First Demonstration of Single-ended OTDR GFE Identification and Correction over a 100-km Hollow-core Fiber Link

Hollow-core fibers (HCFs) are being deployed at an accelerating pace in scenarios where every microsecond matters—high-frequency financial trading, AI cluster interconnects, and cross-border low-latency backbones. In these latency-critical applications, the operational reliability of HCF links is as decisive as raw transmission performance. Optical time-domain reflectometry (OTDR) is the established tool for fiber diagnostics, yet HCF‑OTDR traces are fundamentally challenged because local gas pressure variations at splices induce kilometer-scale backscattering bumps that severely obscure weak loss and reflection signatures. In this work, these backscattering features are referred to as gas-filling events (GFEs). Bidirectional two-way OTDR processing provides an effective physical reference for HCF trace correction because it separates direction-independent backscattering-coefficient variations from direction-dependent accumulated loss. However, its field deployment requires access to both fiber ends, temporally consistent gas states, and accurate trace registration, which are inconvenient for in-service long-haul HCF operation and maintenance (O&M). To overcome this bottleneck, we propose the first single-ended, event-aware GFE identification and correction scheme based solely on one-way OTDR trace data. The scheme models GFEs as wide-scale near-Gaussian bumps, adaptively localizes GFEs, shields embedded reflection-peak regions, applies loss-aware branch selection, and estimates the broad GFE component while reducing the risk of absorbing observed event-like features into the fitted bump. To the best of our knowledge, this is the first experimental demonstration of comprehensive single‑ended GFE correction over a challenging, deployment-relevant 100‑km HCF link comprising 20 spliced segments with multiple GFEs and complex event superpositions—a link length and architecture directly comparable to recently deployed cross‑border financial trading routes. The proposed scheme identifies and corrects all observable GFE-induced uplift regions, achieving a normalized GFE correction degree exceeding 94% and maintaining a shape correlation above 90% with the bidirectional GFE correction baseline, while avoiding the negative peaks and oscillatory residues that can appear in dual-trace combination near strong reflections. This work circumvents the two-end-access requirement of conventional GFE correction and delivers a robust, artifact-free diagnostic tool for scalable, high-precision operation and maintenance of future latency-intensive HCF networks.