Unified matched-filter and excess-power pipeline for gravitational-wave detector characterization
Abstract
We present sgnax, an open-source pipeline for gravitational-wave detector characterization that produces matched-filter and excess-power transient triggers within a single streaming dataflow graph. Built on the Stream Graph Navigator (sgn) framework, sgnax combines the multi-rate sine-Gaussian matched filtering of snax and the multi-resolution Q-transform of omicron in a Python-native package with shared configuration and output formats. Strain and auxiliary channels share data ingestion and whitening, reducing per-channel processing costs by a factor of 2.8 from one to 32 channels for the Q-transform. A full day of 16 kHz strain is processed in 14 minutes; four processes sharing an NVIDIA V100 analyze 792 auxiliary channels at 1.11 times real-time throughput. The same configuration supports offline and online operation through frame caches, shared memory, and the arrakis distribution service. Matched-filter triggers have approximately five seconds of end-to-end latency, while Q-transform triggers are available within tens of seconds. Injection studies recover 99.4% of recoverable sine-Gaussian signals and show broadband burst recovery consistent with established generators. On 24 hours of LIGO strain, sgnax agrees with production omicron to 0.1% in total trigger count, with per-trigger signal-to-noise ratios agreeing at the percent level across frequency bands. Auxiliary-channel comparisons recover 80% of the snax loud-feature population per time bin. The validation also reveals a multiband amplitude error in production snax that inflates reported signal-to-noise ratios below 25.6 Hz by factors of sqrt(2) to 2.
Related
- Agentic AI for Gravitational Wave Data Analysis: A Head-to-Head Comparison of Coding Agents Executing a Matched Filter Pipeline on Einstein Telescope Simulated Data
- Characterization of integrating-sphere systematic effects for photon calibration of the KAGRA gravitational wave detector
- Detector orientation matters: Consequences for gravitational-wave background detectability
- Polarized Anisotropic Stochastic Gravitational Wave Background Search with Ground-Based Detector Networks
- Inferring Properties of Rotating Core-Collapse Supernovae with Gravitational-Wave Detector Networks
BibTeX
@misc{yarbrough2026unified,
title = {{Unified matched-filter and excess-power pipeline for gravitational-wave detector characterization}},
author = {Zach Yarbrough and Olivia Godwin and Derek Davis and Gabriela González},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2608.09804},
eprint = {2608.09804},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.09804},
}
RIS
TY - UNPB
TI - Unified matched-filter and excess-power pipeline for gravitational-wave detector characterization
AU - Zach Yarbrough
AU - Olivia Godwin
AU - Derek Davis
AU - Gabriela González
PY - 2026
DA - 2026/10/27
PB - arXiv
DO - 10.48550/arxiv.2608.09804
UR - https://arxiv.org/abs/2608.09804
AB - We present sgnax, an open-source pipeline for gravitational-wave detector characterization that produces matched-filter and excess-power transient triggers within a single streaming dataflow graph. Built on the Stream Graph Navigator (sgn) framework, sgnax combines the multi-rate sine-Gaussian matched filtering of snax and the multi-resolution Q-transform of omicron in a Python-native package with shared configuration and output formats. Strain and auxiliary channels share data ingestion and whitening, reducing per-channel processing costs by a factor of 2.8 from one to 32 channels for the Q-transform. A full day of 16 kHz strain is processed in 14 minutes; four processes sharing an NVIDIA V100 analyze 792 auxiliary channels at 1.11 times real-time throughput. The same configuration supports offline and online operation through frame caches, shared memory, and the arrakis distribution service. Matched-filter triggers have approximately five seconds of end-to-end latency, while Q-transform triggers are available within tens of seconds. Injection studies recover 99.4% of recoverable sine-Gaussian signals and show broadband burst recovery consistent with established generators. On 24 hours of LIGO strain, sgnax agrees with production omicron to 0.1% in total trigger count, with per-trigger signal-to-noise ratios agreeing at the percent level across frequency bands. Auxiliary-channel comparisons recover 80% of the snax loud-feature population per time bin. The validation also reveals a multiband amplitude error in production snax that inflates reported signal-to-noise ratios below 25.6 Hz by factors of sqrt(2) to 2.
ER -