LEVIATHAN SYSTEMS
← Back to Glossary

What Is Ferrule?_

In fiber optic connectors, the ferrule holds the stripped fiber in a centered bore and provides a polished end-face for low-loss light transmission. Ferrules are typically made of ceramic, metal, or plastic, with ceramic being most common for single-mode and high-performance multimode applications. The ferrule's geometry—including its outer diameter, concentricity, and end-face curvature—directly affects insertion loss and return loss.

Technical Details

Ferrules are precision-machined to tolerances on the order of microns, with the fiber epoxied or crimped inside the bore. The end-face is polished to a specific radius (e.g., PC, UPC, or APC) to minimize back-reflection. Connector types like LC, SC, and MPO use ferrules of different diameters (1.25 mm for LC, 2.5 mm for SC, and multi-fiber arrays for MPO). Proper cleaning and inspection of ferrules are critical to avoid contamination that can cause signal degradation or permanent damage to mating interfaces.

How Leviathan Systems Works with Ferrule

In our GPU rack builds, we handle hundreds of MPO ferrules per rack for the scale-out InfiniBand/Ethernet fiber runs, and each ferrule must be inspected and cleaned before mating to avoid costly rework. Ferrules on NVLink copper spine cables are not applicable, as those use electrical connectors without fiber alignment components.

Appears In

Transceiver Breakout & Splitter Cabling for GPU FabricsOptical Transceiver Handling & Cleanliness on the FloorNCCL All-Reduce Validation as a Cluster Acceptance GateInfiniBand NDR/XDR vs RoCE: What Changes for the Cable PlantPre-Terminated Trunks vs Field Termination for GPU FabricDAC vs AOC vs AEC: Choosing GPU Rack Interconnect CablesSingle-Mode vs Multimode for AI Fabric: The Cable-Plant DecisionFiber Labeling at Scale: A Naming Scheme That Survives a 50,000-Link ClusterCable Pathways & Containment for GPU Rooms: Overhead Tray, Ladder, Fiber RunnerPatch-Panel & Cassette Design for GPU Halls: Breakout, Density, ServiceabilityMPO Trunk Planning & Fiber-Count Math for GPU HallsAMD MI300X vs NVIDIA H100: Infrastructure & Deployment DifferencesNVIDIA GB300 NVL72 Explained: Specs, Power, and What It Takes to DeployDell vs Supermicro vs HPE GPU Servers: A Deployment ComparisonNVIDIA H100 vs H200 vs B200: What Changes for DeploymentData Center Structured Cabling Standards: TIA-942, TIA-606-C, BICSICommon GPU Deployment Mistakes — and How to Avoid ThemHow Long Does GPU Cluster Deployment Take?GPU Commissioning & Acceptance: What to Demand Before You Sign OffSite Readiness Before the GPUs Arrive: Power, Cooling, Floor, PathwaysGPU Rack Assembly: What Drives the CostSpectrum-X vs InfiniBand: What's Different for the Cable PlantPre-Power Inspection: The Walkdown Before Energizing a GPU HallNCCL Bandwidth Validation: Proving a GPU Fabric Before ProductionOTDR & Insertion/Return-Loss Testing for GPU Cluster FiberMPO Polarity (Method A/B/C) for GPU Fabric — and the #1 Cause of Dead LinksFiber Cleaning & Inspection SOP for AI Interconnects (IEC 61300-3-35)