Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

SMART meeting, 30 Jul 202

Present:

  • Ramesh Bhat
  • Keegan Smith
  • Nick Swainston
  • Sam McSweeney
  • Zhongli Zhang
  • Susmita Sett
  • Marcin Sokolowski
  • Isaac Colleran
  • Mengyao Xue
  • Ryan Shannon
  • Willem van Straten

Update overview (Ramesh):

  • Survey ramping up
  • 3 student projects connected to SMART survey
  • Have been awarded time on OzStar (1.5 million hours), looking ahead to Garrawarla
    • 2-3 times speed increase on OzStar (compared to Galaxy)
  • Adopted strategy of “quick look, first pass” pipeline constrained by PhD timelines
    • Limited DM steps, 5-10 mins of data
    • New pulsar (J0036-1033) gives us hope!
    • Also important for “quick” turnaround of data processing so that fields can be re-observed if necessary
    • 96 known pulsars detected with Keegan’s pipeline
    • 5 fields have been processed (although not sifted for candidates)
  • Alternative search strategies are being explored (cf Susmita & Marcin’s imaging pipeline)
  • ADACS proposal initially turned down, but CIC resources have been allocated
  • We have to take advantage of all remaining Phase 2 compact configuration time for data collection
    • There will be a proposal call later this year
  • At least 60% data is in the can
    • By the end, ~3PB, 0.5 million beams, 7 million core hours for 1 round of processing
  • Follow-up of J0036-1033 possible with ~10 existing archived VCS observations (both phase 1 and phase 2, extended and compact)
    • For localisation, regridding/re-beamforming straightforward task w/out need for re-observing
    • Sub-arcminute localisation unlocked
    • Efforts to understand polarimetry -- for new pulsar (and more generally)
    • Attempts to detect new pulsar in imaging
      • So far unsuccessful
      • Suggests that pulsar is “low luminosity pulsar”
      • Proposal has been put in for Parkes, and will be put in for GMRT (deadline tomorrow)
    • Will be basis for short paper

Nick’s update:

  • Just about finished processing (quick look) 9 observations.
  • Been through 1000’s of candidates
  • New Nextflow pipeline:
    • Has improved efficiency, uses containers
    • Can now process 1 obs per week
    • Installed on Garrawarla, Ozstar, and (almost) at SHAO
  • ML part of pipeline helps a lot, but still produces many obvious noise candidates
  • Working to decentralise candidate ranking (so that it’s not just Nick doing it)
  • Also working on Single pulse candidates
    • Likes LOFAR single pulse software/pipeline
    • Have tried SHAO’s pipeline (STEP?) -- almost working, but indications that it may not be the best choice
    • >15 candidates per beam → lots of candidates!!

Marcin’s update:

  • Pipeline:
    • Offline correlation (1 sec integration), RFI flagging, calibration, and other standard MWA processing, output = Stokes images from XX, YY
  • Looking for J0036-1033 in imaging
    • RMS 12 mJy (I) and 2.5 mJy (V) images show nothing convincing at pulsar’s location
    • Pointing towards pulsar is <5mJy
    • Also, GMRT image of same area of sky shows nothing (which gets down to 2.5 mJy/beam)
    • Can see other pulsars in the field with same imaging techniqu

Sam’s update:

  • LOFAR’s tree classifier
    • DM curve, profile, freq-vs-phase, time-vs-phase features
  • Generative adversarial networks
    • Isaac’s ongoing project
  • All available ML tools will feed into database app (with web interface) for human inspection.
  • Working with CIC to develop database/web app.

Discussion:

  • Fancy tricks for localisation? Follow up with Ryan’s FRB localisation techniques with ASKAP (Ryan to email Nick some Python code). May get down to ~1/10 of a (tied-array) beamwidth in precision.
  • Imaging can help weed out nights when ionosphere is very active and moves sources around
  • Test gridding method on known pulsar position to get errors
  • Not likely for Nick’s pulsar to be in a sidelobe -- sidelobe detections have characteristic frequency dependence.