[daip] [!7226]: AIPS - strange scaling of visibility amps & weights w/ DOBAND=1

[Michael Bietenholz]

Michael Bietenholz updated #7226
--------------------------------

strange scaling of visibility amps & weights w/ DOBAND=1
--------------------------------------------------------

           Ticket ID: 7226
                 URL: https://help.nrao.edu/staff/index.php?/Tickets/Ticket/View/7226
           Full Name: Michael Bietenholz
               Email: mbieten at yorku.ca
             Creator: User
          Department: AIPS Data Reduction
       Staff (Owner): -- Unassigned --
                Type: Issue
              Status: Open
            Priority: Default
                 SLA: NRAO E2E
      Template Group: Default
             Created: 08 September 2015 05:34 PM
             Updated: 08 September 2015 05:34 PM
                 Due: 10 September 2015 05:34 PM (2d 0h 0m)
      Resolution Due: 16 September 2015 05:34 PM (8d 0h 0m)



I think something is going funny with the scaling of visibility weights with DOBAND=1. The uploaded small test visibility file has 32 channels.  There is one BP table, which has amplitudes near 1 (at least as plotted out by POSSM) towards the center of the band (BCH=16), and two CL tables, the second of which has the full amplitude calibration with gain factors not near 1.0.  There is only one time-stamp.

I couldn't get UVPRT to print the weights scaled so that I could see what was going on (even when I only printed one vis).  However, LISTR, OPTY=MATX obliged.

I would expect that, if calibration scales the visibility amplitudes by A, then the weights should scale by 1/A**2.  With DOCALI=1 the visibilities and weights do seem to scale this way, but when I set DOBAND=1 there was almost no change in visibility amplitude (expected near the centre of the band), but the weights changed by a factor of 10.

Here's what I got from LISTR,OPTY=MATX, for Stokes RR, antennas 3,10 (EF, NL) BIF=5; EIF=5; BCH=16; ECH=16; DOBAND, DOCALI either +1 or -1 (one channel picked out more or less at random). There is only one visibility, so no averaging should be going on.


no CL, no BP:    amp(DPARM=0): 5.431E-04;     weight (DPARM=14,0): 9.999E-01
 [uncalibrated weights are near unity, as expected]

CL2, no BP:      amp(DPARM=0): 7.924E-02;      weight (DPARM=14,0): 4.780E-05
 [so far so good, amp increases by ~145x, so weight decreases by ~145**2]

CL2, BP1          amp(DPARM=0): 7.577E-02;      weight (DPARM=14,0): 4.612E-06
 [so applying BP1 causes visibility amps to change only a little bit, as is expected since the BP table amps are near unity for this channel; however, it causes the weights to change by ~10x]

I also tried no CL, BP1
CL2, no BP:      amp(DPARM=0): 5.193E-04;      weight (DPARM=14,0): 9.999E-01
 [so here the visibility amplitudes change by ~4%, but weights do *not* change by the expected ~8%]

I looked also at baseline 2-10 (BR-NL), and found again that upon applying BP1, the weights scale differently than 1/amp**2, although in that case the scaling was quite close (not a factor of 10 different like for baseline 3-10).

Am I confused or is something funny going on?

(31DEC15, MNJ from ~Aug 30)

------------------------------------------------------
Staff CP:  https://help.nrao.edu/staff
-------------- next part --------------
A non-text attachment was scrubbed...
Name: test_gb076.fits.gz
Type: application/gzip
Size: 3181850 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/daip/attachments/20150908/957a5d5a/attachment-0001.bin>