slide 1: starts with 90.91, not fitting for synchrotron curvature (because it isn't in the sky; more on that later)
2,3: Finds the best fit model; residual in many bands look very reasonable.
Applies the same model to 90.92. Finds worse chi^2 and more residuals. Residuals in 21 and 25 GHz are opposite in intensity. Similar behavior in other bands. Key point: in real data, such residuals and poor chi^2 would indicate that the model is incomplete.
Discussion: what would happen if you don't have that broad of a frequency range? Suggestion: start with 90.91 with a smaller frequency range. See what you get. Then apply that (with smaller frequency range to 90.92. See what you get.
Blackwell Rao estimator gives biased results even for input CMB - not clear what's going on. Need to debug.
MR shows results with NILC and CMILC with three realization r=0. So far result looks good. There is reduction of foreground residuals as one projects out foreground modes. The increase in noise is not significant yet.
foregroundstelecon20200917.txt · Last modified: 2020/11/02 08:03 by hanany