[evlatests] Another S-band power setup failure

[Rick Perley]

    I have found another instance where the power setup at S-band was 
grossly in error, resulting in serious compromising of the observation 
goals. 

    In this case, the source and calibrator are near dec = -10, so we 
are in a 'bad neighborhood', and trouble can be expected.  The B/D side 
set up more or less correctly, with the PSum values (for a central 
subband) ranging from 7 to 15 counts.  (The nominal level is ~ 14 -- 
anything within 50% of that is considered o.k.). 
    But the A/C side was spectacularly in error, with Psum values 
varying from 0.1 to 4, with nearly all of them less than 1.  That's a 
factor of 15 or more too low!  Unsurprisingly, the data are seriously 
degraded, with the noise nearly doubled. 

    Almost certainly, the problem is that the antennas were pointed too 
close to one of the satellites which radiates strongly in subband 2  
(2180 to 2200 MHz is where the power lies), or subband 3 (2320 -- 2350  
MHz -- this is the Sirius/XM band) at the moment that 'set and remember' 
was doing its thing.  This is the second failure in the set/remember 
procedure that I've found in the 15 S-band databases that I've 
calibrated so far.  (Granted -- the first failure, which I reported on 
earlier this week, is bizarre, and likely has a different cause).  
Nevertheless, two out of 15 is too high, and I think we need to find a 
way to prevent these serious failures.

    For S-band, I suggest two approaches:  (1) Avoidance:  The gain 
setup procedure should be done in a 'safe' direction.  Such directions 
exist (to the NW or NE), but this will often require a fairly long slew, 
hence lost observing time.  (2) Review:  When the system power (either 
PSum or an analog measure) is out of range by a significant value (I'd 
suggest a factor of two), an alert should be generated, and the 
procedure repeated.  (This of course brings up the question of how this 
should be done, and what is to be done if the next try also results in 
an unacceptable value).  Alternatively, default gain setting should be 
utilized.  I prefer this approach -- even stupid values wrong by a 
factor of two are better than what we are occasionally seeing now.