OPENILC

Internal linear combination for cmb data analysis

NILC (Needlets Internal Linear Combination)

This repository provides tools for performing Needlets Internal Linear Combination (NILC). Below is an overview of the key components and guidelines for using them effectively.

Components

1. Source Code

• nilc.py: The main module to perform NILC.

2. Configuration Files

• ./needlets:

  • Defines the cosine needlets configuration.
  • Important Rules:
    • Each row's lmin and lpeak must match the previous row's lpeak and lmax.
    • The last lmax should match lmax in your NILC settings.
  • Resolution Note:
    Ensure your nside is greater than lmax / 2 because different needlet bins may have varying resolutions. The calculations for lmax are valid only when lmax < 2 * nside.

• ./bandinfo.csv:

  • Experimental configuration file that includes parameters like frequency and beam.
  • Ensure lmax_alm is greater than your NILC lmax.
  • Usage: This file is primarily for testing the NILC effect without beam corrections.

• ./bandinfo_beam.csv:

  • Experimental configuration file for cases considering beam corrections.
  • Important Rules:
    • Set lmax_alm carefully to avoid numerical errors at large ell values (e.g., for beam=60, lmax_alm ~ 500).
    • Ensure to use the bigger band-limited lmax (when you do deconvolve and convolve your map) than lmax_alm

3. Tutorials

• test_nilc.py:
  A basic (though currently incomplete) tutorial on using NILC.

• test_nilc_beam.py:
  A tutorial demonstrating how to perform NILC with beam considerations.

Usage Guidelines

1. Set up ./needlets configuration:
   Ensure consistency between lmin, lpeak, and lmax across rows and align with your NILC parameters.

2. Verify nside:
   It should be at least lmax / 2 for accurate results.

3. Check beam parameters:
   When working with ./bandinfo_beam.csv, account for potential numerical errors at high ell.

4. Run Tutorials:
   Use the test_nilc.py and test_nilc_beam.py scripts to understand the process and verify your setup.

KNOWN ISSUE

• large memory usage in NILC
• slow rate when calculating beta covariance matrix in NILC

FUTURE PLAN

• use pytest
• add other ILC

HIGH PERFORMANCE COMPUTATION

SHT can be replace by other package or use healpy build from sources:

• see https://github.com/healpy/healpy/blob/main/INSTALL.rst#generating-native-binaries first
• download cfitsio version 4.5.0, then run ./configure --disable-curl;make;make install
• download healpix version 3.8.3, configure with your cfitsio fitsio.h and cfitsio lib to build healpix cxx
• add PKG_CONFIG_PATH for cfitsio and healpix
• pip install --no-binary healpy healpy