QUBIC

Minutes:

• pixellisation: bilinear interpolation
  - this interpolation improves the convergence for intensity and makes the polarization converge
  - TOD convergence seems to be understood
  - we have to choose a nsub that we consider as "good"
  --> which impact on the computation time? (for 5000 pointings, nsub = 4, old method takes 5.0s and bilin interp takes 16.1s) in bilin interp we build 4 projection operators instead of one
  --> simulate for different nsubs and reconstruct for different nsubs, not necessarily the same (go to r ?), at least 2 nsubs (1% and 5% difference with 104 nsub for 30 pointings?) and check if we see a difference

• atmosphere:
  - increasing PCG iterations reduces apparent leakage between CMB and atmosphere I and QU
  - there remains a shift in intensity maps to be understood (flat leakage between CMB and atmosphere)