Benchmark imaging

[AlecThomson]

We need to:

1. Determine best default parameters for wsclean imaging of cubes.
2. Benchmark the time taken per field for imaging.

[tjgalvin]

On point 1 these were the settings I believe I found most useful. These include the -force-mask option I added to wsclean.

I found doing this as two separate passes proved to be most stable in terms of finishing and not ballooning out the memory. There was also a strange quirk with the mask thresh in the Q/U case. You will see it is 1.65. I think that there is some type of squaring going on in this when wsclean evaluates the cut. in line with the square channel joining option.

wsclean 
       -abs-mem 100 
       -mgain 0.80 
       -force-mask-rounds 6
       -nmiter 25 
          -niter 50000000 
          -local-rms 
          -auto-mask 3.75
          -local-rms-window 60 
          -auto-threshold 0.5 
          -name "${outname}" 
          -size 7500 7500 
          -scale "2.5asec" 
          -weight briggs 0.5 
          -pol I 
          -use-wgridder\
          -join-channels 
          -channels-out 36 
          -minuv-l 200 
          -data-column DATA
          "${MS}" 

wsclean 
       -abs-mem 100 
       -mgain 0.80 
       -force-mask-rounds 6 
          -nmiter 7 
          -niter 50000 
          -local-rms 
          -auto-mask 1.6
          -local-rms-window 25 
          -auto-threshold 0.5 
          -name "${outname}" 
          -size "${size}" ${size} 
          -scale "2.5asec" 
          -weight briggs 0.5 
          -pol QU 
          -use-wgridder 
          -join-polarizations 
          -squared-channel-joining 
          -join-channels 
          -channels-out 36 
          -minuv-l 200 
          -data-column DATA
          "${MS}"