[evlatests] Band Switch Failure STatistics (upper limits)

[Rick Perley]

    The previous report dealt with changes in PDif seen when we change 
bands -- changes from prior values are often due to a switch failing to 
latch correctly, resulting in extra loss.   This failure mode is not 
critical, in that the switched power calibration system is very 
effective in correcting for the change in gain. 

    More critical is the class of failure where the switch fails to 
close at all -- resulting in no fringes.  I've reviewed the data from 
the Tuesday test, and counted the number of times fringes failed to 
appear after a band change. 

    From the total of 27 antennas * 113 commanded band changes * 4 IFs = 
12200 'change events', I find there are a total of about 52 failures to 
provide any fringes -- (0.4% failure rate).  I counted these quickly, 
and may have missed a few -- let's call it 0.5%.  Nearly all of these 
affect all four IFs at the same time.  However, not all of these should 
be blamed on switch failures -- a failure of the subreflector to rotate 
properly, or the antenna to point properly (failed reference solution?) 
would also result in all four IFs failing to provide fringes. 

       The few failures I noted are well distributed amongst the various 
bands.  No preference for a single band to fail is seen. 

    A few antennas demonstrated particularly bad behavior -- probably  
not related to switch failures:

    a) Antenna 13 provided fringes only on one scan (the first) at K, A, 
and Q bands -- all others were 'dead'.  I did not include this in the 
statistics above. 
    b) Antenna 27, in IF 1 (A and C) failed to provide fringes on about 
half the scans at L and S bands only.  Curiously, the failures at L-band 
were matched with zero PDif values -- but not at S-band!  (No 
explanation comes to mind). 
    c) Antenna 16 had a change of delay of a whopping 16 nsec, halfway 
through the run -- seen at all bands, but in IF 'C' only.