[evlatests] P-band (non)-linearities -- Very Good News!
Rick Perley
rperley at nrao.edu
Wed Oct 2 13:02:24 EDT 2013
I reported last Friday that a clear relation between P-band
non-linearities and power delivered to the quantizers was found, in the
(unsurprising) sense that the higher the power delivered to the
quantizers (and re-quantizers), the greater the apparent loss of
visibility when observing the powerful source Cygnus A.
Ken and I discussed the issue. It seemed clear that the problem
lies in the requantizers -- these have 8 levels each side of zero. With
a normal, cold-sky, setup, the rms level is set to about 2.7 levels.
Hence, voltages above (or below) 3-sigma will fall off the sample grid,
and be represented as the maximum (or minimum) levels. rather than their
true values. Fine for the cold-sky source. But when we go to Cygnus A,
which approximately quadruples the power, the 1-sigma voltage is now at
3.7 levels, so voltages beyond 2-sigma are now off the grid.
To test this, we ran an experiment last night, with three different
setup strategies:
1) Set up the power (both the T304 attenuators *and* the
requantizers, to their normal levels) on Cygnus A. Then observe the
calibrator 3C380, and use these gains to calibrate Cygnus A.
2) Set up on 3C380, then observe Cygnus A. Calibrate as above.
3) A hybrid arrangement: Setup the T304 powers on Cygnus A, and the
requantizers separately on Cygnus A and 3C380. (Apparently, we can
remember each ...) Observe both sources, and calibrate as above.
Results:
The results from setups #1 and #2 are perfectly clear:
Setup #1 provides exactly correct flux density of Cygnus A (to 1.5%
of the Baars value) with the assumed value of 3C380 = 40.0 Jy. (This is
at 336 MHz). There are no discrepant antennas.
Setup #2 provides Cygnus A fluxes 'all over the map' -- all of them
too low. We are clearly in a severe (digital) compression regime.
Some sort of error occurred on setup#3 -- we'll investigate this
further, later.
Investigation of the PSum values provides extra, useful information:
All antennas, except one, had the digital (PSum) power for the
setup-source (Cyg A for #1, 3C380 for #2) equal to 14.5 to 15 counts.
This is in stark contrast to the P-band flux density experiment, run
early last month, where the powers varied from 10 to more than 50
counts. The difference appears to be in this new test, the requantizers
were set separately, following the setting of the T304 attenuators.
The one discrepant antenna is ea28, which set itself up (on both
polarizations) to 28 counts -- two times too high.
Ken will probably add some description of the setup regimen, with
recommendations of how the system setup should be done in the future at
this band (and perhaps also at L and S bands).
More information about the evlatests
mailing list