tauclean: pulse broadening deconvolution

Python 3 implementation of the pulse profile scattering deconvolution (CLEAN) code described by Bhat et al. (2003). The technique is based on the extensively used CLEAN algorithms used for synthesis image reconstruction.

Included in package are two scripts:

• simulate - A simulation tool to artificially create scattered pulse profiles

• tauclean - The deconvolution program which will attempt to reconstruct the intrinsic, unscattered pulse profile. It can:

  • search the pulse broadening time scale (tau) parameter space and provide metrics to judge the best model; or

  • deconvolve a specified PBF model (with a known tau) from the data

Usage

simulate

Primarily for testing purposes, the simulate script will go through the nominal steps taken in the CLEAN code itself to produce a scattered pulsar profile by convolving the user-described intrinsic pulse profile with a specified scattering kernel.

An example use case would be:

simulate -n 1024 -p 500 -m 500 -a 12 -w 10 -k thin -t 20.0

which will create a 1024-bin profile (500 ms period) with a single Gaussian component profile centered at bin 500 with an amplitude of 12 and a width of 10 bins. This "intrinsic" profile is then convolved with a thin screen scattering kernel with a pulse broadening time scale (tau) of 20 ms to produce the final scattered profile. There are also options to specify the desired signal-to-noise ratio of the output profile, as well as describe residual dispersion smearing incurred when using incoherent dedispersion (in the above case, the default is to assume coherently dedispersed data, thus there will be no residual smearing).

tauclean

The tauclean script is the user interface to the deconvolution code. The primary output of this script is the reconstructed ("intrinsic") pulse profile. It follows the method laid out in Bhat et al. (2003). tauclean can operate in two modes:

• search (-s) - where the used does not a priori know what the pulse broadening time scale is, but can estimate the range of possibilty.

• deconvolve (-t) - where the user does know the pulse broadening time scale and just wants to deconvolve the profile to reconstruct the intrinsic pulse profile.

An example, where the user knows that the pulse broadening time scale is 20 ms, and has good evidence that the thin scattering screen is suitable, would be something like:

tauclean -t 20.0 -k thin -p 500 -o 440 900 --coherent [file]

where the period (-p) is 500 ms, the "on-pulse" region ( -o, defining the bounds of where the scattered power extends) is between bins 440-900, and the data have been coherently dedispersed (--coherent).

If the user did not know the pulse broadening time scale, then the above command would change to:

tauclean -s 15.0 25.0 1.0 -k thin -p 500 -o 440 900 --coherent [file]

where now the code will trial tau values from 15 ms to 25 ms in steps of 1 ms, producing a reconstruction (plus other diagnostics) for each trial. Only in this case are the "figures of merit" produced, which can be used determine which of the trial values are a better representation of the data. tauclean will invoke multiple processes (with the multiprocessing module) when searching.

Credit

If you made use of tauclean for your research, please give credit by

• Citing the original technique description paper: Bhat et al. 2003, ApJ, 584, 2

• Linking this GitHub page as a footnote or in the acknowledgments (and you could consider adding me as a co-author)