[daip] CLCAL: SIMP extrapolation

[Michael Rupen]

Dear DAIP,

   in a VLA data set with a scan on the target source before any scans on the
calibrator, CLCAL/SIMP derives a phase which slopes significantly with time
for the length of the first (pre-calibrator) target source scan.  2PT
correctly assigns a constant phase to that scan.

It looks to me like this might be a problem in SN2CL. 
- 2PT sets weights WWT1,WWT2 for the immediately preceding and succeeding
   entries in the SN table, with 0.0 weight given to one of these if you are
   extrapolating rather than interpolating.  This weight is then used
   in deriving the phase in the CL table, so extrapolations are (correctly)
   held to a constant phase.
- SIMP uses a similar weighting scheme for the amplitude, but not for the
   phase.  The phase is instead linearly interpolated between PHA and PHB,
   regardless of whether A and B actually exist.

See below for details with my data set.  I've checked that this occurs
for various source/calibrator pairs.

Thanks for having a look --

            Michael

================================================================================

--------
DETAILS:
--------

* AIPS 31DEC06 run on RHEL Dell laptop (mithril)

* Data set is
    20060517.X BAND.12
   which can be found (with lots of tables, sorry) as
     /home/scamper/MPR1/DAIP_DATA.FITS
   [note this includes PT, but I don't see that that should matter]

* Delete tables:
   inname '20060517' ; incla 'X band' ; inseq 12; indisk 1
   inext 'sn'; inver 3 ; extde
   inext 'cl' ; inver 5 ; extde

* default CALIB
   inname '20060517' ; incla 'X band' ; inseq 12; indisk 1
   calsour '20077+40298',''
     docal 1 ; gainuse 1
     uvra   0,   0
     wtuv 0.0
     ante 0
     flagver 1
     solint  0.1667 ; solmode 'p'
     soltype 'l1'
     cparm 0 ; cparm(5) 0
     aparm 0 ; aparm(6) 2 ; minamper 10 ; minphser 10 ; cparm(3) 1 ; cparm(4) 1
     refant 10 $ (E8)
     doflag -2.5
     snver 3
   go calib
   ==> SN/3 -- phases quite reasonable

* default CLCAL  $ SIMP
   inname '20060517' ; incla 'X band' ; inseq 12; indisk 1
   snver 1 ; gainver 1 ; gainuse 5
     calsour '20077+40298',''
     calcode ' '
     sour '20324+40575','CygX-3',''
     soucode ' '
     freqid 1
     refant 10 $ (E8)
     interpol 'simp'
     cutoff 10
   go clcal; wait clcal

   ==> CL/5
     Use LISTR or SNPLT to show that the pre-calibrator CL phases indeed
       show significant slope with time
     Use VPLOT to confirm that this introduces a slope into previously-flat
       phases in the data

* default CLCAL  $ 2PT
   inname '20060517' ; incla 'X band' ; inseq 12; indisk 1
   snver 1 ; gainver 1 ; gainuse 6
     calsour '20077+40298',''
     calcode ' '
     sour '20324+40575','CygX-3',''
     soucode ' '
     freqid 1
     refant 10 $ (E8)
     interpol '2PT'
     cutoff 10
   go clcal; wait clcal

   ==> CL/6
     Use LISTR or SNPLT to show that the pre-calibrator CL phases are now flat
       show significant slope with time
     Use VPLOT to confirm that this leaves the source data nicely
       flat with time

================================================================================