1

National Radio Astronomy ObservatoryNRAO Users Committee Meeting – May 20-21, 2008

Common Astronomy Software Applications (CASA)

Crystal Brogan and CASA Team

Given by: John Hibbard

2

CASA (Common Astronomy Software Applications)

• What is it?– Suite of applications for the

reduction and analysis of radio-astronomical data (derived from the former AIPS++ package)

– Algorithms written in C++– Interface python/ipython– Plotting in matplotlib– Qt based Viewer &

tablebrowser

• Key Differences from aips++– Python/Ipython interface provides

powerful and well known scripting language

– Task interface to functionality– Full in-line help– Smaller package size– Faster startup time– Performance improvements– Smaller memory footprint– Expanded functionality– Scientist-written documentation

3

CASA Staff– Devel. group leader: Brian Glendenning (Interim)

• Joe McMullin transitioned to ALMA Test Lead in Chile Oct. 2007• Permanent replacement being sought• McMullin is now one of the primary consumers of CASA

– Development team• Currently 11 FTE

– NRAO (EVLA/ALMA) 7.5 + 1 fixed term employee (until 12/31/08)– Canada 2, NAOJ 0.5 (split w/ Pipeline)

• Planned additions– NRAO Operations: McMullin replacement

– NRAO/NAASC: 2 (1 Hired, 1 Offer/Declined)

– East Asia ARC: 2 (Hired)

– ALMA EU Computing: 2

4

CASA Scientific Oversight– Internal scientific oversight

• CASA Project Scientist: Steve Myers• ALMA Subsystem scientist: Crystal Brogan• EVLA Scientist: Gustaaf van Moorsel• NRAO E2E Project Scientist: Ed Fomalont• NRAO Users Group (~ 10 scientist in CV and SOC)• ALMA EU/EA ARC Scientists• ALMA AIV & Commissioning (starting, led by Joe McMullin,

CASA single dish focus Lewis Knee)

“CASA Cabal”:

Meets weekly

Determines priorities

Sets development targets

5

Response to UC Questions

• “We had requested continued updates of release details including new tools and algorithms being developed. Can CASA now support the incorporation of single dish spectral observations with interferometric data in order to properly deal with missing short baselines?”

6

Beta ReleaseRelease cycle: updates quarterly, major releases annually Beta Release - Nov. 15, 2007

Focused current limited user support capacity on: Training ARC and construction staff who require CASA for their

work Initial community input from ALMA & NRAO science advisory

committees CASA being used at the ATF to analyze data Beta Release (Patch 1) – April 1, 2008 Beta user group widened to include several ALMA post-docs

and EU ARC node scientists Beta Release (Patch 2) – June 15, 2008

Improvements aimed at supporting CASA tutorials at NRAO synthesis summer school

A CASA Advisory Group (CAG) will be created with ~15 members drawn from committed external users to provide more formal and constant feedback

CASA publically available from my.nrao.edu (person must read and accept first a “Beta release policy statement”). We reserve the right to disable if load becomes debilitating to development

7

Initial Beta feedback • Insufficient platform support

– Adding platforms is not hard, but time consuming adds testing overhead (1-2 days/platform/patch)

• MacOS X/Japanese version has problems• Trouble installing on system with previous AIPS++ installation• Need more image analysis functionality• Various plotting improvements needed • Single dish improvements (more integrated tasks, etc)• Documentation must be improved• Incorrect third party WCS & unhelpful error message• Need deconvolution with user supplied beam• Finding restoring beam in Miriad FITS image

8

Patch 1 Improvements• 64-bit support for RedHat Enterprise• Non-root linux installation• Polarization calibration• Image analysis (immath,imcontsub,imstat)• Significant single dish taskification• Improved plotting speed• Assorted bug fixes

Patch 2 Improvements• More flexible calibration options for multiple spectral windows• Combination of “clean” and “mosaic” tasks• Syntax based control of global/local parameter behavior• ALMA simulator more robust, user friendly• Viewer based region file creation• Image plane source and spectral Gaussian fitting capability (enhancement will be needed)• Logger and terminal messaging improvements • More Linux platforms (SUSE, Ubuntu, Debian)• Assorted bug fixes

9

Brief Summary of Current Capabilities(after Patch 2 release)

• Data Import– VLA (EVLA) archive– External ALMA and EVLA fillers complete– UVFITS (BIMA, CARMA, SMA, ATCA etc, though some subtleties with headers etc)

• Flagging – UV-plot based including time, channel averaging – Viewer flagging– Manual flagging

• Calibration– Polarization– VLA flux density calibrator images– Spline fitting– Flexible combination of multiple spectral windows

• Imaging– Single field and mosaics– Interferometer plus single dish combination (3 methods, one experimental)– Multi-scale clean (somewhat experimental)– Interactive clean boxing– Analysis includes image math, statistics, image plane fitting, spectral fitting

• Operating Systems– RHE 4, 5 (32 and 64 bit)– Fedora 6, 7, 8 (32 bit, some 64 bit),– SUSE, Ubuntu, Debian (and perhaps others with non-root install) – Intel Mac OSX 10.4.7 and later

10

Production of Scientific Images in CASA Becoming Routine

An extended radio counterpart of TeV J2032+4130 in the Cygnus OB association. VLA 3.6 cm continuum 5 point mosaic, D configuration, multi-scale clean. Butt et al. (2008)

Green contours show SMA 12CO (2-1) integrated intensity superposed on a GLIMPSE 8 µm image of the Infrared Dark Cloud (IRDC) G19.3+0.07. Six-pointing SMA mosaic imaged in CASA – calibration of SMA data coming soon. Brogan et al. (in prep).

11

B-field vectors in Jupiter magnetosphere. VLA 6 cm full stokes continuum, D configuration, with polarization and gain calibration and imaging in CASA

Production of Scientific Images in CASA Becoming Routine

12

almasimos: ALMA Simulator

R. Indebetouw

13

An Important CASA Milestone

• ALMA data format ASDM CASA filler completed

• CASA routinely being used to reduce data at the ALMA Test Facility at VLA

• This summer CASA will produce FITS files from raw EVLA data format for initial correlator testing in AIPS

Edited & Calibrated (gain, bandpass, & flux) spectrum of 4 ATF ASDMs

Zoom in on amplitude and phase showing good agreement between 4 datasets

14

Planned Future developments• Number of supported platforms

– Key problem with porting is matplotlib/tk; fussy about host OS

• Improve UI (GUIs)

• Satisfactory speed demonstrated with large datasets (“Terabyte Initiative”)

• EVLA– Wide-band, Widefield imaging (requires extensive effort)– WIDAR correlator data support– UV-editing (needs to be more automated; sophisticated)

• ALMA– Heterogeneous array imaging (baseline based PB correction)– More sophisticated system temperature/opacity corrections– More sophisticated simulation capabilities – Algorithms for baseline, delay, and pointing determination

» Low priority; provided by TelCal subsystem in ALMA

• Post observation corrections (baseline length etc.)

15

Planned Future developments (2)• Plotting speed and flexibility• Viewer flagging and autoflagging improvements• Viewer improvements

– Annotations; Channel maps; Better postscript output

• More sophisticated image analysis tasks– Make them interactive

• Improvements to calibration solution visualization

16

CASA Terabyte Initiative Begun• Led by S. Bhatnagar• Start with 10% = 100 GB (raw data) size

– Corresponds to 24 minutes of peak data rate or 4 hours of expected average

– 15M visibilities, 512 channels, VLA B-array (10 km) at 6cm– 4kx4k output image– Run time 16 hours

• 65% I/O, 35% CPU (imaging)

• 2 TB I/O load

• Timing broadly consistent with AIPS (22h)– Details of major/minor cycles make exact comparisons tricky

• I/O major limiting factor More than parallelization is needed to solve this problemHigh performance computing investigations ongoing

17

The Way Forward• These development targets will receive highest priority in coming cycles

• Patch 3 will come after a longer than usual development cycle (@ Oct. 1) to allow some of these harder problems to be addressed

• In future, one of the scientific developers will effectively “sit out” of the more mundane development activities to allow for a steady flow of advanced algorithm development

– This CASA developer will also be more closely tied to commissioning efforts

• It is essential to find a permanent replacement for Joe McMullin; please convey to any potentially interested colleagues

• A new CASA Advisory Group will provide focused feedback on package in addition to daily input from general users – let us know if you want to be on it!

• General user feedback will continue to be submitted via the CASA helpdesk, there will be a NAASC study to determine the best long term helpdesk solution