[alma-config] Initial CLEANed image, and parameters used

[John Conway]

Hi Steve,

 I'm running off now to attend the the IAU in Manchester now- so I have
only had chance to give a quick eyeball at the new CLEAN simulations. 
Sounds like this is a great piece of automated software(!!)
congratulations.

I havn't has chance to look at any of the  AIPS parameters - one
point though is that to be useful  you need to make a difference image
between  the CLEAN image and the true model convolved with the same
restoring  beam. The construction errors at most will only be at a level
of 100:1  or 1000:1 and so will only be visible in the error maps not in
the  direct CLEAN images (although these levels seem small the may be
scientifically important when looking at the morphology of weak 
emission or making line ratio maps). I think 
after we have looked at a set of simple error maps then we can later
decide to  compute more complex error estimates
like fractional error maps etc etc.

I agree with Min that a telecon would be good soon. Wednesday
23rd is OK for me (anytime on or after after Monday Aug 21st
is OK when I'll be back here). If possible it would be good 
to have Steve attend, hopefully this doesn't conflict with your
holidays, but I also think it would be good to discuss
progress as soon as possible.  At that telecon it would 
be good to discuss details about the models, whether we need
to revise the simulation strategy, and what error maps metrics
etc we want. I think therefore it might be good for instance 
to run through for the K arrays the A,C,E simulations for all models, 
with the  reconstruction error maps plotted, this will be good to let 
us know we are doing something useful. It would not harm to 
take a couple of the most complex images at C array and run it 
through for K and C style arrays and see what the level 
of difference (if any) in the reconstruction errrors is. This
sould not be a definitive test between the arrays styles but just 
to get a first order of what the differences if any are.

 Cheers
  John


P.S I suspect it might be hard to interpret the pure A-array difference
images because of missing short spacings. Perhaps both style of arrays
would benifit from shorter than 80m baselines and/or we must 
add a 15minute snapshot of C-array data like Min did in his simulation.
We can discuss all this at the telecon.

One reason these simulations are important I think they will probably
remind us again  about the 
importabce of shorts spacing for all arrays, something that we 
might lose sight of if we imply worry about dirty beam or simple uv
coverage metrics, its always good to have simulations of realistic 
sources to show were not off track - which is what this 
exercise is all about.  


On Fri, 11 Aug 2000, Steven wrote:

> 
> An initial run of CLEAN has been performed on the KCM51_4+25B7 image, and
> the results are shown on the pages indexed below in contour and greyscale
> form.
> 
> http://stevenheddle.co.uk/ALMA/SIL/CLtrial/node1.html
> 
> 
> The dirty image as already presented is shown on the left and the CLEAN
> image shown on the right. The parameters used in IMAGR are listed
> immediately below, with main points including gain 0.1, number of iterations
> 20000, and CLEAN box equal to the image size as displayed. which in effect
> means that the A array has a 1024x1024 CLEAN box, and for the C and E arrays
> the box size scales as the inverse of the pixel scale.
> 
> With the benefit of hindsight, it would have been better to have presented
> initial results for the A arrays, and I will work on CLEANing an image for
> each member of the SIL for imaging using an A array as the next step and
> presenting these results prior to running the Big Batch Job. However I would
> be grateful if you could still examine the AIPS parameters used (listed on
> the pages indexed above) to see if they are both appropriate and agreeable.
> 
> As regards the simulations to date, the 12000 odd lines of the batch file
> used to perform this is attached as a zip file, if anyone wishes to see
> exactly what has been done. The unknown factor referenced there is the
> nuvco3mm parameter file which I use to initialise everything to a known
> state. This file contains the default parameters of a new, clean AIPS
> installation, modified as follows:
> 
> INFILE '   M51'
> OUTNAME ' '
> OUTCLASS ' '
> OUTSEQ 0
> OUTDISK 0
> DOTV -1
> NMAPS 0
> SMODEL 0
> DPARM(5)=1
> DPARM(6)=4
> FUNCTYPE 'LN'
> INNAME ' '
> INCLASS 'NUVCO'
> QUAL -1
> SELBAND 1000
> SELFREQ 2.99793E+05
> FREQID -1
> DOCALIB -1
> DOPOL -1
> BLVER -1
> FLAGVER -1
> DOBAND -1
> BPVER -1
> STOKES ' '
> BCHAN 1
> ECHAN 1
> CHINC 1
> BIF 1
> EIF 1
> SMOOTH 0
> UVRANGE 0
> UVTAPER 0
> CPARM(1)=12
> BADDISK=2
> CELLSIZE(1)=0.03
> CELLSIZE(2)=0.03
> CMETHOD 'DFT'
> FACTOR 0
> APARM(6)=0
> APARM(8)=1
> APARM(9)=1
> APARM(10)=0.05
> BPARM 0
> CPARM(1)=12
> NFIELD 1
> GUARD(1)=-1
> GUARD(2)=-2
> UVBXFN 1
> XTYPE 5
> YTYPE 5
> IMSIZE(*,1)=1024,1024
> FLDSIZE(*,1)=1014,1014
> UVWTFN 'NO'
> NITER 0
> GAIN 0.25
> MINPATCH 51
> RASHIFT(1)=0
> DECSHIFT(1)=0
> NBOXE(1)=1
> WIN(*,1,1)=6,6,1019,1019
> UVWTFN 'NA'
> ROBUST 5
> (NUVCO, referenced here and in the batch file, is simply UVCON in the
> 15APR99 installation upgraded to be the same as the 15OCT99 version)
> 
> Final point is that when running the CLEANing, thousands of errors about the
> header not containing pointing informations are flagged, which doesn't seen
> to affect the progress of the CLEANing, but did contribute to a message file
> of 4.7 million lines which was clogging things up . This did concern me for
> running the full 180 CLEANs (rather than the 5 attempts so far), but I guess
> I can slip a CLRMSG into the CLEANing  batch file every few maps.
> 
> Are the dirty maps etc. already posted OK? It seems to have gone very quiet
> on the mailing list.
> 
> Cheers,
>     Steven.
> 
> 
> 
>