Correlation of reconstructed lensing template to ideal for 06b.00
December 14 2019, Clem Pryke
This post is similar to Correlation_of_reconstructed_lensing_template_to_ideal II.
We have made DSR config map based sims as set 06 | 20191101_dc06_specsmaps
Julien Carron ran his basic QE lensing template reconstruction on these using the 95GHz maps (only), and for the "Pole deep" coverage pattern (06b), and Gaussian uniform foreground (06b.00). He is using 200<ell<3000 to perform the reconstruction.
Julien provided recons from the signal+noise+foreground (comb) maps, and also from the signal only maps (llcdm). Following previous practice I take the comb-llcdm as the noise realizations. The plots below show the Q map for the first realization - looks sensible - the right sky area and noise increasing towards the edge of the patch.
I next take the cross correlation of the B-modes versus the ideal template (where ideal is lensed-unlensed inputs). The plot below shows the results averaging over the first 10 realizations.
We should focus on the ell<200 part which is relevant for r limits. Note that the recon from signal only is not the "best the QE method can do" - it is still using the same weights as the signal+noise recon so is disguarding some high ell information. However, we know from Correlation_of_reconstructed_lensing_template_to_ideal II that the algorithmic ceiling is 0.8-0.9.
I request that Julien run his recon on the 06c and 06d coverage patterns and for the 07 and 09 foreground models.