// Wireless Fiber · An Open Systems Inc. Division

Fiber speed.
Moving target.

A trackside optical corridor that delivers up to a continuous 160 Gbps to trains at line speed, built entirely from components shipping today. No trenched fiber, no spectrum, no moving parts. Four 40 Gbps laser links, two at each end of the train, alive at all times and handed pole to pole at full line speed, 350 km/h and beyond.

2 km pitch

PAIR A head terminal, dual 40G forward to next pole

PAIR B tail terminal, dual 40G aft to previous pole

handoff = terminal swap · zero blackout

40 Gbps
per optical link, today
4 links
two per train end
160 Gbps
bonded aggregate
0
moving parts

// The Problem

The last unconnected mile moves at 350 km/h

A modern train is a rolling data center: forward-facing cameras, condition monitoring on every axle, signaling telemetry, and a thousand passengers holding screens. It generates terabytes per day. It connects like a phone from 2015.

Cellular was built for handsets, not for a kilometer of steel moving through its cells at line speed. Trenching new fiber along the right of way costs more per kilometer than most corridors can justify, and even where lineside fiber exists, it stops at the ballast: it never reaches the train itself. The gap is not bandwidth in the ground. The gap is the last 2.5 meters, at speed.

Cellular / FRMCS to train

10s–100s of Mbps, shared

Trenched lineside fiber

high cost per km, never reaches the train

Satellite

latency + capacity ceilings

Wireless Fiber corridor

40–160 Gbps, continuous, to the vehicle

// The System

Two terminals. A line of poles. A handoff that never drops.

The architecture has three parts, and the third one is the trick.

01

Poles at roof height

Lightweight trackside poles, 2.5 m from the train centerline, carry optical terminals mounted level with the train roof. Matched height means the vertical pointing angle is near zero at every range: the entire problem collapses to a single steering axis.

02

Dual links at each end

Every train carries a terminal at each end, and every terminal runs two independent 40 Gbps optical chains, matched by two chains on the pole face it works. The head pair looks forward, up to 2 km to the next pole. The tail pair looks aft. All four links carry traffic simultaneously.

03

Handoff is a terminal swap

As the train reaches a pole, the head pair releases it and the tail pair collects it astern seconds later, at the same shallow angle. The other end's pair holds its own pole throughout. No beam ever has to chase a target past its window, and connectivity never blacks out.

Design doctrine: never fight the geometry. Conventional tracking systems chase targets through wide angles with fast mechanics. This system is arranged so the target barely moves: a pole 2 km ahead sits 0.07 degrees off the axis of travel, and by the time the angle gets difficult, responsibility has already passed to the other end of the train. The budgets are set by geometry, not speed: they hold unchanged from freight running to 350+ km/h high-speed service, and the corridor adds only microseconds of latency, because there is no radio stack between the train and the fiber.

WF-OPT-001 · SYSTEM CONCEPT

// The Geometry

The physics is on our side

At 2.5 m lateral offset, grazing geometry does most of the engineering. From 100 m out to 2 km, the azimuth to the pole sweeps just 1.4 degrees down to 0.07. The whole working link lives inside a 3 degree cone.

012345602505007501000steering window ≤ 3°drop / acquiredegdistance ahead of pole (m)
FIG 01 · azimuth to pole vs distance, 2.5 m lateral offset, matched-height mounting · beyond 100 m the pole never leaves a ±1.5° cone

The steering system is a stabilizer, not a tracker. The train's trajectory consumes about 1.5 degrees across an entire pole pass. Car-body sway and yaw contribute as much or more. So the beam steerer spends its life canceling vibration around a nearly fixed pointing solution, fed forward from inertial sensing and a surveyed pole database, with a beacon loop cleaning up the residual at multi-kilohertz rates.

WF-OPT-001 · POINTING BUDGET

The train spans part of the gap. Link range is terminal to pole, not train to pole: with poles at 2 km pitch, the head and tail links share only the pitch minus the train length between them. A 400 m trainset never needs more than 1.6 km of reach and averages 800 m. A 1.5 km freight consist never needs more than 500 m, and one of its terminals is always within burst range of a pole. Longer trains make the corridor easier, not harder.

WF-OPT-001 · CONSIST GEOMETRY

// The Terminal

One steered aperture per end. Nothing moves.

Twenty years of continuous rail vibration is the death of mechanisms. So the terminal contains none: no gimbal, no mirror, no motor. Beam steering is done electronically, in silicon, by a solid-state metasurface that redirects light in about 100 microseconds.

FIG 02 · transmit chain, one per train end · receive path: 100 mm telescope, wavelength demux, FEC tuned for atmospheric turbulence

Why zero moving parts is the spec

A mechanical pointer on a railcar is qualified to survive vibration. Surviving qualification is not the same as surviving two decades of continuous broadband excitation with no maintenance access, across a fleet. When nothing moves, the reliability problem migrates to solder joints, ingress, and mounting: standard rolling-stock electronics practice, engineered against EN 61373, with no mechanism lifetime anywhere in the failure tree.

Why 1550 nanometers

The 1550 nm C-band carries the entire telecom component ecosystem: mature lasers, amplifiers, and receivers at commodity cost, wavelength multiplexing for capacity growth, and the most favorable eye-safety physics of any band, which is what makes a 1 W transmitter workable in a rail corridor under IEC 60825 assessment with active interlocks.

// The Trackside

The pole never steers. It only switches.

Here the constraint that defines railways becomes an optical design tool: the train can only be on the track. The target locus is a known one-dimensional curve, so the pole's beams are not angle sectors, they are range bins along the track. Each pole face carries two parallel chains of six fixed 1 W transmitters, one chain per train link, each statically aimed at its bin. The only control action at the pole is chip select.

constant ~4 m spot at every bin edge → constant received powerB1 50–100.B2 100–200.B3 200–400.B4 400–800.B5 800–1600.B6 1.6–2 km.POLE2.5 m offsetTRAIN · B3 ACTIVENEAR BINS: WIDE BEAMSshort range affords wide divergence, which also covers the bin's azimuth spreadFAR BINS: NARROW BEAMS2 mrad at 2 km, aimed 0.08° off the track axis; the target barely moves in angle
FIG 03 · pole beam plan, lateral scale exaggerated · log-spaced range bins, ratio 2 per bin · one bin lit at a time, two during handoff
Pole range bin plan
BinRange (m)Azimuth aimDivergenceSpot @ far edgeDwell @ 350 km/h
B61600–20000.08°2.0 mrad4.0 m4.1 s
B5800–16000.13°2.5 mrad4.0 m8.2 s
B4400–8000.25°5.0 mrad4.0 m4.1 s
B3200–4000.5°10 mrad4.0 m2.1 s
B2100–2001.0°20 mrad4.0 m1.0 s
B150–1001.9°40 mrad4.0 m0.5 s
Every bin closes the same budget: 1 W → constant spot → flat received power across the entire 50 m to 2 km sweep

Constant spot

Divergence is graded per bin so the spot at each bin's far edge is the same 4 m. Geometric loss is identical everywhere: one link budget, no receiver gain transients, anywhere in the approach.

Trivial switching

Five chip switches per pole pass. Even at 350 km/h the fastest handoff window is half a second against sub-millisecond switch time: nearly three orders of magnitude of margin, driven by beacon signal vote at the pole and position handover pole to pole over the lineside fiber.

Graceful failure

A failed chip removes one range bin: the link closes later in the approach, nothing else changes. Fog removes bins far to near by the same mechanism. One switching logic handles component failure and weather identically.

FIG 04 · one pole pass at 350 km/h · the train's other link pair holds against the neighboring pole throughout: continuity never depends on one pole

Both ends of the link get simpler. Each pole carries two mirror-image faces, and each face carries a parallel pair of six-bin chains: twenty-four fixed transmitters per pole, serving the approaching head pair on one face and the departing tail pair on the other. A failed transmitter costs one bin on one chain of one face, and the partner chain covers through it. Every element that could have been a precision tracking mechanism has been replaced by geometry and a switch.

WF-OPT-002 · POLE BEAM PLAN

// Capacity

40 gigabits per link. Today's parts.

Every link carries four independent 10 Gbps channels through one aperture. Two links per train end, each matched by its own trackside chain: 80 Gbps per end, 160 across the train. Shipping components, no roadmap.

CH1 · 1550 nm

10 Gbps

CH2

10 Gbps

CH3

10 Gbps

CH4

10 Gbps

Per link 40G

Four 10G channels through a single 1 W aperture. Shipping silicon: no development risk anywhere in the data path.

Per end 80G

Two independent chains per terminal, matched by dual chains on the pole face. Bond the pair, or run it 1+1 protected at 40G.

Per train 160G

Four links live across head and tail. Bond everything for 160 Gbps aggregate, or hold full 1+1 protection at both ends and run a guaranteed 80G.

What a continuous 160 Gbps means in practice: a ten minute run moves 12 terabytes. A full day of operation drains every camera, every sensor, and every passenger session on the consist with room to spare, in both directions. And growth stays a channel-rate card swap on the same optics, the same poles, the same corridor.

WF-OPT-001 · SERVICE MODEL

// Availability

Weather shortens the link. It doesn't sever it.

The range to the working pole sweeps continuously from 2 km down to zero on every pass. That geometry turns weather from an outage into a narrower window.

Clear air

17 dB of link margin at full 2 km range: 10G receivers buy extra sensitivity. The link closes everywhere on the corridor, both directions, all channels.

MODE: FULL RATE, ALL FOUR LINKS

Heavy rain

The margin absorbs roughly 8 dB per kilometer of attenuation, which covers heavy rainfall at full pole pitch.

MODE: FULL RATE, ALL FOUR LINKS

Dense fog

No 2 km optical link survives dense fog. But the link still closes inside 300 to 400 m of each pole, so the train buffers between poles and bursts on approach, and priority traffic rides the front of each burst window. Long consists barely notice: with a 1.5 km train at 2 km pitch, one terminal is always within burst range of a pole, and coverage never gaps at all. Safety signaling stays on the operator's existing train control layer, untouched.

MODE: BUFFER + BURST

Pure optics, layered anyway. There is no radio in this system. Resilience comes from four independent optical links, buffer-and-burst behavior that degrades with weather instead of failing, and the existing lineside fiber rings behind the poles, with ring protection switching beneath everything. Safety-critical signaling remains where operators already run it, on their existing train control systems: this corridor is the capacity layer, and it is engineered so that no single link, chain, or fiber path failure isolates the train.

WF-OPT-001 · AVAILABILITY MODEL

// Deployment

Pole pitch is set by sightline, not physics

A beam running parallel to the ground at roof height for 2 km must clear everything in the corridor at that height. So deployment is a survey discipline, governed by five rules.

  1. R-01

    Full pitch on straight runs

    Open, straight territory takes the full 2 km pole spacing, where the link budget, not the sightline, is the binding constraint. And because link range is terminal to pole rather than train to pole, freight territory running long consists can stretch the pitch further, by roughly the consist length.

  2. R-02

    Poles on the outside of curves

    On curved territory the consist itself blocks the inside chord. Poles sit on the outside of curves at reduced pitch, sized so at least one terminal always holds line of sight.

  3. R-03

    Tighter pitch through obstructions

    Cuttings, overbridges, signal gantries, catenary territory, and vegetation each shorten the usable sightline. Pitch compresses locally rather than the system failing globally.

  4. R-04

    Survey first, then build

    The first field task on any corridor is a line-of-sight survey at terminal height: a LiDAR scan and desktop exercise on rights of way that are already mapped. The survey output is the pole plan.

  5. R-05

    Mid-mounts when the math asks

    Where stretched pitch, short trainsets, or fog statistics leave the coverage arithmetic thin, an optional mid-consist terminal with forward and aft apertures adds two more links: another 80 Gbps, a third connection point never more than half a pitch from a pole, and burst windows twice as frequent in fog. Coverage scales on the vehicle, in a depot, not in trackside civil works.

Backhaul: the fiber is already there. Rail corridors carry lineside fiber on protected rings, and the poles tap those rings rather than requiring new construction. The corridor adds no trenching: it is the last 2.5 meters on top of fiber the railroad already owns, terminating in standard 100GE handoffs at the operator's aggregation points.

WF-OPT-001 · CORRIDOR ENGINEERING

Nothing to license. Railside RF is a heavy process: spectrum filings, interference studies, coordination with PTC, GSM-R, and FRMCS systems, multiplied by every jurisdiction a corridor crosses, with renewal risk forever after. An optical corridor files nothing. Light is license-free worldwide, regulated by eye safety rather than spectrum authorities, and it cannot interfere with the radio systems the railroad already depends on. Deployment moves at construction speed, not docket speed.

WF-OPT-001 · REGULATORY MODEL

// What It Unlocks

Built for the railroad that's coming

UC-01 · TRACK SAFETY

Prevent the disaster before the next train

Every revenue train scans the track it rides: lidar, forward vision, geometry and vibration signatures, streamed at full rate and analyzed before the following service arrives. Broken rail, buckling, washout, obstruction: found by the last train, fixed before the next. Inspection stops being a quarterly special train and becomes something the timetable does dozens of times a day, for free.

UC-02 · LIDAR SCANNING

A digital twin that refreshes itself

Full-resolution lidar of track, catenary, structures, and vegetation rides along on every pass, and the corridor carries the point clouds off in real time. The network becomes a living survey: clearances verified, encroachment tracked, and every asset measured continuously instead of annually.

UC-03 · AI TRAINING DATA

The fleet becomes a data engine

Synchronized video, lidar, and sensor streams are exactly the multimodal corpus that rail autonomy and defect-detection models train on, and today it is discarded at the source for want of a pipe. With the corridor, every run feeds the models that protect and automate the next one.

UC-04 · SENSORS + TELEMETRY

Condition monitoring at full fidelity

Today's rolling stock telemetry is decimated to fit the pipe. With the pipe removed as a constraint, every axle, bearing, pantograph, and brake streams raw, and predictive maintenance models train on complete data instead of samples.

UC-05 · CAMERA SECURITY

Every car on a live feed

Onboard CCTV stops being a recorder that gets reviewed after the fact and becomes a live feed: every car, full resolution, monitored in real time, with forward and rear video preserved off the train the moment it is captured. Evidence that cannot be lost with the vehicle.

UC-06 · LINK SECURITY

A beam you'd have to stand in

A 2 mrad optical link has no sidelobes to sniff and nothing to jam from a distance: intercepting it means physically standing in the beam, on railroad property, at roof height. No RF signature, a deterministic fiber-fed path behind the poles, and encryption end to end on top.

UC-07 · AUTONOMY

Autonomous and remote operations

Remote supervision, teleoperation fallback, and full-resolution forward video need corridor-grade uplink, not best-effort cellular. A continuous 40 Gbps per link makes the vehicle observable in real time, and corridor latency is measured in microseconds, not the tens of milliseconds a radio access network adds.

UC-08 · WIFI + INTERNET

Passenger connectivity that actually works

A trainload of passengers sharing cellular capacity is the connectivity most riders know. A 160 Gbps trunk to the consist makes onboard access feel like home broadband, tunnels and cuttings aside.

The use cases compound. Trains scan the track, the corridor moves the data, models learn the failure signatures, and the next train rides safer rail: capacity, safety, and intelligence feeding each other on every pass. The corridor is not a pipe with applications attached. It is the sensory system of the railroad.

WF-OPT-001 · APPLICATION MODEL

// Reference

System specification

Band

1550 nm C-band, four multiplexed channels per link

channel rate 10G, shipping component set · capacity grows by channel rate on the same optics

Transmit power

1 W (+30 dBm) aggregate, post-amplifier

Beam divergence

~2 mrad

~4 m spot at 2 km against a 100 mm receive telescope

Range

0 to 2 km per link, continuously swept every pole pass

worst-case reach = pole pitch minus train length: a 400 m trainset never exceeds 1.6 km, a 1.5 km consist never exceeds 500 m

Capacity

40G per link · 80G per end (dual links) · 160G bonded aggregate

four links live: bond all, or hold full 1+1 protection at both ends for a guaranteed 80G · upgrades by channel rate, not corridor rebuild

Steering

Electronic, solid-state metasurface, azimuth only, ±3° window

~100 µs response · stabilization via inertial feed-forward + beacon closed loop

Trackside terminal

Dual parallel six-bin chain sets per face, statically aimed, switch-only

two mirror faces per pole, 24 fixed transmitters · log-spaced bins, ratio 2 · constant ~4 m spot, flat received power 50 m to 2 km

Handoff

Dual-terminal make-before-break

head releases at ~50 m, tail acquires astern, second link holds throughout

Latency

Propagation-limited: ~7 µs per 2 km hop

no radio scheduling stack · light in air outruns light in fiber by ~50% per km

Line speed

Full performance to 350+ km/h · budgets set by geometry, not speed

fastest switching window 0.5 s at 350 km/h vs sub-ms switch time · HSR adds aero fairings and clearance-envelope pole placement

Mounting

Roofline train terminals, matched-height trackside poles at 2.5 m lateral offset

optional mid-consist terminal with forward and aft apertures for stretched pitch, short trainsets, or fog-prone territory

Environment

EN 61373 shock and vibration · −40 to +85 °C · sealed heated windows · zero moving parts

Safety

IEC 60825 corridor assessment

interlocks on beacon loss and odometry fault · transmit inhibit zones at crossings and stations

Spectrum

License-free: optical, no spectrum authority filings

no interference with or from PTC, GSM-R, FRMCS · no coordination zones, no renewals · governed by IEC 60825 eye safety instead

Weather posture

Full rate through heavy rain · buffer-and-burst proximity mode in dense fog · priority traffic leads each burst window

Resilience

Pure optics, layered: four optical links (dual per end) over existing lineside fiber rings

ring-protected paths the railroad already operates · no single link, chain, or path failure isolates the train · safety signaling stays on existing train control systems

// Next Step

Pick a corridor. We'll bring the math.

The pilot program starts with a 10 to 20 km corridor segment: we run the line-of-sight survey at terminal height, produce the pole plan and per-segment link budget, and deploy a head-and-tail demonstration on revenue or test rolling stock. Every figure on this page is indicative and gets replaced by your corridor's numbers.