2021-04-01

Phase 2 had compact and extended modes
New Phase 3 has new hardware, new signal chain
Luckily, will still have compact configuration in the short term
Paper on new pulsar is in press
- ICRAR will do a press release + news item, to coincide with publication (in a few weeks)
- Tentative date 20/21 April
- RB thanks all who helped with the paper!
- Lessons: new finds with MWA can rely on Parkes, etc for follow up
- First time pulsar has been timed with MWA -- an important milestone
- New GMRT and Parkes proposals in the works
- GMRT proposal is about scintillation studies at high Galactic latitudes -- successful!
MWA is back online since last Friday.
- Reason: new hardware procurement is delayed a few months
- An opportunity for us to cover more SMART survey obs's
- MWA will be in compact configuration for the rest of the year
- 1PPS is not there yet (as of this week), so timing might be compromised for incoming data over the next few months
Review of observations completed so far (see slides for info-graphic)
- Still in "shallow" search mode. Prospects for new finds are still good!
Shared risk proposal has been submitted
- Bhat, McSweeney, Swainston, Smith, Xue.
Name SM as "SMART Survey Project Scientist"

Data processing:
- "Quick check" 41 obs processed, 93 known pulsars detected
- "Shallow search" 23 obs searched (on Garrawarla and OzStar)
- Benchmarks: Table given in slides
- Used LOFAR's single pulse search software and found RRAT J0410-31
  - Vindication that LOFAR's software works well.
RB: J0410-31 is a HTRU discovery.
- Comforting that back-of-the-envelope calculations suggest that single pulse sensitivity is ~few Jy
RS: 7 sigma pulse? Is that LOFAR's detection sigma?
- NS: Not sure where that number comes from.
- RB: Lots not understood -- e.g. side/grating lobe detection? It's 10 deg away from pencil beam.
WvS: One of the plots was time vs DM, and it shows an interesting trace. What does it mean? Is it expected?
- NS: Not sure, but it's one of the less useful diagnostics in my experience.
- RB: Common theme: using LOFAR software -- maybe buggy, maybe not optimised. Question is, is it worth the effort to understand it very thoroughly?
- RB: could be good for student project!

MWA upgrade means that we will be served coarse channelised data instead of fine channelised data
If we want a homogeneous data set, we will need to do our own offline PFB
Opportunity to change software paradigm from "limited-mode pulsar backend system" to "flexible, exploratory research tool"
Modular design will greatly ease extension development down the line
Script-based interface, where individual steps are translated to a sequence of GPU kernel calls
Development period until June, then testing and commissioning, ready in time for MWA hardware upgrade.
WvS: How many people involved in development?
- SM: Me and Keegan. But it's not a complete rewrite (obviously already have a working beamformer to draw from)
- IM: My online beamformer is being currently developed as well. Perhaps combining our efforts, or at least liaising with each other may be beneficial to both.

Primary objectives:
- New data format
- Time stamping
- Frequency labelling
- Timing stability
- Polarimetric verification and leakage
- Calibration
- Full coherent beam sensitivity
Using Targets:
- J0034-0721, J1752-2806, J2241-5236, J0437-4715, J2145-0750, B0628-28, B0737-40, B0740-28, B1749-28, J0036-1033
Time frame for fifth survey campaign = Sept-Oct (if all goes well, fingers crossed)
Proposal was sent out to SMART collarboration (but only at the same time that it was submitted, due to compressed time line)
RS: How much is dependent on new hardware arriving on time?
- SM: Actually, delays in hardware are good for us.
- RM: Significant delays probably not likely, and in any case doesn't change our long-term plan.

GAN model (Generative adversarial network) not fully working as of the end of 2021
New model being investigated: Transformers
Problems with GAN: didn't work well with large data cubes as inputs
WvS: Two words of caution: (1) Not sure how well these work in noisy regimes, (2) Class imbalance problem
- IC: On (2), GANs offer natural solution by generating "true" candidates

Time scale for system coming online is a 1-2 months.
Will be able to invite collaborators to participate in candidate checking efforts easily.

May want SMART telecons on a more regular basis.
Have to continue scientific output.
- Three papers planned for 2021:
  1. Survey description paper (Bhat et al)
  2. Population study paper (Swainston et al)
  3. Low-frequency profile paper (Smith et al)
WvS: Single pulse timeseries will be available?
- No automation for that right now. Currently would have to process "manually" from raw data.
- WvS: Data outputs might be large, but perhaps only statistical information need be kept.
- WvS: We can revisit this idea if plucky students come along looking for science projects.
- RB: Should prove existence of "user base" first.