[evlatests] More on System Non-Linear performance

[Rick Perley]

    I reported a few weeks ago on an unexpected result, wherein 
observations of Cygnus A, which increases system power by a significant 
factor (factor of 4 at L-band, 3 at S-band, and 2 at C-band) caused the 
PDif values to noticeably decline, typically by 10% or so.  This should 
not happen in a linear system with fixed system gain.  Probably 
importantly, there is no such effect at X-Band.  (The effect has not yet 
been tested for at higher frequencies). 

    I now have more evidence of the effect, taken from the S-band data 
taken on 3C273.  This object is at zero declination, so it rises 
more-or-less in the east, transits through the geostationary belt, and 
sets in the west.  The observations spanned 2 through 4 GHz, and we have 
switched power in each of the 16 subbands (each 128 MHz) wide.  
Satellite radio lies within subband 3, while another strong emitter lies 
in subband 2. 

    The effects of XM radio are easy to see.  As the source approached 
the geostationary belt, PSum increased by up to a factor of 150 within 
subband 3.  The interferor in subband 2 is clearly also in the 
geostationary belt, as the power within that subband increased also, but 
'only' by a factor of about 15.  Unsurprisingly, the PDif values in both 
these two subbands are completely trashed during the time when the 
satellite signals are dominating. 
    In an ideal linear system, these effects should be confined to the 
subbands in which they lie -- so in our case, all the other subbands 
should show PSum and PDif values completely independent of what's 
happening in the trashed subbands.

    But that's not what happened (sad to say).  All the PDif and PSum 
values in the 'good' subbands show strong perturbations which reflect 
the increased power in subbands 2 and 3.  In all cases, the PDif and 
PSum values decline, by up to 30%.  We are seeing strong evidence for an 
overall gain compression.  The effects are the same in all subbands 
(other than 2 and 3). 
    That the compression is present in subbands 9 through 16 (3 to 4 GHz 
tuning, using the B and D IFs) is strong evidence (but not proof) that 
the problem lies upstream of the T304 module. 
    The 150-fold increase in power in subband 3 translates to an 
increment of about 10 over the entire bandwidth at S-band -- about 10 
dB.  This is significant, but well within the specs for the signal 
chain.  Since the synchronous power measures are made before the 
requantizer, we can't blame the latter process for the effects seen. 

    We need to understand -- and correct -- the origin of this problem.