New Paper: Mitigating PSD Drift in Zero-Latency GW Searches

Our new paper introduces the Dual Cutler-Vallisneri Corrections, a novel perturbative framework to eliminate sky-localization and timing biases caused by power spectral density (PSD) drift in real-time gravitational-wave pipelines.

Read Paper (ArXiv) View Code

The Problem: Noise Drift in Real-Time Searches

To enable multi-messenger astronomy, gravitational-wave pipelines must generate zero-latency alerts to catch the early optical counterparts of mergers. This requires matched-filtering strain data against theoretical templates using a “lagged” estimate of the detector’s Power Spectral Density (PSD).

However, detector noise fluctuates continuously. This spectral drift creates a mismatch that warps the matched-filter metric, severely biasing timing estimates and degrading sky-localization.

The Solution: Dual Cutler-Vallisneri Corrections

Building on low-latency methods by Tsukada et al.  [1], we generalized the classic Cutler-Vallisneri formalism to compute metric shifts caused specifically by PSD drift.

The result is a set of analytic, causal, and computationally cheap corrections for the timing and phase biases. These first-order corrections for timing (\(\delta t^{(1)}\)) and phase (\(\delta \phi^{(1)}\)) are power-weighted spectral averages of the whitening phase mismatch:

\[ \delta t^{(1)} = \frac{\int (2\pi f) \Phi_a(f) w(f) df}{\int (2\pi f)^2 w(f) df}, \quad \delta \phi^{(1)} = \frac{\int \Phi_a(f) w(f) df}{\int w(f) df} \]

Here, \(\Phi_a(f)\) is the whitening phase perturbation induced by the drift, and \(w(f)\) is the whitened signal power.

Figure 1: Distributions of the timing and phase biases induced by a 1-week PSD lag across ~160 GWTC-4.0 event-detector pairs. Timing biases exhibit a heavy tail exceeding 200 µs, while phase biases are generally confined to 0.2 rad.

Validation and Impact

We validated this framework using real GWTC-4.0 strain data. As seen above, uncorrected 1-week PSD drift can induce severe systematics. Beyond coordinate shifts, this drift also directly suppresses the recovered signal amplitude.

Figure 2: Distribution of the fractional SNR loss caused by uncorrected PSD drift. The median sensitivity loss of 3-5% translates to a ~15% reduction in overall detection volume.

Applying our mathematical framework successfully restores these parameters. Crucially, this eliminates the artificial broadening of early-warning sky maps, ensuring rapid alerts remain highly accurate for telescopes hunting for electromagnetic counterparts.

Figure 3: Systematic shifts in sky-localization (Right Ascension and Declination) induced by uncorrected PSD drift. While the median shift is modest, severe outliers can experience pointing errors between 5° and 10°.

For the complete derivation, see the full paper  [2].

References

[1]

L. Tsukada, K. Cannon, C. Hanna, D. Keppel, D. Meacher, and C. Messick, Application of a zero-latency whitening filter to compact binary coalescence gravitational-wave searches, Phys. Rev. D 97, 103009 (2018).

[2]

J. Kennington, Dual cutler-vallisneri corrections: Mitigating PSD drift in zero-latency gravitational-wave searches, arXiv:2603.09132 (2026).