AIPS HELP file for FARS in 31DEC25
As of Thu Jan 23 20:49:36 2025
FARS: Task to create Faraday rotation measure images
INPUTS
INNAME First image name Qpol
INCLASS First image class
INSEQ 0.0 9999.0 First image seq. #
INDISK 0.0 9.0 First image disk drive #
IN2NAME Second image name Upol
IN2CLASS Second image class
IN2SEQ 0.0 9999.0 Second image seq. #
IN2DISK 0.0 9.0 Second image disk drive #
OUTNAME First output image name
OUTSEQ -1.0 9999.0 First output image seq. #
OUTDISK 0.0 9.0 First output image disk #
INFILE Input file of weights
DOALIGN -2.0 1.0 Should images be coincident?
(See HELP.)
BLC 0.0 4096.0 Bottom left corner
TRC 0.0 4096.0 Top right corner
See help for the 1st axis
(must be frequencies!!!)
APARM Parameters for algorithm:
1 number of pixels at
half of the Fourier
transform output
The whole number is
2*APARM(1)+1
2 cell size in 1/m^2
0 =>
PI/(4*(Lmax^2-Lmin^2))
Lmax,Lmin-max,min lambda
at the data
3 0 => regular output
1 => output is RMTF
4 0=> CLEANed Fourier
transform
1=> unCLEANed Fourier
transform
5 0=>original(shifted back)
RE/IM are sent out
1=>the shifted RE/IM are
sent out
2=>amplitude and phase of
the data are sent out
(shifted back)
6 is not used
7 0=> convolve the clean
components
1=> no convolve
8 0=> use the Gaussian as
the convolve function
1=> use the Re of RMTF as
the convolve function
9 full width of Gaussian
convolve function, at 0.5
level, in 1/m^2
0 => fit to real RMTF
10 what send to output?
0 => sum of CLEAN and
residual
1 => CLEAN result
2 => residual
GAIN Gain in the CLEAN
NITER Maximum number of clean
components
FLUX Minimum flux of clean
component (Jy)
DOHIST -3.0 1.0 -2 => copy 1st HI only
-3 => copy no HI files
PRTLEV Number of Cleans to print
HELP SECTION
FARS
Type: Task
The measured brightness of images at a set of lambda^2 is
related to the brightness distribution as a function of Faraday
rotation depth through the Fourier transform. To solve the
inverse problem: "Evaluation of the brightness distribution as a
function of Faraday rotation depth using the measured brightness
in the given set of lambda^2" has been called "Faraday Rotation
Synthesis". The task FARS performs this operation.
FARS now writes 4 slice files in each output image. They
reresent the real part, imaginary part, amplitude, and phase of
the RMTF or dirty beam of the synthesis. TVSLICE and other
slice functions may be used to examine these parts of the beam.
The units displayed are based on the image and so they do not
really apply. The phase slice is in degrees and the others
really in normalized beam units.
The RUN file named DOFARS defines a procedure that will do the
initial transposes and find the weights text file, run FARS,
delete the transposed input cubes, and transpose the output
cubes to yield 4 output cubes in all.
Adverbs:
INNAME......First image name. Q-polarization. Standard defaults.
In transposed order with frequency (or frequncy ID) as
the first axis.
INCLASS.....First image class. Standard defaults.
INSEQ.......First image seq. #. 0 => highest.
INDISK......Disk drive # for the first image. 0 => any.
IN2NAME.....Second image name. U-polarization. Standard defaults.
In transposed order with frequency (or frequncy ID) as
the first axis.
IN2CLASS....Second image class. Standard defaults.
IN2SEQ......Second image seq. #. 0 => highest.
IN2DISK.....Disk drive # for the second image. 0 => any.
OUTNAME.....Output image name. Real part. Standard defaults.
Output image class for first image 'FARSre' or 'FARSam'
Output image class for second image 'FARSim' or 'FARSph'
OUTSEQ......Output image seq. #. 0 => highest unique.
OUTDISK.....Output disk number. 0 => highest with space.
INFILE......User-supplied text file defining weights for ALL
spectral channels INCLUDING those < BLC(1). The
sequence of the weights must correspond to the sequence
of frequencies in the input cubes. The weights are read
with a free format and more than one may occur on each
line of the input text file. The weights are read as
floating-point values and any value >= 0 is acceptable.
DOALIGN.....Controls how the two input images are to be aligned
True (>.1) means that the images must agree in
their coordinates, though not necessarily in the reference
pixel position. Alignment is by coordinate values (if
DOALIGN > -0.1) or by offsets from the reference pixel
positions (if DOALIGN <= -0.1). NOTE: all real axes (>1
point) are aligned. If DOALIGN = -2, the headers are
ignored and the images are aligned at pixel (1,1,...).
(see HELP DOALIGN).
BLC.........Bottom left corner of the 1st input image. The other
images are aligned by coordinates (see DOALIGN) on all
axes having > 1 point. The second image may have fewer
real axes than the 1st. The 2 windows must have the same
dimension on the first 2 axes, but the task will select a
smaller window than was specified if needed to overlap the
2 images.
The first axis (must be frequency. Use TRANS if not!!!)
is controlled by BLC, TRC only for input data. The
output size for the first axis is controlled by
APARM(1,2,4)
TRC.........Top right corner of input images. (See BLC.)
APARM.......Parameters needed for algorithm:
APARM(1) number of pixels for half of the Fourier
transform output. The total number is
2*APARM(1)+1 with coordinate zero at the center.
The value of APARM(1) should be chosen in
accordance the expected range of Faraday
rotation measure and value of CELL (APARM(2))
APARM(2) cell size of the outputs, in 1/m^2
The cell is recomended to be less or around
the default value
(pi)/4/(lambda^2max - lambda^2min)
APARM(3) 0 => regular outputs
1 => outputs are RMTF ("1-D dirty beam")
Note that the RMTF is now written as
slice files so this option is normally
not needed.
APARM(4) 0=> CLEANed Fourier transform, using inputs:
NITER, FLUX, GAIN. The CLEAN uses the
shifed (at lambda^2) data but the cleaned
components correspond to the original
lambda^2
1=> the uncleaned Fourier transform is recorded
RE/amp is recorded to the first output
IM/phase is recorded to the second output
APARM(5) The fourier transform (FT) is done using a
lambda^2 coordinate shifted by the mean
(weighted) lambda^2.
0=> The output are images of Real and
Imaginary after applying a phase shift to
undo this offset.
1=> The output Real and Imaginary do not have
any phase shift applied.
This option may allow better
discrimination of different features in
the Fourier transform. In particular,
without Cleaning anyway, the value at the
RM=0 pixel represents the sum of Q in real
and of U in imaginary.
2=> As 0 except that the amplitude and phase
(corrected for the L2mean) of the data are
sent out.
APARM(6) is not used
APARM(7) 0=> convolve the clean components
1 => no convolution; so just the set of the
clean components is sent to the output files
No convolution is forced if uncleaned Fourier
data are sent out (APARM(4)=1)
APARM(8) 0=> use the Gaussian as the convolution
function
1=> use the Re of RMTF as the convolution
function
APARM(9) full width at half maximumof the Gaussian
convolution function, in 1/m^2
0 -> fit the real part of the RMTF.
APARM(10) What send to the output?
0 => sum of CLEAN and residual
1 => CLEAN result
2 => residual
GAIN........Gain in the CLEAN
NITER.......Maximum number of clean components 0 => 1
FLUX........Minimum Clean component (Jy)
The task can subtract the given number of complex
components. On each iteration, the maximum (and its
position) of the spectrum amplitude is determined.
The complex function RMTF is multiplied by the
complex value of the spectrum at the position of the
amplitude maximum and by GAIN. This product function
is shifted to the position of the maximu and
subtracted from the current residual spectrum. This
subtraction process is terminated when the number of
iterations reaches NITER or when the maximum found is
less than FLUX.
DOHIST......Normally the HI file of input 1 is copied to the output
history file and the HI file of the second input is
appended. The history of the Q and U cubes may be very
similar, so to avoid having the file grow exponentially
DOHIST = -2 => copy the first HI file only.
DOHIST = -3 => copy no HI file, write FARS HI only.
PRTLEV......> 0 => Number of Cleans to print. <= 0 -> none.
EXPLAIN SECTION
FARS: Task to solve Faraday Rotation Synthesys problem
DOCUMENTOR: L. Kogan
RELATED PROGRAMS:
The main ideas of the algorithm are described by
B.J. Burn at:
MNRAS, 133, 67, 1966 and
M.A.Brentjens
and A.G.de Bruyn at:
Astronomy & Astrophysics, 441, 1217-1228(2005)
FARS reads the two input cube images which should have identical
structure. The first axis of the cubes must be FREQ or FQID.
If not, the task 'TRANS' should be used to achieve this order.
The first cube should correspond to the Q-linear polarization
component. The second cube should correspond to the U-linear
polarization component. The purpose of the task is to get the
brightness distribution (for each pixel in the RA-DEC plane)
corresponding to the given Fraday rotation depth.
Multi-frequency observations are required to achieve this.
The task forms complex function of wavelength squared (the first
input image is the real part, the second input is the imaginary
part). For each pixel of the RA-DEC plane, FARS carries out the
Fourier transform along the wavelength squared axis. This
Fourier transform converts the observed complex brightness as a
function of the lambda^2 to the output complex function of the
Faraday rotation depth. The Fourier transform can be directed to
the outputs directly or it can be deconvolved with a complex
Clean algorithm. The real part or amplitude of the output is
recorded into the first output file and the imaginary part or
phase of this output is recorded in the second output file. The
input data are determined from the positive set of lambda^2. The
output of the Fourier transform (complex function) is the
convolution of the actual brightness (as a function of the
Faraday depth) and the so called Rotation Measure Transfer
Function (RMTF). The RMTF is a one dimensional Fourier transform
of the lambda^2 sampling function and is recorded in slice files
attached to the output images. The RMTF is similar to the dirty
beam in aperture synthesis. The dirty beam is real function
because the brightness distribution is a real function. RMTF is
a complex function, but it's imaginary part can be zeroed at the
vicinity of zero Faraday depth if we shift the origin of the
lambda^2 set to the median. The more homogeneous the
distribution of the lambda^2 the wider the area where the
imaginary part of RMTF is near zero. FARS carries out all
calculation with the centered data, and then rotates the outputs
by the phase corresponded to the central lambda^2. The output
(RE, IM) can be shifted back to the original lambda^2
distribution or not (or amplitude and phase can be sent to the
output) under control of APARM(5).
It is supposed the images for each of the frequencies are available
The example of the typical image header is following:
AIPS 1: ----------------------------------------------------------------
AIPS 1: Type Pixels Coord value at Pixel Coord incr Rotat
AIPS 1: RA---SIN 2048 12 28 16.418 1024.00 -0.100000 0.00
AIPS 1: DEC--SIN 1024 12 39 58.294 513.00 0.100000 0.00
AIPS 1: FREQ 1 8.0851000E+09 1.00 5.0000000E+07 0.00
AIPS 1: STOKES 1 2.0000000E+00 1.00 1.0000000E+00 0.00
The images for different frequencies should be combined in the cube
using the AIPS task MCUBE. The recommended inputs for MCUBE are
following:
AIPS 1: MCUBE: Task to collect a set of n-dim maps into a (n+1)-dim map
AIPS 1: Adverbs Values Comments
AIPS 1: ----------------------------------------------------------------
AIPS 1: INNAME 'M87X' Input name(name).
AIPS 1: INCLASS 'Q-COS' Input name(class).
AIPS 1: INSEQ 1 Input name(seq. #). 0=>high
AIPS 1: First sequence # in the set
AIPS 1: INDISK 3 Input disk drive #. 0=>any
AIPS 1: IN2SEQ 4 Last sequence # in set.
AIPS 1: IN3SEQ 1 Sequence # increment.
AIPS 1: OUTNAME 'M87XCUBE' Output name(name).
AIPS 1: OUTCLASS 'Q-COS' Output name(class).
AIPS 1: OUTSEQ 0 Output name(seq. #).
AIPS 1: 0 => highest unique
AIPS 1: OUTDISK 3 Output image disk drive #
AIPS 1: 0 => highest with space
AIPS 1: DOALIGN 0 Alignment control parm
AIPS 1: AXREF 1 n+1 axis pixel of map INSEQ.
AIPS 1: 0 => 1
AIPS 1: AX2REF 0 n+1 axis pixel of map IN2SEQ
AIPS 1: 0 => opposite end from INSEQ
AIPS 1: NPOINTS 4 Number of pixels on axis n+1
AIPS 1: DOCONCAT 2 > 1 => make a SEQ.NUM or
AIPS 1:
NPOINTS is the number of combined images, which should be prepared
under the same NAME and CLASS but with different INSEQ=1,2,3,4
Use task FQUBE is the images are not distributed linearly in
frequency. This task adds an FQ table to the output image
listing the actual frequency of each plane in its cube.
The output cube image should have the following axes in the image
header:
AIPS 1: ----------------------------------------------------------------
AIPS 1: Type Pixels Coord value at Pixel Coord incr Rotat
AIPS 1: RA---SIN 2048 12 28 16.418 1024.00 -0.100000 0.00
AIPS 1: DEC--SIN 1024 12 39 58.294 513.00 0.100000 0.00
AIPS 1: FQID 4 1.0000000E+00 1.00 1.0000000E+00 0.00
AIPS 1: STOKES 1 2.0000000E+00 1.00 1.0000000E+00 0.00
AIPS 1: FREQ 1 8.0851000E+09 1.00 5.0000000E+07 0.00
AIPS 1: ----------------------------------------------------------------
The axis 'FQID' reports the number of frequencies at the cube.
The axis 'FREQ' reports the reference frequency.
The actual frequencies are the sum of the reference frequency and
the offsets given in the FQ table created by MCUBE or FQUBE.
FARS requires another order of the axes. So the AIPS task TRANS should
be applied to get it. An example of TRANS inputs is:
AIPS 1: TRANS: Task to transpose a subimage of an up to 7-dim. image
AIPS 1: Adverbs Values Comments
AIPS 1: ----------------------------------------------------------------
AIPS 1: INNAME 'M87XCUBE' Input name(name).
AIPS 1: INCLASS 'Q-COS' Input name(class).
AIPS 1: INSEQ 1 Input name(seq. #). 0=>high
AIPS 1: INDISK 3 Input disk drive #. 0=>any
AIPS 1: OUTNAME 'M87XTRANS' Output name(name).
AIPS 1: OUTCLASS 'Q-COS' Output name(class).
AIPS 1: OUTSEQ 0 Output name(seq. #).
AIPS 1: 0 => highest unique
AIPS 1: OUTDISK 3 Output image disk drive #
AIPS 1: 0 => highest with room
AIPS 1: BLC *all 0 Bottom left corner of image
AIPS 1: 0 => entire image
AIPS 1: TRC *all 0 Top right corner of image
AIPS 1: 0 => entire image
AIPS 1: TRANSCOD '31245' New axis order in terms of
AIPS 1: input axis numbers
AIPS 1: BADDISK *all 0 Disks to avoid for scratch
So the final image cube should have the following axes in the image
header:
AIPS 1: ----------------------------------------------------------------
AIPS 1: Type Pixels Coord value at Pixel Coord incr Rotat
AIPS 1: FQID 4 1.0000000E+00 1.00 1.0000000E+00 0.00
AIPS 1: RA---SIN 2048 12 28 16.418 1024.00 -0.100000 0.00
AIPS 1: DEC--SIN 1024 12 39 58.294 513.00 0.100000 0.00
AIPS 1: STOKES 1 2.0000000E+00 1.00 1.0000000E+00 0.00
AIPS 1: FREQ 1 8.0851000E+09 1.00 5.0000000E+07 0.00
AIPS 1: ----------------------------------------------------------------
The same set of steps should be carried out to get the second linear
polarization input image for FARS.