WSClean Version 2.3

Released 2017-03-01

Summary: The most prominent new feature is local RMS thresholding using either an automatically generated RMS map or by providing an RMS map. When using multi-scale clean, a component list using Gaussians can now be requested. This release also fixes some larger bugs that had been introduced in the Hogbom/generalized Clark clean transition, so I recommend to upgrade.

New features:

  • Local RMS thresholding, options -rms-background, -rms-background-window and -rms-background-method (SF #102, SF #105).

  • Use channel weights when fitting spectral polynomials. This helps in particular to properly ignore missing channels, which would formerly cause bad fits (SF #107).

  • Option -save-component-list for outputting the Gaussian component lists in multi-scale clean.

  • Improve speed of multi-scale clean, by setting the clark masking threshold lower than the stopping threshold in the subminor loop optimization.

  • Added a -multiscale-shape parameter to either select Gaussians or a tapered squared function.

  • WSClean now ships with an example program wsuvbinning to produce a csv file with binned visibilities, properly normalized to values in Jy (requested by C. Spingola).

  • Fits files now contain a sum-of-visibility-weights keyword WSCVWSUM which can be used to calculate the nr vis per uv bin (this is required to use the wsuvbinning tool).

Bugfixes:

  • Rectangular images (i.e., width != height) caused cleaning to fail.

  • Crash when performing full XX/XY/YX/YY imaging, causing missing MFS files. Reported by T. Franzen.

  • Bug in -intervalsout; interval index would not increase, causing each next index to overwrite the previous index.

  • Segfault when chosing an odd image size and enabling beam correction, reported by M. Iacobelli. Caused by FFTW not properly supporting unaligned r2c fft.

  • Segmentation fault when auto-masking with multiple output intervals.

  • Some improvements to the IDG connection.

  • Work on a polynomial channel fitter as described in the WSClean multi-scale clean paper, which in the end will allow better multi-frequency fitting. This is not finished, see ticket (SF #110).

  • On some platforms, parsing any integer command line parameter resulted in an exception (fixed between 2.2.0 and 2.2.1).