[evlatests] X-Band Beats

[Rick Perley]

    I reported yesterday (at the bottom of a message about drop-outs) on 
a strong
beat phenomenon seen on the EVLA antennas at two frequencies:  8185, and 
8785 MHz.

    I took a much closer look at the data today.  The results:

    A) This is indeed a EVLA-only phenomenon.  All EVLA to VLA baselines 
are clean.

    B) The 8185 MHz beat is between the astronomical signal and a signal 
whose phase
is unchanging between the two antennas.   We know this because the beat 
phenomenon
has a period *exactly* equal to the natural fringe rate.   Noting that twice
the 4096 MHz LO is 8192 -- which lies within the passband at 8185 MHz, 
it is suggested
that there is a significant leakage of LO power.  If it is a CW, it must 
be quite strong,
as the modulation of the beat signal is near 100% (i.e., we get as much 
correlation in
the LO as we do in the astronomical broad-band signal). 
    I checked with Vivek, who has long ago searched for such LO signals 
in the
bandpass, but he reports that he found nothing at 8192 MHz.    Perhaps 
we should
use Mike's spectral snaphot program to look for spikes?  (The correlator 
is out of
commission for most of the day, so single-antenna tests are all we can 
get). 

    C)  The 8785 MHz beat must have a different origin.  Although it 
clearly has a
periodicity, the period has no obvious relation to the period of the 
natural fringe.  For
example, on baseline 13 x 14, whose U-coordinate is 3.5 Klambda, we 
would expect a
5-second beat.  The apparent period is 0.8 seconds (or shorter -- hard 
to tell with 0.4
second integration).  Baseline 13 x 18 has a U-coordinate of -12 
KLambda, which would
predict a 1.5 second period.  What we see is a beautiful cyclic beat of 
7 second period. 
Baseline 14 x 18 has a U-coordinate of -15 KLambda , so a natural period 
of 1.2 seconds
would be expected.  We see no obvious periodicity at all, but instead a 
ratty amplitude
and phase, which would be expected if the actual beat has a period very 
much less than
the 0.4 second integration.    Is this some sort of aliased broadband 
signal?