Joint Plenary for Forecasting/Reference Design

Notes

Questions: What plans for instrument systematics modeling on the decadal timescale and beyond? What instrument parameters values should we use, guided by simulations & experience?

• Complementary performance-based and ab initio forecasting
• Write down specs for eg. bandwidth, NET, etc, justified by demonstrated performance
• Need single reference document for all of these (wiki page? google sheet?)
  • Organize by subsections of CDR reference design chapter
  • Incorporate requirements & achieved values from S3 experiments
  • Include links to S3 references
  • "Are we done? What next?"
• Cascade diagram progressively decomposing each element
• For many boxes there are uncontroversial numbers that can be used as the starting point
• On which parameters do we need to do better than S3 to achieve our goals?
• CDT framework provides maxima for eg. additive contamination
  • see CDT report appendix A, wiki links & forecasting paper in progress
• Track propagation of systematic from source to power spectra to identify scaling with eg. number of detectors
• What's going to be hard?
  • Beams, beams, beams & beams
  • Cross-talk
  • Bandpass mismatch
  • Polarization calibration
  • Readout electronics stability
• Big arrays won't necessarily integrate down as we might hope
• Prioritize effects that need to be addressed - things that can either be checked-off fast or are highest risk/reward
• Need to avoid telecon proliferation
  • Small group to lay out spreadsheet (Zeesh, Colin, ... ) & call telecon(s) only as needed
  • Add tab to John Ruhl's spreadsheet & require login to edit
• Detector fab questions:
  • Current approach scales from B/K - how representative of S4 fab will this be? Yield, scatter, ...
  • Need flow down to factors that inform NET, including spread (with frequency)
  • Use best B/K as credible target
• Identify things could be legitimate improvements
  • Base temperature
  • Increased duty cycle from dilution refrigeration
  • Hold these as margins?
  • Detector count is the special case!