[evlatests] Adventures with Switched Power (continued ...)

[Rick Perley]

    A 4-hour observation was taken last night at 6cm, with the purpose 
of further exploring the characteristics of our switched power system, 
and its application to visibility data. 

    The test was in two parts:

    A)  Two hours of alternating between two sources at low elevation:  
one rising, and one setting.  The flux density of each was about 1 Jy.  
Each observation was 10 minutes long.  Slewing took up three minutes for 
each. 
   
    B) Two hours of tracking 3C147 (~6 Jy), at high elevation. 

    OSRO1 mode was employed.  The frequencies employed were 4.5 and 6.5 
GHz. 

    Results are grouped by subject. 

    1) On-Line flagging. 

    A remarkable discovery was made:  If one observes a source below the 
elevation limit (8 degrees), the on-line flagging does not flag the 
data!  This is apparently because the system, recognizing the source to 
be too low, simply commands the antennas to go to 8 degrees elevation -- 
which they do without error.  Having reached the commanded position, no 
flags are set.  (Hence -- the data taken while the antennas are moving 
to that position are flagged -- but not the data when the antennas have 
reached this position). 
    Given the uncertainties in source positions which accompany the 
dynamic scheduling, I think it important that this condition be 
recognized, the data data flagged. 
    Otherwise -- the on-line flags seem to work quite well.  Some 
apparently good data were flagged, but this is certainly a tiny minority. 

    2)  Switched power. 

    With few exceptions, the 'Differenced Power' (PDif) are very clean.  
The notable exceptions are:
    *  Antenna 12, IF 'B':  Pdif, Psum and Tsys show a square wave 
pattern, period a few tens of seconds.
    *  Antenna 25, IF 'B':  Save as above, but much stronger (10% effect 
in total power). 

    3) Elevation Effects. 

    With 'set and hold', which keeps system attenuation constant, and 
presuming (hoping?!) that the electronics are superbly stable (not a 
requirement, but certainly an advantage), the measured PDif should not 
change over the 4 hour run.  But it does ...
    For most antennas, the change in PDif between the long, high 
elevation observation of 3C147 and the lowest elevation (8 degrees) was 
quite small -- typically 5% or less, with the PDif almost always smaller 
at the low elevation.  There is a clear elevation-dependent effect 
here.  This would not be a problem if the visibility amplitudes changed 
by a similar amount (barring real atmospheric absorption), and the 
application of the PDif corrects the visibility amplitudes.  See below 
for this result. 
    Some antennas show a very much larger effect, which is always 
different for the two polarization, but the same for the two 
frequencies.  These are:
    * Antenna 1:  PDif declines by 20% in RCP, and by 7% in LCP. 
    * Antenna 10:  PDif declines by 9% in RCP, but not at all in LCP.
    * Antenna 21:  Pdif declines by 5% in RCP, and 12% in LCP.
    * Antenna 24:  PDif declines by 17% in RCP, and only 3% in LCP.
    * Antenna 28:  PDif declines by 11% in RCP, but increases by 4% in 
RCP. 
    Note that we do not know at this time whether this change is due to 
an elevation dependence of the system gain, or is due to an elevation 
dependency of the switched power source. 

    4) Unfortunately, my attempt to apply the switched power to the data 
has not been successful.  The program 'TYAPL' somehow makes all the data 
from 3C147 disappear -- even though the visibility count remains right.  
(That is, the data are present, but cannot be 'seen'...).  The 
low-elevation sources mostly come through, and a *preliminary* estimate 
is that all but the lowest observation is properly corrected by the 
application of PDif.  (And I suppose it possible that the 8 degree 
observation is low due to real atmospheric absorption -- and effect that 
PDif cannot detect). 

    5) A few antennas show periodic patterns in the PDif and PSum values:

    * Antenna 10 shows a triangle-wave variation, of amplitude 4% and 
period 1 hour.
    * Antenna 14 shows a small (couple percent) 'kink' in RCP, and a 
much larger (7%) 'bump' in LCP. 
    * Antenna 18 shows a larger version of the triangle pattern:  1.25 
hour period, ~8% amplitude (pk-pk) on three of the four IFs, and a 
spectacular 25% amplitude in RCP of IF#2. 
    A small number of antenna show apparently sinusoidal variations 
(typically 2%), which are very likely smaller versions of the 'triangle' 
patterns reported above. 
    One should be able to discriminate between a real gain change, and a 
change in the added switched power by examining the ratio (i.e., Tsys).  
I have not yet done this -- will do so shortly.