[evlatests] Scan-based sensitivity changes, and other odd effects

[Rick Perley]

    Review of the antenna-based amplitudes -- from correlation 
coefficients -- has shown some very curious (albeit small) effects which 
are clearly EVLA based. 

1)  Scan-based changes in system temperature. 

    For some antenna-IFs, we are seeing scan-based changes in 
sensitivity, which are correlated with changes in system temperature.  A 
good example is that of antenna 18 at X-band.  Through the 2-hour run, 
the correlation coefficients are seen to lie in two different states, 
separated by about 3% (as seen by an unchanging antenna), or about 6% in 
terms of system temperature.  The two states always last for the full 
scan length -- there are no changes within any 5-minute scan.  The two 
IFs I had in the run behaved quite differently.  Investigation of the 
Tsys showed exactly what is needed to explain the loss in coherence -- 
two different system temperatures, differing by 2 to 3K -- exactly the 
magnitude needed! 

    The only other antenna which shows this clearly at X-band is antenna 
26A. 
    At L-band, however, the effect is more commonly observed:  16 
(again, the two IFs have different ups and downs), 19 (especially in A), 
21C (a single change of 3%), 23C (a single change of 5%), 24C (multiple 
changes of 2%) 26A (similar to 24C).  In all cases, a corresponding 
change of the appropriate sign was seen in Tsys.  However, in these 
cases, the magnitude of the L-band system temperature change is not 
sufficient to correct the amplitudes completely.  For both 21C and 24C, 
the post-Tsys correction shows a residual effect of about 1%, in the 
sense that the reported rise in Tsys is not large enough to explain the 
entire change in sensitivity. 

    So -- what is causing scan-based small (few K) changes in Tsys?  Ken 
has wondered if the T304 attenuator setup might do this.  If it did, it 
would not be seen in the B and D IFs, where we used fixed value of 
attenuation.  Unfortunately, I cannot judge this, as only IFs A and C 
were utilized.  On the other hand, I wouldn't have expected a change in 
system temperature from minor change in attenuator settings so deep in 
the signal path -- the preceding RF and IF gain should be enough to 
reduce the additional noise to negligible levels -- shouldn't it? 

2) 'Weird Effects' at L-band on antennas 11 and 17. 

    The correlation coefficients seen on some EVLA antennas are behaving 
in a very strange manner which is tightly connected to system 
temperature changes.  A good example is antenna 11.  The correlation 
coefficients for both IFs show exactly parallel, essentially random, 
changes of about 3%.  These changes (unlike those described above) occur 
both within scans, and between them.  (Note, however, that behavior at 
X-band is close to normal).  Examination of the Tsys values shows that 
there are exactly parallel changes in system temperature -- but in IF 
'C' only!  Hence, the Tsys-corrected values of visibility look quite 
good in IF 'C', but retain the characteristics described above in IF 
'A'.  Clearly, our measurements of Tsys in IF 'A' at L-band are not valid. 

    Similarly, antenna 17A shows remarkable changes in correlation, over 
a timescale of about one hour (5%), much more strongly seen in IF 'A' 
than in 'C'.  The measured Tsys values mirror these changes quite well 
(and I note the values are 50 to 65K -- two times too high), but the 
post-Tsys visibility stability, although much better than that of the 
correlation coefficients, still shows residual effects.