[daip] [!6945]: AIPS - SAD: spurious sources, with PRTLVL debug results

[NRAO Helpdesk]

Drew Medlin updated #6945
-------------------------

       Staff (Owner): Eric Greisen (was: -- Unassigned --)

SAD: spurious sources, with PRTLVL debug results
------------------------------------------------

           Ticket ID: 6945
                 URL: https://help.nrao.edu/staff/index.php?/Tickets/Ticket/View/6945
           Full Name: Robert Gray
               Email: roberthansengray at gmail.com
             Creator: User
          Department: AIPS Data Reduction
       Staff (Owner): Eric Greisen
                Type: Issue
              Status: Response Overdue
            Priority: Default
                 SLA: NRAO E2E
      Template Group: Default
             Created: 23 July 2015 02:14 PM
             Updated: 23 July 2015 02:14 PM
                 Due: 17 November 2015 12:00 AM (4d 7h 4m)
      Resolution Due: 04 August 2015 02:23 PM 


Hi Eric,

I'm working on the problem of apparently spurious 'sources' generated by SAD which we discussed some months back, and attaching runs with the PRTLEV debug code you suggested.   

The problem only affects maybe one out of 10 cubes, and usually generates only one spurious feature - often near the edge, and often 100+ sigma - and I can check the spectrum and reject it as spurious.  But now I've encountered a cube with MANY spurious features scattered in frequency and position, and I can't inspect and reject them all manually so I've got to get this figured out.  

The attached Excel file shows two runs using PRTLEV, one on a cube where a spurious source is found, and a second run on that channel after extracting it from the cube using SUBIM where the spurious source seems to be found but rejected (reasons '    11   ' which I think are ‘maximum interior ... excursion’ although I’m not sure exactly what that means or how /why it would affect one run an not the other).    

I suspected that the problem might be in my SRCS proc that repeatedly invokes SAD (using a different RMS for each channel from IMSTAT and five decreasing thresholds from 3 sigma down to 1), so I stripped off all of that stuff to make a bare-bones SRCSX which simply invokes SAD repeatedly over all channels and it STILL has the spurious source problem (that code is below). 

Any suggestions you have for resolving the problem would be much appreciated!  

Best wishes,

Bob Gray

AIPS 1:  1 PROC SRCSX( SIGMAMIN, CH1, CH2, IMAGESIZ )
AIPS 1:  2 SCALAR CHAN, CH1, CH2, SIGMAMIN, IMAGESIZ, IMAGEBDR
AIPS 1:  3 SCALAR CPARM1, CPARM2, CPARM3, CPARM4, CPARM5
AIPS 1:  4 INEXT 'MF'; INVERS -1; EXTDEST;
AIPS 1:  5 OUTVERS -1;
AIPS 1:  6 IMAGEBDR = 10
AIPS 1:  7 IF IMAGESIZ = 512 THEN
AIPS 1:  8 NGAUSS=400
AIPS 1:  9 END
AIPS 1: 10 IF IMAGESIZ =1024 THEN
AIPS 1: 11 NGAUSS = 800
AIPS 1: 12 END
AIPS 1: 13 IF IMAGESIZ =2048 THEN
AIPS 1: 14 NGAUSS = 1600
AIPS 1: 15 END
AIPS 1: 16 FOR CHAN = CH1 TO CH2
AIPS 1: 17 BLC IMAGEBDR IMAGEBDR CHAN
AIPS 1: 18 TRC IMAGESIZ-IMAGEBDR IMAGESIZ-IMAGEBDR CHAN
AIPS 1: 19 TASK 'SAD'
AIPS 1: 20 GO SAD
AIPS 1: 21 WAIT SAD
AIPS 1: 22 END
AIPS 1: 23 INEXT 'MF'; INVERS -1; EXTDEST;
AIPS 1: 24 FINISH
>
>SRCSX(3, 1, 8192, 512)


------------------------------------------------------
Staff CP:  https://help.nrao.edu/staff