[evlatests] WIDAR Tests of 27 August

[Rick Perley]

    With the return of WIDAR, Michael took some data on our favorite 
point source, 0217+738 on Aug 27.
The source was observed at very low elevation (20 degrees!) -- not a 
good idea IMHO.  See later notes ...

    Observations were taken at X and C bands.  I report here on the X 
band only, as the Cband data had some sort of translation problems. 

    All 12 antennas fringed in all four correlations.  In general, data 
quality is very good.  No 'wobbles' are seen.  But lots of other, likely 
more minor, issues were found.  Read on ...

    0) The archive claims the observation is an hour long.  But only 12 
minutes of data were actually present.  Michael states that some board 
malfunctioned.  Why should this completely terminate the observation (so 
far as the archive is concerned)? 
    1) Drop-outs are present.  A small number -- << 0.1% of the 
visibility amplitudes and phases are identically zero.  All four 
correlations are affected, independently. 
    2) The 10-second interval 'drop-downs' are completely absent!  (Yea!)
    3) Antenna 3, in LCP on sub-bands 3 and 4, has a sharply tilted 
bandpass shape:  the bandpass slopes sharply downwards with frequency in 
subband 4, and oppositely in subband 3.  Subbands 1 and 2 look fine.
    4) Antenna 8 has a llarge-amplitude sinusoidal bandpass curve in LCP 
on all four sub-bands.  There is exactly one cycle for this -- with 
inverted phase for half the cycle.  It looks like a (very) bad lag to 
may eyes (or all lags are bad but one). 
    5) Essentially all antennas show small amplitude and phase 
oscillations in both polarizations.  The frequency scale is always the 
same, corresponding to about 7 meters free-space.  Sure sounds like a 
reflection between subreflector and feed/receiver to me.  Some 
antenna-IFs are worse than others -- 9 in LCP is particularly notable.  
    6) The amplitude gains show curious gain fluctuations which are 
different for each antenna, but essentially identical in all subbands 
and polarizations for any given antenna.  The fluctuation timescale is 
fairly short -- 10 to 20 seconds.  The fluctuations are not sinusoidal.  
I don't think these are 'wobbles'.   There is no effect on the phase.   
Some antennas show very little of this effect:  1 (W16) and 24 (W14) 
show very little effect.  Antennas which might be expected to be 
shadowed are *much worse*.  The amplitude variations for some exceed 10% 
!!!!!  I suspect shadowing and/or cross-coupling is involved here.  
Alternately, we have a serious pointing issue (but I think this much 
less likely). 
       *** I strongly urge that we NOT observe low elevation sources in 
these tight configurations.  We have enough things to dig out of these 
data without worrying about shadowing effects ***
    7) Antenna 23 had a large amplitude loss upon return to the 
calibrator.  (There were only two such observations, due to problem (0), 
above).  The change in amplitude gain was about 20%. 
    8) In general, the LCP amplitudes are a little lower than RCP from 
the same antenna/subband -- but not always. 
    9) Delays on the RCP side scatter typically about 10 -- 20 nsec 
about the average.  The spread is *much* larger on the LCP side -- many 
tens of nsec, with some antennas (1 and 2 especially) much further away 
from the mean -- over 100 nsec for these two. 
   
    I calibrated the data, using a 2-point A&Phase average, integrating 
over the ~central half of the bandpass, and applying the mean bandpass 
solutions.  (The bad LCP subbands from antennas 3 and 8 were flagged out). 

    Imaging tests showed the following.  I had only 7 minutes of good 
on-source data -- one cannot expect much with so limited an integration. 

    1) Imaging with a single, central channel, with 1 to 4 IFs, in 
either polarization, produced maps which appear noise-limited.  The DR 
is a modest 1200. 
    2) Noise reduced as expected for channel integrations up to about 16 
or perhaps 32. 
    3) Images made with more than 32 channels do not follow the sqrt(BW) 
noise expected, and indeed get worse when more ~200 channels are 
included.  This can (likely) only be caused by a variable bandpass.  I 
did not further investigate this.  The max dynamic range I got was about 
11,000.  This is very good for a short observation -- I honestly don't 
know what to expect here...
    4) Application of a closure correction had no discernible effect.  
The reported closure levels are very low -- a few times 0.01% in ampl 
and phase -- probably the noise limit, as only 7 minutes of data were 
available to calculate these. 
    5) I turned on AIPS' closure reporting statistics -- these are 
consistent with the numbers given above.  No antennas or subbands 'stood 
out' in the statistics.  All seems good here. 
    6) The residual gains, after the 2-point calibration reported above, 
showed that the LCP shows considerably greater scatter than RCP -- by 
about a factor of 3.  Yet the images in LCP are only slightly poorer 
than RCP -- most of which can likely be explained by the absence of 
antennas 3 and 8.  I have no easy explanation for this ...

    That's about all I can squeeze out of this, in the 2 hours available 
on a Friday afternoon ...

    Rick