[evlatests] SY table values

[Rick Perley]

    Bill Cotton modified the 'BDF2AIPS' program to prevent SY table 
values being recorded when there are no valid visibility data.  This is 
very effective in removing nearly all (and perhaps all) of the 
outrageous PSum and PDif values which so cluttered the editing process 
associated with reviewing these tables, prior to their application. 

    Perusal of the SY table values in the original database (the one 
filled without Bill's update) shows the nature of the problem: 

    I looked at a sequence of observations that changed the band in the 
following sequence:  L -> X -> CLo -> CHi -> S -> L, noting changes in 
the values of PDif and PSum at each of the transitions points.  Note 
that the SY table is recorded at 1-second intervals, while the 
visibility data were averaged in the CBE to 5 seconds.  This difference 
may be important. 

    A) From L to X:  The SY table for the L-band data showed no change 
after the last valid visibility record.  Indeed, the SY table values 
continued on for 11 seconds past the actual end of the scan.  The 
following X-band file contains 8  records (1 second each) preceding the 
first valid visibility data, the first of which is  greatly discrepant 
form the following values.  The other 7 values show a smooth transition 
in PSum, presumably associated with subreflector rotation, and no change 
in PDif. 
    B)  From X to C-Lo.  There are two SY table records following the 
last X-band visibility record, and both are outrageous by any criterion 
-- with PSum and PDif 10 to 100 times too high.  There are about 12 SY 
table entries in the C-Lo band file which precede the first valid 
visibility record, the first two of which are greatly discrepant, and 
the rest showing some sort of smooth change, but with a 'jump' to a new 
value at the same time as the first valid visibility. 
    C) From C-Lo to C-Hi.  No change in sideband is expected here.  
There are about 4 records in the SY table of C-Lo beyond the end of the 
scan, but no abrupt change in either PDif or PSum.  The C-Hi database 
has two records preceding the first valid visibility -- these are not 
extreme, although clearly incorrect. 
    D)  From C-Hi to S.  There are two SY records in C-Hi after the end 
of the visibility data, the last of which easily qualifies as 
'outrageous', with a values of Psum about 1% of the correct value.  The 
S-band dataset (curiously!) shows no SY data in advance of the first 
visibility.  However, the first eight values are clearly very incorrect 
-- a factor of up to ten wrong.  These were not removed by Bill's 
updated filler, as the visibilites were non-zero-integer.  However, 
these visibilities are clearly wrong, and were flagged out by the 
on-line system. 
    E)  From S to L:  SY tables values in the S band database were 
present 2 seconds past the end of the visibilities, but only the last 
one of these was unusual, and only by a small value.  There were only 
two SY tables values in the L-band database in advance of visibility 
data, the first of which was very discrepant in PSum. 

    Summary:  The largest discrepancies accompany change of subband.  
But all frequency changes are accompanied by erroneous values in the SY 
tables.