Ekinops Dark Fiber Learning Path
Stage 4

Optical Transport

The building blocks that turn lit fiber into real services. Once you can name each component, say what it does, and say where it sits on the path, any Ekinops design becomes readable — and any fault becomes a place to look rather than a mystery.

How to read this page

Every component below does exactly one of four jobs: convert a signal (transponder/muxponder/regen), combine or split wavelengths (mux/demux, OADM/ROADM), add power (EDFA/Raman), or carry management (OSC). Keep sorting each part into one of those four buckets and the whole system stays legible.

Client side vs line side

Mental model

Every transport device has two faces. The client side faces your world — routers, switches, servers, SAN — usually with familiar grey short-reach optics (SR/LR pluggables at a standard nominal wavelength like 1310 nm, all the same colour, cheap, interchangeable). The line side faces the optical network — a specific coloured DWDM wavelength on a precise ITU-grid channel, often generated by a tunable laser so one card can be set to any channel. The device's whole job is to bridge those two worlds.

End-to-end optical transport path, client side versus line side Left to right: Router A on the grey short-reach client side connects to a transponder or muxponder that performs optical-electrical-optical conversion. Its coloured line side feeds a multiplexer, then a booster EDFA, launching many wavelengths onto the dark fiber. At the far end a pre-amplifier and demultiplexer feed a far transponder and Router Z, back on the grey client side. The client side uses interchangeable grey optics; the line side uses a specific coloured DWDM wavelength on the ITU grid. CLIENT · grey LINE SIDE · coloured DWDM λ (ITU grid) CLIENT · grey Router Aclient Transponder /Muxponder(O–E–O) Mux BoosterEDFA Pre-amp+ Demux Transponder Router Zclient dark fiber · many λ
End-to-end path: a grey client feeds a transponder/muxponder whose coloured line side runs through mux and amplifier onto the shared fiber, then demux/pre-amp and a far transponder back to a grey client. Text version below.

Why grey on the client

Client optics are commodity. The router doesn't know or care that a transport box is downstream — it just lights a standard short/long-reach pluggable. That keeps the client world cheap, swappable, and vendor-neutral.

Why coloured on the line

The line side must land on an exact ITU channel so dozens of wavelengths can share one fiber without colliding. Tunable lasers mean you stock one card type and provision its channel, instead of stocking 96 fixed-colour cards.

Transponder vs muxponder

Both are O-E-O (optical → electrical → optical) converters that adapt a grey client to a coloured line wavelength. The difference is purely how many clients ride one wavelength.

Transponder — 1 client → 1 line λ

Takes one client and converts it to one line wavelength. Simplest possible client-to-λ adapter. Use it when a service already fills the wavelength (e.g. a 100G client onto a 100G line) or when you deliberately want one customer alone on a lambda.

Muxponder — N clients → 1 line λ

Aggregates several lower-rate clients into one higher-rate line wavelength (commonly with OTN mapping underneath). Fewer wavelengths burned for the same traffic — critical when fiber or channel count is the scarce resource.

MUXPONDER AGGREGATION (example: 10 x 10G clients -> one 100G line wavelength) client side (grey) line side (one coloured λ) ┌───────────┐ │ 10G #1 │──┐ │ 10G #2 │──┤ │ 10G #3 │──┤ ┌────────────────────┐ │ 10G #4 │──┤ │ MUXPONDER │ │ 10G #5 │──┼───────►│ map 10x10G into │══════════► 100G λ │ 10G #6 │──┤ │ one OTU4 (~111.8G │ one DWDM channel on │ 10G #7 │──┤ │ incl. FEC + OTN │ the ITU grid │ 10G #8 │──┤ │ overhead) │ │ 10G #9 │──┤ └────────────────────┘ │ 10G #10 │──┘ └───────────┘ Other common shapes: 4x10G->40G, 4x25G->100G, N x 100G -> 400G. Exact client counts / mixes depend on the specific module -> verify against Ekinops docs/BOM.
Common mistake

Assuming a muxponder is "free" aggregation. Every client you add shares the fate of that one wavelength — cut the line λ and all N clients drop together. Aggregation trades wavelength efficiency for a larger blast radius. If two of those clients need independent survivability, they may belong on separate wavelengths (or separate protected paths), not stacked in the same muxponder.

1R / 2R / 3R — amplify vs regenerate

Signals degrade two different ways over distance: they lose power, and they lose quality (noise accumulates — OSNR drops — and chromatic dispersion smears the pulses). These need different fixes. The "R" count tells you how much rebuilding a device does.

ClassWhat it doesDomainFixes power?Fixes OSNR / dispersion?
1R — re-amplifyBoosts power only (an amplifier, e.g. EDFA)Analog / opticalYesNo — amplifies the noise along with the signal, adds its own ASE noise
2R — + re-shapeBoosts and squares up the pulse shapePartly electricalYesPartly — cleans shape but not timing jitter; rare on its own in modern DWDM
3R — + re-time (regen)Re-amplify + re-shape + re-time: a full O-E-O rebuildElectrical (O-E-O)YesYes — resets OSNR and dispersion to like-new; the signal is regenerated from decoded bits
AMPLIFIER (1R) REGENERATOR (3R, O-E-O) clean ───►■■■──► louder+noisier clean ──►[OEO]──► clean again (analog gain, decode bits, retransmit noise rides along) a brand-new signal good for power loss. needed when a span exceeds does NOT reset OSNR. optical reach (OSNR floor / dispersion limit hit).
Key idea — "alien wavelength"

An alien wavelength is a wavelength generated by a third-party transponder (not the line-system vendor's) that the DWDM line system carries transparently end to end — the line system amplifies and switches the light without terminating it. Powerful (mix vendors, put coherent pluggables straight in routers) but it means the line system can't see inside that signal, so OSNR and reach planning for the alien λ are on you, not the line vendor's turnkey tool.

EDFA vs Raman amplification

Both add optical power to all channels at once without O-E-O. They differ in where the gain happens and what it buys you.

EDFA (erbium-doped fiber amplifier)

  • C-band, roughly 1530–1565 nm.
  • Typical gain ~15–25 dB; amplifies the whole band in one shot.
  • Adds ASE noise (noise figure ~4–6 dB) — so it lifts power but slowly erodes OSNR each hop.
  • Needs gain-flattening / tilt control so channels at the band edges aren't left high or low.
  • Gain is lumped in a box at the site.

Raman amplification

  • A high-power pump laser injects into the transmission fiber itself.
  • Gain is distributed along the fiber, not lumped — it effectively amplifies before the signal has fully decayed, so it improves OSNR and extends reach.
  • More complex and has real laser-safety implications (high pump power on the live fiber).
  • Often paired with EDFAs on the longest spans, not instead of them.
AMPLIFIER POSITIONS on a span (Tx site) ................ (Rx site) Tx ─►[ BOOSTER ]═══════════[ INLINE ]═══════════[ PRE-AMP ]─► Rx post-Tx, high mid-span, makes up just before Rx, launch power onto accumulated span lifts the weak the fiber loss on long hauls signal for a clean decode Raman: pump injected INTO the span fiber (typically counter-propagating from the far end) to add distributed gain and lift OSNR before the pre-amp.
Common mistake

Believing amplifiers "fix" a marginal link. They do not touch the OSNR floor or dispersion — only power. Stacking EDFAs on a link that's already noise-limited just makes a louder bad signal and can drive receivers into overload. When reach (not power) is the wall, the answer is a 3R regenerator or a more capable coherent transponder, not more gain.

OSC — the optical supervisory channel

The OSC is a dedicated out-of-band management wavelength (often around 1510 nm or 1625 nm, outside the amplified C-band traffic) that carries management, comms, and telemetry between optical nodes. Crucially it is dropped, terminated, and regenerated at every node — because it must be readable at each site — while the amplified traffic band passes straight through the amplifiers untouched.

OSC vs TRAFFIC through a 3-node line system NODE A NODE B NODE C ┌────────┐ fiber span ┌────────┐ fiber span ┌────────┐ │ │═══════ traffic ═══════════ traffic (amplified, passes ══════► │ │ band λs │ EDFA │ through) │ │ │ add │ │ amps │ │ drop │ │ OSC ──┼──── OSC λ ──────┼─► drop │──── OSC λ ────────┼─► drop │ │ │ (~1510/1625) │ + re- │ regenerated, │ │ └────────┘ │ gen ──┼── re-added ──────►└────────┘ OSC is terminated & └────────┘ fresh at B reborn at EACH node; Traffic band is amplified straight through; traffic band is NOT. only the OSC gets read + rebuilt per hop.
Why the OSC matters operationally

The OSC is how you reach a remote shelf whose traffic path is broken. If a span's traffic is down but the OSC is up, you can still log into the far node, read alarms, and drive a switch. If the OSC itself is down between two sites, you are managing blind past that point — that's often your first clue to which span failed.

OADM vs ROADM

OADM — fixed add/drop

Optical Add/Drop Multiplexer. Passive filters drop and add a specific, hard-wired set of channels at a site; everything else expresses through. Cheap, simple, no power — but the channel plan is set in the glass. Changing which λ drops here means a truck roll and new filters.

ROADM — reconfigurable add/drop

Reconfigurable OADM. A WSS (wavelength-selective switch) steers any channel to any direction/port under software control. You provision add/drop/express remotely — no re-cabling. This is what makes mesh optical networks and remote wavelength turn-up possible.

CDC — the ROADM flexibility vocabulary
  • Colorless — any wavelength can go to any add/drop port (the port isn't wired to one fixed colour).
  • Directionless — an added/dropped wavelength can be sent to any line direction, not one pre-cabled way.
  • Contentionless — no internal blocking; the same wavelength can be added/dropped in multiple directions at once without colliding inside the node.

A node's degree = the number of line directions it connects. A degree-2 node sits on a line (east/west); a degree ≥3 node is a mesh branch point where a ROADM chooses which direction each wavelength takes.

DEGREE-2 (in-line) DEGREE-3 (mesh branch, ROADM/WSS) West ═[ROADM]═ East North │ West ═══[WSS]═══ East │ (add/drop) two directions three+ directions: any λ -> any direction chosen in software

Topologies

Point-to-point

A ══════ Z

Two sites, one span. Simplest and most common — DCI, back-to-back offices, a single leased-lambda service. Protection here means a whole second diverse span.

Linear w/ add-drop

A ══ B ══ C ══ Z ▲ ▲ drop drop

A chain of sites; intermediate nodes drop/add some channels and express the rest. OADM or ROADM at the middle sites.

Ring

A ── B │ │ D ── C

Sites on a loop (2-fiber or 4-fiber). Survives a single cut by sending traffic the other way around. Efficient shared protection.

Mesh (ROADM degree ≥3)

A───────B │╲ ╱│ Multiple ROADM branch points. Any node pair can be │ ╲ ╱ │ connected over more than one route, so a single cut │ ╲ ╱ │ reroutes wavelengths onto a surviving path in software. │ ╱ ╲ │ Maximum flexibility + resilience; highest cost/complexity. │ ╱ ╲ │ D───────C

Mesh needs ROADMs of degree ≥3 at the branch points. It's the optical equivalent of a routed core with multiple paths: expensive, but a cut becomes a reroute rather than an outage.

Component reference table

The one table to keep handy when reading a design. "Where it sits" tells you which part of the path it lives on; "what to check" is your first operational move when something's wrong.

ComponentWhat it doesWhere it sitsWhat to check
TransponderConverts one client signal to one line wavelength (O-E-O)Between client gear and the line systemClient + line Tx/Rx power, correct channel, client service up
MuxponderAggregates several clients into one higher-rate wavelengthBetween multiple client ports and one line portEach client's status, aggregate line rate, OTN mapping/tributary config
Mux / demuxCombines (mux) / separates (demux) wavelengths on one fiberAt the fiber junction of the line systemPer-channel port assignment, insertion loss, clean connectors
EDFA (amplifier)Optically boosts power of all C-band channels at once (1R)Booster (post-Tx), inline (mid-span), pre-amp (pre-Rx)Input/output power, gain, tilt/flatness, not driving Rx into overload
RamanDistributed gain via pump laser into the span fiber; lifts OSNRInto the transmission fiber, usually longest spansPump power, OSNR gain, laser-safety interlocks, fiber cleanliness
Regenerator (3R)Rebuilds a degraded signal (re-amplify + re-shape + re-time)Mid-path where optical reach/OSNR runs outWhether the problem is power (amp) or quality (regen), pre/post-FEC errors
OADMFixed add/drop of specific channels; rest express throughIntermediate/terminal site on the lineWhich channels drop vs express, port mapping, insertion loss
ROADM (WSS)Software-reconfigurable add/drop and per-λ routing; CDC/degreeFlexible / mesh branch sites (degree ≥3)Wavelength routing config, per-channel power balancing, blocked/contended paths
OSCOut-of-band management λ, dropped + regenerated at each nodeA dedicated wavelength (~1510/1625 nm) alongside trafficOSC link up between adjacent sites, reachability of remote shelves

Common use cases

DCI — data-center interconnect

Very high capacity per fiber, low latency, short-to-metro reach. Increasingly done with 400G ZR / ZR+ coherent pluggables placed straight into the routers (the router is the transponder, an alien-wavelength model) or with dedicated transponders/muxponders.

Watch: per-fiber capacity ceiling, latency budget, and whether router pluggables or the line system own OSNR/reach planning.

SAN extension

Fibre Channel (storage/replication) over WDM between sites. Distance drives buffer-to-buffer credits (too few and FC throughput collapses over distance) and the choice of sync vs async replication. Rule of thumb: ~5 µs/km one-way, so ~1 ms round trip per 100 km (approximate).

Watch: latency vs the array's sync-replication limit; BB-credit sizing; timing preservation.

Carrier wavelength service

Hand a customer a transparent lambda — a point-to-point pipe they light with their own optics. Often an alien wavelength to your line system.

Watch: the demarc and who owns OSNR/reach; what client rates the λ must transparently carry; monitoring you can offer when you can't see inside the customer's signal.

Common mistake

Reaching for an amplifier when you actually need a regenerator. An EDFA raises power and its own noise — if signal quality (OSNR) is already poor, more amplification won't fix errors. Degraded power → amplify. Degraded quality / reach exceeded → regenerate (or go coherent).

Field checklist — reading a transport design

1. A design shows one line wavelength carrying ten 10G clients. Transponder or muxponder — and what's the operational catch?

Muxponder — it aggregates several lower-rate clients into a single higher-rate line wavelength (10×10G→100G/OTU4). The catch: all ten clients share the fate of that one wavelength, so a line-λ failure drops all of them together.

2. A span's traffic is degraded. The optical power is fine but the pre-FEC error rate is climbing. Add an amplifier or a regenerator?

Regenerator (3R). Power is adequate, so the problem is quality/OSNR or dispersion, which an amplifier can't reset — it would only amplify the noise. 3R (or a more capable coherent transponder) rebuilds the signal.

3. Why is the OSC dropped and regenerated at every node while the traffic band just passes through the amplifiers?

Because the OSC must be readable at each site to manage that node, so it's terminated and reborn per hop. Traffic only needs power along the way, so it's amplified straight through. This is also why you can often still reach a remote shelf over the OSC when its traffic path is down.

4. What does a node's "degree" tell you, and what's the minimum degree for a mesh branch point?

Degree = number of line directions the node connects. Degree-2 is in-line (east/west); a mesh branch point needs degree ≥3, with a ROADM/WSS choosing each wavelength's direction in software.