[daip] CMETHOD in IMAGR

[Eric Greisen]

Lynn D. Matthews writes:

 > Meanwhile, my colleague Ciriaco Goddi has been imaging large fields (8192)
 > using the old (2005) version of AIPS. In the course of investigating why
 > certain channels of the data cube had such hideous sidelobes, we
 > discovered that forcing IMAGR to use either CMETHOD='DFT' or 'GRID' both
 > produced good images (DFT somewhat better than GRID). However, letting
 > IMAGR to its own devices (i.e., CMETHOD='') resulted in ugly sidelobes and
 > much higher noise. A sample comparison of the results of IMSTAT for a box
 > near a bright source are given below.

      I find this hard to believe and suggest that something else has
also changed.  If the Cleans are identical then ' ' means that some
subtractions are DFT and some grid without interpolation.  'GRID'
means that they are all grid without interpolation.  Since this is a
bad thing with strong objects near the edge of the field, the ' ' can
only be better or the same depending on whether the strong sources at
the edge fell mostly in the DFT bits or the GRID bits.  I suppose that
getting it right initially with DFT and then making a gridded error
will change the Clean path and so might actually be worse but I would
be suprised if that were the rule rather than the exception.

As you say below, there is nothing much I can do about it in old
versions.

Eric Greisen

 > 
 > Since both NEW and TST aips now force a DFT for large images, this may now
 > be a non-issue. However, I don't understand things well enough to know
 > whether for smaller images with sources near the map edge, the "hybrid"
 > approach that IMAGR defaults to in NEW and TST might still in some cases
 > produce inferior cleaning. I'm also wondering if it's worth re-imaging
 > some of the large cubes I produced earlier last year, before the "Forced
 > DFT" was implemented.

     If there were significant sources outside the middle of such
large fields, you certainly should consider this.  I would be
disturbed about any of the IMSTATs below and would investigate what
the image looks like in detail.

Eric


 > Lynn
 > 
 > CMETHOD=''
 > 
 > AIPS 1: Mean=-6.9370E-04 rms= 4.7006E-02 JY/BEAM   over  403840. pixels
 > AIPS 1: Maximum= 2.5575E-01 at  4049  1920    1    1    1    1    1
 > AIPS 1: Skypos: RA 05 35 14.50782210  DEC -05 22 30.659257
 > AIPS 1: Skypos: IPOL  43115.939 MHZ
 > AIPS 1: Minimum=-2.0024E-01 at 3890 2425    1    1    1    1    1
 > AIPS 1: Skypos: RA 05 35 14.50835444  DEC -05 22 30.634007
 > AIPS 1: Skypos: IPOL  43115.939 MHZ
 > AIPS 1: Flux density = -5.9147E+00 Jy.   Beam area =  47.36 pixels
 > 
 > CMETHOD='GRID'
 > 
 > AIPS 1: Mean= 3.8284E-04 rms= 2.1941E-02 JY/BEAM   over  403840. pixels
 > AIPS 1: Maximum= 9.7174E-02 at  4023  1868    1    1    1    1    1
 > AIPS 1: Skypos: RA 05 35 14.50790915  DEC -05 22 30.661857
 > AIPS 1: Skypos: IPOL  43115.939 MHZ
 > AIPS 1: Minimum=-1.2267E-01 at 4016 1941    1    1    1    1    1
 > AIPS 1: Skypos: RA 05 35 14.50793258  DEC -05 22 30.658207
 > AIPS 1: Skypos: IPOL  43115.939 MHZ
 > AIPS 1: Flux density =  3.2643E+00 Jy.   Beam area =  47.36 pixels
 > 
 > CMETHOD='DFT'
 > 
 > AIPS 1: Mean= 1.0113E-03 rms= 2.4188E-02 JY/BEAM   over  403840. pixels
 > AIPS 1: Maximum= 1.0001E-01 at  4123  2191    1    1    1    1    1
 > AIPS 1: Skypos: RA 05 35 14.50757434  DEC -05 22 30.645707
 > AIPS 1: Skypos: IPOL  43115.939 MHZ
 > AIPS 1: Minimum=-1.0683E-01 at 4016 1941    1    1    1    1    1
 > AIPS 1: Skypos: RA 05 35 14.50793258  DEC -05 22 30.658207
 > AIPS 1: Skypos: IPOL  43115.939 MHZ
 > AIPS 1: Flux density =  8.6223E+00 Jy.   Beam area =  47.36 pixels
 > 
 > _______________________________________________
 > Daip mailing list
 > Daip at listmgr.cv.nrao.edu
 > http://listmgr.cv.nrao.edu/mailman/listinfo/daip