Pure b

[ellie-gleave]

added options for mask-based purification and pure B-mode sims

in params file (paramfile_pure_b.yaml) these are pure_B and pureB_sims, respectively

[kwolz]

Hi @ellie-gleave, in your first version of this PR you implemented an option to null the E-mode maps at the preprocessing stage. Why is this needed? Should we consider keeping it in the pipeline?

[ellie-gleave]

@kwolz Yes that's still needed—it's to generate a comparison which has 'perfect' purification

[kwolz]

@ellie-gleave Okay, would it be reasonable to make this option only available to mock data and (covariance) sims? To test purification, one can still look at those. The TF validation and estimation sims should not be analytically B-purified (meaning using the NaMaster option) since we found the resulting transfer function to be unstable.

[kwolz]

Summary of latest changes in mask_handler.py:

• Removed parser option self_assemble for simplicity, but assemble binary and final analysis mask from nhits map. This has been validated using B-mode purification. No other input masks are needed.
• User can pass file path to custom nhits map to pipeline by paramfile argument 'input_nhits_map. If input_nhits_map is left empty, download the nominal SO SAT BB nhits map.
• Custom nhits map is automatically validated by comparing its smoothness with that of the nominal analysis mask. If first and second spin derivatives are too large, output a warning.

[Bai-Chiang]

Thanks, this solves #31. We should also remove --self_assemble argument in run_all.sh.

[kwolz]

May I ask where could I find documentation about the spin derivatives? I don't know what it is. Thanks.

The canonical reference to this is, I believe, https://arxiv.org/pdf/0903.2350.pdf The first and second spin derivative are, up to a factor, the second and third term on the rhs of eq (24).

[chervias]

I think there is a bug when using a custom nhits map. Here you simply load the nhits mask instead of building one. No new nhits map is created. In the next script, mocker.py, it will try to load the hits map here, which will be a function on metadata_manager.py. Here it will try to load the hits map, which calls the variable self.nhits_map_name, which has the string defined under nhits_map in the param file, which was not created in the first place, so the script fails.

[Bai-Chiang]

@chervias I made some quick fix to use my custom hitmap https://github.com/simonsobs/SOOPERCOOL/compare/pure_b...Bai-Chiang:SOOPERCOOL:read_hitmap If you think it's fine I could create a PR.

[chervias]

Yes, @Bai-Chiang can you PR into this branch please?

[Bai-Chiang]

33

[chervias]

Thanks @Bai-Chiang the bug regarding using a hits map from the user is fixed.

[kwolz]

The latest commit fixed the case handling of the two parameter file arguments pure_B and tf_est_pure_B. This is now ready for review to merge into main.

[chervias]

I tested with my own hits map, in both m_filterer and toast cases. It looks good to me. I would merge.