Radio frequency interference monitoring and mitigation for astronomy with phased array feeds

Abstract

Despite progress in radio frequency interference (RFI) algorithms and techniques in radio astronomy, much work remains to implement these techniques as part of a holistic strategy in routine, large-scale telescope operations. The characterisation of RFI using flagged (discarded) data compiled from over 5000 observations (1500 h) using the ASKAP radio telescope is presented. The average amount of flagged data due to radio frequency interference across the routinely-used ‘clean’ con- tinuum science bands is 3%. Subspace projection spatial filtering techniques applied at the beamformer have been implemented in the ASKAP codebase for a self-generated interferer. We find that the unwanted signal can be suppressed by at least 31 dB using oblique projection to the noise floor of an 880 s integration, with a typical degradation of 1.2 K in system-temperature-over-efficiency (Tsys/η) to that 1 MHz channel. Interferometric measurements of system equivalent flux density showed an increase of no more than 4%. Furthermore, no significant change to the gain in the main beam was confirmed using holography within ±0.4 % on average about the half power point. The final chapter presents the first steps in extending the static case to moving RFI. A set of tools is under development to identify RFI-affected channels, estimate their RFI subspace and forward predict the performance of oblique projection. Subspace estimation using the median absolute deviation is extended by considering the eigenvalues of neighbouring frequency channels. Forward predictions of the performance of the mitigation are made using the correlation of maxSNR beamformer weights with individual spatial signature vectors, the Y-Factor, and the singular value decomposition of the projection matrix based on a small population of ASKAP array covariance matrices. This is the first time RFI and its directionality have been analyzed at this scale using actual telescope data.