SLAC-2017:FSM Large angular scales

From CMB-S4 wiki
Revision as of 19:13, 28 February 2017 by Flauger (talk | contribs)
Jump to navigationJump to search

Back to SLAC-2017 main page

Foregrounds, Systematics and Modeling: Large Angular Scales

Charge for this session: "plan for flowing down the L1 science requirements to L2 measurement requirements, consistent with the CDT time scale. We suspect that your work will include identifying tasks appropriate for splitting into working groups, and organizing regular telecons for these working groups."

Post session talks here.

  • Systematics/Instrument inputs -- John Kovac
    * Band selection (X=2): John/Denis, Adrian/Charlie, Jo and Reijo
        * idea is NOT final optimization, but only to cover representative bands to inform CDT Strawman Concept 
    * Inclusion of systematics (X=3):  [John coordinating]
        * idea is NOT final systematics verification (i.e. effect-by-effect), only to capture representative challenges
            * "Unknown systematics residual" modeled as additive contamination (harmonic space), fractional level linked to N_l
            * Bandpass uncertainties
            * Pol angle uncertainty?
    * Noise models and bands for delensing survey (X=4)
        * adopt identical per-detector N_l assumptions, with higher ell_knee?  Beamsize?
        * at some point, do we use real (S3) noise maps as basis of scaling?
CDT Instrument systematics have the following leads and groups:
 overall coordination: John K
  Beams             --- Bill, Steve, Mike, John K
  Calibration      --- polarization angles,    Brian
                                intercalibration between different angular scales,   John C and Tom
                                temporal aspects       Ed
  Modulator systematics --- Adrian, Brian
  Time response              --- Ed
  • Analysis -- Raphael Flauger
   - Longer term goal: Turn science requirements into measurement requirements with the help of data challenges
   - Currently a number of groups have volunteered to participate in the data challenge
       - David Alonso
       - Colin Bischoff, Victor Buza, Justin Wilmert
       - Hans Kristian Eriksen, Unni Fuskeland, Ingunn Wehus
       - Josquin Errard
       - Raphael Flauger
       - Dongwan Han, Neelima Sehgal
   - Shorter term goal: Validate Fisher forecasts with DC1.0 simulations
       - Some first (and preliminary) results for DC1.0 from
           - Colin Bischoff, Victor Buza, Justin Wilmert
           - Josquin Errard
           - Raphael Flauger
       - DC1.0 analyses should be completed and refined, performance of different algorithms should be compared, etc. 
   - What should be provided for the next data challenges?
       - Currently the same set of simulations are used to calibrate estimators, noise bias, covariance matrix, etc. and for analyses. Should we move to one (larger and potentially less complex) set of simulations for noise bias, covariance matrix, and another (smaller) set to test algorithms, study biases, etc.?

  • Wrap-Up/Discussion -- Lloyd Knox

Notes from session

Julian - Data challenges

NS: what motivates sigma_r = 5e-4? - not now

RF advocates adding nonzero r before Stage 3

CP cautions that more complex foregrounds are not necessarily more correct or realistic and could push us to regions of instrument parameter space that are not feasible to build

CL: more important part of the progression is toward ability to put in foregrounds in simulated maps that are not baked into the component separation

LK: we're not going to get through all steps in JB's slide "Plan in Progress" so which should we prioritize?

JK: can anyone do sims with full de-lensing right now (or before October)?

NS: de-lensing has progressed a lot, but how to connect those groups with r forecasters? need a component separator for the lensing groups lots of discussion of code-sharing and keeping things modular

John K. - Foregrounds, Systematics, and Modeling: Large Angular Scales

"Unknown systematics residual" modeled as additive contamination in harmonic space

CP: isn't this just putting an arbitrary bias on r?

Albert S. - [missed this]

Raphael - Analysis

how are the analyzers supposed to deal with the unknown systematics?

JK: we're not going to tell you

Action items/Next steps

Summarize action items here