Adaptive algorithms for low-latency cancellation of seismic Newtonian-noise at the Virgo gravitational-wave detector

TL;DR

This paper investigates the performance limits of adaptive, time-variant filters for canceling seismic Newtonian noise in gravitational-wave detectors, demonstrating their superiority over static filters through a Virgo site case study.

Contribution

It introduces a framework to analyze the limitations of time-variant noise cancellation and provides a proof-of-concept showing adaptive filters outperform static Wiener filters in seismic noise reduction.

Findings

Adaptive filters significantly outperform static Wiener filters.

Residual noise remains above the statistical error bound.

Adaptive filter architecture influences cancellation effectiveness.

Abstract

A system was recently implemented in the Virgo detector to cancel noise in its data produced by seismic waves directly coupling with the suspended test masses through gravitational interaction. The data from seismometers are being filtered to produce a coherent estimate of the associated gravitational noise also known as Newtonian noise. The first implementation of the system uses a time-invariant (static) Wiener filter, which is the optimal filter for Newtonian-noise cancellation assuming that the noise is stationary. However, time variations in the form of transients and slow changes in correlations between sensors are possible and while time-variant filters are expected to cope with these variations better than a static Wiener filter, the question is what the limitations are of time-variant noise cancellation. In this study, we present a framework to study the performance limitations of time-variant noise cancellation filters and carry out a proof-of-concept with adaptive filters on seismic data at the Virgo site. We demonstrate that the adaptive filters, at least those with superior architecture, indeed significantly outperform the static Wiener filter with the residual noise remaining above the statistical error bound.