Skip to content

mrbell/pyrmsynth

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

61 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

pyrmsynth - Python based RM Synthesis code including RMCLEAN

Current version: 1.3.0 Updated on: 2015-06-22

pyrmsynth performs RM-synthesis, either simply by Fourier transformation (to produce a dirty image) or using the RMCLEAN method as described by Heald, et al. (2009). It uses FFTs for the Fourier inversion and, as far as
known to the authors, this is the only RM synthesis software around that does this. The Numpy FFTs are themselves quite fast, but in order to use them, the data first need to be placed on a regularly spaced lambda^2 grid. For this, the data are "gridded" by convolution with a Kaiser-Bessel Window function and sampling at regular intervals, as described in e.g. Beatty, Nishimura, and Pauly (IEEE Transactions in Med. Imaging, Vol 24, No. 6, 2005). This procedure also naturally allows for the handling of non-regularly spaced frequencies.

The gridding procedure, which requires a convolution, is quite slow when implemented in pure Python, thus, it was re-implemented the gridding routines using Cython, which converts python-eque code into C code that can be compiled and imported into Python.

The result is a package that performs fast RM Synthesis and RM CLEAN imaging while still providing the flexibility of a Python interface.

pyrmsynth contains an application rmsynthesis.py for processing the lines of sight in a "stack" of sky images, i.e. many polarized sky images at different frequencies. This application was written with LOFAR processing in mind, but should be generally useful as long as your images are provided as a set of FITS files generaged by CASA (or something compatible).

pyrmsynth also can be used as a more generic library for writing your own RM synthesis applications. The rm_tools sub-package contains efficient classes for RM synthesis and RM CLEAN. You can easily write your own scripts to do file I/O and use this package to do the actual RM synthesis computations in an efficient manner.


Building:

Dependancies: cython, gsl.

To build the cython component, cd to the rm_tools directory and run:

python setup.py build_ext --inplace

Add the rm_tools directory to your PYTHONPATH and put rmsynthesis.py somewhere in your PATH. E.g.:

echo 'export PYTHONPATH=$PYTHONPATH:/path/to/rm_tools' >> ~/.bashrc
cp rmsynthesis.py ~/bin

Code usage:

python rmsynthesis.py

The rmsynthesis.py software, for the moment, works on sets of FITS files. Each FITS file contains images from a single sub-band, or some other subset of the observed frequencies. As a default, the code assumes all Stokes parameters to be saved in one FITS file. There is an additional option that allows for the handling of separately save Q and U FITS files. Simply put all FITS files in a single directory and the software will read them all in, stack them into a single data cube, and perform RM synthesis along each line of sight.

User defined frequency weights can be included by providing a text file in which each line containes a weight to be applied to each frequency. The name of this file MUST be "weight.txt".

A spectral index can be provided to the code, either in form of an average global value, or in form of an additional FITS file containing a spectral index estimate. The FITS file needs to be specified in the parameter file.

The software reads in a parameter file. An example templatecan be found in the file rmsynth.par. All of the parameters listed in the sample file must be included, unless explicitly stated in the parameter file. A description of the various options is included in the comments in the .par file.

In addition to the parameter file, there are a couple of options that you can set when running the code. Type rmsynthesis.py -h if you need help. Right now, the options are

Options:

--version show program's version number and exit

-h, --help show this help message and exit

-p, --plot_rmsf Plot the RMSF as soon as it is computed.

-V, --stokes_v Produce a Stokes V cube after reading the fits files.

-s, --separate_stokes Indicate that the Stokes Q and U input images are stored in separate FITS files.

-f, --freq_last Indicate that NAXIS4 is the frequency axis.

-r, --rest_freq Indicate that the frequency for an image is given in the RESTFREQ header keyword

For more detailed information, please refer to the pyrmsynth wiki (https://github.com/mrbell/pyrmsynth/wiki/Using-rmsynthesis.py). Any bugs and issues can be reported to the developers via the github issue tracker.


Simulation tool:

Pyrmsynth comes with a simple simulation file, to be found in the simulation sub-directory. It takes a user-provided FITS file as a header template, a simple model text file, and a number of user-defined command line inputs to produce a simple model data set that can be used for testing purposes.

pyrmsynth is licensed under the GPLv3.