[evlatests] L-Band Sensitivity dependence on just about everything

[Rick Perley]

    I ran a test yesterday evening designed to give us some insight into 
what factors influence the well-established dependence of EVLA 
sensitivity on frequency. 

    To recap:  We have previously found that the 1385 MHz sensitivity is 
considerably better (by up to a factor of 40%) than the sensitivity at 
1385 MHz.  This effect is seen in all four EVLA antennas which are 
outfitted at L-band, and it is not a function of the particular IF.   
Interchanging the AC and BD frequencies interchanges the sensivity also. 

    The experiment run yesterday was designed to test for any dependence 
on the values utilized by the two LOs involved:  the L301, and the 
L302.   There are (according to Walter's tloser program) seven legal 
setups, which will tune the EVLA to 1385 and 1485 MHz:

    For 1385.0 MHz, with 12.5 MHz BW, we have:

Frequency band requested = 1378.750000 to 1391.250000 MHz which is L band

All legal tunings (MHz):
L301 = 11648  L302 = 13149.250000
L301 = 11904  L302 = 13405.250000
L301 = 12160  L302 = 13661.250000
L301 = 12416  L302 = 13917.250000
L301 = 12672  L302 = 14173.250000
L301 = 12928  L302 = 14429.250000  <- default
L301 = 13184  L302 = 14685.250000

    While for 1485.0 MHz, the L302 is decreased by 100 MHz for the same 
L301 values. 

    For both frequencies, the default L301 is 12928. 

    I observed 3C286 at high elevation (above 80 degrees)  at AC   = 
1485, BD = 1385 MHz, with 12.5 MHz BW, for each of the 7 legal 
combinations.  I then reversed the frequency assignments, so AC = 1385, 
and BD = 1485.  This sequence to repeated, to check repeatability (and 
overcome any possible late starts) -- there were thus 28 individual 
observations. 
    The observations were made in spectral line mode, in order to avoid 
the well-known sensitivity losses in continuum due to mis-matched 
bandpasses. 
    I used the well-tested calibrated AIPS weights as an indicator of 
the antenna sensitivity.  It is important to remember that the AIPS 
weights is proportion to (efficiency/Tsys)^2.  Thus -- a high value is 
GOOD! 

    Results:

    An amazing wealth of information (mostly confusing!) was garnered.  
In short, we see sensitivity dependencies on frequency, antenna, IF, and 
LO values. 

    I can find five major points to summarize here:

    1)  The strongest effect of all is that already reported -- the 
sensitivity of all EVLA antennas is much better at 1385 than at 1485 
MHz.  As always, when the frequencies are interchanged between IFs, the 
sensitivities are interchanged also.  Except for the interesting 
IF-dependent effects noted below, the AIPS weights for IFs B and D at 
1485 MHz are the same as for A and C, when the frequencies are the same. 
    The AIPS weights ratio between the frequencies is typically nearly a 
factor of two (meaning, a sensitivity ratio of sqrt(2), or 40%).  The 
range of the 1385/1485 AIPS weight ratio varies from 1.4 to 2.0 (other 
than for antenna 16 in the LCP side -- a special case).  Thus, the 
sensitivity ratio varies from ~15% to 40%, between these frequencies. 
    However, it should be noted that the VLA *also* shows this same 
effect -- more sensitive at 1385 than at 1485!  However, the effect is 
not nearly as strong:  the ratio of the mean aips weights between these 
frequencies is 1.25 -- for a sensitivity ratio of about 1.12 -- or 12%. 

    2)  There is a wide spread of sensitivity amongst the EVLA antennas, 
particularly at 1485 MHz.   However, this is at least partially due to 
the antenna-IF sensitivity to the particularly LO frequencies chosen, as 
I will describe below.  (Indeed, at 1385 MHz, if one could specify which 
L301/L302 to use as a function of antenna-IF, the resulting 
sensitivities are the same for antennas 13, 14 and 16.  This is not so 
true at 1485 MHz, where considerable spread remains.  Antenna 18 is 
special.  Read on). 
    We must remember that the VLA antennas *also* have a wide spread in 
sensitivity -- a factor of two in AIPS weights at these frequencies.  
This spread is not because of one or two clearly misbehaving antennas  
-- there's a wide spread, period. 


    3)  Antenna 18 is easily the best EVLA antenna at L-band.  
Unfortunately, the RCP side was not operating last night, but for the 
LCP side, the AIPS weights for antenna 18 place is well above the mean 
for VLA antennas at both frequencies!  (Remember that these observations 
are at very high elevation -- for 'normal' elevations, antenna 18 will 
easily be the best in the array).  Now, antenna 18 does show the 
sensitivity dependence on frequency, but it is relatively small -- 1.4 
in aips weights, or 20% in sensitivity.  Not much greater than the VLA 
dependence. 

    4) Most intriguing is the dependence of sensitivity on the L301/L302 
selection.  In short, every antenna-IF has its own 'personality' -- and 
this personality is not dependent on whether the frequency was 1385 or 
1485 MHz. 
By personality, I mean each antenna has a different (and very, very 
different for some) dependence on the L301/L302 selection. 
    In general, sensitivity improves with increasing L301 (which also 
means decreasing L302, since specifying one, and the observing 
frequency, fixes the other).  There is little difference in sensitivity 
between the highest three or four L301 values, and a rapid decline for 
lower L301 values.   But some antenna-IFs have almost no sensitivity 
dependence on L301/L302, while for others it is catastrophic, and for 
some, it's inverted -- (lower L302 improves sensitivity).  Some highlights:
       a) For antenna 13, all four IFs, the 2nd lowest value of L301 
(11904) gives the best sensitivity.  (Note that for the lowest value, 
11648 MHz, antenna 13 gives no fringes).  However, the sensitivity 
dependence of these four IFs on L301/L302 is very flat -- about 5% 
across the whole tuning range. 
       b) antenna-IFs 16A, 16B, and 14B have no dependence at all on the 
L301/L302 values -- and give good fringes for all 7 legal tunings. 
       c) On the other hand, 16D and 14C have a spectacular variation of 
sensitivity with L301/L302:  Essentially no sensitivity at all for the 
lowest 4 L301 settings, with a rapid improvement for the higher three.  
The highest legal setting of 13184 MHz gives a 50% improvement in AIPS 
weights over the next highest (which is also the default) 0f 12928 MHz. 

    5) The system temperature varies significantly between these L301 
settings, and in a manner which reflects, nearly exactly, the 
sensitivity trends noted above.  ***However*** -- if Tsys were the only 
factor influencing the sensitivity, (and the variations reported by the 
on-line system are correct), then the AIPS weights should scale with 
1/Tsys^2.  But the sensitivity variations are far stronger than this.  
For example, ant-IF 14A, at 1485 MHz, has the Tsys increasing from 35 
for L301 = 13184 to 55 at L301 = 11648, but the aips weight decreased 
from 88 to 10 -- factor of 9 which is far higher than (55/35)^2.  The 
same IF at 1385 MHz saw the Tsys increase from 33 to 68 as the L301 ran 
from 13184 to 11648, while the aips weight decreased from 120 to 9.  
    It's probably important to emphasize that the Tsys variations 
reflect the sensitivity variations almost without exception -- but the 
dependency is far stronger than quadratic. 

    6) Finally (yes, I said 5 above, and this is the 6th point):  the 
variation of Tsys with frequency seems to give no hints of what is 
happening.  For antenna 13, on all IFs, for example, Tsys = 43 at 1485 
MHz, and 54 at 1385 MHz.   (That is, Tsys dropped when the frequency was 
changed from 1385 to 1485 Mhz -- the reverse of what we might expect, 
given the sensitivity ratio, which is particularly large for this 
antenna, a factor of 2 in the aips weights.)  For antennas 14 and 16, 
Tsys increases slightly (by < 10%) when the frequency is changed from 
1385 to 1485 Mhz.  For antenna 18, the Tsys was stuck at 2.8 degrees for 
all IFs, all frequencies, all of the time. 

    I hope something in the above list of remarkable phenomena triggers 
a 'gotcha' thought amongst the readers.