 News |
FAQ
|
| Search
|
Home
|
| Getting Started | Documentation | Glish | Learn More | Programming | Contact Us |
|
| Version 1.9 Build 1367 |
|
T.J. Cornwell, NRAO
1998 January 19
Progress continues towards a beta release in March 1998 and a limited public release in June 1998.
In Single Dish support, we continued our support of the use of AIPS++ by the Green Bank Telescope engineers. For example, opacity measurements at Green Bank are now analysed in AIPS++; see:
http://www.gb.nrao.edu/~jmcmulli/tipper/tipper.html.
The dish program continues to accumulate new capabilities and refinements, such as sine-wave fitting, persistence of results, improved data presentation, etc.
In Synthesis support, we started parallel development by a number of people, overseen and coordinated by Athol Kemball. Such parallel development is a good test of the completeness of the synthesis code. Kemball started on an implementation of cross-calibration. This required some relatively small modifications to the philosophy of the existing calibration software, as well as some changes to the class interface. This work is proceeding satisfactorily and we anticipate usable applications within the next couple of months. On the imaging side, the work has been split in a number of directions: Cornwell was to work on the user interface, Marson on the component models (especially important for WSRT processing), Briggs on wide-field imaging, and Brouw on a design for mosaicing. The first two have proceeded satisfactorily. The command line user interface for the synthesis code is now that of version 2 of the AIPS++ Distributed Object system. Closer integration with the rest of the system has lead to a number of very nice features, such as the ability to specify phase tracking centers in a range of coordinate systems e.g. ra-dec, az-el, galactic, super-galactic, position of the sun or planets, etc. Component lists are now editable via a graphical user interface written by Noordam, and integrated into the system by Marson during his recent visit to WSRT. Part of Briggs' work on wide-field imaging was taken over by Cornwell, who implemented multi-field cleaning, including very general phase-shifting. True wide-field imaging, similar to that in the SDE dragon program, now only awaits support for geometry conversion, which Brouw is scheduled to work on. Brouw's work on a mosaicing design required a demonstration of mosaicing using the existing classes, something that is now in progress.
For visualization of synthesis data, Noordam has continued to work on his Glish-based visualizer, with help from Wieringa to write C++ support classes. Testing at NFRA shows that this application is still too slow and uses too much memory. Hence some tuning is needed. Similar tuning was needed in the development of the image analysis application, where after tuning, the speed was comparable to that of existing code in other systems, with the addition benefit that other, non-traditional, access patterns could be used at relatively little cost (a key requirement for AIPS++ code).
In Measures, we continued adding capabilities to the system:
In Glish support, our main activity was to issue version 2.6 (the first major release outside the project for about 3 years and the first since NRAO took over control of Glish). This required an overhaul of the 250 page manual, reflecting many changes and additions in capability that have been made during this time. Following the release, we proceeded to start the addition of string and file input/output handling. In AIPS++ Infrastructure, our work continues to be driven by the needs of applications development. We have performed the long-awaited overhaul of the Lattice classes. The overhaul, made after gaining considerable experience writing code with the existing classes, has simplified use of the classes. We have also added support for sub-lattices, and a LatticeExpr class, allowing numerical processing with Lattices. We expect that the latter will soon be made available to users as an Image calculator.
As an important part of the infrastructure, we continue to develop a graphical user interface for standard AIPS++ Distributed Objects, based upon the tk widgets now bound to Glish. Part of this is a facility, now available, to track and report visually on memory usage by the various clients started by AIPS++. In addition to the GUI, we have also developed a design for a simple parameter-setting CLI that will allow users to bypass Glish for some operations. We plan to implement this in the next quarter.
In FITS support, we did the following:
In Image Analysis and display, we started initial applications development using the image display library, a joint undertaking of ATNF (Tom Oosterloo) and BIMA/NCSA (John Pixton). Both Tom and John continue to work on debugging the library (Tom on his own time since he has left the project), responding to bugs reported by Harold Ravlin of BIMA/NCSA. We expect this development effort to increase in the next few months when David Barnes joins the ATNF AIPS++ group.
In the System area, we finished a port to the IRIX native compiler, and investigated the new egcs compiler (Cygnus version of the FSF g++ compiler). We implemented a scheme for consolidating templates into groups, thereby cutting down on build time. As part of this effort, a number (25%!) of dead unused templates were removed from the system, also improving the build time.
In the area of Parallelization, Young transferred to an NCSA-funded position, dedicated to parallelization of AIPS++ code. Briggs and Roberts of NCSA are also targeted to work on parallel algorithms. While this work is primarily targeted for massively parallel computers, it should also benefit inexpensive systems with a small number of processors. If networking speeds are sufficient, we will allow transparent access of NCSA hardware from Charlottesville and the AOC. The first parallel applications will be available for the first limited public release this summer.
In Documentation, we added one person, Kate Weatherall, devoted 50% to AIPS++ documentation. She has worked on technical editing of the Glish manual, a proposal to restructure the AIPS++ Web site, and, jointly with Bob Hjellming, a plan for an on-line, HTML-based AIPS++ Newsletter. We are proceeding with the goal of releasing a first edition of the Newsletter by February 1.
In Management, we started the Quality Assurance Group as described in the AIPS++ Operations Plan (AIPS++ Note 205). This group headed by Ralph Marson currently consists of three people: Code Cop (Marson), Rules Boss (Brouw), and Chief Tester (Noordam). The mission of the QAG is to act as independent means of assessing and ensuring quality in products of AIPS++.
Testing of the last release by our testers has been much less that we had hoped. We are now embarked upon preparing the next beta release (1.0), which we believe must have both improved user interfaces and more functionality to attract new users. We anticipate having working versions of various standard GUIs within a month. In addition, we are working on a simple parameter-setting shell, that we expect will satisfy the needs of users for whom the GUI is overkill. Turning to applications, the development of functionality in synthesis processing is now sufficiently rapid to give us some confidence that the flag ship application (imager, renamed to sky) will attract new users.
The release schedule is now:
While there are other demands on the project, most importantly the pressing need to support the use of AIPS++ at WSRT, we believe that it is important to adhere as closely as possible to this release schedule. Previous delays have been ascribable either to slow development or to a change in strategy provoked by testing.
In Single Dish, we will add a number of features to the dish program in preparation for testing:
In Synthesis, we will develop mosaicing within the sky application, add sky support for components, and complete an image geometry conversion machine that is needed for wide-field imaging. Depending on the rate of progress on the conversion engine, we may be able to implement wide-field imaging. We will continue to improve the user interface, and make refinements as suggested by our alpha-testers. We will produce an initial cross-calibration capability, according to the design put forward by Kemball.
We will provide support, as needed, for the effort now underway to get WSRT/TMS Setup functions implemented in AIPS++. This will be a two-pronged effort: some functions will be implemented soon using the basic tools available in Glish, and some will be later implemented using the cross-calibration framework now being worked on by Kemball. We will work to improve the performance of the Glish-based visualizer developed by Noordam and Wieringa.
In Image Analysis and Display, we will develop an image calculator based on the new LatticeExpr classes. We will increase development of image display applications using the new image display library.
In Infrastructure, we will complete the auto-GUI and parameter-setting shells described above. We will implement a state-saving API for Glish functions. We will add support for region specification to the Lattice and Image classes. We will develop an image coordinate conversion machine as needed for wide-field imaging.
In Measures, we will implement user-specified corrections to emphemerides, as needed, for example, for comet observations. We will continue the integration of Measures throughout the system, most particularly via the TableMeasures classes and in the Glish binding of Distributed Objects.
In Glish support, we will complete the addition of support for string and file handling.
In the System area, we will implement a way of tagging good builds of the development system for preservation for use by alpha testers. We will enable checkin and checkout of Tables. We will investigate ways of restructuring the unix directory layout of AIPS++.
In the area of Parallelization, we will start work towards parallel deconvolution, gaussian fitting, and visibility gridding applications.
In Documentation, we will release a first edition of the AIPS++ Newsletter. We will restructure the AIPS++ Web site, with the goals of making navigation easier and improving the presentation of information thus found.
In Management, we will hold the second meeting of the Scientific and Technical Advisory Group, scheduled for February 9-10 in Socorro.
We will prepare for the third beta release, scheduled for March.
This quarter the ATNF had 3 people working in AIPS++. These are Neil Killeen (25-75%, also local manager), Wim Brouw (90%),) and Mark Wieringa (30%). Tom Oosterloo left the project and is now working for ESO in Italy. He has, however, been able to do some AIPS++ work. David Barnes, who will join us in February 1998, has been able, in between his thesis work, to work on some enhancements for the Parkes' Multibeam system. The output of this quarter has been somewhat lower for the regular team members. Our revered leader spent about half the quarter on a, partly busman's, holiday, while Mark was kept busy with introducing the several new computing hires in Narrabri. Wim spent a large fraction of his time organising and running the 1kT international technical Workshop. All three also took some end-of-year holiday and attended an ATNF staff retreat.
In general the one active system (Solaris/gnu) gave little trouble over the quarter. Unfortunately the trials (by Mark Calabretta) to get the SGI native port and the, recently announced, egcs free software port that was supposed to run under IRIX working were unsuccessful till now. The egcs system was not even built itself, while the SGI native port, even after the arrival towards the end of the quarter of the makedefs from NCSA, did also not succeed. The latter's failure could be due (as could maybe even the egcs port's failure) to the 64 bit port that was attempted. The 32-bit ports could maybe succeed, but would, a.o. things, entail the building of all support libraries in 32-bit mode (currently they are available in 64-bit mode: the machine's design width).
Neil visited Dwingeloo on his holiday, and worked fruitfully with Ger to get the Lattice expressions working to a stage were they could be checekd in. Ralph visited Epping for 5 minutes, and discussed with Wim the option of having regular QAG phone discussions. They will start towards the end of January. Tom visited the ATNF for an ESO Symposium.
Most of Mark's aips++ time was spent this quarter on the MeasurementSet DO functionality required for Jan Noordam's data visualizer. He implemented:
Wim spent all but no time on the mosaicing software, which is his next major aips++ task. He spent time, especially earlier in the quarter, on:
-0; abs(-0)
+0; and catered for
it in some output formatting
Neil's time was spent on
Although formerly not attached to any AIPS++ task, he has continued his work on some aspects of the display library.
The basic classes (like canvasses and the classes that do administration for these canvasses) are implemented now and to a large extent have also been debugged. The functionality offered at the moment are the Canvasses (with very extended control over colourmaps) and a number of classes that 'organise' the display (worldcanvasholder, animation, zooming). Also the basic control mechanism (restriction) is implemented and tested. One display class is implemented that enables display of pseudocolour image sequences of datacubes. A test program that displays movies of a datacube has also been written.
David will join the ATNF in February 1998. He made almost no actual contribution to the AIPS++ code in the quarter, but made changes needed to make the Multibeam software accommodate the internal changes in the FFT server and ArrayMath code (i.e. the speed up of the median method). It may, further, be interesting to note that extensive improvements were made in the Multibeam processing throughput, both due to internal aips++ improvements, and some due to time that he has spent programming. In summary, these changes have offered a total speed up of around 40%, so we are now running easily in realtime, using only one instance of the process-ing software. Secondly, the entire system is now far more robust, again due to internal aips++ changes and our own modifications; the frequency of LiveData crashes has dropped easily below one per 50 hrs, where it used to be one per two hours!
During the past quarter we had 6 people (3.5 FTE) working on AIPS++: Dick Crutcher (20%), John Pixton (50%), Harold Ravlin (100%), Doug Roberts (75%), Dan Briggs (100%), and Peter Teuben (5%). Accomplishments during the past quarter are described below under each person.
Dick Crutcher served as local AIPS++ project manager and as BIMA director in charge of our participation in AIPS++. During this quarter negotiations with NCSA about continuing funding for AIPS++ were concluded. Our long-term situation is the following. The Illinois Astronomy Department will fund Ravlin, the University of Maryland funds Teuben, and NCSA will continue funding for Roberts and Briggs. For now, Pixton continues at the 50% level, with his display library knowledge to be passed on to a new person in the NCSA visualization group during the next few months. Beginning next October, the level of effort for the display library will drop to maintenance from the development level. How AIPS++ visualization and image analysis work will progress with the loss of Osterloo and Pixton to the project is a matter of considerable uncertainty.
John Pixton has focused on the AIPS++ display library. His main effort was to finish those parts of the library for which he was responsible and fix bugs reported by Harold Ravlin, who has started working with the library. He also carried out a small study of the Visualization Tool Kit system (which will be the basis of most of NCSA's future visualization software development) and its possible compatibility with the AIPS++ Display Library.
Harold Ravlin administers our AIPS++ workstations, which takes about 1/3 of his time. He continued to spend some of his time on Aipsview development, maintenance, and bug fixes. New features included: (1) Made the title, xlabel, and ylabel movable on profile plots in the same way that they are movable for image plots. (2) Added the ability to display linear axes as offsets from a user selectable position within the image. (3) Added 'captions', extra labels (like the X & Y labels) the user can position on the image window. (4) Added support for profile locking. (The user can 'freeze' a profile so it isn't updated when the cursor changes). (5) Added support for FITS files with CD matrix keywords. (6) Started porting Aipsview to the SGI Origins. (There is still a problem with compiling under the 7.2 C++ compiler). He also began working with the new AIPS++ Display Library, and reported a number of bugs to Pixton, who fixed them. He also installed the new version of glish on our Suns, but got compile errors on the IRIX system.
Doug Roberts finished porting the AIPS++ system to the NCSA Origin2000 computer using the SGI native C++ compiler. He attempted to install AIPS++ under g++ on the Origin2000, but the gnu software had been improperly installed by NCSA and he was unable to proceed. He has been editing the Aipsview manual to reflect functional changes that Harold Ravlin has been putting into Aipsview. He has tested the AIPS++ calibration, imaging, and deconvolution routines against the same ones in MIRIAD by processing the same BIMA data set completely through the two systems; he found them to give the identical results within the noise. He has also written a users' manual on processing BIMA data through AIPS++ to help BIMA MIRIAD users transition to using AIPS++. The spectral-line Beta version of AIPS++ has been installed on the Illinois Astronomy computer system. He has given two talks to the BIMA group here at Illinois on the background of AIPS++ and specifically how to use it to process BIMA data; the talks include a demonstration script to calibrate, image, and deconvolve BIMA data in AIPS++.
Daniel Briggs arrived in Champaign-Urbana in mid-August, joined the Radio Astronomy Imaging group, and established computer presence on various local machines. He brought up the SDE package on the RAI SGI Onyx as a testbed for development efforts on the Origin 2000. He trained in Socorro from 20-Sep-97 to 11-Nov-97 in basic AIPS++ design methodology and infrastructure specifics. This resulted in finding and documenting significant number of system problems in process, obtaining fluency in glish as demonstrated by the minor modules sh.g, stopwatch.g, and improvements to catalog.g. He started work on the geometrical conversion engine, HGEOM. This is a fundamental part of many imaging algorithms needed by AIPS++ and by the wide field imaging task in particular. He wrote a first cut at a parallel infrastructure document to coordinate parallel development and code control efforts within the greater AIPS++ framework. Finally, he attended a number of parallel-processing meetings and consultations with the NCSA performance engineering group, in preparation for development of pilot parallel standalone AIPS++ demonstration programs.
Teuben continued to work on the bimafiller, in particular filling of the mosaicing tables, and proper support of multi-source visibility files. These should be ready for the next beta release.
The local project members are Ger van Diepen (GVD), Michael Haller (MXH), Jan Noordam (JEN, local manager), Friso Olnon (FMO).
Because of the WSRT upgrade, AIPS++ synthesis reduction lies more on the critical path for NFRA than for most other consortium partners. In particular, the various WSRT setup programs must be implemented in AIPS++ in the first half of 1998.
Commissioning of the new WSRT backend (DZB) and on-line software (TMS) has finally started in earnest this quarter. The uv-data are stored directly into an MS by means of a 'dzbfiller'. From there, the uv-data is usually converted to a format that can be read by existing packages for inspection and further reduction. The role of the uvf visualiser has been smaller than was hoped, because of its slowness. This problem is being addressed, both in hardware and software.
Friso Olnon developed a prototype class for total power normalization of WSRT data and their inclusion in the DZB and WSRT fillers. The class has been released to the DZB/TMS group for further development.
Friso checked in the first version of ms2fits. This NFRA application program converts the MeasurementSets produced at the WSRT into FITS files with binary-table extensions. The resulting FITS file is meant for local use (archiving) only, and not necessarily readable by AIPS. Essential is that no header or data information is lost in the transfer. The application will be reviewed and tested in cooperation with the JIVE and TMS/DZB people.
Ger van Diepen has provided support to the TMS group, in particular in explaining how to use tables and TaQL.
NFRA places a high priority to the use of parametrised model components as an integral part of the Source Model. Jan Noordam has checked in a gui for user interaction with source model components, using the basic componentlist DO by Ralph Marson. The latter has made a productive two-week visit to Dwingeloo in december, which resulted in a joint proposal for the further development of components during 1998. Although stage I of this plan has been completed, we will only be able to start using components after completion of stage Ia.
Friso Olnon cut short the upgrade of the wsrtfiller, because of the urgency of other WSRT-specific work, and because of the upcoming upgrade of the MeasurementSet. The checked-in filler fills a MeasurementSet from WSRT archived data, which are accepted by the AIPS++ beta-test processing modules. For continuum (DCB) datasets the filler only writes a single subband (out of 8) for the moment. That subband of course can be chosen. The filler does not yet handle multiple pointing positions (for mosaicing), and it does not store IF data.
Ger van Diepen spent most of his time on the Lattice and Image classes. A large fraction of the Lattice code has been restructured. As a result the code is simpler and it is easier to add other classes to the Lattice framework. Two major additions have been done:
As a spin-off the Array classes have been changed. The major change was the removal of the variable origin. It used to give a lot of worries and hardly any benefit.
The glish interface to the table system has been improved. It is possible to get (or put) a slice of an array in a table.
Michael Haller completed Aips++ module TableMeasures and is waiting for it to be reviewed. He has also worked on a proposal for new a new AIPS++ module BTrees. This has been suspended in favour of work on urgent local (NFRA) applications
``Regular'' maintenance of the local AIPS++ site has cost Friso Olnon less than half a day per week. Some problems with the user documentation have not yet been solved (or seriously tackled).
Since begin December, the responsibilities of local site manager have been taken over by Michael Haller. He has also updated the local Linux systems.
Several NFRA Software Engineering Group (SEG) workshops have been presented by AIPS++ group members. Their subjects were C++ code and how to use debuggers and tools like purify.
Eelke Klein, a student from the HIO Leeuwarden, has configured QAC++ and created a few scripts for it. It'll make it possible to use QAC++ on a regular basis to inspect the code and to derive some metrics.
The core NRAO AIPS++ group is now Cornwell (100%), Garwood (100%), Glendenning (100%), Kemball(100%), Marson (100%), McMullin (100%), Schiebel (100%), Uphoff (100%), Weatherall (50%), and Young (100%). In addition, as described above, Murphy of the CV computer group also contributes. Peter Barnes will join us in June 1998 to work in the synthesis initiative.
Tim Cornwell's primary responsibility is Project management. His coding activities focus around the synthesis package. In the last quarter, he has:
Bob Garwood's primary responsibility is to oversee the single dish work which currently is focussed on the DISH environment as well as supporting the GBT. Over the past 3 months (Oct through December) he has done the following work:
Brian Glendenning works on general AIPS++ development and management in a large number of areas. In the past quarter he has:
Athol Kemball directs the synthesis development effort. During this quarter he worked on the following items:
A design was produced for an extension of the current synthesis code to support full cross-calibration. This term is used to imply the ability to derive calibration information from a selected subset of uv-data, which can be transferred, interpolated and applied to another selected subset of uv-data. This includes the possibility of incremental calibration. This document has been distributed and discussed within the AIPS++ project as well as with some beta testers.
Work was started in the later part of this quarter on the implementation of a prototype cross-calibration system consistent with the design discussed above. This work has concentrated on the implementation of the key components of the cross-calibration system. This includes calibration selection and interpolation, as well as the formalization of calibration table conventions. This work has proceeded by modifying the existing imager classes directly in a manner that is consistent with future expansion to more advanced calibration features.
A significant effort was made towards a revised MeasurementSet definition to fully support VLBI processing, recent synthesis requirements, and general changes suggested since the initial definition. A draft document has been distributed within the project and the proposed changes are under general discussion. Implementation is planned for February this year.
A basic proposal was initiated for an interactive, non-GUI parameter setting shell, which could be invoked at the CLI to execute DO methods. This work is continuing and will be finalized early in the next quarter.
Work has continued on the classes developed for AIPS/AIPS++ interoperability in the previous quarter and will continue into the next quarter. This has involved revisions to allow the code to be included in the next public release and planning for expanded functionality.
Ralph Marson works on synthesis code and as code cop. During the last quarter, he worked on:
These improvements along with the update to current AIPS++ coding standards, and the switch to a faster FORTRAN FFT library (FFTPACK is now used exclusively rather than selectively) have resulting in speed improvements by a factor of two or three for one dimensional FFT's. One dimensional FFT's are at the heart of the routines which do transforms on large multi-dimensional arrays.
Joseph McMullin's main responsibilities are managing the Green Bank installation of AIPS++ and supporting AIPS++-related GBT needs.
Darrell Schiebel basically spent the entire quarter working on Glish. In general, this involved fixing bugs, adding new functionality, and preparing a public release of Glish 2.6. This release was the major milestone achieved during the quarter. There has not been a public release of Glish since shortly after AIPS++ adopted it for use within the project, and there have been many improvements since then ( see http://www.cv.nrao.edu/glish/manual for the entire list ). In preparing for the release, Darrell had to overhaul Glish's 250 page manual, ensure that the auto configuration scripts (based on GNU's "autoconf") worked properly with minimal user effort, fix all of the major outstanding bugs, and create a home page for Glish (http://www.cv.nrao.edu/glish/). Glish 2.6 was released in November. It was important to have a public release so external users (primarily at MIT and ODS) could be brought more into sync with the version of Glish which AIPS++ uses. Since the NRAO is now the official maintainer of Glish, it was also important to produce a release emanating from NRAO.
Some of the bug fixes and functionality improvements which Darrell added to Glish during the quarter are:
Beyond these major changes, Darrell also worked on many other bugs during the quarter. These involved things like fixing memory leaks, fixes to Tk widgets, fixes to internal event handling under specific circumstances, improvements to built-in functions, fixes to ``const'' variables, and misc. consistency fixes.
Jeff Uphoff works on single dish processing and also the AIPS++ system. In the last quarter, he did:
Kate Weatherall joined us November 14 to work on documentation for AIPS++. She has:
http://www.nrao.edu/~kweather/monkey/newsletter/newsletter.html
Wes Young has worked in a number of different areas. Over the last quarter he:
The principal activity for the next quarter will be support of beta testing, and preparation of the next release.
In Single Dish support, we did continue development of the Single Dish analysis program, with the goal of making this part of the third beta release. We did continue work towards the formalization of the SDFITS standard, providing a draft standard and some example SDFITS files.
In Synthesis support, we did increase our activity considerably:
In Measures, we did complete the addition of support for Measures in Tables.
In Glish support, we did continue to fix bugs as they occur in testing. We did issue release 2.6 of Glish.
In AIPS++ Infrastructure, we did continue to track the needs of applications programmers. We did not finish the improvements to the table browser. We did not integrate history and logging. We did not finish the DO Version 2 GUI support including support for plug-in components. We did complete optimization and cleanup of the FFTServer class.
In Visualization and Image Analysis, we did do some minor testing and application development using the display library.
In the System area, we did not complete the port to the KAI compiler.
In Documentation, we did not yet make progress on the intermediate level documentation requested by our beta testers (cookbooks and migration guides). We did review our Web presence.
In Management, as the first stage in implementation of the Operations plan, we did set up the Quality Assurance Group.
In this section, we give the names of people in the various AIPS++ groups and the nominal fraction of time allocated to AIPS++.
The ATNF group is: Neil Killeen (25-75%), Wim Brouw (90%), and Mark Wieringa (30%). Tom Oosterloo's position is to be filled by David Barnes in mid February.
The BIMA/NCSA group is: Dick Crutcher (20%), Dan Briggs (100%), John Pixton (50%), Harold Ravlin (100%), Doug Roberts (75%), and Peter Teuben (5%).
The NFRA group is: Ger van Diepen (100%), Michael Haller (100%), Jan Noordam (25%), and Friso Olnon (50%).
The NRAO group is: Tim Cornwell (100%), Bob Garwood (90%), Brian Glendenning (100%), Athol Kemball (100%), Ralph Marson (100%), Joe McMullin (100%), Pat Murphy (30%), Darrell Schiebel (100%), Jeff Uphoff (100%), Kate Weatherall (50%) and Wes Young (100%).
Thus, in aggregate, we have 24 people contributing just over 17 FTEs to the AIPS++ Project. Of these, 12 are employed by NRAO, and contribute 9.6 FTEs. The numbers for the other partners are: ATNF 3 and 1.45, BIMA/NCSA 5 and 3.10, NFRA 4 and 2.75.
