01.01 sim input maps - first try

From CMB-S4 wiki
Jump to: navigation, search

Mar 27 2017, Clem Pryke

This posting is a follow on from Mar 23 posting.

Instead of Gaussian dust/sync I try adding the PySM maps which someone (Julian?) has provided in the NERSC system at /project/projectdirs/cmb/data/generic/galactic/pysm. We have defined sky model XX.01 to be PySM in "vanilla" mode a1d1f1s1 - see the PySM paper and GitHub for some description.

Here is the full sky map masked to |b|>10, filtered to beam size, cut to ell>30 and rotated to celestial coords. The crude cut on galactic latitude is to try to prevent ringing from the ell cut causing trouble - see below. I propose that we work in Celestial coords as is common for ground based experiments - although I expect there may be push back on this from those with space heritage.

20170327 fig1.gif

Masking down to 3% sky patch and adding in noise and LLCDM we get something which looks indistinguishable from the Gaussian foreground version in the previous post:

20170327 fig2.gif

Taking the spectra of the above with anafast we get the following. TT looks identical to the previous post. The dust level in BB is clearly a bit higher than the Gaussian model set to the level found in the BK field. However, there isn't necessarily any conflict in this as the BK field is 3 times smaller than the field being used here. It looks like my attempt to cut ell<30 is not working cleanly and I'd appreciate any tips on how to improve the performance. [Note added 5/9/17: I experimented with masking down to a 5% region before going to alms and it gets rid of the low ell response - should improve this part for round 2.]

20170327 fig3.gif

One note: PySM is "Planck locked" at low ell. So even if we had multiple realizations of the small scale randomized component this wouldn't help much. I previously thought this was a big problem. However Colin has now shown that for Gaussian sims and the current BK noise regime at least, the multicomponent cross spectra method delivers r values that have the same mean and spread when the foreground realization is held fixed at a single realization. Not sure in what circumstances this will break down.