Astronomical Image Processing System (AIPS)

The NRAO Astronomical Image Processing System (AIPS) is a software package for interactive (and, optionally, batch) calibration and editing of radio interferometric data and for the calibration, construction, display and analysis of astronomical images made from those data using Fourier synthesis methods.

1.5 GHz VLBA image of SS233. (Credit: Amy Mioduszewski, NRAO)

AIPS has been the principal tool for display and analysis of both two- and three-dimensional radio images (i.e., continuum "maps" and spectral-line "cubes") from the NRAO's Very Large Array (VLA) since early in 1981. It has also provided the main route for self-calibration and imaging of VLA continuum and spectral-line data. It contains facilities for display and editing of data in the aperture, or u-v, plane; for image construction by Fourier inversion; for deconvolution of the point source response by Clean and by maximum entropy methods; for image combination, filtering, and parameter estimation; and for a wide variety of image and graphical displays. It records all user-generated operations and parameters that affect the quality of the derived images, as "history" files that are appended to the data sets and can be exported with them from AIPS in the IAU-standard FITS (Flexible Image Transport System; see newsgroup sci.astro.fits format. AIPS implements a simple command language which is used to run "tasks" (i.e., separate programs) and to interact with text, graphics and image displays. A batch mode is also available. The package contains over 7 Megabytes of "help" text that provides on-line documentation for users. There is also a suite of printed manuals for users and for programmers wishing to code their own applications "tasks" within AIPS.

In partnership with the scientific community, we:

Design and development of the package began in Charlottesville, Virginia in 1978. It presently consists of over 4300 files containing 1.46 million lines of text. These comprise over 400,000 lines of documentation and on-line help in over 1300 files, and almost a million lines of text in over 2300 Fortran and C source files. It contains over 350 distinct applications "tasks," representing well over 60 person-years of effort since 1978.

AIPS is maintained by a very small group. They are responsible for the code design and maintenance, for documentation aimed at users and programmers, and for making the code available to non-NRAO sites. Since its release under the Free Software Foundation's General Public License in mid-1995, its availability via the internet has been made considerably easier, and for that one release (15JUL95), about 150 sites downloaded the software, and conservative estimates of the number of machines running this one version of AIPS alone indicate about 450. NRAO currently offers AIPS installation kits (ready-to-run binaries) for most of the currently available UNIX systems, with updates available semi-annually. The number of sites running some version of AIPS is probably in excess of 250, possibly more than 300.

Astronomical observations at radio wavelengths allow scientists to address fundamental questions about our Universe such as:

In 1983, when AIPS was selected as the primary data reduction package for the Very Long Baseline Array (VLBA), the scope of the AIPS effort was expanded to embrace all stages of radio interferometric calibration, both continuum and spectral line. The AIPS package contains a full suite of calibration and editing functions for both VLA and VLBI data, including interactive and batch methods for editing visibility data. In 1996/7, considerable effort was expended to ensure that AIPS would be capable of handling the data from Orbiting VLBI satellites such as VSOP. For VLBI, it reads data in MkII, MkIII and VLBA formats, performs global fringe-fitting by two alternative methods, offers special phase-referencing and polarization calibration, and performs geometric corrections, in addition to the standard calibrations done for connected-element interferometers. The calibration methods for both domains encourage the use of realistic models for the calibration sources and iterated models using self-calibration for the program sources.