[evlatests] Power Levels at P-band

[Rick Perley]

    Ken and Huib have recently exchanged messages about the power levels 
seen in our P-band data. 

    My long run -- 12 hours observing on calibrator sources -- taken 
early this month, provides (I think) useful additional information and 
clues as to what is wrong.  

    This run, for scheduling reasons, comprised six separate observing 
scripts.  The diagnostic advantage of this was that we can see how the 
system gains ('set and remember') change for each of the six 
observations, as well as look at what the levels were. 

    The setup for this experiment used 16 subbands, each 16 MHz wide.  

    The result is there is both a large range in power levels, and a 
large range in power level differences amongst the six setups.  Some 
antennas were at the right level for all setups, some were at the wrong 
level for all, and some changed their powers by large fractions (up to 
an order of magnitude!).  The behavior is so diverse that a table is 
needed.  Below I show the power level ranges (in 'counts') for each 
antenna and polarization.    Note that the desired level is 14 counts, 
and anything more than a factor of two higher or lower is considered 
very undesirable.    These numbers come from the the 'PSum' values, 
determined by the correlator station board, for subband number 8 (right 
in the middle of the passband). 

Antenna   'R' = Horizontal    'L' = Vertical
----------------------------------------------------
    01          32 -- 38                 18 -- 30
    03            3 -- 12                   2 -- 18
    04            8 -- 15                  38 -- 55
    05            3 -- 10                  10 -- 16
    06           10 -- 16                  8 -- 12
    07               18                     15 -- 19
    08           20 -- 28                  8 -- 13
    09             8 -- 20                 26 -- 32
    10             5 -- 11                  2 -- 20
    11           12 -- 15                22 -- 26
    12           14 -- 18                 4 -- 10
    13            9 -- 17                  12 -- 14
    14            9 -- 20                      22
    15           8 -- 24                      28
    16           5 -- 16                    8 -- 22
    17              35                           4
    18              20                        24 -- 30     
    20              18                        8 -- 18
    22           10 -- 16                 15 -- 28
    23            4 -- 24                   5 -- 12
    24              14                       12 -- 20
    25             6 -- 8                    16 -- 20
    26             5 -- 20                  7 -- 16
    27            12 -- 16                13 -- 20
    28             18                           14
---------------------------------------------------------------

    So what is happening here?  There are two problems -- incorrect 
levels, and widely changing levels.  They may, or may not, be related.

    Ken and I discussed some explanations:

    1) High levels.  These are the easiest to explain.  The T304 power 
meters are reacting to a much wider bandpass than the correlator 
subbands (which is where the PSum values are determined).  The T304 
input power meter sees everything the front end delivers, while the 
output power meter sees a 1 GHz-wide piece of this.  If, in the 1 GHz 
band visible to the output power meter, the actual sky signal occupies 
only (say) 50%, the power seen in a subband at the station board will be 
a factor of two higher than nominal. 

    2) Low levels.   This simplest explanation is that there is 
something strong in the overall passband, but outside the spectral range 
visible to the correlator, so that the T304 power detectors are driving 
down the gain, so the power visible to the subband power detectors is 
now too low. 

    3) Variable gains.  The obvious explanation is RFI (and this must 
play a role at some level).  But it's hard to explain the variation 
amongst the antennas.  Many are completely stable, while nearby antennas 
show the greatest variability.  RFI -- especially external RFI -- cannot 
reasonably 'single out' an antenna, while ignoring the antenna 100 
meters away.   Note, for example, that the most variable antennas -- 
ea03, ea10, and ea23, were located at E10, E04, and W18, respectively, 
while the three most stable antennas -- ea07, ea17, and ea28, were at 
E12, E14, and E16.  (Two other quite stable antennas, ea18 and ea24 are 
at E18 and W04 -- the stability grouping at the end of the east arm may 
be significant ...). 

    One wonders if the high, or variable, power levels has any relation 
to our 'PDif Compression' problem.  This data set is ideal to answer 
this, as power levels vary by well over an order of magnitude between 
antennas, and by nearly that much for some individual antennas.  The 
unhappy (but not surprising, to me) answer is:  There is *no* 
relationship at all between the presence or absence of this compression 
and the power levels seen by the station board. 
    The most glaring example of this is for ea17, which has -- by far!-- 
the higher (claimed) gain compression when observing Cyg A or Cas A:  
the compression is about 35% for both polarizations.  But the power 
level on the 'R' side was 35 counts (way too high), while it was 4 
counts (way too low) on the 'L' side.