ASKAP Science Data Processor software - ASKAPsoft Version 0.24.0

Full description

ASKAPsoft, the ASKAP Science Data Processor, provides data processing functionality, including:

* Calibration
* Spectral line imaging
* Continuum imaging
* Source detection and generation of source catalogs
* Transient detection

ASKAPsoft is developed as a part of the CSIRO Australian Square Kilometre Array Pathfinder (ASKAP) Science Data Processor component. ASKAPsoft is a key component in the ASKAP system. It is the primary software for storing and processing raw data, and initiating the archiving of resulting science data products into the data archive (CASDA).

The processing pipelines within ASKAPsoft are largely written in C++ built on top of casacore and other third party libraries. The software is designed to be parallelised, where possible, for performance.

ASKAPsoft is designed to be built and executed in a standard Unix/Linux environment and core dependencies must be fulfilled by the platform. These include, but are not limited to, a C/C++/Fortran compiler, Make, Python 2.7, Java 7 and MPI. More specific dependencies are downloaded by the ASKAPsoft build system and are installed within the ASKAPsoft development tree. Specific to the Debian platform, after a standard installation of Debian Wheezy (7.x) the following packages will need to be installed with apt-get:

* g++
* gfortran
* openjdk-7-jdk
* python-dev
* flex
* bison
* openmpi-bin
* libopenmpi-dev
* libfreetype6-dev
* libpng12-dev

More information regarding the building, installation and running of the software can be found in the README file in the root of the file structure that forms this collection.

Source code can be accessed via the links in Related Materials section.

-----
The new release 0.24.0 contains numerous fixes and updates to both the ASKAPsoft processing software, and the pipeline scripts.

Key pipeline improvements are aimed at improving the speed and reliability of the overall processing:

The preprocessing stages (splitting, bandpass calibration, flagging, averaging, as well as the spectral preprocessing tasks of selfcal gains application and continuum subtraction) can be performed in parallel on short time chunks.
The self-calibration task can be split up into multiple slurm jobs, one for each of the imaging tasks, and one for each loop of source-finding/modelling/calibration. This avoids the occasional slurm time-out error that arises swapping between the tasks within a single slurm job.

There are also a large number of changes to the pipeline default parameters. These largely provide more relevant parameters for imaging recent 36-antenna observations. There is, in particular, a new single parameter that can turn on/off the spectral processing. It is recommended the list of new pipeline defaults is studied carefully and config files adjusted accordingly before using the new pipeline version.

For the processing software, the key changes are with the spectral imaging option of imager. It is now more robust for how the channels are distributed across the workers, and can better handle writers that have few or no valid channels (due, for instance, to frequency-frame conversion). The interface to the frame conversion has changed, with an additional 'lsrk' option to go alongside 'topo' and 'bary', and the interfaces to provide the input and output channels has changed slightly. The recording of the restoring beam, both in the image header and in the beamlog text file, is now more reliable.

There are also a number of other bug fixes, relating to bandpass calibration, mssplit, and mosaicking of already-mosaicked fields.

Please see the complete 0.24.0 release notes for more information.