[evlatests] Noise at sub-band edges

[Rick Perley]

    A question was raised today about the noise increase near sub-band 
boundaries. 

    The excellent 6-cm data taken Thursday evening offers the following. 

    The data are calibrated in the standard way:  FRING, BPASS, CALIB, 
CLCAL.    Only a single delay for the entire 90 minutes has been 
utilized.  The BPASS solutions are individualized (one for each 90 
second scan).  The CALIB phases are solved and interpolated individually 
for each 1-second integration (with a CL table of the same interval).  
The CALIB amplitudes are computed on a 1-minute average. 

    I calculated the noise dispersions using UVHGM, for the weakest of 
the four sources in the file, on the imaginary component only.   No 
smoothing has been applied. 

    All four sub-bands, and both polarizations behave identically.  The 
data below are from subband 4, RCP only.   There are 256 channels in 
each subband, and the channel width is 500 kHz. 

    Channel         Dispersion          Normalized
--------------------------------------------------------------
    2                        .605 Jy                1.78
    5                        .472                     1.39
    10                      .379                     1.12
    15                      .353                     1.04
    20                      .345                     1.02
 (all channels through to #230 are the same, at 0.340 Jy )
    230                   .343                      1.01
    235                   .349                      1.03
    240                   .355                      1.04
    245                   .369                      1.09
    248                   .395                      1.16
    250                   .421                      1.24
    253                   .497                      1.46
    255                   .590                      1.74
-------------------------------------------------------------

    I don't have the expected noise increase function handy, but suspect 
the values above will be pretty close to what we expect ...