Transient detection with the MWA Tara Murphy on behalf of the MWA Transients KSP LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 2 Outline 1 The MWA 2 Transient detection 3 Current status Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 3 The MWA Transients team ˆ Roger Cappallo (Haystack) ˆ Shami Chatterjee (ATNF) ˆ Judd Bowman (Caltech) ˆ Bryan Gaensler (USyd) ˆ Lincoln Greenhill (CfA) ˆ David Kaplan (UCSB) ˆ Justin Kasper (CfA) ˆ Daniel Mitchell (CfA) ˆ Miguel Morales (Washington) ˆ Tara Murphy (USyd) ˆ Steven Ord (CfA) ˆ Billy Robbins (USyd) ˆ Brian Schmidt (ANU) ˆ Randall Wayth (CfA) . . . and other members of the MWA Scientific Collaboration. . . Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 4 The only science in this talk Image ref: Shami Chatterjee Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 5 The Murchison Widefield Array ˆ MIT, CfA, RRI, ATNF, UMelb, USyd, UWA, Curtin, ANU. . . http://www.haystack.mit.edu/ast/arrays/mwa Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 6 MWA – vital statistics ˆ 512 tiles comprising 16 diapoles each ˆ This gives 130 816 baselines ˆ Distributed over 1.5 km, with outliers out to 3 km ˆ Frequency: 80 − 300 MHz ˆ Bandwidth: ∼ 31 MHz ˆ Spatial resolution: 2 ˆ Spectral resolution: 1024 channels (30 kHz) ˆ Time resolution: 0.5 sec raw, 8 sec calibrated ˆ Field of view: 15 − 15◦ ˆ Point source sensitivity (1 mJy in 1 sec, 0.34 mJy in 1 hr) Ref: Lonsdale et al., 2009, submitted Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 7 The MWA has excellent UV coverage. . . Image ref: Lonsdale et al., 2009, submitted Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 8 . . . and is in a great location Image ref: CSIRO Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 9 The MWA has 3 key science projects 1 Epoch of Reionization (EoR) 2 Solar, Heliospheric, and Ionospheric science (SHI) 3 Transients science 4 Galactic and Extragalactic science (GEG) Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 10 MWA transient science packages ˆ All Sky Monitor (ASM) ˆ Search phase space: direction, timing, freq., dispersion measure ˆ Look for single pulse off-on-off ˆ Blind search in primary field of view ˆ Power of two image binning 8s → 16s → . . . → days ˆ Transient Lightcurve Analyzer (TLA) ˆ “Watch list” of single pixel positions ˆ Light curve saved at these positions at full time resolution ˆ ∼ 102 pixels in the FoV ˆ Allows more complex offline analysis of light curves ˆ Short & Long Term Synoptic Surveys ˆ Dedicated, periodic observations of the whole sky ˆ Beamformer Light Curve (BLC) Ref: Cappallo et al., Kerastari Meeting, 2007 Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 11 ‘Fast’ and ‘Slow’ transients with the MWA ˆ Raw data flows from correlator to RTS at 160 Gbit/s (time averaged to 40 Gbit/s) ˆ The Real Time System (RTS) will do calibration and imaging ˆ Output will be 16 GB image every 8 seconds (1024 × 1024) pix × 1024 chan × 4 pol ˆ Integrated images archived every 10 minutes∗ ˆ ‘Fast’ transients < 8 seconds ˆ ‘Slow’ transients > 8 seconds ˆ Longer timescales can be investigated using archival data Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 12 Beamformer light curve ˆ 512T will have voltage beams — perfect for targetted observations ˆ 32T unlikely to have voltage beams, but raw data is recorded ˆ Plan to perform blind searches in raw data ˆ Computationally expensive =⇒ GPU-based dedisperser ˆ Technological demonstrator as well as scientific interest Randall Wayth and others Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 13 The Transient Lightcurve Analyser ˆ We can’t archive all the imaging data ˆ ASM/pipeline will detect transients and save relevant data ˆ The TLA saves data on N sources at full resolution ˆ At 1 kB/s/src, N=100 is 3 TB after one year of observing ˆ Scheduler will maintain a list of priorities ˆ The TLA will run in continuous piggybacking mode Ref: David Kaplan, Justin Kasper Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 14 The ‘TLA’ in action Image ref: David Kaplan Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 15 Benchmarking a transient detection pipeline Light curves Data from RTS Source finding Classification Extract light curves Transient detection Archive Store Event handling Store 8 s Queue ˆ 2.16 GHz machine / 2 GB RAM (my laptop!) ˆ Implemented in Python ˆ Uses Miriad for most tasks ˆ Standard MySQL database ˆ VOEventNet for publishing Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 16 Extracting light curves from 8 second images Light curves Data from RTS Source finding Classification Extract light curves Transient detection Archive Store Event handling Store Queue ˆ Find all sources above flux cutoff ˆ Source fitting is probably too slow ˆ Store fluxes for all detected sources ˆ At least as a buffer ˆ Database growth rate: ∼ 48 Bytes per source (RA, Dec, flux, error, time-stamp) ∼ 1000 − 10000 sources per 8 s ∼ 16 GB of memory =⇒ ∼ 3 − 30 days to fill ˆ Need spatial indexing scheme Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 17 We are currently testing detection algorithms Light curves Data from RTS Source finding Classification Extract light curves Transient detection Archive Store Event handling Store Queue ˆ Experimenting with different algorithms (e.g. OIS) ˆ Source finding vs. direct image manipulation ˆ Testing on suite of simulated data Billy Robbins Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 18 Classifying transient sources Light curves Data from RTS Source finding Classification Extract light curves Transient detection Archive Store Event handling Store Queue ˆ Initial classification using MWA lightcurves ˆ More comprehensive classification using external resources Brian Schmidt ˆ Exclude known sources — TLA ˆ Pass to event handler ˆ VOEvent packet for each transient ˆ For each transient store ˆ Postage stamps ˆ Light curves Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 19 Detecting simulated transients from MAPS Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 20 Transient statistics from archival data ˆ Molonglo Observatory Synthesis Telescope ˆ 843 MHz, 45” resolution ˆ 3 major surveys since 1980’s: ˆ SUMSS – Mauch et al. 2003, MNRAS, 342, 1117 ˆ MGPS-1, MGPS-2 – Murphy et al. 2007, MNRAS, 382, 382 ˆ Aim to add limits to Bower snapshot rates analysis Image ref: Keith Bannister Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 21 First MWA transient test event received. . . Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 22 First results from MWA 26T - Pictor A Image ref: Randall Wayth Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 23 First results from MWA 26T - Puppis A Image ref: Randall Wayth Murphy MWA Transients LOFAR Meeting, 17th Dec 2008 

The MWA Transient detection Current status 24 Australian Square Kilometre Array Pathfinder (interloper) ˆ EoIs in Survey Science Projects closed Dec 15th ˆ VAST: An ASKAP Survey for Variables and Slow Transients (Murphy, Chatterjee et al.) ˆ A Fast Transients Survey with ASKAP (Dodson, Macquart et al.) http://www.atnf.csiro.au/projects/askap Image ref: Paul Bourke & CSIRO Murphy MWA Transients LOFAR Meeting, 17th Dec 2008