Skip to content

Latest commit

 

History

History
80 lines (62 loc) · 5.79 KB

Stages.md

File metadata and controls

80 lines (62 loc) · 5.79 KB

Pipeline stages

BBPower is made up of the following stages:

1. BBPowerSpecter

Stage summary

BBPowerSpecter takes a set of maps (with their masks and other metadata), and computes all possible power spectra, cross-correlating different data splits, frequency channels and polarization channels. It does that for the data and for a number of simulations.

Inputs

  • splits_list: a txt file containing a list of file paths. Each file should be a fits file containing 2 maps (Q and U) per frequency channel. Each file should correspond to one data split (i.e. a map made from only part of the data).
  • masks_apodized: a fits file containing one map per frequency channel. Each map should be the apodized mask to be used on all maps corresponding to that frequency channel.
  • bandpasses_list: a txt file containing a list of file paths. Each file should be a txt file containing the bandpass for one of the frequency channels (in the same order in which the maps are stored in the files above). Two columns: frequency and transmission.
  • sims_list: a txt file containing a list of paths. Each path should lead to a directory containing a sky simulation. The simulation maps should be structured in the same way as the data splits contained in splits_list.
  • beams_list: a txt file containing a list of file paths. Each file should be a txt file containing the beam for one of the frequency channels (in the same order in which the maps are stored in the files above). Two columns: multipole ell and beam_ell.

Outputs

  • cells_all_splits: a fits file containing all possible auto- and cross-spectra between all different data splits, frequency channels and polarization components (E and B). The files are in SACC format, and contain also the bandpass and beam information, as well as all the bandpower window functions.
  • cells_all_sims: a txt file containing a list of files. Each file is the fits file containing the power spectra for a given simulation, in the same format as cells_all_splits.

Parameters

  • bpw_edges: a txt file containing a list of integers defining the edges of the bandpowers.
  • purify_B: set to True if you want to use B-mode purification.
  • n_iter: iter parameter for spherical harmonic transforms.

2. BBPowerSummarizer

Stage summary

BBPowerSummarizer takes all the cross-split spectra generated by BBPowerSpecter for data and sims and calculates the total coadded spectra, coadded spectra using only pairs of different splits, the noise power spectra, and all null power spectra. It does so for the data and all sims. The sims spectra are then used to compute the covariance matrix of the data spectra.

Inputs

  • splits_list: see BBPowerSpecter inputs.
  • bandpasses_list: see BBPowerSpecter inputs.
  • cells_all_splits: see BBPowerSpecter outputs.
  • cells_all_sims: see BBPowerSpecter outputs.

Outputs

  • cells_coadded: a fits file containing all cross-frequency and cross-polarization power spectra coadded over pairs of different data splits (i.e. these spectra should not have any noise contribution). The files will be in SACC format, and will contain also the covariance as well as all bandpasses, beams and bandpower window functions.
  • cells_coadded_total: same as cells_coadded, except this file will contain the spectra computed by coadding over all pairs of splits (including auto-correlations). These will have a noise bias.
  • cells_noise: this will contain the noise power spectra (in the same format as cells_coadded), calculated as the difference between cells_coadded_total and cells_coadded.
  • cells_null: this will contain all combinations of cross-split/frequency/polarization power spectra that should be compatible with zero. Also in SACC format.

Options

  • nulls_covar_type: type of covariance matrix for null spectra. Choose diagonal, block_diagonal or dense.
  • nulls_covar_diag_order: number of secondary diagonals to store if the previous option is diagonal or block_diagonal.
  • data_covar_type: same as nulls_covar_type for the coadded data files.
  • data_covar_diag_order: same as nulls_covar_diag_order for the coadded data files.

3. BBCompSep

Stage summary

This stage takes a set of coadded multi-frequency polarization power spectra and evaluates their likelihood with respect to a multi-frequency CMB+foreground model. In its standard form, it will run an MCMC of the likelihood and store the parameter chains. It can also be used to calculate Fisher matrices, minimize the likelihood or simply evaluate it at a single point. We only document the standard form here.

Inputs

  • cells_coadded: see BBPowerSummarizer outputs.
  • cells_noise: see BBPowerSummarizer outputs
  • cells_fiducial: a fits file containing multi-frequency and multi-polarization power spectra for a model that should roughly resemble the data (e.g. the model used to generate the simulations in the previous stages). This is only needed if you're using the Hamimeche & Lewis likelihood.

Output

  • param_chains: a numpy npz file containing the parameter MCMC chain.

Parameters

See test/test_config_sampling.yml

4. BBPlotter

Stage summary

This stage generates a webpage containing a set of plots illustrating the main pipeline results.

Inputs

  • cells_coadded_total: see BBPowerSummarizer output.
  • cells_coadded: see BBPowerSummarizer output.
  • cells_noise: see BBPowerSummarizer output.
  • cells_null: see BBPowerSummarizer output.
  • cells_fiducial: see BBCompSep input.
  • param_chains: see BBCompSep output.

Outputs

  • plots: a directory containing all the output plots as png files.
  • plots_page: an HTML file containing links to all of those plots.

Parameters

See test/test_config_sampling.yml