The impact of beam deconvolution on noise properties in CMB measurements: Application to Planck LFI

TL;DR

This paper analyzes how beam deconvolution affects noise in CMB measurements, deriving covariance matrices and validating results with Planck LFI data simulations to improve noise modeling.

Contribution

It introduces a low-resolution noise covariance matrix for deconvolution residuals and compares destriping methods, enhancing noise understanding in CMB data analysis.

Findings

Residual noise is well modeled by the covariance matrix.

Full destriping reduces residual noise but introduces data correlations.

The white noise approximation is effective at high multipoles.

Abstract

We present an analysis of the effects of beam deconvolution on noise properties in CMB measurements. The analysis is built around the artDeco beam deconvolver code. We derive a low-resolution noise covariance matrix that describes the residual noise in deconvolution products, both in harmonic and pixel space. The matrix models the residual correlated noise that remains in time-ordered data after destriping, and the effect of deconvolution on it. To validate the results, we generate noise simulations that mimic the data from the Planck LFI instrument. A $\chi^2$ test for the full 70 GHz covariance in multipole range $\ell=0-50$ yields a mean reduced $\chi^2$ of 1.0037. We compare two destriping options, full and independent destriping, when deconvolving subsets of available data. Full destriping leaves substantially less residual noise, but leaves data sets intercorrelated. We derive also a white noise covariance matrix that provides an approximation of the full noise at high multipoles, and study the properties on high-resolution noise in pixel space through simulations.