Bad realizations in gsync/gdust sims

From CMB-S4 wiki
Jump to: navigation, search

November 9 2018, Clem Pryke

Ben noticed that there are some realizations of model 04.00 (Gaussian sync/dust) which are bad fits in the parametric re-analysis [1]. It was then realized that these are clearly visible in the final pager and should have been noticed earlier [2]. Further the exact same pattern of badness can be seen in the corresponding 02.00 plot [3].

I traced the problem to the gsync/map and gdust/map files. The corresponding alm files from which they were generated are OK. Somehow the alm read-in failed resulting in the E&B alm's being all zero - and in a few cases only the E. We can detect these bad realizations by reading in each map in turn and taking the rms of the Q and U maps. The scatter plots below show the results. There are some realizations which have zero for Q and U rms and some which have a suppressed value. I make a simple statistic to find these as the red line in each panel.

181109 f1.gif 181109 f2.gif

The complete list of bad realizations with zero based indexing (same as the filenames) is below. These need to be screened out of all .00 re-analysis results so far, and also .07 since this is built from the same maps. Model .05 is built from the alms and is not effected as we can see in the first link above. Before any future round of sims these maps will be regenerated to properly fix the problem.

   56
  130
  131
  135
  143
  152
  206
  275
  309
  315
  347
  357
  367
  372
  387
  388
  410
  443
  452
  477
  480
  497
  512
  577
  579
  596
  620
  632
  704
  743
  782
  907
  925
  935