2021-09-09

Owned by Ramesh Bhat
Last updated: Sept 17, 2021

Present:

  • Sam McSweeney (SM)
  • Ryan Craft (RC)
  • Susmita Sett (SS)
  • Nick Swainston (NS)
  • Marcin Sokolowski (MS)
  • Shi Dai (SD)
  • Ramesh Bhat (RB)
  • Willem van Straten (WvS)
  • Ryan Shannon (RS)

Agenda:

  1. Project overview and general updates (Ramesh)
  2. Processing update and candidate sifting (Nick)
  3. New system  - status and commissioning (Sammy)
  4. ML exploration, using SGAN and similar (Isaac)
  5. Database migration plans to ODC (Sammy/Nick)
  6. Briefing on papers in prep. (Sammy/Nick/Ramesh) 
  7. Discussion time (all)

Discussion:

  • Ramesh’s welcome

    • Introductions by SD and RC

  • Project Overview (RB)

    • First pulsar discovery published in ApJL (Swainston et al. 2021)
      • Discovery after 1% of data processed
      • Dilpreet’s artwork featured
      • Good media coverage
      • Thanks to everyone for their help to reach this milestone!
    • Data collection update
      • Since last meeting, 10/14 observations from 4th campaign taken
      • Then… SYSTEM FAILURE!
      • Total survey now ~75% complete
      • Will have to wait for MWA upgrade to continue
    • 2nd pulsar discovery
      • J0024-1931, “Keegan’s pulsar”
      • Discovered in last week of May
      • Went for GMRT and Parkes proposals, but initial “rough” localisation made follow-up difficult.
      • Pulsar hunting sessions revealed a stronger detection in another beam! Original detection in grating lobe
      • July GMRT proposal included updated localisation
    • MWA G0057 proposal for 2021A (commissioning proposal)
      • Shared-risk
      • Full time allocation
      • Will be testing new software pipeline with it
    • “Pulsars and Fast Transients” made an official Science Working Group
    • Funding for 1.1 PB storage under “datasets of national significance” (Lead: Greg Sleap)
    • OzSTAR allocation of 250 kSU for Q3-Q4 2021
    • Data Central to take over completion of  Webapp
    • Parkes proposal = full time allocated
    • 3 papers currently in drafting stages
    • More meetings - monthly? (e.g. 2nd Thursday at UT 8:00?)
    • Discussion:
      • WvS - Proposed monthly meeting time and cadence works for me
      • RS & SD - ditto

  • NS’s update:

    • Recap of data processing progress
      • 123 known pulsars detected
      • 35 obs searched
      • 7 obs classified
      • 2 pulsars discovered
    • WvS: Why 10 mins?
      • NS: Motivated by quick look
      • RB: Limited by available kSU on OzSTAR, also there’s only 1 Nick
    • SD: Running single pulse search?
      • NS: Yes, using the one from LOFAR, including ML
      • RS: Seen Keegan’s pulsar in single pulse search?
      • NS: No, not yet.
      • RB: It should be there, but the pipeline has some issues to sort out 

  • Sam’s update:

    • Software paradigm has shifted since last time:
      • Creating a library instead of a scripting language
      • Called “vcsbeam”
      • Incremental development by cleaning up the existing code, so that easier to test for correctness
      • Heavily relying on new mwalib library, provided and supported by MWA Ops team.
      • Library-isation almost complete
    • Offline (forward) fine PFB
      • Kernels have been written, not yet tested
      • Have simultaneous legacy & MWAX data now to test it against (as of a couple of weeks ago), although the data may have problems because running both the legacy system (for the last time before it was retired!) and the new MWAX system in parallel put a lot of stress on the system.
      • Commissioning of the MWAX system still in the near future!
    • Added MPI for channel splicing will improve efficiency of HPC use.

  • Isaac’s update:

    • Honours project is using a semi-supervised GAN.
    • Retraining portions of the (and perhaps the whole) network proving  necessary.
    • Currently sitting near ~70-80% accuracy.
    • Gaining confidence that we’ll have a usable product by the end of the project (~November) that is at least as good as the current LOFAR system that we’re using.

  • Sam’s 2nd update:

    • Keegan’s pulsar shows extremely interesting sub-pulse drifting behaviour
    • Clearly has multiple drift modes, but P3 changes (gradually) so much during individual modes that characterisation of the modes by P3 (or drift rate) is not possible.
    • Actually have many more positive detections in archival MWA data spanning a long time period -- getting a timing solution seems possible
    • Excellent example of the benefits of longer dwell times -- to catch the bright intermittent ones!

  • Sam’s 3rd update:

    • There has been an excellent agreement of the theoretical grating lobe prediction and the tied array beams in which B2327-20 was detected
    • This may open up new strategies for finding true candidates by looking for candidates in nearby beams that have similar periods and DMs.