[evlatests] Results from P-band test run

Rick Perley rperley at nrao.edu
Fri Aug 16 17:48:46 EDT 2013


    I ran a short test last evening, with the primary purpose of 
determining whether using narrow subbands at P-band is a useful thing to 
do. 

    The motivation arises from the switched power.  The current wide 
subbands (128 MHz) are good if one wants to avoid the digital filter 
'notches'.  However, if one wants (or needs) to utilize the switched 
power values (for, say, tracking gain changes in the electronics), a 128 
MHz-wide subband is a real problem because:
         a) the PSum and PDif values, being total power, are hugely 
sensitive to RFI contained within the subband.  The two 128 MHz-wide 
subbands each contain *a lot* of RFI
         b) the value of the Tcals will surely change dramatically as a 
function of frequency -- the 128 MHz-wide defaults are a large fraction 
of the observing frequency, so I would expect this variation to be 
significant.  Narrower subbands will permit a reasonable interpolation 
of Tcal over frequency (as we do at the higher frequency bands).

    For last night's experiment, I chose two setups:

    16 x 16MHz subbands, covering 236 through 492 MHz, with 250 kHz 
resolution
    8 x 32 MHz subbands, covering the same range with the same resolution.

    Both worked flawlessly.  I observed 3C295 and 3C48.

    Examination of the PSum and PDif values for the first setup shows 
nice, stable values for the subbands that are absent strong RFI.  These 
'clean' subbands (not perfectly clean -- but really close) are 4 through 
8 (AIPS-style counting):  284 through 364 MHz.  RFI is really, really 
bad in subbands 1, 2, 15 and 16 (236 to 268 and 460 to 492 MHz).  The 
remaining subbands all have significant useful spectral spans -- any 
good RFI-flagging algorithm should easily clean up the spectra.  (I 
didn't bother to do this , for these tests). 

    All antennas were in the array, except ea21, which is in the barn.  
All antennas fringed, except ea19, which has no receiver.  Of the 26 
fringing antennas, good fringes were obtained from all except ea01, 
which has virtually no signal in 'RCP'.  (Quotes, because I don't yet 
know if this is 'H' or 'V').  *** Dan:  we can resolve this if you tell 
me which polarization -- horizontal or vertical -- is not working from 
ea01***). 

    The stability of the correlations (amplitude and phase) is wonderful 
('to die for'!). 

    I used the known values for the two target sources to generate the 
'absolute' (non-normalized) bandpasses.  The bandpass amplitudes are 
remarkably flat from 240 to about 450 MHz, above which many antennas 
begin to lose signal.

    Many antennas show a rapid bandpass oscillation (standing wave) of 
spectral period ~3.2 MHz.  This is seen in both amplitude and phase, and 
corresponds to a (free-space) length of 150/6.4 = 47 meters.  This looks 
to me like an impedance mismatch in the cable junctions.  The antennas 
with the most prominent effects are:  7R and 25R.  Although the 
oscillation can be seen in about half of the others, it is quite minor.  
ea25R has a (much larger) slower oscillation as well -- of period 38 
MHz.  If due to a reflection this corresponds to a free-space length of 
~ 4 meters.
Two antennas have large (10 dB) and irregular bandpass variations -- 
ea04L, 07L.

    Polarizations are swapped on ea03, 09, 14 and 26.  I'm also pretty 
sure they are swapped in ea01 (the near-dead 'RCP' channel makes this 
determination slightly dicier). 

    High cross-polarization (> 20%) is seen in ea05, 06, 09, 11, 12 and 
14.  The worst (by far) are ea12 and ea14 -- 40% or more. 

    Stable, non-zero, PDif values are seen on all antennas except:  
ea01, ea10, ea11 ea13 and ea25.  All of these have good PSum.  On ea02, 
the PDif values are good in 'RCP', but are variable and weak in 'LCP'. 
    Curiously, a number of antenna-IFs were set to too low a power (as 
seen in PSum) -- all the really low ones are in the 'RCP' side.  This is 
also seen in the visibilities from these antennas. 

    A curious effect is seen when examining the 'CD' table ('Cal Device' 
-- where the TCal values are stored in AIPS).  For the 16-subband 
database, I find that only the first four subbands have Tcal values 
entered -- all the rest have been left at the 'native' values of 1.0. 
    Non-unity Tcal values (for subbands 1 through 4) are provided for 
antennas ea02, 03, 06, 08, 09, 11, 18, 20, 26 and 27. 

    The calibrated visibility spectra are lovely -- showing the expected 
'horn' effect at short spacings expected from background confusion. 

    In short -- this system really works well!  Far better than the old, 
narrowband one. 





More information about the evlatests mailing list