QUBIC

Minutes:

• JAX (Tom):
  - works the same way as numpy, but allows for just-in-time compilation
  - overall much faster if repeated operations

• Preconditioner:
  - no major changes

• Projection operators (Leonora):
  - the exploding values might be due to plotting the mean of the relative difference
  - the frequency of sub-bands was computed in log spacing, no one remembers why (imitate physical bands that have a width proportional to the central frequency?)
  --> try with linear spacing if there is still an error

• answer from Sigurd Kirkevold Naess on QUBIC forecast with realistic atmosphere (Tom)
  - hardware changes not possible
  - ACT says that the atmosphere changes on time scales of about 50 seconds
  - the instrument gathers information on numerous parts (a large fraction) of the sky at the same time, with different detectors

• Moon (Alexandre):
  - try to recover the negative features around the Moon with a Butterworth filter on a fake Moon and correct for the effect on the real map
  - look at the positions of the TES' squid in the squid boxes (4 x 8 x 4 squids per box) where the cross-talk happens between neighbour squids
  - impose priors on gaussian fitting (m.limits[""]?)
  - interactive alternatives to matplotlib: try plotly (close to matplotlib), bokeh and seaborn