Joint EMU/POSSUM Discussion #2

  • 10.00am-12.30pm AEDT, Mon 14 Feb 2011
  • Faraday Room, Radiophysics Laboratory, CSIRO Marsfield
  • Teleconference line: +61 2 6276 6622, Meeting ID: 8718


  • Julie Grant
  • Andrew Hopkins
  • Bryan Gaensler
  • Martin Meyer
  • Tim Robishaw
  • Emil Lenc
  • Larry Rudnick (via phone)
  • Shane O'Sullivan
  • Ray Norris
  • Shea Brown
  • Elaine Sadler (via phone, until 11.00am)
  • Catherine Cress (via phone)
  • Ron Ekers
  • Ettore Carretti
  • Ilana Feain (from 11.20am)


1000-1020Gaensler/BrownCoordination of milestones: POSSUM milestones, EMU milestones
1020-1050Sadler/MeyerOptions for low frequencies via FLASH and DINGO: DINGO presentation
1050-1120Lenc/RobishawCoordination of simulations
1120-1150O'Sullivan/LencCoordination of calibration
1150-1215NorrisCoordination of catalogues and data products: POSSUM data requirements, EMU data requirements
1215-1230Gaensler/NorrisDeliverables and date for next meeting

Coordination of Milestones

Not much overlap between submitted POSSUM & EMU milestone documents

  • POSSUM document has a lot of focus on RM analysis
  • Common EMU/POSSUM deadline: by June 2011, need to tell CSIRO what our data products will be by June 2011 and also need to finalise our quality control procedures by June 2011

What's not in the current milestones?

  • Need close coordination for next simulation run
  • EMU has milestone for dynamic range issues; need input from / coordination with POSSUM. Might want to simulate polarisation leakage? POSSUM to give rough measure of how well we expect to calibrate out


  • BETA observing strategy? ATCA pilot observations?
  • Submission of final survey strategy

  • Hopkins: does POSSUM care how EMU decides how to do the survey?
    • Weighting of beams for best polarisation may not be optimal for field of view, sensitivity (being explored by Tony Willis)
    • Calibration overhead associated with polarisation calibration
  • Norris: BETA observing (not otherwise on agenda)
    • Proposal to cover Circinus and Fornax fields with ATCA. But for EMU and POSSUM, not ideal fields because they have a lot of bright sources. EMU already has CDFS, for which 4 deg^2 has already been closely studied in total intensity and polarisation (via Chris Hales). Surrounding 30 deg^2 is covered by NVSS. When BETA observing starts, would like to do CDFS as well as Circinus and Fornax. (Don't need to re-observe CDFS field with ATCA!)
    • Source extraction: can DUCHAMP be modified to handle polarisation? Jeroen Stil working on this.
    • Regular item needed in future meetings on source extraction
  • Rudnick: Will EMU be putting together a milestone table that looks like the POSSUM one?

Options for Low Frequencies via FLASH and DINGO


  • Sadler: summary slide
    • FLASH-wide is 25000 deg^2, 700-1000 MHz, 2 hours per field (50% the sensitivity of EMU)
    • FLASH is not worried about uniformity of sensitivity across the field, because only targeting bright sources; EMU presumably needs uniform sensitivity
    • FLASH does not need polarisation calibration; POSSUM would want to carry extra overheads associated with this
  • Two models:
    • Just make the continuum data available to whomever wants it
    • Change survey parameters to make continuum data useful
  • Ilana Feain is in all of EMU, POSSUM and FLASH, and can coordinate this
  • Norris: crazy not to use the continuum data, but don't know what overheads are. For EMU, probably not much, just need to set up tiling; for POSSUM, could be substantial.
  • Ekers: to estimate overheads, first need EMU science case for lower frequencies. Will be confusion limited? Won't use the survey itself, will use it to calculate spectral indices of sources you've already found.
  • Sadler: Does POSSUM need the whole sky? FLASH asked for 10% of their survey time allocated early, to assess detection rate and refine future survey parameters. Suggest that any coordination with POSSUM (or even EMU) take place after we get this initial 10%.
  • Sadler: for BETA, FLASH will mainly look at calibration strategies, and will provide test sources that we have observed with ATCA. Only test observations for bandpass calibrations and observations of known observers, don't expect to detect anything due to lack of sensitivity. In the meantime, plan should be to keep talking and to share our memos / technical outputs.
  • Ekers: could FLASH do 850-1150 MHz to remove gap that POSSUM will otherwise incur in frequency?
    • Sadler: no, need to go as low as possible to maximise detection rate.
    • BETA experiment: consider both 700-1000 MHz and 850-1150 MHz to see what's best for polarisation?
    • Ettore: expect that 700 MHz will not be good for polarisation; 850 MHz will be a lot better.
  • Norris: In continuum mode, we have 10 arcsec resolution using long baselines. In HI, default is to only use central antennas. (Similar issues for WALLABY and EMU)
    • Sadler: we know where our HI sources are for FLASH, so don't need whole field, can possibly image at full resolution


  • Deep: 150 deg^2, 1120-1420 MHz, 500 hours/pointing; 150 deg^2 will not be a contiguous patch
  • Ultra-deep: 60 deg^2, 990-1290 MHz, 2500 hours/pointing; 60 deg^2 will not be a contiguous patch
  • These fields will have substantial multi-wavelength coverage from GAMA (VISTA, UKIRT, VST, SDSS, GALEX, AAT, Herschel, WISE, GMRT)
  • Choice of survey fields: probably five fields for deep survey (plus another two fields if ultra-deep not embedded in deep survey; five fields total if ultra-deep embedded in deep survey), not yet finalised. DINGO has been told performance will not be ideal within 10 deg of celestial equator. Most GAMA-II fields are at 0 dec or -30 dec.
  • Clear interest in continuum from some DINGO team members, but no-one working on continuum within DINGO
  • Compare specs of DINGO to original specs of EMU-Deep: DINGO has similar sensitivity (~1 uJy), wider field of view, reduced frequency coverage
  • Norris: Need one person who can be responsible for coordinating discussion of overlap between EMU and DINGO.

Coordination of Simulations


  • Lenc: most of our simulations have been looking at deconvolution issues, gridding. Haven't looked at FPA simulations as yet.
  • Current simulations use only flat spectrum sources. Next step is to add SED info from SKADS database.


  • Robishaw: will deliver report to ASKAP Computing Group on 16 Feb that will detail first set of ASKAP simulations, and utility for POSSUM.
  • May 2010: EMU and POSSUM were given 20 deg^2 simulation of same area, but with different numbers of pixels. EMU got a single band-averaged Stokes I map. POSSUM got I,Q,U in 32 channels.
  • Was difficult to determine what went into simulations. Jeroen Stil provided an input catalogue with a variety of source types, all with different numbers of components. About 2.3 million of these were simulated in 20 deg^2 field.
  • Comparison of input catalogue to POSSUM simulation and measured output: AGN and starburst galaxies match in RM, normal galaxies do not have correct RMs.
  • Performed RM synthesis of simulation cube: dirty Q and U cubes are 30 GB each.

Coordination of EMU and POSSUM simulations

  • Want to aim for a single EMU/POSSUM input for next round of simulations (report due on 1 June)
  • Will receive both a single-plane Stokes I image and a (I,Q,U) spectral cube
  • Ekers: Want simple simulations. Complicated simulations are a huge amount of work and are not that useful. Keep input simulations simple and don't worry about trying to put in every realistic input effect. But definitely simulate the effect of an FPA. (Although simulations won't handle this properly!)
  • Gaensler: main issue for joint coordination of EMU/POSSUM work on simulations is getting from final image to initial source list?
  • Hopkins: EMU effort on source finding is focused both on point sources and extended sources as two separate problems
  • Gaensler: two POSSUM products, POSSUM Polarisation Catalogue (PPC) and POSSUM Polarisation Atlas (PPA). PPC essentially starts with Stokes I and asks what polarisation is. PPA starts with blind catalogue of polarisation and asks what Stokes I is.
  • Ekers: do RM synthesis at polarisation of peak Stokes I pixel, or worry about potential offset polarisation? Suggest keep it simple for main catalogue and focus on polarisation properties of the peak - this is what previous pol catalogues have always ended up doing. But perhaps have a flag in POSSUM Polarisation Catalogue (PPC) that says peak polarisation pixel is not the same as peak Stokes I pixel?
  • Bottom line for coordination: need single input catalogue for POSSUM & EMU for next round of simulations.

Coordination of Calibration

  • O'Sullivan: summary slide
  • Concurrent pipelines: beam-dependent gains, one for bandpasses, one for polarisation
  • Global Sky Model: predict forward (updating in real time)
    • Need spectral index and polarisation information
  • ATCA project C2437: accurate full pol properties for 10 bright calibrators
    • Not enough for Global Sky Model - use S-PASS sources?
  • Beam weighting schemes (Tony Willis): fitting gaussian to overall beam reduces leakage, but suffer 10%-20% loss of sensitivity at centre (much worse at edges)
  • For EMU, will likely have bright sources in sky model fitted, subtracted and then added in again at end. For POSSUM, don't want Q and U subtracted out.
  • Feain: Do either EMU or POSSUM want the (time dependent) beam weights stored? This hasn't been stated as a requirement previously.
  • Ekers: Bottom line is that you can't be guaranteed high pol performance around the edge of the PAF. This would affect tiling pattern for EMU and WALLABY.
  • Meyer: Brad Warren is the person responsible for computing WALLABY's optimal tiling pattern.
  • Hopkins: Depends on how bad performance at edge is: if only 10% level, no worse than general variations over whole survey
  • Rudnick: issue is whether poor pol performance is at edge of each beam, or edge of whole field of view? Likely to have both issues. If the instrumental is bad at the edges of each FPA pointing, you might want to cut off each one for beam forming for POSSUM differently than for EMU.

Coordination of Catalogues and Data Products

  • POSSUM and EMU both want u-v data
  • POSSUM wants 300 channels while EMU just want polynomial solution
  • Quality control process: POSSUM further down the track on this so far than EMU.
  • Feain: a lot of quality control will need to wait for BETA and ASKAP data. But will have 30 deg^2 ATCA test fields to work on for quality control
  • Norris: POSSUM and EMU both plan to have final reports on quality control and cataloguing by June 2011. I and Lisa Harvey-Smith can easily merge the POSSUM and EMU reports.
  • Ekers: do you want to retain weather monitoring information? For AT20G this was hugely useful after the fact for quality control assessment.
  • Norris: By June, want a joint EMU/POSSUM quality control process set up in a rudimentary fashion, summarised in a joint document.
  • Norris: Also want quality control for ATCA pilot observations developed on same time scale

Deliverables and Date for Next Meeting

All of the following to have reports at next meeting

  • June: joint quality control and data requirements, and initial quality control for ATCA observations (Norris, Harvey-Smith)
  • June: coordinated input catalogue for next round of simulations (Hopkins, Brown, Johnston-Hollitt, Stil, Heald)
  • May/June: coordinated EMU & POSSUM milestones for Jul-Dec 2011
  • Next EMU/POSSUM meeting: Identify particular person responsible for talking to DINGO & FLASH about coordination on continuum, and develop science case document for EMU/POSSUM needing low frequencies (Ilana Feain will begin thinking about this)
  • May: first input on how polarisation weighting affects sensitivity & field of view, and discussion then can occurs between POSSUM/EMU/WALLABY on potential compromises/overheads that this implies (Tony Willis)
  • Next meeting: Wed 13 Apr, 10.00am-12.30pm, at ATNF.