[ Basic Info | User Guide ]
Basic Information on invert
Purpose: Transform multi-pointing visibility data into a map
Categories: map making
INVERT is a MIRIAD task that forms images from visibilities.
INVERT can form continuum images or spectral line cubes. It can
generate images/cubes for several polarisations, as well as
handling multi-frequency synthesis and mosaicing observations.
INVERT can also form complex-valued images from non-Hermitian
data (e.g. holography data). Appropriate point-spread functions
can also be generated.
Input visibility data files. Several files can be given.
Output map (image) file name. Each output file consists of a
single polarization/Stokes parameter. If several different
pols/Stokes images are being made, then several file names
should be given. No default.
Output beam (point-spread function) file name. The default is
not to make a beam.
The size of the output dataset. The default is to image out to
primary beam half power points. For options=mosaic, an image of
this size is made for each pointing before a linear mosaic
operation is performed.
Image cell size, in arcsec. If two values are given, they give
the RA and DEC cell sizes. If only one value is given, the
cells are made square. The default is about one third of the
resolution of the resultant images.
When not mosaicing, this gives the sky position to shift to the
center of the output images. The position is specified as an
offset (in arcsec) from the observing center. The default is to
perform no shifting.
When mosaicing, this gives the sky coordinate (RA and DEC) of
the reference pixel in the imaging process. The value can be
given in the form hh:mm:ss,dd:mm:ss, or as decimal hours and
degrees. INVERT applies appropriate shifts to make this
location fall on a pixel. The default is a central observing
This determines a gaussian taper to apply to the visibility
data. It specifies the FWHM of an image-domain gaussian --
tapering the visibility data is equivalent to convolving with
this image-domain gaussian.
Either one or two values can be given, in arcsec, being the FWHM
in the RA and DEC directions. If only one value is given, the
taper is assumed to be symmetric. The default is no taper.
The signal-to-noise ratio will be optimised in the output image
if this parameter is set to the FWHM of typical image features
If you are more accustomed to giving this parameter in the uv
plane (as AIPS requires), then:
fwhm(image plane) = 182 / fwhm(uv plane)
where the image plane fwhm is measured in arcseconds, and the uv
plane fwhm is measured in kilowavelengths.
Sidelobe suppression area, given in arcseconds. This parameter
gives the area around a source where INVERT attempts to suppress
sidelobes. Two values (for the RA and DEC directions
respectively) can be given. If only one value is given, the
suppression area is made square. The default is to suppress
sidelobes in an area as large as the field being mapped.
Note that uniform weighting can produce images with spuriously
high noise, especially for mfs imaging. It is recommended
to use robust~0.5 if sup is non zero or unset.
The suppression area is essentially an alternate way of
specifying the weighting scheme being used. Suppressing
sidelobes in the entire field corresponds to uniform weighting
(so the default corresponds to uniform weighting). Natural
weighting gives the best signal to noise ratio, at the expense
of no sidelobe suppression. Natural weighting corresponds to
SUP=0. Values between these extremes give a tradeoff between
signal to noise and sidelobe suppression, and roughly correspond
to AIPS "super-uniform" weighting. [A better way to move between
these extremes is to leave sup unset and vary the robust
parameter from -2 to 2.]
Brigg's visibility weighting robustness parameter. This
parameter can be used to down-weight excessive weight being
given to visibilities in relatively sparsely filled regions of
the $u-v$ plane when using uniform weighting. Most useful
settings are in the range [-2,2], with values less than -2
corresponding to very little down-weighting, and values greater
than +2 reducing the weighting to natural weighting.
Sidelobe levels and beam-shape degrade with increasing values of
robustness, but the theoretical noise level will also decrease.
The default is no down-weighting (robust=-infinity).
Standard line parameter, with the normal defaults. In
particular, the default is to image all channels. See the help
on line for more information.
The line parameter consists of a string followed by up to
four numbers, viz:
where 'linetype' is one of "channel", "wide", "velocity" or
Line type of the reference channel, specified in a similar to
the line parameter. Specifically, it is in the form:
Before mapping, the visibility data are divided by the reference
channel. The default is no reference channel.
This allows a subset of the uv data to be used in the mapping
process. See the Users Manual for information on how to specify
this parameter. The default is to use all data.
Standard polarisation/Stokes parameter selection. See the help
on stokes for more information. Several polarisations can be
given. The default is "ii" (i.e. Stokes-I, given the assumption
that the source is unpolarised).
This gives extra processing options. Several options can be
given (abbreviated to uniqueness), and separated by commas:
nocal Do not apply gains table calibration to the data.
nopol Do not apply polarisation leakage corrections.
nopass Do not apply bandpass table calibration to the data.
double Normally INVERT makes the beam patterns the same
size as the output image. This option causes the
beam patterns to be twice as large.
systemp Weight each visibility in inverse proportion to the
noise variance. Normally visibilities are weighted
in proportion to integration time. Weighting based
on the noise variance optimises the signal-to-noise
ratio (provided the measures of the system
temperature are reliable!).
fsystemp Like systemp, but use frequency dependent Tsys.
You need to run atrecal before invert to create the
systempf variable containing the Tsys spectrum.
Atrecal requires autocorrelations to be present.
This option only works in combination with the
mfs Perform multi-frequency synthesis. The causes all
the channel data to be used in forming a single map.
The frequency dependence of the uv coordinate is
thus used to give better uv coverage and/or avoid
frequency smearing. For this option to produce
useful maps, the intensity change over the frequency
band must be small. Set the 'line' parameter to
select the channels that you wish to grid.
sdb Generate the spectral dirty beam as well as the
normal beam, when MFS processing. The default is
only to create the normal beam. If the spectral
dirty beam is created, this is saved as an extra
plane in the beam dataset.
mosaic Process multiple pointings, and generate a linear
mosaic of these pointings. For single pointings
to be combined with linmos you can use this to
specify a common reference position with the
offset parameter. Observations using OTF mosaicing
always need to specify this to ensure the moving
beam is handled properly.
imaginary Make imaginary image for non-Hermitian data
amplitude Produce a image using the data amplitudes only. The
phases of the data are set to zero.
phase Produce an image using the data phase only. The
amplitudes of the data are set to 1.
sin Label the output map and beam as a SIN projection.
Default is NCP unless non-east-west baselines are
present or the field centre is within 3 deg of the
celestial equator (because NCP blows up near the
equator). Note that this option simply changes
ctype1 and ctype2 in the header, the translation
only being correct to first order about the field
centre. A similar result could be obtained by
running 'puthd' on the output map, e.g.
puthd in=<map>/ctype1 value=RA---SIN
puthd in=<map>/ctype2 value=DEC--SIN
and likewise for the beam.
This determines the algorithm to be used in imaging.
Possible values are:
fft The conventional grid-and-FFT approach. This is the
default and by far the fastest.
dft Use a discrete Fourier transform. This avoids aliasing
but at a hugh time penalty.
median This uses a median approach. This is generally robust
to bad data and sidelobes, has a even larger time
penalty and produces images that cannot be deconvolved.
NOTE: Dft and median modes are not supported with
NOTE: This parameter should be used with caution! See the Users
Guide for more information on its applicability.
When forming spectral cubes, INVERT normally insists that all
channels in a given visibility spectrum must be good before
accepting the spectrum for imaging. This keyword allows this
rule to be relaxed. It consists of two parts: a tolerance and
a method for replacing the bad channels.
The tolerance is a value between 0 and 1, giving the fraction of
channels that INVERT will tolerate as being bad before the
spectrum is totally discarded. The default is 0, indicating
that INVERT will not tolerate any bad channels. A value of 1
indicates that INVERT will accept a spectrum as long as there is
at least one good channel.
The replacement method is either the value `zero' or
'interpolate', indicating that the bad channels are either to be
replaced with 0, or to be estimated by linear interpolation of
two adjacent good channels. See the Users Guide for the merits
and evils of the two approaches. The default is 'zero'.
Revision: 1.16, 2012/09/18 01:52:21 UTC
Generated by email@example.com on 24 Sep 2012