[evlatests] Some Curious Results from Yesterday's WIDAR testing

[Rick Perley]

    I have completed the calibration and imaging tests from yesterday's 
WIDAR data.  For review:

    Source = 0217+738 -- a 4.3 Jy perfect point source at both X and C 
bands.
    Bands:  Both X and C
    Scheme:  Alternating between X and C, with 5 minutes on source for 
each, for 2 hours. 
    N(sub)   Four sub-bands, each 128 MHz wide, with 1024 channels, in 
RCP. 
    TimeAvg:  1 second

    Essential Results:

    X-Band. 

    1) As noted before, no 180 degree phase jumps -- no phase jumps of 
any kind.  Amplitude jumps noted, mostly of the 1 db variety already 
noted. 
    2) Bandpass calibration shows extraordinarily stable bandpass 
shapes.   Solutions were made for each 5 minute integration. 
    3) A 'continuum-equivalent' database, averaging of 961 channels, 
and  utilizing the BP solution was made.  Standard self-cal was run on 
this 'continuum' data. 
    3) A four-sub-band image (after self-cal) provided a 165,000:1 image 
-- without baseline calibration.   The image noise is exactly as 
predicted from the single-channel histrogram width (0.65 Jy) reduced by 
the sqrt rt. of the number of channels (961), IFs (4), and number of 
visibilities (147600). 
    4) A baseline calibration was performed (averaged over the full 2 
hour extent).  Solutions were all smaller than 1 part in 10^4, usually 
10^5. 
    5) An image with the baseline calibration applied was made -- a 
marginal improvement was noted. 

    I could find no baselines with oscillatory phases of the few seconds 
period variety.  Baseline 19 x 24 (see below) *might* show the effect, 
but it's not obvious. 
   

    C-Band.

    1) Again, no phase jumps, but many amplitude jumps, of the same 1 dB 
variety.
    2) Unlike X-band, the bandpass solutions are *NOT* stable -- many 
antennas showing sinusoidal differentials (in both amplitude and phase) 
compared to the mean.  These variations are steady over the length of a 
scan. 
    3) The same 'continuum equivalent' database was made as at X-band.  
Self-cal was run.
    4) The image looks quite poor, with 'lumps and bumps' seen. 
    5) BLCAL was run -- the solutions are much larger than the X-band 
ones, with many baselines having closure corrections at or higher than 
0.1%. 
    6) An image made with the BL solutions applied looks as good as the 
X-band image -- dynamic range of 165,000:1, with a 1 mJy background 
source clearly visible.  (Peak on map is 4300 mJy). 

    Perusal of the visibilities showed only a single baseline which 
shows the infamous phase oscillations:  19 x 24.  The period is about 
3.0 seconds throughout the hour.  No 'stationary' point was noted.  
There is no obvious amplitude oscillation on this baseline.  The 
amplitude of the phase oscillation is not large -- about 1 degree, 
pk-pk.  It clearly averages out in the image.
   
    So -- we have a new mystery:  Why do we have undulatory variable 
bandpasses and notable closure errors at C-band, when X-band data, taken 
at the same time on the same antennas in the same mode, is clean?