A method of RFI mitigation based on spatial filtering and its real-time system

With the construction and application of large-scale astronomical receiving arrays, the high-speed signal streams they generate pose significant challenges for efficient data processing.Consequently, there is a pressing need for a real-time computing system designed for large-scale receiving arrays and weak signal processing.Radio frequency interference (RFI) mitigation and its real-time implementation are among the critical technical issues that need to be addressed.These issues not only affect the extraction of signals of interest but also impact the calibration of beamforming factors.To this end, a method of RFI mitigation based on spatial filtering has been designed and implemented on a dedicated algorithm acceleration core.Spatial filtering is an effective method for array interference mitigation.The Australian Square Kilometer Pilot Array (ASKAP) conducted interference reduction experiments based on spatial filtering and subspace projection method while applying phased array feed to actual observations, and achieved promising experimental results.However, ASKAP is a synthetic aperture telescope with wider spaced receiving units, making it relatively less sensitive to RFI.This paper proposes an RFI mitigation strategy for astronomical phased array with larger scale and closer spacing.The main algorithmic workflow is as follows: First, perform data preprocessing, including calibration, bandpass and baseline removal.Next, calculate the covariance matrix of the signals, and then compute the eigenvalues and eigenvectors.Then, an interference projection matrix is used to construct the spatial filter.Finally, select RFI based on a predefined threshold.Additionally, block processing and compression techniques will be employed to enhance efficiency.This method supports parallelization and will be integrated into a heterogeneous computing platform named high-throughput data pipeline parallel computing system.In this system, the CPU primarily handles complex logical operations such as process control, while the GPU leverages its multiple computing cores to optimize and accelerate tasks such as covariance matrix computation, matrix eigenvalue decomposition, and matrix multiplication within the spatial filtering algorithm.In the future, this system will process array signals from more than 200 links.