[evlatests] P-Band Sensitivity

[Rick Perley]

     A few spare moments last week were used to determine the basic 
sensitivity of the P-band antennas.  I observed 3C286, and an adjacent 
'blank' field with high time/frequency resolution:  1 second and 62.5 
kHz.  The idea is to reduce the (B.t) product to a value small enough 
that thermal noise dominates the 'confusion noise' -- the rms visibility 
contribution from all sources visible to the interferometer.

     There were  six antennas at the ends of the E and W arms -- all 
others were in the 'D' configuration.  It was quickly found that good 
calibration was only possible restricting the (u,v) range to baselines 
longer than 1 K-lambda -- i.e., excluding all D-configuration 
baselines.  When this was done, the BP solution was quite good.  To keep 
things simple, and since I didn't care about the phase solutions, a 
single 'absolute' BP solution was made -- this effectively gain 
calibrates the data without need for CALIB.  I used the Scaife-Heald 
scale for setting the flux density of 3C286. This basic procedure was 
checked by plotting the calibrated visibility for 3C286 -- the 
amplitudes were at the correct levels.

     For the purpose of determining the basic sensitivity, I selected 
only those spectral channels lying in RFI-free zones.  After Hanning 
smoothing, then BP calibration, the noise histograms were generated for 
these RFI-free regions, again using only the longest spacings.   For 
nearly all chosen frequency zones, the histograms are perfectly 
gaussian.  The widths are typically 5 to 10 Jy.  There will be small 
correction for confusion visibilities -- I haven't bothered to do this yet.
     A more realistic way to generate the sensitivity is through direct 
imaging.  I tried this, but the (u,v) coverage is so bad that the images 
were quite useless.

     I converted these gaussian widths to SEFD via:

     SEFD = sigma.sqrt(2.DeltaT.DeltaNu)/0.95

     where DeltaT = 1 sec, and I took DeltaNu = 100 kHz (close to the 
appropriate width for Hanning smoothed data with the basic frequency 
width of 62.5 MHz).   The factor 0.95 is to account for the efficiency 
of the digital correlator.

     The results are shown in the attached plot, showing the SEFD as a 
function of frequency from 245 through 475 MHz.  The subband boundaries 
are shown as vertical red lines -- I excluded the far edges (most of 
subbands #1 and #16), as the sensitivities are greatly degraded at those 
frequencies.
     The MUOS band is shown -- which explains why there are no 
measurements in subband#9.

     The general picture is as has been noted in the past:  The best 
sensitivity lies between 290 through 360 MHz (corresponding to the 
original design range for the dipoles).  Between 380 and 450 MHz the 
sensitivity is fairly uniform, at a level about 50% worse than in the 
central.  The choppiness of the measured sensitivities as a function of 
frequency in this region is probably caused by RFI effects -- it's hard 
to perfectly avoid all the sporadic RFI in this zone.
     An interesting curiosity are the two 'humps' in sensitivity near 
325 and 345 MHz.  These seem to align with the subband boundaries, but 
are far too broad to be explained by the subband sensitivity rolloff.
     The rapid rise in noise at the high end of subband#8 above 355 MHz 
is real, and doesn't appear to be caused by the proximity of the MUOS 
emissions.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: PBandSensJune2014.png
Type: image/png
Size: 14202 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20140625/fb8f5c2e/attachment.png>