Child pages
  • 24 August, 2016
Skip to end of metadata
Go to start of metadata

post-LSST2016 weekly meeting.

 

Scott:

  • Lots of development on GalSim.  Should soon have all of the physics effect PhoSim has.
  • Broker community needs something to test on.  If we push hard on AlertSim we can get people to adopt it.
  • Hack sessions were very productive.

 

Rahul:

  • Deblending: DM is looking for people to test 'our' deblending software.  Community is developing independent deblending algorithms.  Forecasting effects on weak lensing.
  • Survey strategy whitepaper: no clearly defined cadence for Deep Drilling Fields.  Community does not believe OpSim is incorporating white paper suggestions for Deep Drilling Fields.  Kem thinks this is not the case (we are incorporating community requests into OpSim).  Kem talked to Neil Brandt about this.  May have clarified this issue.

Chris:

  • Community/Project communication still not optimal.  Not enough time for everyone to talk to everyone else.  Should the split be more explicit: first two days are project meeting; second two days are science meeting; explicitly point out who needs to stay for science meeting.  Basically: there were too many project-only sessions during last half of the week.
  • End-to-End GalSim: a lot of work being done.  Need to figure out how to unify it.  People are even thinking about how to do raytracing through the telescope optics.
    • Descriptions of layout of optics and electronics are stored in many different places throughout our software.  Should consolidate that.
    • Need to think about how to package LSST-specific modules for GalSim.
  • DESC would like rotational dithers.  Adding it as an afterburner to OpSim (random rotation after each filter change).  In the future, there may be a request to put rotational dithers into OpSim.
    • Kem: we know that this is coming.

Kem:

  • Project people who should have been at SOCS session were busy in review sessions.
  • Observing Strategy whitepaper seems to be making good progress.
    • Deep Drilling whitepaper contributors do not seem to be following up on their requested simulations.  People are not clear where to find our simulations.  Will connect with Neil Brandt and Will Clarkson to make sure they can find those simulations.
    • Will re-run requested simulations with newest version of OpSim.
      • Peter: should we withold new simulations until community provides metrics for Deep Drilling?
      • Rahul: it is easy to evaluate whether a simulations meets a request.
    • Gregory Dubois-Felsman is worried that community thinks Deep Drilling exposures should be processed differently from WFD exposures.
  • CCB meeting on 30-second exposures: Thinks its a good idea.  Robert Lupton seems confident that we can identify cosmic rays in a single exposure.  Need to look into radioactivity limits on the optics, camera body, etc.  Will need to start running OpSim with single 30-second exposures.

Andy:

  • Worried that the annual review will continue to intersect with AHM.  Kem thought Beth said that would not happen again.
  • Better meeting than previous project meetings.
  • Simulations were not over-scheduled.  We were free to go to science collaboration sessions.
  • Hack sessions were good.  Tim Jenness did a lot of work and was able to effectively organize hackers.
  • Good idea to use community to organize science sessions.  People wanted to be there.
  • Seems to be a disconnect between what Science Community wants and what they think we are doing (e.g. we can probably already give them the 'rolling cadence' they want).  Science Community does not see survey scheduling as an optimization problem.
  • PhoSim has code to translate descriptions of our camera system into files used by our software.  We need to re-implement it (the PhoSim version is not publicly available).
  • Concerned about delivery timeline of OpSim v4 scheduler.  We are making commitments to Science Community that the next version comes in November.  If that slips, Science Community will be frustrated.
  • Concerned that DDS only runs on Red Hat Linux.
  • Is SAL development going to be fast enough to support what we need?  Follow up with Francisco.  Make sure he is getting what he needs.
  • Observing Strategy whitepaper will use baseline cadence; not baseline with new skybrightness.
  • Offered to give talk in review session.  Reviewers have no concerns about simulations.  They were quite pleased with our progress.

Cathy:

  • There is definitely a gap in communication between project (esp OpSim) and community regarding what simulations have been done.

Michael:

  • Sees 30 second exposure request as a way to prevent us from designing a camera that can only do two 15 second snaps.

Peter:

  • Can cache skybrightness model in OpSim.  Needs 6GB of memory for each simulated survey year.  It is a factor of 1000 faster than just directly calling skybrightness from OpSim at each pointing.
    • Michael: not able to discuss incorporating this until next month
    • Peter: need to validate fidelity of interpolation on skybrightness
  • Tim's hack sessions were fantastic.
  • Management wasn't there for workplace culture.  That's problematic.
  • Jonathan Sick's documentation talk was great.  We should have documentation sprints.
  • Nice to see fewer plenaries than last year.

 

Actions:

  • Re-run Deep Drilling proposals from community with latest OpSim code.  There are about 4 of them.
    • Kem: we should probably run something with all of the proposed Deep Drilling fields in a single simulation.
    • Andy: put these in a separate table on the same website where we have the Tier-1 runs.
    • Rahul will put together metrics to evaluate Deep Drilling requests
  • Re-implement PhoSim code that generates/translated descriptions of camera and optical system.
  • Validate Peter's full-sky skybrightness code (i.e. the interpolation assumptions Peter made) and incorporate it into SOCS.
  • Go through Observing Strategy whitepaper.  Identify sections that are missing metrics.
  • Observing Strategy whitepaper group needs to know what kind of rolling cadence they can do in November (with OpSim v4).  We need to write down what we are developing and what we will be able to support.  They will write down their use cases and use those as testing.
  • How deep does next-generation all sky camera need to go to get transparency maps.
  • Get AlertSim into the stack and available for broker-developers to run.
  • No labels