On the Signal Processing Operations in LIGO signals

TL;DR

This paper critically examines LIGO's gravitational wave data processing, revealing that the matched filter often produces false positives due to signal processing errors, and proposes normalized cross-correlation as a more reliable alternative.

Contribution

It identifies flaws in LIGO's matched filter implementation and demonstrates that normalized cross-correlation avoids false peaks, improving gravitational wave detection reliability.

Findings

Matched filter misfires with high SNR peaks for noise and sine waves.

Normalized CCF does not produce false peaks with noise or sine wave signals.

LIGO's current processing can mistake noise for genuine signals.

Abstract

This article analyzes the data for the five gravitational wave (GW) events detected in Hanford(H1), Livingston(L1) and Virgo(V1) detectors by the LIGO collaboration. It is shown that GW170814, GW170817, GW151226 and GW170104 are very weak signals whose amplitude does not rise significantly during the GW event, and they are indistinguishable from non-stationary detector noise. LIGO software implements cross-correlation funcion(CCF) of H1/L1 signals with the template reference signal, in frequency domain, in a matched filter, using 32 second windows. It is shown that this matched filter misfires with high SNR/CCF peaks, even for very low-amplitude, short bursts of sine wave signals and additive white gaussian noise(AWGN), all the time. It is shown that this erratic behaviour of the matched filter, is due to the error in signal processing operations, such as lack of cyclic prefix necessary to account for circular convolution. It is also shown that normalized CCF method implemented in time domain using short windows, does not have false CCF peaks for sine wave and noise bursts. It is shown that the normalized CCF for GW151226 and GW170104, when correlating H1/L1 and template, is indistinguishable from correlating detector noise and the template. It is also shown that the normalized CCF for GW151226 and GW170104, when correlating H1/L1 and template, is indistinguishable from correlating H1/L1 and bogus chirp templates which are frequency modulated(FM) waveforms which differ significantly from ideal templates. Similar results are shown with LIGO matched filter, which misfires with high Signal to Noise Ratio(SNR) for bogus chirp templates.