Software correlators as testbeds for RFI algorithms

TL;DR

This paper demonstrates that software correlators, due to their flexibility and ease of modification, serve as effective testbeds for developing and testing RFI detection algorithms in radio interferometry, exemplified by enhancements to the DiFX correlator.

Contribution

The paper shows how software correlators can be easily adapted to test new RFI detection techniques, exemplified by adding kurtosis detection to DiFX, revealing previously unknown short-duration RFI.

Findings

Software correlators are flexible and quick to modify.

Adding kurtosis detection to DiFX reveals new short-duration RFI.

Software correlators are effective for testing RFI mitigation algorithms.

Abstract

In-correlator techniques offer the possibility of identifying and/or excising radio frequency interference (RFI) from interferometric observations at much higher time and/or frequency resolution than is generally possible with the final visibility dataset. Due to the considerable computational requirements of the correlation procedure, cross-correlators have most commonly been implemented using high-speed digital signal processing boards, which typically require long development times and are difficult to alter once complete. "Software" correlators, on the other hand, make use of commodity server machines and a correlation algorithm coded in a high-level language. They are inherently much more flexible and can be developed - and modified - much more rapidly than purpose-built "hardware" correlators. Software correlators are thus a natural choice for testing new RFI detection and mitigation techniques for interferometers. The ease with which software correlators can be adapted to test RFI detection algorithms is demonstrated by the addition of kurtosis detection and plotting to the widely used DiFX software correlator, which highlights previously unknown short -duration RFI at the Hancock VLBA station.