OSKAR-1: Beamforming simulation for the SKA aperture array
B. J. Mort, Oxford e-Research Centre (benjamin.mort at oerc.ox.ac.uk)
F. Dulwich, Oxford e-Research Centre (fred.dulwich at oerc.ox.ac.uk)
S. Salvini, Oxford e-Research Centre ( stef.salvini at oerc.ox.ac.uk)
K. Zarb Adami, Oxford Astrophysics (kza at astro.ox.ac.uk)
M. E. Jones, Oxford Astrophysics (mike at astro.ox.ac.uk)
A. E. Trefethen, Oxford e-Research Centre (anne.trefethen at oerc.ox.ac.uk)
S. Rawlings, Oxford Astrophysics (sr at astro.ox.ac.uk)
Update (18/6/2012): The status of OSKAR
What is the SKA?
The SKA (Square Kilometre Array) will be the most powerful radio telescope ever built. Currently being designed by a large international collaboration, it will be operational by 2020. The SKA will enable researchers to answer some of the most fundamental astrophysical questions.
The SKA will be an aperture-synthesis interferometer, consisting of many connected elements (stations) spread out over large distances (thousands of kilometres for the SKA). Each station will include an aperture array of up to 100,000 omni-directional antennas: their signals will be combined digitally to form beams in multiple directions at the same time (a beam is a pointing direction with associated width or field-of-view).
More information can be obtained from the SKA web page.
OSKAR at a Glance
The EPSRC-funded OSKAR project is a collaboration between Oxford Astrophysics and the Oxford e-Research Centre. The computational challenges posed by beamforming for very large aperture arrays, such as envisaged in SKA stations, are formidable, and, before OSKAR, no tools were available to carry out any in-depth studies. In a nutshell, OSKAR is a computational tool for the investigation of the scientific and engineering concepts underlying beamforming for the SKA station aperture arrays.
OSKAR allows a hierarchical beamforming strategy, for any number of levels, from one upwards. For example, for a two-level, stations would be organised in tiles, or sets of antennas. Tile beams would be formed first, the same on every tile, and then these would be combined to form station beams. In any case, only a subset of frequencies/beams would be used in practice: all the unwanted ones would be removed from the computational pipeline as early as possible. In the case of one-level, the antenna signals would be combined directly into station beams.
The weights used to combine signals into beams depend on the tile/station geometry, required corrections, etc. and vary much slower than the sampling rate. Hence, they can be computed separately at a relatively low cost.
OSKAR provides two main modes of operation: simulation and beam pattern computation. In the first mode, we simulate the functioning of the aperture array for a given sky and required beam/beams over a period of time (including Earth rotation). In the second mode (beam pattern) we study the response of a beam to object outside its line of direction: this is essential information for all radio telescopes.
OSKAR consists of two separate software components, a front-end and a back-end.
- OSKAR GUI ("front-end")
- This is implemented using portable libraries, hence it is cross-platform, i.e. it can be run on any system, from Linux to Windows; on users’ PCs, for example. Through it, users set up all the simulation parameters, including the model sky, the array layout and the beams of interest. The OSKAR GUI also allows the visualisation of the output beams.
- OSKAR Computational Engine ("back-end")
- This can run on a supercomputing cluster. It is based on MPI and enables the user to model any conceivable array configuration with sets of beams at multiple levels. Great care has been taken to abstract the communication layer and isolate it from the processing modules to achieve maximum flexibility.
Because of its highly modular design, OSKAR should be viewed as a framework allowing experiments with different antenna responses, station configurations, novel beamforming algorithms, external data corruptions (noise), calibrations/corrections, etc.
The computational pipeline can be checkpointed to allow hot-restarting at any level. This also allows components of OSKAR to be ported to and tested on alternative architectures such as GPUs, FPGAs and massively multi-core processors without the need of full-scale hardware and of porting the whole of OSKAR.
- User information (guide with tutorials, animations)
- Software (binaries and source tarballs)
- Software information (developer guide, APIs)
- Users Guide
- This includes instructions on how to obtain, install and run OSKAR, and some tutorials.
It is available both as PDF (2.3 MB) or HTML.
- Animation (Tutorial)
- An animation of an OSKAR full run is available (requiring Flash). Both versions below use a reduced colour palette for size reasons.
View an 800 x 600 animation.
View a 1024 x 768 animation.
- Source Code
- The source code for the front-end and the back-end can be downloaded here:
Download the OSKAR front-end
Download the OSKAR back-end
- Linux Repositories
- Binary packages are available from specific OSKAR repositories. The link below allow you to add such repositories. For installation, please consult the User Guide.
Ubuntu: 8.04, 8.10, 9.04, 9.10
- Windows front-end binary
- The front-end is available for Windows XP/Vista/7.
- Developers Guide
- This contains more information about the file formats and data structures used by OSKAR.
It is available both as PDF or HTML.
- API Documentation
- The API reference documentation provides a detailed description of the software for developers. It is available both in Hyperlinked PDF and HTML.
OSKAR GUI (front-end) PDF, HTML.
OSKAR simulator (back-end): PDF, HTML.
WikiThe OSKAR Wiki site has more information about OSKAR.
The OSKAR software is released under a modified BSD licence. It consists of a number of components: