[evlatests] Another curious transition issue

[Rick Perley]

    I used an hour of maintenance time last night to  test EVLA and VLA 
sensitivities across all of L-band.  The tunings ran from 1025 through 
2200 MHz in steps of 20 MHz, with the BD IF set 10 MHz above the ACs. 

    The results for the EVLA at frequencies below about 1250 MHz were 
unusual, in that the A and C IFs shows highly variable sensitivities 
(varying with frequency) while the B and D IFs degraded smoothly as the 
frequency decreased. 

    Jim Jackson immediately suspected the power leveling circuitry, 
although it was not immediately obvious how this might be responsible, 
given that the leveling is done with the power detected in a full 1 -- 2 
GHz bandpass which is output from the T304. 

    I thus made up a short file to run this afternoon, while Ken watched 
what was happening as we tuned through this area.  This exercise 
clarified the problem ...

    The essential issue is that in order to fit the slice of frequency 
space which we are going to send to the old (narrow-band) correlator, 
the L301 and L302s must shift the spectrum so the desired frequency 
slice fits into the fixed 64 MHz-wide window of Mike's D-to-A box.  This 
window is (to my way of thinking) located at about 1830 MHz, so when we 
wish to observe at a very low frequency, the overall front-end spectrum 
is shifted way, way over.  This means that the total power 'seen' by the 
output TP meter in the T304 is very much reduced, as the front-ends 
provide very little power at (real) frequencies below 1.2 GHz, inducing 
the T304 to decrease the attenuation to keep the total power at the 
agreed upon level.  This is appropriate for the 8-bit sampler, but is 
not good downstream, because the spectral power density has been greatly 
increased.  So when the signal passes through the D-A window, the total 
power is now too high, and overdrives the backend A/D whose job it is to 
keep the total power in the range the VLA's 3-bit sampler needs.  This 
causes a dramatic loss in correlator coefficient (readily seen).   

    For the B/D IFs, the situation is essentially reversed.  Here there 
is to power autolevelling -- levels are set by a table.  As we tune 
downwards in frequency, the actual power delivered to the 8-bit sampler 
declines -- which is of little consequence to that device.  Apparently, 
the reduced power getting through the D-A's window is within the range 
that the VLA's backend A/D can handle -- although Ken noted that the 
gain was maxed out for these frequences. 

    This is another 'transition' issue, but will only affect those brave 
enough to observe at frequencies below 1.2 GHz.  Currently, only antenna 
14 has an OMT which lets useful signals through to the correlator (some 
very nice plots showing this will be in an upcoming memo) at these 
frequencies.  

    I'll add here that remarkable observation that I got good, stable 
fringes at all frequencies between 945 and 1250 MHz in this afternoon's 
tests -- despite the multitude of aircraft up there!  More on this later.