Manpage of impol
Section: User Commands (1)
Return to Main Contents
impol - Compute polarized intensity and position angle from Q and U
IMPOL computes the total linearly polarized intensity
(optionally debiasing it) and position angle images from
Stokes Q and U images. Position angle is positive N -> E.
Upto three values; the Q, U and I images, respectively.
The I image is only needed if you want to compute the
fractional polarization image as well or if you want to
blank the output based upon an I S/N ratio.
Wild card expansion is supported.
Up to two values; the output polarized intensity image and
optionally, its associated error image (which will be constant).
Default is no output images.
Up to values; the output fractional polarization image
and optionally, its associated error image. You need
to input an I image to keyword "in" for this.
Up to two values; the output position angle image and optionally,
its associated error image (which will not be constant). These
will be in degrees (but see OPTIONS=RADIANS),
Default is no output images.
Up to 2 values; the mean standard deviation of the noise in
the Q & U images (i.e. one number for them both), and the
standard deviation of the I image.
These are required for debiasing (Q,U only), or for generating
output error images, or for blanking the output. Try to make the
Q,U value as accurate as possible for the debiasing.
Perhaps measure it from a V image
No default for sigma_QU, sigma_I defaults to sigma_QU
Up to 2 values. The first is the S/N ratio, P/SIGMA_QU, below
which the output images are blanked (see also options=zero below).
It is generally recommended that an SNCUT of at least 2 is used.
The second value, which is only valid when you have input an I
image and sigma, is the S/N ratio, I/SIGMA_I, below which output
images are blanked (defaults to no I based blanking)
The default is 2.0 and 0.0
The output images are blanked if the error in the position
angle image (degrees or radians depending on OPTIONS) is greater
than this value. This is active even if you don't output
the PA image. Note that there is no equivalent for the output
error of the POLI image because the error is constant and
equal to SIGMA. Keyword SNCUT essentially takes care of this.
The default is no position angle error based blanking.
After computing the position angle image, rotate the position
angles back to zero wavelength by an amount specified by
RM (rad/m**2). Better to use IMRM to generate the rotation
measure and zero wavelength position angle images.
Default is 0.0
Task enrichment options. Minimum match is active.
"bias" If computing polarized intensity, do NOT remove the Ricean
bias in the image. By default, the bias is removed to first
order with P = sqrt(P_obs**2 - sigma**2) You should have
a very good reason for using this option (e.g. a detection
experiment). See VLA memo no. 161 by Patrick Leahy for
more details of bias removal.
"zero" When the output pixel is clipped because the debiasing
fails (P**2 may become negative), setting OPTIONS=ZERO
will cause the output polarized intensity image (not the
position angle image) pixel to be set to 0.0 rather than
being masked out. This is very important if you are
interested in doing statistics on a region of low
polarized intensity S/N ratio. If you leave the region
masked rather than zeroed, you will bias the statistics
in that region -- zero is a better estimate of the pixel
than just excluding it from the statistics (provided the
clip level is sufficiently small). Residual bias in the
statistical results from the area then depend upon how
well the bias remover works and at what level clipping
was performed. See VLA memo no. 161 by Patrick Leahy.
"radians" Output the position angle image in radians instead
"relax" Only warn about image axis descriptor mismatches
instead of giving a fatal error
PGPLOT device on which to draw a plot showing the effect of bias
in polarized intensity images. It plots true polarized intensity
versus the bias, which is the estimated polarized intensity minus
the true polarized intensity. Three estimators are shown;
observed, first order, and maximum likelhood. It is assumed
that sigma_P = 1 in these plots. Because these plots are drawn
following a monte carlo simulation of some 15,000 trials of the
noise, you will need to be patient. You can just make this bias
plot without actually working on any data if you wish. See also
VLA memo no. 161 by Patrick Leahy.
Default is no plot.
- PERSON RESPONSIBLE
This document was created by
using the manual pages.
Time: 18:35:38 GMT, July 05, 2011