QUBIC

Minutes:

• FMM comments:
  - agree to use only one get_dict function
  - some classes need to hardcode a special dictionnary for which some parameters are invariant and it makes sense
  - the mixed Qubic and Planck TOD are a heterogeneous file with Qubic TOD in Watt for each second and Planck TOD in microK for each sky pixel -> it should be written somewhere, for example with a plot TOD function

• projection operators:
  - old QubicSoft H created nicely converging TOD (QubicMultibandInstrument), uses Trapezoid Operator and sub band edges
  - new H is not working (QubicMultibandTrapezoidalIntegration), doesn't use Trapezoid Operator and sub band centers
  - changing edges for centers breaks the convergence
  - trapezoid operator is well coded in pyoperators (see p126 in Mathias' thesis)
  - there is a mismatch between bands edges (Nsub + 1 points used for the trapezoidal integration) and the bands centers (Nsub points used as the frequency of the sub bands)
  - so why is it working when edges are used both for computation and as the frequency of the sub band?
  - it works when centers are defined as left edges or as edges of Nsub - 1
  --> check how the freq_nus are used in the code

• Qnoise:
  - working for DB, but UWB still needs a correction of its shape operators
  --> rewrite the code with the new get_dict once it is coded
  --> change the way the drawings are made to ensure the same random state whether or not we draw detector / photon noise

• randomness:
  - OK with the idea of np.random.SeedSequence when needed, and np.random.rng_default

• atmosphere:
  - atmosphere reconstruction works with atmosphere only
  - atmosphere and CMB simulated in two different coordinate systems, but leakage from atmosphere to CMB
  - maybe due to a problem in the coordinates
  --> plot a single point (or a line) in galactic and local coordinates and see if it is where intended
  - maybe due to small coverage at edges of local patch (because moving)
  --> change the npointing strategy, for example start with the random pointings at one side and go from there (in the direction of the local movement with respect to galactic)
  - these artifacts are also seen in a polychromatic simulation
  --> try reducing the time needed for one pointing
  - when starting the PCG at the true solution, the chi2 is 0 and when starting close, the result is good

• common missions: to be then extended to the rest of the collaboration, maybe with privated messages
  - create one get_dict that creates the dictionnary from a template with updated parameters (assign_default_values or something), see Steve's suggestion in Rocket Chat DA chat and dict.get("key", default_value)
  - (create a plot TOD function that explains the difference between Qubic and Planck TOD)
  - add documentation for pickle files (see http://qubic.in2p3.fr/wiki/pmwiki.php/DataAnalysis/Intro2QUBIC for a tutorial on FMM and CMM)
  - find how freq_nus are treated in the code (Qinstrument, Qacquisition and others like Qbeams), is it as centers or as edges?
  - change the parameter type to instrument_type (UWB or DB)
  - change how parameters are implemented in classes to be sure not to overwrite the dictionnary if not wanted
  - check how the photon noise is implemented, if all the functions are used and compare with previous version
  - correct the reshape operators in UWB mapmaking