Wayside Fiber: Three Layers, Three Failure Modes
A three-part series on the fiber plant beneath modern railroad signalling: where the cable physically runs, how the network on it talks, and who keeps the record true. One hypothetical corridor, carried through all three parts.
Ask a signal engineer what the railroad runs on and you will hear about track circuits, interlockings, and vital processors. All true. But walk into any modern wayside house and follow the equipment back far enough, and everything lands in the same place: a yellow jumper, a distribution panel, and a buried cable full of glass. The vital comms between control points ride it. The crossing health monitors ride it. The defect detectors, the event recorders, the cameras, the maintainer's phone: all of it, one cable. The wayside fiber network is no longer a utility bolted onto the signal system. It is the substrate the signal system rides on.
That substrate can fail three different ways, and the three ways have almost nothing in common.
It can fail physically. A backhoe, a lightning-shattered pole line, water in a splice enclosure, or simply a loss budget that never closed. Physical failures are loud. You know within seconds, and the fix is a crew, a splice trailer, and a long night.
It can fail logically. The glass is perfect and the network is useless: a spanning tree that reconverges slower than a vital processor's patience, a duplicate IP address that surfaces two years after somebody cloned a config, a flat subnet that quietly ran out of room. Logical failures are confusing. They look like hardware, they get chased like hardware, and they are never hardware.
And it can fail institutionally. Nothing breaks at all. The cable outlives the person who documented it, the as-builts drift eighteen months behind the plant, the strand assignment lives in a spreadsheet on a laptop that gets reimaged, and one day a routine repair takes four times longer than it should because nobody living knows which tube feeds what. Institutional failures are silent, and they are the ones that turn the other two kinds from incidents into ordeals.
Three layers, three failure modes, three very different disciplines for preventing them. That is the series.
The corridor
Everything in these three parts happens on the same hypothetical railroad: a 49.5-mile corridor with six locations along it, from the office at Waverly to the junction at CP Chester, tied together by a 48-strand single-mode trunk, plus one deliberate exception to "single cable" that Part 1 pays for in full: a small second cable on a physically separate route, because a safety-critical ring is only a ring if one dig-in cannot take both directions of it.
The running example: one corridor, six locations, 49.5 miles. Every diagram in the series comes back to it.
Why hypothetical? Two reasons. First, real fiber topology is security-sensitive information; the routes, splice points, and hut locations of an operating railroad's comms backbone are exactly the things you do not publish. Second, and honestly more important: a controlled example teaches better. On a real corridor the interesting decisions are buried under a century of accumulated exceptions. On this one, every strand, every address, and every dollar is visible, so when we spend one, you see the balance change.
The names and mileposts are invented. The engineering is not.
The three parts
Part 1: Where It Is is the physical plant: why signalling went to glass in the first place, what is actually inside a 48-strand cable, how a midspan access works, and how a loss budget quietly fails at the far end of the corridor. It ends with a link LED lit at every location and nothing able to reach anything.
Part 2: How It Talks is the logical plant: which devices get to share a wire with vital equipment, how an addressing schema encodes the corridor into four octets, and a race you have probably never watched on purpose: ring reconvergence against a vital processor's stale-data timer. It also pays off a small disaster this paragraph is quietly planting: somewhere in Part 2, a replacement recorder comes out of its box already owning another location's address.
Part 3: Who Keeps It True is the institutional plant: why the capital grant that built the corridor funds nothing that keeps its records accurate, how an OTDR baseline doubles as the condition assessment your asset management program keeps asking for, and why a documented change process is the only real difference between a record and a rumor.
The cable will last thirty years. Read in order.