Does coherent signal discrimination improve faster than glitch frequency grows with detector additions?

# Coherent Signal Discrimination vs. Glitch Growth in Multi-Detector Networks

Your question cuts to a real tension in gravitational-wave astronomy: as we add detectors, we boost our ability to distinguish real signals from noise, but we also multiply the number of glitches (instrumental artifacts) that can masquerade as signals.

The Signal-Side Advantage

Adding detectors does improve discrimination in your favor—initially. The BayesWave burst-search pipeline quantifies this via the log signal-versus-glitch Bayes factor, which increases with detector count [AG-2024.03-419]. The intuition is straightforward: a genuine gravitational wave should produce coherent, physically consistent signals across all detectors simultaneously, while a random instrumental glitch in one detector looks uncorrelated to the others. More detectors = stronger lever to distinguish the real from the fake.

The Glitch-Side Problem

But here's the catch: non-Gaussian noise transients (glitches) don't stay rare as networks expand. The frequency of glitches grows with the number of detectors [AG-2024.03-419], and this growth directly undermines the discrimination advantage. Worse, glitches are structurally tricky—they "can mimic or mask burst signals," not just add random noise. This reduces overall detection confidence [AG-2024.03-419].

Does Discrimination Win?

The honest answer from the archive is: it's complicated, and the race appears close enough that it requires careful management.

On the positive side, new discriminators are improving. Recent work shows that chi-square statistics built from singular-value decomposition of real glitch populations can distinguish binary black hole signals as effectively as older sine-Gaussian approaches [AG-2025.07-142], suggesting the toolkit for glitch rejection is advancing. And when glitches are far from the signal in time-frequency space, they don't bias your inferences at all—meaning you don't need to mitigate every glitch [AG-2025.06-690].

On the negative side, glitch rates vary unpredictably by morphology (some follow Poisson statistics, others don't), and detection tools like iDQ only catch ~39% of known glitches even with improvements [AG-2024.12-134]. For third-generation detectors like the Einstein Telescope, the problem will intensify because signals last longer in the sensitivity band, creating more opportunity for overlap [AG-2024.11-444].

Bottom Line

Discrimination capability improves faster than glitch frequency in the controlled regime (two to three detectors, clean signal-glitch separation), but the advantage narrows as networks grow and overlap becomes inevitable. This is why active glitch mitigation—not just better discrimination—will remain essential for next-generation observatories.