[daip] Re: Multi-Resolution Clean in AIPS

[Eric Greisen]

Greg Ball writes:

 > I am trying out the Multi-Resolution Clean in AIPS (IMAGR) to try and 
 > recover extended emission, and I would really like some guidance on how it 
 > should be used.  I understand that it is experimental so maybe there are 
 > no definitive answers, but I would appreciate any suggestions you can 
 > make, or references you can give (I have looked up Wakker and Schwarz.)

         Their algorithm is rather different and should not be much of
a guide.

 > 
 > Although there are some suggestions in the AIPS help of values to choose 
 > for the MR Clean related adverbs, I cannot find discussion of how to 
 > choose the WGAUSS values, except that including 0 for a point source is
 > recommended.  Surely WGAUSS values larger than the structures in 
 > my image are useless, but are they harmless, or will they tend to 
 > overclean?  How might one choose a value for NGAUSS (if not the 
 > obvious value, 2) and how to space the WGAUSS values?

I just make a guess to start with.  You must always have 0 - the sky
always has unresolved objects.  Often I make a quick image with just
this to see what the beam size is.  Then I make 2-4 more at sqrt(10)
steps although it is a bit of a guess since they will be fattened some
by the data sampling.  Then I start the Clean.  The first imaging
tells you a lot:  If there are nothing but point sources, then the
peak flux at each resolution should be the same.  If there is a lot of
extended emission, then the peak flux will go up as the beam size
increases (until you get larger than one source component).  One
should then include only resolutions where the peak goes up from the
next lower WGAUSS.  The most important parameter is IMAGRPRM(11) which
should be chosen to more or less flatten the playing field - i.e. add
prefernce to the lower WGAUSS just enough so that one Cleans all
resolutions instead of only the fattest one.  I usually have set the
rest of the IMAGRPRMs to zero.  The FGAUSS values are used to cut off
Cleaning at some point point.  Often I do it more than once to see
when I would like to have quit with the highest WGAUSS being stopped
at higher flux (Jy/beam) and the 0 width being allo9wed to go quite
deep in the end.  The FGAUS may go down as you iterate with Self-Cal.

An example is in our HUGE Y2K test:

It does at an early stage
         OVERLAP=2; NGAUSS=4; WGAUS=0,0.7,2.1,6.3; DO3D=1
         FGAUS=0, 0.013, 0.070, 0.250; UVBOX=0; UVTAP=0
         IMAGRP=0; IMAGRP(11)~0.52,0,0.1,0.3,0.1,80; FACTOR=0
and at a later stage
         NGAUSS=4; WGAUS=0,0.7,2.1,6.3; DO3D=1; UVBOX=0
         FGAUS=0, 0.005, 0.025, 0.120; BMAJ=0; UVTAP=0
         IMAGRP=0; IMAGRP(11)~0.52,0,0.1,0.3,0.1,80

>From my line experiment HI file
   203  IMAGR WGAUSS(  2) = 3.60000E+01
   204  IMAGR WGAUSS(  3) = 1.08000E+02
   205  IMAGR WGAUSS(  4) = 3.24000E+02
   206  IMAGR FGAUSS(  1) = 2.00000E-04
   207  IMAGR FGAUSS(  2) = 5.50000E-04
   208  IMAGR FGAUSS(  3) = 2.60000E-03
   209  IMAGR FGAUSS(  4) = 7.80000E-03

   231  IMAGR IMAGRPRM(11) = 0.62000 / MRClean control
   232  IMAGR IMAGRPRM(12) = 0.00000 / MRClean control
   233  IMAGR IMAGRPRM(13) = 0.00000 / MRClean control
   234  IMAGR IMAGRPRM(14) = 0.00000 / MRClean control
   235  IMAGR IMAGRPRM(15) = 0.00000 / MRClean control
   236  IMAGR IMAGRPRM(16) =      0. / MRClean control

The most important thing is to be experimental with it - watch it on
the TV with DOTV true throughout.

Eric Greisen