This demonstration, requested by the Computing Division of ALMA under the general requirements of the ALMA Science Advisory Committee, and in the frame of a general re-use analysis of existing off-line analysis software, is aimed at determining whether AIPS++ can satisfy the off-line and pipeline data reduction needs of ALMA. The main questions that are naturally asked are:
How AIPS++ can be adapted to reduce data of an instrument for which it was not initially designed?
How long is the learning curve for developers who have sufficient experience in the processing of millimeter data, but no experience at all the AIPS++ programming environment?
Can we perform an end-to-end experiment on actual, real-life millimeter-wave spectroscopic data?
The demo should perform an end-to-end analysis of Plateau de Bure data inside AIPS++. This should include the following steps:
Read the raw data in an agreed FITS format (this could be the foreseen data format for the ALMA Test Interferometer).
Perform atmospheric calibration to convert data into TA* scale
Use atmospheric path length corrected data
Reduce band pass calibration obtained in a strong celestial point source
Reduce phase and amplitude phase calibration using quasars
Transfer the phase calibration from 3mm wavelength observations to 1mm wavelength observations.
Image continuum and spectroscopic data
Deconvolve single field images
Besides the software and data products, the principal deliverable from test is a report containing short, motivated answers to the well-defined questions posed above. If any answers are undecided, then the pro's and con's should be clearly exposed.
The proposed duration of the test is 9 months. The goal is to have an answer by April 2002, and report to the ALMA project (Computing Division).
The team will consist of:
Two AIPS++ developers from NRAO, working part-time on the project, who would visit IRAM for periods of several weeks at a time over the course of the test.
An IRAM software engineer working half time on this project, who could visit NRAO for several weeks to become familiar with AIPS++ development and use. The proposed person at IRAM is Dominique Broguière.
Two or more IRAM astronomers with good astronomical experience of Plateau de Bure data reduction, they should provide a continuous expertise of how the PdB data is handled for actual astronomical projects. They should have the responsibility of selecting the test data.
Representatives of the SSR will be involved in the test monitoring and evaluation.
IRAM will provide some logistic help for stays in Grenoble. NRAO will provide same for stays in Socorro.
The following milestones are envisioned for the test:
Phase I (end November 2001):
Agree on the export FITS format for visibility data [1,2].
Provide a copy of the CLIC/GILDAS code to the AIPS++ project [1]
Select two representative IRAM datasets covering the data reduction scope outlined above. [1]
Provide these uncalibrated data in the agreed FITS format.
Reduce in CLIC/GILDAS [1]
Reduce the test data in CLIC/GILDAS as described in the data reduction scope above.
Export interim copies of the data after each calibration step in the agreed FITS format.
Export the calibration solutions for each step.
Export the final images in FITS format.
Provide a basic text log of the CLIC/GILDAS reduction sequence
Provide a point of contact to answer queries regarding CLIC or the reduction sequence.
Reduce in AIPS++ [2]
Reduce the same test data in AIPS++ as described in the data reduction scope above.
Provide documented AIPS++ scripts for end-to-end reduction of each test dataset.
Formally compare the final image product and calibration solutions between the packages.
Phase II (end April 2002):
Reduction of other data
Select additional IRAM datasets meeting the data reduction scope defined for the test. [1]
Reduce these data in AIPS++ using the scripts provided from Phase I. [1]
Provide support in correcting any AIPS++ defects arising from these tests [2]
Inter-compare with CLIC/GILDAS results. [1,2]
Optional phase III (extending beyond April 2002):
Evaluate the scalability of AIPS++ from IRAM to ALMA data
Measure the performance of AIPS++ on simulated ALMA datasets using 64 antennas.
Notes:
[1] - IRAM/ALMA task
[2] - AIPS++ task
[1,2] - Joint task
The following communications mechanisms will be used:
Monthly teleconference between relevant people. Date to be set by agreement.
Archived, public mail exploder: aips2-alma@nrao.edu (to be set up by AIPS++). We recommend that most if not all email go via this exploder.
Agreed by:
Tim Cornwell, AIPS++ Project Manager
Robert Lucas, Chair ALMA SSR
Brian Glendenning, Head ALMA Computing
2001-06-18