Poster Session Abstracts
Graph Intelligence for Benchmarking Optical Networks Akanksha Ahuja, University of Cambridge Topology Bench advances benchmarking in wavelength-routed optical networks. Our work has three key contributions. First, we provide the most extensive open-access dataset, comprising 105 georeferenced real-world core networks and 270,900 synthetic topologies. Second, we integrate structural, spatial, and spectral metrics to capture the graph-based characteristics of optical networks. Third, we provide a systematic framework to select representative sets of topologies based on unsupervised learning to organise real networks into objectively defined clusters. These three contributions provide three key benefits.
Novel Optimisation Problems Enabled by Fast Coherent GN Model Network Simulation Michael Doherty, PhD Student, University College London
Simulators for optical networks are designed to provide accurate estimates of network performance under varying traffic conditions. This allows for optimized allocation of network resources, e.g. routing, bandwidth, and launch
powers. However, existing simulators either provide accurate physical layer models that are slow to compute (e.g. GNPy) or use a basic physical layer model to reduce computation time (e.g. optical-rl- gym). We present our network simulator, XLRON-GN, that provides accurate per channel GSNR estimation through the closed-form ISRS GN model and GPU-based computation, accurate to 0.2dB over C+L bands. It provides sufficiently high throughput (>1000 new lightpath connections per second) to enable new optimization problem settings. We demonstrate per-channel launch power optimization in a dynamic C+L elastic inter-DC network. This is the first time launch power has been optimized for all paths in a network. We apply reinforcement learning to the problem and benchmark other optimisation techniques, demonstrating significant capacity increase..
Scan to watch poster preview.
AI-Driven Earthquake Early Warning Using Optical Fiber Networks Hasan Awad, Politecnico di Torino
Scan to watch poster preview.
Our research leverages a groundbreaking use of the entire traffic-carrying optical fiber terrestrial network for early earthquake detection and epicenter localization. This Artificial Intelligence driven approach focuses on analyzing polarization changes in light traveling through optical fibers
Quantum Data Centre: Requirements and Architecture Emilio Hugues Salas, BT
to distinguish between external events, primarily detecting the arrival of primary earthquake wave, that precedes the destructive wave by tens of seconds, in a noisy environment. Tested in real-world scenarios over a 38 km fiber cable deployed in the city of Turin, Italy, this system successfully offers urban areas near the epicenter a 21 to 57 seconds alert time window before severe shaking occurs. These early warnings enable prompt earthquake countermeasures and swift initiation of emergency plans to mitigate the humanitarian and economic impacts, particularly in densely populated regions. This technique is cost- effective and reliable, requiring no dedicated or expensive equipment, and can be utilized to sense other external events beyond seismic activity. .
We discuss different options for low-complexity multipath interference (MPI) mitigation that effectively suppresses MPI noise in experimental 112GBd PAM4 signals. We remove the DC offset before and after feed-forward equalization(FFE), in combination with post filter and different
memory maximum likelihood sequence estimation (MLSE). We add simulations on the feasibility of 400 Gb/s PAM4/PAM6 transmission under disturbance by MPI.
5G fronthaul with analog radio over fiber link
Scan to watch poster preview
Krishna S Kumar, Avirata Defence Systems Analog radio over fiber (ARoF) link for 5G new radio (NR) fronthaul for the FR2 band is presented with design and performance analysis based on numerical simulations. The link is designed to operate in the frequency range of 24.25 GHz to 27.5 GHz with a gain of 0 dB ± 1 dB. The
MPI-Penalty-Free S-Band Transmission over G.654.E- Compliant Fibres Romulo Aparecido, UCL
simulation model includes the impact of the various sources of noise and nonlinearity in the link that affect the link performance in terms of noise figure, saturation and spurious-free dynamic range (SFDR). The designed link exhibits an SFDR of >95 dB/Hz2/3 with an input 1 dB saturation power (P1dB) of -3.6 dB and a noise figure of 34 dB. The simulation model is capable of scaling to higher or lower frequencies of operation and is also extendable to multichannel DWDM ARoF links.
G.654.E-compliant fibres feature ultra-low loss and a large effective area (~125 μm² compared to the standard 80 μm²), reducing nonlinear interactions during propagation. However, their maximum allowed cable cutoff wavelength of 1530 nm raises concerns for S-band transmission
(1460–1530 nm) and potential multipath interference (MPI). For the first time, we experimentally demonstrated MPI-penalty-free S-band transmission over G.654.E fibres, including Corning® Vascade® EX2500 and Corning® TXF® fibres, by comparing their performance to a variable optical attenuator (VOA) with equivalent loss. Launch power was optimised for maximum SNR in the linear regime, limiting impairments to transceiver noise, ASE noise, chromatic dispersion (fully compensated digitally), and MPI. If present, MPI would introduce an SNR difference between fibre and VOA transmission. We transmitted 22 channels (1474–1525 nm) over 80 km and 160 km for 16 GBaud and 64 GBaud rates. SNR differences between fibre and VOA were below 0.4 dB, confirming penalty-free S-band operation.
Scan to watch poster preview.
Scan to watch poster preview.
14
Made with FlippingBook flipbook maker