Searching for fast radio transients with the VLBA Adam Deller, 28 June 2011 

The V-FASTR experiment  Pathfinder for the CRAFT high time resolution survey on ASKAP  Commensal observation on the Very Long Baseline Array, all projects  Membership is a CRAFT subset  Curtin: Steven Tingay (PI), Randall Wayth  NRAO: Walter Brisken  ASTRON: Adam Deller  NASA JPL: David Thompson, Kiri Wagstaff, Walid Majid 

The Very Long Baseline Array • 10 x 25m dishes • 8,000 km baselines • Frequency range covering 0.3 - 86 GHz • Synthesized beam ~mas; coherent search not possible! 

The V-FASTR experiment Existing (V-FASTR adds few % overhead) 1 3U server; ~$6k (2x6 core, big disks) 

DiFX/VLBA-specific challenges  Incoherent filterbank spectra (< 1 MHz freq. resolution) comes out as UDP packets from multiple processing nodes @ ms time resolution:  Some packets get lost  Data must be sorted into time order  During re-ordering, other unwanted signals such as strong RFI and noise calibration must also be excised 

Data processing frequency time raw filterbank data bandpass, tcal corrected data Also: boxcar averaging to match longer pulse widths 

V-FASTR properties  RFI robustness (different sites); detection thresholds can be lowered (just like ARTEMIS on LOFAR)  Ability to accurately localize transients:  Detection implies sufficient S/N to fringe fit  Even S/N ~ 8 gives delay precision <10 ns  With 8,000 km baselines, that gives sky localization to sub-arcsecond precision  Imaging ability (scattering, short term motion / near field) 

V-FASTR properties  Baseband data still readily available (like hours of “transient buffer board”); can be automatically captured and reprocessed at higher time resolution  Competitive sensitivity to other transient searches:  Within primary beam 9 x VLBA antennas = 43 m equivalent for the incoherent search, but 75 m equivalent when coherently combined in follow-up analysis  Out of primary beam considerably more sensitive than e.g. Parkes multibeam 

Initial V-FASTR results  Single pulse detections of B0329+54 and the Crab 

Initial V-FASTR results  Single pulse detections of B0329+54 and the Crab The single detected Crab giant pulse 

Initial V-FASTR results  Automated dump of baseband data and reprocessing at higher time resolution + generating cross-correlations for imaging Raw visibilities (no dedispersion) 

Initial V-FASTR results  Automated dump of baseband data and reprocessing at higher time resolution + generating cross-correlations for imaging Post-correlation dedispersed (ready to image) 

Initial V-FASTR results  Imaging of B0329+54 from a single pulse, no external calibration (position correct to several arcsecond accuracy) Cleaned image from ~5ms data 

Processing challenges / progress  VLBA is not designed for high quality single dish work - gain variations, local RFI rarely interfere with VLBI but are obvious to V-FASTR  Flagging known RFI a priori is useful  (Somewhat) surprising results so far:  “Robust” estimator, where several of the highest and/or lowest value antennas are discarded, strongly outperforms simple sum  High pass filter (tau ~ 0.1 second) eliminates slow gain variations at cost of reduced sensitivity to longer transients 

Processing challenges / progress 

Processing challenges / progress 

Processing challenges / progress 

Current status  Two papers in press:  Experiment description, sensitivity estimates  Algorithm evaluation  Multi-purpose proposal approved for repeated 330 MHz observations of M81/M82 field:  Deep P-band continuum + slow transients (image) + potential fast transients  Long time series of stable setup  Now running full-time on VLBA 

Future work  Improved follow-up analysis  Automatic higher time resolution recorrelation of the dumped baseband data  Trial coherent dedispersion (pre-correlation) for candidates which are unresolved in time  User-friendly visualisation of the results (waterfall plots, frequency averaged dedispersed data, DM sensitivity plots)  Automated dissemination of candidate results (daily email digest, statistics) + automated update to V-FASTR webpage  Detection/limits paper (~3 months obs)