UCSD-2019: Flowdown: Science to Measurement Requirements

From CMB-S4 wiki
Jump to navigationJump to search

Link back to agenda


Charge from SOC

  • Define what simulations, and other forecasting tools, are still needed before CD-1, the schedule and plan for these simulations and other forecasting tools, and when any inputs are needed to achieve a complete connection between science and measurement for CD-1.
  • Coordinate carefully with science and analysis working groups to understand the assumptions they are making and inputs they are using.

Additional tips for making this session maximally useful

  • People listed below are being asked to prepare for this discussion by outlining, as well as they can, their response to the charge, and highlighting the questions they have about completing a response to the charge, and more generally, questions they have that will help them have a productive meeting in their Friday afternoon parallel session.
  • List systematic effects and suggest plan for writing down appropriate requirements, or questions that may need communication with the Technical Council and coordination with them.
  • Read the Science to Measurement Requirements chapter of the DSR in advance to know current state of our art, if you have not already.


Remote attendance

Zoom link



  • We need requirements specified so people can build without consulting with each other constantly. Too many people for that consultation to be possible.
  • "Good enough" is! "Better" costs more.
  • science goals --> measurement reqs --> instrument reqs --> mission reqs, plus feasibility and cost loop back.
  • Pryke: we've got foregrounds and we don't know what they are. How do we deal with this fundamental degree of uncertainty?
  • Lawrence: I'm in complete agreement with that point. We have to make some allowances for the fact we don't know. One way we do that is to have a substantial margin on science requirements: a factor of two in some cases.
  • Ed Wollack: It's not a new problem.
  • Clem Pryke: but split bands for conservatism was not judged to be a valid argument. We needed to demonstrate it led to improvements, when we don't know what the foregrounds are like!


  • Four science goals that drive design: r, light relics, clusters, GRBs. (Other science goals are not design drivers)
  • Leads to two surveys: ultra-deep hybrid resolution and deep/wide high resolution
  • Worked example of science to measurement flowdown for deep/wide survey. Science goals lead to sky coverage, beam size, map depth, observing cadence.
    • Still need to think more about systematics (esp. beams), elevation minimum, margin for science targets.
    • Frequency coverage is based on Colin Hill calculations that optimized several science goals... but none of these were actually "galaxy clusters". Is this good enough, or does it need to be revisited?
    • Increasing sigma(Delta Neff) requirement from 0.03 to 0.033 makes measurement much easier. Can we do this (but maybe keep a goal of 0.03)?
    • Do we need to connect measurement optimizations to cost models? Otherwise, we always want more angular resolution and deeper maps.
  • Discussion of beam systematics and the need to simulate them:
    • For Neff science goal, planet measurements will measure necessary beam parameters at high significance... so long as they don't vary frequently
    • For r / delensing, beam systematics (like far sidelobes) will be important. Still have a problem of how much to believe physical optics models, ground models, etc.
  • Example DESI CD-1 requirements document -- Only 19 pages, all flows down from BAO science goal. Sample variance pushes to large sky area, setting errors on D(z) and H(z). This is posted on the wiki
  • Supposed to have draft baseline requirements document by end of 2020. Lots of work needed by June 2020!
  • Lawrence suggests keeping Neff as sole driver of angular resolution, and dropping galaxy clusters as a driver of this. (Private comment to Knox after the session).


  • Colin reviewed forecasting done so far in r forecasting group.
  • Kimmy sketches potential questions that require addressing by CD-1:
    • Investigate real de-lensing
    • Incorporate model-loss approximation in phi reconstruction
    • Consider impact of instrumental systematics
    • Determine whether three above items affect choice fo siting for telescopes.
    • Once experiment configuration set, we ask... (missed this one)
  • Kimmy on CD-1 simulation needs:
    • Astrophysical:
      • Galactic foregrounds
      • Extragalactic foregrounds (not included so far... necessary?)
    • Instrumental
      • Gain variations, beam: shape mismatich, shape variations, bandpass variations, polcal variation,
      • More?
  • Analysis needs
    • Delensing: beyond quadratic estimator
    • Component separation (and foreground cleaning)
      • Existing tools include cross-spectra parametric fit and ILC (interests in developing more?)
      • For phi reconstruction need dedicated component separated map as input to phi reconstruction
  • HOw we plan to organize this effort: using data challenges
    • We will have more detailed discussion in the low-ell BB parallel
    • Regarding data challenges
      • So far the r forecasting working group had generated needed sims/... (missed what was here)


  • What analysis tools need to be developed? Lots exist already.
  • What sims are needed?
    • We want sims with beam systematics, bandpass systematics?
  • We can run a cluster finder over set of multifrequency maps and have interesting results on this by next collaboration workshop. (See Nick's slides for more details on this statement).