Simulations for forecasting r performance
Here is a selection of codes and sim maps:
Ben Thorne, David Alonso, Jo Dunkley
Spits out Nside=256 fits maps by running 'python main.py main_config.ini'. See the readme for any required dependencies.
Code already includes a few options for each component, and can be easily extended to add new inputs.
Suggested starting points and ini files for each (noise level to be adjusted):
- over-simple 's1d1a1' = power-law synchrotron, single component dust, minimal spatial variation of indices, no polarised AME
- towards more realism 's4d3a2' = spatial steepening of the synchrotron index away from plane and curvature, 1 component dust with spatially varying dust index sigma=0.3, 2% polarised AME.
- towards more realism 's4d4a2' = spatial steepening of the synchrotron index away from plane and curvature, 2 component dust, 2% polarised AME
Current issues: The version of code on github (v0.3) doesn't yet have small-scale realizations added, so the polarization in low S/N regions is still noisy and all maps are smoothed to 1 degree resolution. This means that power spectra of e.g. the B2 region are noisy and so Victor's method using power spectra won't instantly work.
Solution: could try Victor's method on higher S/N regions as a test of how well things work. Or wait for the small-scale Gaussian sims (~2 weeks). Pixel-based methods will work a little better on these current v0.3 sims.
N.B. - this code is still under development. Please let us know about any issues/bugs you spot! And we welcome contributors!