Planck was within 20% of original design. We could degrade our numbers by ~20%. But this ignores the new set of issues with TES arrays (e.g. uniformity and yield).
Suggestion: reduce our numbers by a Planck factor (20%) and add a factor for yield/uniformity.
AT: requirements must be met for flight. If ground tests put you close then costs and work grow to ensure meeting requirements. Seen past missions where requirement is factor of 2 lower.
SH: add yield too, 20% degradation may be low.
AT: cost analysis assumed 90%, so reasonable place to start. (SH: 90% seems fine)
SH: So what is the additional degradation? How about sqrt(2)? AT: 1.5 or 1.6?
AK: lay out chain of logic. show where numbers are coming from. so assumptions
SH: This goes to costers and decadal. goals slightly different. For decadal we need to show high quality science. For costers need to ensure we're not making to aggressive assumptions which drive costs.
AT: 2 stories. Science objective is X, need Y sensitivity to reach that. Prediction is A, that is below Y by some margin which is enough for …