HEIG Meeting (online) - Thu, Sep 14 2023 @ 13:00 UTC

• Attendees :
  • Mark Cresitello-Dittmar, Mathieu Servillat, Catherine Boisson, Francois Bonnarel, Raffaelle D'Abrusco, Ian Evans, Janet Evans, Matthias Fuessling, Tess Jaffe, Bruno Khelifi, Mireille Louys, Jutta Schinabel, Fabian Schussler
• Regrets :
  • Ada Nebot

Zoom

Actions

• JE: Send doodle poll for a mtg in ~1 month (Oct 19 @13:00 was a possibility; Oct 11 bad)
• CTAO: prepare a presentation/discussion on Instrument Response Function (IRF) data and how it relates to the event data (for next mtg)
• IE: Prepare a presentation/discussion on ARF & RMF and how it relates (for next mtg)
• DM project leads: Make an event list & supporting products available to the group to review; try to map that data to the Cube model using the Chandra example Mark presented
• Mark CD: Follow up with a group or 2 to review their data and model - offer input and also take in possible changes to current cube thinking
• Mathieu: What makes an event list unique? ie: What content should users expect to exist in an Event List instance? Let's try to answer that at the next mtg
• All: Plan for side meeting at the next interop in Nov - hybrid

Meeting Notes

• MCD presentation of the IVOA Dataset & Cube model and as an example, the Chandra event list mapping to the model
• Q&A Highlights: Good discussion and participation overall
  • General discussion that the model can map to specific event lists for each project - let's try the exercise and see if we're missing something??
    • BK: event lists containing time, energy, direction
    • BK: why a separation of Image and SparseCube?
      • MCD: They are structurally different, NDImage contains pixel axes, while the SparseCube is a collection of Points. They both share the same DataAxis elements in the model, so interaction/combination should be possible. We will need to see the example.
      • Variable size bins?? - yes
      • FB: it may be possible to describe a DataProduct with some sparse and some binned axes as a SparseCube with a DataAxis of NDImage type.
    • ...
  • Discussion on Dataset (header info) & DataProduct (the Data itself) & use of those terms in the model
    • MCD: there are somewhat contradictory statements in the current model which stem from conflating the concept of Dataset as a modeled object and the serialization of that object.
      • Dataset is only the metadata
      • a dataset may be a single file (eg: Image, Spectrum), or a tar file or directory containing multiple files
    • MS: at CTA, the DataProduct is considered to contain both the metadata, and data.
    • This topic can be discussed more, perhaps in the DM group with a goal to find concensus on names and relations between 3 objects
      • The collection of metadata: currently Dataset
      • The structural definition of the data: currently DataProduct
      • The container holding both: currently not modeled.
  • Discussion on Supporting products and how they fit in (IRF, ARF, RMF) - examples for next mtg
    • BK: we need to include the instrument response function, it is very important to the interpretation of the event list
      • IE: described how Chandra uses a rather simple IRF and performs the conversion from PHA to Energy during processing, storing energy as a static column.
      • MCD: possibly can use Transform
        • IE: asserted that the process for the conversion is rather complex, and further requires Spectra.. so may not be well suited for Transform.
      • TJ: notes that HEASAR standards for Spectra require a 'channel' axis.. is this supported
        • IE: the Chandra pha/pi columns are equivalent to channel
      • BK: for them, the relation between IRF and is much looser.. may be 1-1 with an event list or 1-* (or *-1 ?.. I (MCD) don't remember what was said exactly).
    • TJ: question if the model has been compared against the OGIP standard in full
      • MCD: Work to date has focused on the Chandra event list example, Ian says Chandra meets the OGIP standard. We should review OGIP more generally and a HEASARC example.
        • post-meeting side note: if we map an OGIP standard to the Cube model, then any implementation which understands an OGIP compliant serialization is a Cube model implementation.
  • MS: Raised Question - What are the key Measurements expected in High Energy Event Lists? eg: Time, Position, Energy
    • Further discussion for next mtg
  • MCD: Raised Question - Do our various event lists include errors?
    • BK: have optional error elements
    • JS: neutrino data has positional errors usually, sometimes energy
  • Other questions/topics raised in the presentation not yet discussed.
    • We need to refine modeling of non-measurement types of data from example products (eg: quality bits)
    • Do we need to model the concept of 'data subspace'.. for Chandra, the GTIs are a component of this concept as are any REGION filters applied to the event list.
    • Do we need to support virtual data in version 1?
      • this was not discussed, but the content of the general discussion leads me to believe that this is needed.