ea04, ea11.  (This group is only slightly disjoint from group B 
-- they could be considered a single group.)

     D) ea05, ea09, ea20, and ea28

     E) ea10, ea13, ea17, ea21, and ea27.

     One antenna belongs to no group -- ea15.  It is highly 
cross-polarized against everybody.

     So there is some characteristic which is causing this grouping.   I 
(desperately) hope the easiest explanation is the answer:  That the 
members of each group share a common dipole orientation.  Using group A 
as reference, the cross-polarizations of the others are 5 -- 15% (in 
voltage).  Dipole misorientations of less than 10 degrees can do this.

     But there may be some other curious correlation, unknown to me, 
responsible for these groupings.  Any ideas?  (It's not related to the 
arm these antennas are on).

     One more interesting fact:  The AIPS program 'PCAL' perfectly 
solved for, and removed these cross-polarizations, when used in its 
'continuum' mode, and when the polarization of 3C147 was specified as 
zero (which is certainly a good approximation).