[evlatests] L-Band Phase Stability and Polarization

[Rick Perley]

    I used 3 hours' of Dynamic  Time last evening to stress the system 
again, this time with an eye to determining the polarization 
characteristics across the full interim bandwidth (1200 to 2000 MHz). 

    Observations were made in polarization mode 'PB', to use the `BD' IF 
pair, as I understand that 26 A and C are not yet back in operation at 
L-band.  The BW = 12.5, giving 781 kHz resolution. 
    I chose 7 frequencies to sample:  1240, 1360, 1485, 1612, 1725, 
1825, and 1950 MHz.  Observations were made of 3C84 and 3C138 (the 
latter to set the B-D phase difference), and as a monitor of whether the 
derived polarization is right.  Ten observations, at each frequency, 
were made of 3C84, and a single observation of 3C138 was made at the 
end.  I cycled through these frequencies, from lowest to highest, with 2 
minutes of duration for each observation.  So, any given frequency was 
sampled once every 14 minutes throughout the 3 hours. 

    Thanks to Rob Long, for correcting the RCP wiring problem in antenna 
17, just prior to the start of this run. 

    All eight EVLA antennas fringed nicely in both polarizations at all 
frequencies, with the following exceptions:

    Antenna 13 failed to fringe at 1360 MHz on every observation (both 
polarizations).
    Antenna 26 failed to fringe on about half of the observations at 
1360 and 1485 MHz. 

    All VLA antennas fringed at all frequencies between 1240 and 1725 MHz. 

    My choice of 1612 MHz proved a poor one -- considerable RFI was 
seen.  All others were 'clean'. 

    The test ran from ~5:30 to 8:30 PM, MST.

Results:

    A)  Phase Behavior. 

    A different (and perhaps new?) form of global phase jump has been 
found.  All EVLA antennas, referenced to any VLA antenna, are showing 
phase jumps *of precisely 180 degrees, regardless of the frequency!!*.  
There were between one and 6 such jumps at each frequency.   The jumps 
always occured between scans -- no discontinous change in phase was seen 
within any scan. 

    All EVLA antennas have a very strong phase gradient w.r.t the VLA 
antennas of about -85 degrees per hour.  The opposite polarizations are 
identical.  I cannot determine the difference between IF pairs, as only 
one was observed.
    The EVLA antennas' phase drift w.r.t an EVLA antenna is much 
reduced, but still very evident.  When referenced to EVLA antenna 14, 
the residual phase drift is approximately proportional to both frequency 
and distance from antenna 14. 

    Vivek believes the drifts are largely due to RT phase not being 
applied, and is independently trying to confirm this. 

    Only one EVLA antennas has odd short-term phase behavior:  Antenna 
24, at all frequencies and in both IFs, has occasional 10-second long 
jumps in phase, of magnitude ~90 degrees.  After 10 seconds, it returns 
to the previous value.  This jumps happens infrequency -- about once per 
2-minute observation. 

    B)  Amplitude Behavior.

    Nothing unusual or worrisome was noted. 

    C) Polarization. 

    As a reminder, the EVLA antennas are all outfitted with the 
quadrature hybrid polarizer (which converts the linear output of the OMT 
to a circular basis), while the VLA antennas retain the old polarizers.  
In addition, EVLA antenna 14 has the new OMT -- all others retain the 
old VLA OMT. 

    The primary goal of this test was to see if decent polarimetry could 
be done in the extended frequency range.  The cross-polarizations were 
determined using the AIPS program PCAL, using both VLA and EVLA antennas 
as references (as the cross-polarizations determined as differential, 
w.r.t  a reference antenna).  Below are comments for each band.  A more 
sophisticated analysis will be done at a later time by George 
Moellenbrock. 

    In short -- good solutions were found for all antennas, at all 
frequencies.  The derived polarization for the sources looks right.   No 
statements on stability of the solutions can be derived from this work. 
   
    1240 MHz:   With VLA1 as reference, the EVLA antennas show very 
little difference from VLA antennas in their X-polarization 
percentages.  Antennas 16, 17, and 18 are all quite high (10% on LCP 
side), but we note that many VLA antennas are at 8%.   Little changes 
when EVLA14 is the reference.
       Good solutions were determined, and the derived polarization of 
3C84 (zero, to within the noise), and 3C138 was obtained. 

    1360 MHz.  With VLA1 as reference, VLA antenna polarization are low 
-- 2 -- 3 %, while some EVLA ones are higher -- typically near 10%.  
With 16 as reference, ant14 is very high (12%), and 18 and 23 are quite 
good, 3 -- 4%.  Other EVLA antennas lie in between. 

    1485 MHz.  With VLA1 as reference, VLA antennas are low -- typically 
3%.  EVLA antennas are typically 6%.  With 14 as reference, all other 
EVLA antennas are quite highly polarized -- typically 10%.  With 18 as 
reference, all antennas are typically 6 - 8% polarized, except 14 and 17 
with are over 10%. 

    1612 MHz.  With VLA 1 as reference, VLA antennas are at the 1 -- 2% 
level, (except 28, which is 10% polarized -- clearly a special problem), 
while EVLA antennas are typically 4%.  With various EVLA antennas as 
reference, the results are mixed -- this always makes the VLA antennas 
show higher polarization, but some EVLA antennas are worse, and some 
better, depending on what the reference is. 

    1725 MHz.  With 1 as reference, typical VLA polarization are 2 to 
5%.  EVLA polarization are 5 to 7%.  With an EVLA reference, the 
situation described at 1612 Mhz is seen. 

    1825 MHz.  Only the EVLA antennas work here.  Polarizations are high 
-- 8 to 10%. 

    1950 MHz.  Quite high polarizations -- 10 to 20% with 14 as 
reference.  Making 16 the reference dramatically drops the polarization 
on all higher numbered antennas -- to typically 8%, while 13 and14 stay 
very high. 

    D)  Imaging.

    I made images at all frequencies, looking for signs of significant 
non-closing errors.  For all frequencies except 1612 MHz (where RFI is 
clearly an issue), the images are very nice -- a peak to rms noise ratio 
of 20,000 or better.  Given that each of these images included only 20 
minutes of data, over a duration of slightly more than two hours, and 
that 3C84 is quite a complicated object at this resolution and 
frequency, I think these images very nice indeed, and see no evidence 
for non-closing errors.