Clean
From CASA Guides
Deconvolve an image with selected algorithm
This is the main clean deconvolution task. It contains many functions
1) Make 'dirty' image and 'dirty' beam (psf)
2) Multi-frequency-continuum images or spectral channel imaging
3) Full Stokes imaging
4) Mosaicking of several pointings
5) Multi-scale cleaning
6) Interactive clean boxing
7) Use starting model (eg from single dish)
vis -- Name of input visibility file
default: none; example: vis='ngc5921.ms'
imagename -- Pre-name of output images:
default: none; example: imagename='m2'
output images are:
m2.image; cleaned and restored image
With or without primary beam correction
m2.psf; point-spread function (dirty beam)
m2.flux; relative sky sensitivity over field
m2.flux.pbcoverage; relative pb coverage over field
(gets created only for ft='mosaic')
m2.model; image of clean components
m2.residual; image of residuals
m2.interactive.mask; image containing clean regions
To include outlier fields:
imagename=['n5921','outlier1','outlier2']
outlierfile --- Text file name which contains image names, sizes, field
centers
field -- Select fields in mosaic. Use field id(s) or field name(s).
['go listobs' to obtain the list id's or names]
default: = all fields
If field string is a non-negative integer, it is assumed to
be a field index otherwise, it is assumed to be a
field name
field='0~2'; field ids 0,1,2
field='0,4,5~7'; field ids 0,4,5,6,7
field='3C286,3C295'; field named 3C286 and 3C295
field = '3,4C*'; field id 3, all names starting with 4C
spw -- Select spectral window/channels
NOTE: This selects the data passed as the INPUT to mode
default: =all spectral windows and channels
spw='0~2,4'; spectral windows 0,1,2,4 (all channels)
spw='0:5~61'; spw 0, channels 5 to 61
spw='<2'; spectral windows less than 2 (i.e. 0,1)
spw='0,10,3:3~45'; spw 0,10 all channels, spw 3,
channels 3 to 45.
spw='0~2:2~6'; spw 0,1,2 with channels 2 through 6 in each.
spw='0:0~10;15~60'; spectral window 0 with channels
0-10,15-60
spw='0:0~10,1:20~30,2:1;2;3'; spw 0, channels 0-10,
spw 1, channels 20-30, and spw 2, channels, 1,2 and 3
selectdata -- Other data selection parameters
default: True
>>> selectdata=True expandable parameters
See help par.selectdata for more on these
timerange -- Select data based on time range:
default: (all); examples,
timerange = 'YYYY/MM/DD/hh:mm:ss~YYYY/MM/DD/hh:mm:ss'
Note: if YYYY/MM/DD is missing date defaults to first
day in data set
timerange='09:14:0~09:54:0' picks 40 min on first day
timerange='25:00:00~27:30:00' picks 1 hr to 3 hr
30min on NEXT day
timerange='09:44:00' pick data within one integration
of time
timerange='>10:24:00' data after this time
uvrange -- Select data within uvrange (default units meters)
default: (all); example:
uvrange='0~1000klambda'; uvrange from 0-1000 kilo-lambda
uvrange='>4klambda';uvranges greater than 4 kilo lambda
antenna -- Select data based on antenna/baseline
default: (all)
If antenna string is a non-negative integer, it is
assumed to be an antenna index, otherwise, it is
considered an antenna name.
antenna='5&6'; baseline between antenna index 5 and
index 6.
antenna='VA05&VA06'; baseline between VLA antenna 5
and 6.
antenna='5&6;7&8'; baselines 5-6 and 7-8
antenna='5'; all baselines with antenna index 5
antenna='05'; all baselines with antenna number 05
(VLA old name)
antenna='5,6,9'; all baselines with antennas 5,6,9
index numbers
scan -- Scan number range.
default: (all)
example: scan='1~5'
Check 'go listobs' to insure the scan numbers are in order.
mode -- Frequency Specification:
NOTE: See examples below:
default: 'mfs'
mode = 'mfs' means produce one image from all
specified data.
mode = 'channel'; Use with nchan, start, width to specify
output image cube. See examples below
mode = 'velocity', means channels are specified in
velocity.
mode = 'frequency', means channels are specified in
frequency.
>>> mode='mfs' expandable parameters
Make a continuum image from the selected frequency
channels/range using Multi-frequency synthesis
algorithm for wide-band narrow field imaging.
nterms is the number of Taylor terms to be used to
model the frequency dependence of the sky emission.
nterms=1 is equivalent to assuming no frequency
dependence. nterms=2 is equivalent to the
Sault-Wieringa algorithm (AandAS, 1994)
reffreq is the reference frequency about which the
Taylor expansion is done.
>>> mode expandable parameters (for modes other than 'mfs')
Start, width are given in units of channels, frequency
or velocity as indicated by mode (note: only nearest neighbour
interpolation is available at this time).
nchan -- Number of channels (planes) in output image
default: 1; example: nchan=3
start -- Start input channel (relative-0)
default=0; example: start=5
width -- Output channel width in units of the input
channel width (>1 indicates channel averaging)
default=1; example: width=4
interpolation -- Interpolation type of spectral data when gridded on
the uv-plane
default = 'nearest'
HOWEVER, 'linear' is recommended
outframe -- velocity reference frame of output image (for mode='velocity' or 'frequency')
Options: ,'LSRK','LSRD','BARY','GEO','TOPO','GALACTO','LGROUP','CMB'
default: ; same as input data; example: frame='bary' for Barycentric frame
veltype -- (for mode='velocity') velocity definition
Options: 'radio','optical','true' (='relativistic')
veltype='true' (or equivalently 'relativistic')
for velocity defined without approximations using the relativistic expression
default: 'radio'
examples:
spw = '0,1'; mode = 'mfs'
will produce one image made from all channels in spw
0 and 1
spw='0:5~28^2'; mode = 'mfs'
will produce one image made with channels
(5,7,9,...,25,27)
spw = '0'; mode = 'channel': nchan=3; start=5; width=4
will produce an image with 3 output planes
plane 1 contains data from channels (5+6+7+8)
plane 2 contains data from channels (9+10+11+12)
plane 3 contains data from channels (13+14+15+16)
spw = '0:0~63^3'; mode='channel'; nchan=21; start = 0;
width = 1
will produce an image with 20 output planes
Plane 1 contains data from channel 0
Plane 2 contains date from channel 2
Plane 21 contains data from channel 61
spw = '0:0~40^2'; mode = 'channel'; nchan = 3; start =
5; width = 4
will produce an image with three output planes
plane 1 contains channels (5,7)
plane 2 contains channels (13,15)
plane 3 contains channels (21,23)
psfmode -- method of PSF calculation to use during minor cycles:
default: 'clark': Options: 'clark','clarkstokes', 'hogbom'
'clark' use smaller beam (faster, usually good enough);
for stokes images clean components peaks are searched in the I^2+Q^2+U^2+V^2 domain
'clarkstokes' locate clean components independently in each stokes image
'hogbom' full-width of image (slower, better for poor
uv-coverage)
Note: psfmode will be used to clean is imagermode =
imagermode -- Advanced imaging e.g mosaic or Cotton-Schwab clean
default: imagermode=: Options: , 'csclean', 'mosaic'
default => psfmode cleaning algorithm used
>>> gridmode= expandable parameters
The default value of has no effect.
>>> gridmode='widefield' expandable parameters
Apply corrections for non-coplanar effects during imaging
using the W-Projection algorithm (Cornwell et al. IEEE JSTSP
(2008)) or faceting or a combination of the two.
wprojplanes is the number of pre-computed w-planes used for
the W-Projection algorithm. wprojplanes=1 disables
correction for non-coplanar effects.
facets is the number of facets used. W-Projection is done
for each facet.
>>> gridmode='aprojection' expandable parameters
Correct for the (E)VLA polarization squint using the
A-Projection algorithm (Bhatnagar et al., AandA (2008)).
cfcache is the name of the directory to be used to cache the
convolution functions. These functions can be reused again if
the image parameters are unchanged. If the image parameters
change, a new cache must be created (or the existing one
removed).
painc (in degrees) is the Parallactic Angle increment used to
compute the convolution functions.
>>> imagermode='mosaic' expandable parameter(s):
Image as a mosaic of the different pointings (uses csclean style too)
mosweight -- Individually weight the fields of the mosaic
default: False; example: mosweight=True
This can be useful if some of your fields are more
sensitive than others (i.e. due to time spent
on-source); this parameter will give more weight to
higher sensitivity fields in the overlap regions.
ftmachine -- Gridding method for the image;
Options: ft (standard interferometric gridding), sd
(standard single dish) both (ft and sd as appropriate),
mosaic (gridding use PB as convolution function)
default: 'mosaic'; example: ftmachine='ft'
if imagermode mosaic is chosen and ftmachine is mosaic,
heterogenous arrays like Carma or Alma are recognized
and the right Primary Beam (depending on the size of the dish)
is used for each baseline.
scaletype -- Controls scaling of pixels in the image plane.
(Not fully implemented...for now only controls
what is seen if interactive=True...but in the future will
control the image on which clean components are searched)
default='SAULT'; example: scaletype='PBCOR'
Options: 'PBCOR','SAULT'
'SAULT' when interactive=True shows the residual
with constant noise across the mosaic. If
pbcor=False, the final output image is NOT
corrected for the PB pattern, and therefore is
not "flux correct". Division of SAULT
<imagename>.image by the <imagename>.flux image
will produce a "flux correct image", can also
be acheived by setting pbcor=True.
'PBCOR' uses the SAULT scaling scheme for
deconvolution, but if interactive=True shows the
primary beam corrected image; the final PBCOR
image is "flux correct" if pbcor=True.
>>> imagermode='csclean' expandable parameter(s):
Image using the Cotton-Schwab algorithm in between major cycles
cyclefactor -- Change the threshold at which the deconvolution
cycle will stop, degrid and subtract from the visibilities.
For poor PSFs, reconcile often (cyclefactor=4 or 5);
For good PSFs, use cyclefactor 1.5 to 2.0.
Note: threshold = cyclefactor * max sidelobe * max residual.
default: 1.5; example: cyclefactor=4
cyclespeedup -- Cycle threshold doubles in this number of
iterations
default: -1;
example: cyclespeedup=3
try cyclespeedup = 50 to speed up cleaning
multiscale -- set of scales to use in deconvolution. If set,
cleans with several resolutions using hobgom clean. The
scale sizes are in units of cellsize. So if
cell='2arcsec', a multiscale scale=10 = 20arcsec. First
scale should always be 0 (point), we suggest second on
the order of synthesized beam, third 3-5 times
synthesized beam, etc. For example if synthesized beam
is 10" and cell=2", try multscale = [0,5,15]. Note,
multiscale is currently a bit slow.
default: multiscale=[] (standard CLEAN using psfmode algorithm,
no multi-scale). Example: multiscale = [0,5,15]
>>> multiscale expandable parameter(s):
negcomponent -- Stop component search when the largest scale has
found this number of negative components; -1 means continue
component search even if the largest component is
negative. default: -1; example: negcomponent=50
smallscalebias -- A bias toward smaller scales.
The peak flux found at each scale is weighted by
a factor = 1 - smallscalebias*scale/max_scale, so
that Fw = F*factor.
Typically the values range from 0.2 to 1.0.
default: 0.6
imsize -- Image pixel size (x,y). DOES NOT HAVE TO BE A POWER OF 2
default = [256,256]; example: imsize=[350,350]
imsize = 500 is equivalent to [500,500]
If include outlier fields, e.g., [[400,400],[100,100]] or
use outlierfile.
Avoid odd-numbered imsize.
cell -- Cell size (x,y)
default= '1.0arcsec';
example: cell=['0.5arcsec,'0.5arcsec'] or
cell=['1arcmin', '1arcmin']
cell = '1arcsec' is equivalent to ['1arcsec','1arcsec']
NOTE:cell = 2.0 => ['2arcsec', '2arcsec']
phasecenter -- direction measure or fieldid for the mosaic center
default: => first field selected ; example: phasecenter=6
or phasecenter='J2000 19h30m00 -40d00m00'
If include outlier fields,
e.g. ['J2000 19h30m00 -40d00m00',J2000 19h25m00 -38d40m00']
or use outlierfile.
restfreq -- Specify rest frequency to use for output image
default= Occasionally it is necessary to set this (for
example some VLA spectral line data). For example for
NH_3 (1,1) put restfreq='23.694496GHz'
stokes -- Stokes parameters to image
default='I'; example: stokes='IQUV';
Options: 'I','IVQU','IQUV','RR','LL','XX','YY','RRLL','XXYY'
niter -- Maximum number iterations,
if niter=0, then no CLEANing is done ("invert" only)
default: 500; example: niter=5000
gain -- Loop gain for CLEANing
default: 0.1; example: gain=0.5
threshold -- Flux level at which to stop CLEANing
default: '0.0mJy';
example: threshold='2.3mJy' (always include units)
threshold = '0.0023Jy'
threshold = '0.0023Jy/beam' (okay also)
interactive -- use interactive clean (with GUI viewer)
default: interactive=False
example: interactive=True
interactive clean allows the user to build the cleaning
mask interactively using the viewer. The viewer will
appear every npercycle interation, but modify as needed
The final interactive mask is saved in the file
imagename_interactive.mask. The initial masks use the
union of mask and cleanbox (see below).
>>> interactive=True expandable parameters
npercycle -- this is the number of iterations between each clean
to update mask interactively. It is important to modify
this number interactively during the cleaning, starting wiht
a low number like 20, but then increasing as more extended
emission is encountered.
chaniter -- specify how interactive CLEAN is performed,
either by stepping through channels or do jointly for all channel
default: chaniter='joint';
example: chaniter='channel'; step through channels
mask -- Specification of cleanbox(es), mask image(s), and/or
region(s) to be used for CLEANing. As long as the image has
the same shape (size), mask images from a previous
interactive session can be used for a new execution. NOTE:
the initial clean mask actually used is the union of what
is specified in mask and <imagename>.mask default: [] (no
masking); Possible specification types: (a) Explicit
cleanbox pixel ranges example: mask=[110,110,150,145] clean
region with blc=110,100; trc=150,145 (pixel values) (b)
Filename with cleanbox pixel values with ascii format:
example: mask='mycleanbox.txt' <fieldid blc-x blc-y
trc-x trc-y> on each line
1 45 66 123 124
2 23 100 300 340
(c) Filename for image mask example: mask='myimage.mask'
(d) Filename for region specification (e.g. from viewer)
example: mask='myregion.rgn' (e) Combinations of any of the
above example: mask=[[110,110,150,145],'mycleanbox.txt',
'myimage.mask','myregion.rgn']
If include outlier fields, then mask need to be specified in
nested lists:
e.g. mask=[[[110,110,150,145],'myimage.mask'],[],[20,20,40,40]]
(A clean box with [110,110,150,145] and a image mask for main field,
no mask for 1st outlier field, 1 clean box for second outlier field.)
uvtaper -- Apply additional uv tapering of the visibilities.
default: uvtaper=False; example: uvtaper=True
>>> uvtaper=True expandable parameters
outertaper -- uv-taper on outer baselines in uv-plane
[bmaj, bmin, bpa] taper Gaussian scale in uv or
angular units. NOTE: uv taper in (klambda) is roughly on-sky
FWHM(arcsec/200)
default: outertaper=[]; no outer taper applied
example: outertaper=['5klambda'] circular taper
FWHM=5 kilo-lambda
outertaper=['5klambda','3klambda','45.0deg']
outertaper=['10arcsec'] on-sky FWHM 10"
outertaper=['300.0'] default units are meters
in aperture plane
innertaper -- uv-taper in center of uv-plane
[bmaj,bmin,bpa] Gaussian scale at which taper falls to
zero at uv=0
default: innertaper=[]; no inner taper applied
NOT YET IMPLEMENTED
modelimage -- Name of model image(s) to initialize cleaning. If
multiple images, then these will be added together to
form initial staring model NOTE: these are in addition
to any initial model in the <imagename>.model image file
default: (none); example: modelimage='orion.model'
modelimage=['orion.model','sdorion.image'] Note: if the
units in the image are Jy/beam as in a single-dish
image, then it will be converted to Jy/pixel as in a
model image, using the restoring beam in the image
header
weighting -- Weighting to apply to visibilities:
default='natural'; example: weighting='uniform';
Options: 'natural','uniform','briggs',
'superuniform','briggsabs','radial'
>>> Weighting expandable parameters
For weighting='briggs' and 'briggsabs'
robust -- Brigg's robustness parameter
default=0.0; example: robust=0.5;
Options: -2.0 to 2.0; -2 (uniform)/+2 (natural)
For weighting='briggsabs'
noise -- noise parameter to use for Briggs "abs"
weighting
example noise='1.0mJy'
npixels -- uv-cell area used for weight calculation
example npixels=1
Default = 0
superuniform: 0 Means 3x3 cells for weighting
the cell weight is proportional to the weight of
the 3x3 cells centered on it.
superuniform = F means 1x1 cell for averaging weights.
briggs/briggsabs: 0 is similar to 1x1 cell weight.
1 may? be similar to 3X3 cells.
Only npixels 0 or 1 recommended
restoringbeam -- Output Gaussian restoring beam for CLEAN image
[bmaj, bmin, bpa] elliptical Gaussian restoring beam
default units are in arc-seconds for bmaj,bmin, degrees
for bpa default: restoringbeam=[]; Use PSF calculated
from dirty beam.
example: restoringbeam=['10arcsec'] circular Gaussian
FWHM 10" example:
restoringbeam=['10.0','5.0','45.0deg'] 10"x5"
at 45 degrees
pbcor -- Output primary beam-corrected image
default: pbcor=False; output un-corrected image
example: pbcor=True; output pb-corrected image (masked outside
minpb) Note: if you set pbcor=False, you can later
recover the pbcor image by dividing by the .flux image
(e.g. using immath)
minpb -- Minimum PB level to use default=0.1;
The flux image is used to determine this
except for the case of mosaic with ft='mosaic'
where the flux.pbcoverage image is used.
example:
minpb=0.01 Note: this minpb is always in effect
(regardless of pbcor=True/False)
calready -- if True will create scratch columns if they are
not there. And after clean completes the predicted model
visibility is from the clean components are written to the ms.
async -- Run asynchronously
default = False; do not run asychronously
======================================================================
HINTS ON CLEAN WITH FLANKING FIELDS
1. Decide if the images will be specified directly in the
inputs or with an outlier file. For more than a few fields,
an outlier file more convenient.
Direct Method:
cell = ['1.0arcsec', '1.0arcsec']
imagename = ['M1_0','M1_1','M1_2]
imsize = [[1024,1024],[128,128],[128,128]]
phasecenter = ['J2000 13h27m20.98 43d26m28.0',
'J2000 13h30m52.159 43d23m08.02', 'J2000 13h24m08.16 43d09m48.0']
Text file method (in outlier.txt)
imagename = 'M1'
outlierfile = 'outlier.txt'
[phasecenter, imsize ignored]
Contents of outlier.txt
C 0 1024 1024 13 27 20.98 43 26 28.0
C 1 128 128 13 30 52.158 43 23 08.00
C 2 128 128 13 24 08.163 43 09 48.00
In both cases the following images will be made:
M1_0.image, M1_1.image, M1_2.image cleaned images
M1.0.model, M1_1.model, M1_2.model model images
M1.0.residual, M1_1.residual, M1_2.residual residual images
an outlier file more convenient.
Direct Method:
cell = ['1.0arcsec', '1.0arcsec']
imagename = ['M1_0','M1_1','M1_2]
imsize = [[1024,1024],[128,128],[128,128]]
phasecenter = ['J2000 13h27m20.98 43d26m28.0',
'J2000 13h30m52.159 43d23m08.02', 'J2000 13h24m08.16 43d09m48.0']
Text file method (in outlier.txt)
imagename = 'M1'
outlierfile = 'outlier.txt'
[phasecenter, imsize ignored]
Contents of outlier.txt
C 0 1024 1024 13 27 20.98 43 26 28.0
C 1 128 128 13 30 52.158 43 23 08.00
C 2 128 128 13 24 08.163 43 09 48.00
In both cases the following images will be made:
M1_0.image, M1_1.image, M1_2.image cleaned images
M1.0.model, M1_1.model, M1_2.model model images
M1.0.residual, M1_1.residual, M1_2.residual residual images
2. Masks for flanking fields are specified in same way as
in the single output field case, but need extra '[ ]' to
distinguish each field.
mask=[['myregion.rg',[100,100,150,150]],['myimage1.mask'],[]]
would apply masks:
for the first field (main field),
myregion.rg and a box defined by [100,100,150,150] in pixels
for the seconf field (first outlier), myimage1.mask
for the third field (second outlier), no mask (produce a mask for
whole field)
However, if boxfiles are given, ids in the first column of the files
are used to match with fields (using order given imagename or
outlierfile.
So, if the content of a boxfile looks like this,
0 45 66 123 124
1 23 100 300 340
2 20 20 40 40
then [45 66 123 124] is assigned to first field (imagename[0], or first line
of outlierfile).
