[evlatests] L-Band sensitivity, continued ...
Rick Perley
rperley at nrao.edu
Wed Jul 26 19:44:52 EDT 2006
Mike Revnell earlier today circulated a note pointing out that
antenna 13, in particular, as a strong spectral slope. Peter N.
reminded me that in such a case, quantizer noise can significantly
degrade the SNR for those parts of the passband where the spectral power
density is low.
To check if this is the cause of our decreasing sensitivity with
increasing frequency, we used David Sevilla's program (which collects
the FFT spectra that Mike Revnell generates, averages over arbitrary
times, and computes the correct sky frequency) to provide some
near-real-time spectra. We set it up to produce 16384 channels (or, 61
kHz resolution), on IF 'A', for antennas 13, 14, 16, and 18. We then
commanded the array to cycle through the spectral sweep file I've used
over the past two days.
Results:
Antenna 13 indeed has quite a strong slope in its bandpass. The
maximum spectral power is near 1150 MHz, decreasing by nearly 10 dB to
1400 MHz, then becomes flat to about 1600 MHz, and drops by another 8 dB
or so to 1750 MHz.
However, antennas 14, 16, and 18 have quite flat power spectral
densities from 1100 through 1780 MHz (above which there was a filter
cutoff), with maximum deviations of no more than 3 dB about the mean.
So while the slope in antenna 13's spectrum may contribute to the
sensitivity loss, the losses seen in antennas 14, 16, and 18 should not
originate from quantization of the bandpass.
We were easily able to confirm that Walter has correctly computed
the frequencies -- the DMEs, ABQ radars, INMARSAT, GLONASS, and Iridium
all show up in their assigned slots. With the spectral resolution
selected (61 kHz), the maximum was about 20 dB above the sky noise.
With 16384 channels, each spike is diluted by -10*log(16384) = -42 dB,
so each contributes less than 1% to the total system power. There may
be 100 or more channels with RFI at any time, but each contributes very
much less than that maximum -- I judge the total RFI power seen at any
moment to be somewhat less than the total cold sky noise (with an error
bar of about 10 dB!)
More information about the evlatests
mailing list