[evlatests] Content and editing of SY files
Rick Perley
rperley at nrao.edu
Mon Dec 20 18:37:20 EST 2010
My recent perusal of the SY tables provided via AIPS for the Herc A
demo project has revealed some problems, and some potential solutions.
Eric has now provided the capability of having the FG (flag) files
applied to the SY tables so that when running the editing program SNEDT,
one does not have to view the switched power values corresponding to the
antennas/times for which the data have already been flagged. This is a
great help, but unfortunately it is not sufficient.
When running SNEDT, even after applying the FG tables as described
above, the switched power displayed is remarkably bad, with absurd
values (often the mega range) in PSum and PDif being shown. It is a
huge effort to have to flag these tables in order to 'clean up' the
data, prior to its smoothing and subsequent application. This begs the
question of 'why are those bad data there at all?'.
It turns out that these absurd values are all identified with times
for which there are no visibility data at all! Not just flagged
visibilies, but *no visibilities*. I ask again (as I have already asked
this once before) -- why should we have SY tables values corresponding
to times for which there are no data?
I tried a simple, but laborious experiment to demonstrate this: I
manually flagged the X-band data for those times lying between the
listed scans, and for time preceding the first and following the last
scans. The resulting SY table displayed by SNEDT (with the flagging
option turned on) were 'clean as a whistle'! Virtually no table editing
was required -- hugely reducing the amount of manual effort involved in
cleaning up the SY tables.
So -- if it proves to be too difficult to prevent recording SY table
values corresponding to times outside the times for which there are
visibility data, perhaps somebody can write a simple task which does
automatically what I did by hand -- runs UVFLG for all times between
scans. (For my test of this, there were 38 scans, so had to painfully
enter 39 * 8 = 312 integer values ... ugh!)
Something like this is going to be needed if our users are going to
be able to make use of these data!
More information about the evlatests
mailing list