Ekinops Dark Fiber Learning Path
Engineering

Amplification & Safety

Amplifiers are what let a wavelength travel hundreds of kilometres without an electrical rebuild — and they are also, quite literally, the most dangerous thing in the building. This page covers how optical amplifiers work and where they sit, the delicate balance of launch power, and then the safety systems and rules that keep invisible high-power light from hurting people. Read the safety section as if your eyesight depends on it, because someone's does.

How to read this page

First half is engineering: EDFAs, where amplifiers sit on a span, gain tilt, the launch-power sweet spot, and Raman. Second half is safety: automatic power shutdown (APR/ALS), laser hazard classes, and the one rule that never bends — never look into a fiber; meter it.

The EDFA — an optical amplifier

The workhorse amplifier is the EDFA — Erbium-Doped Fiber Amplifier. A short length of fiber doped with erbium is energized by a pump laser; when your signal passes through, the excited erbium atoms dump their energy into it, amplifying every channel in the band at once — no O-E-O, no per-channel handling. It operates in the C-band (roughly 1530–1565 nm).

PropertyTypical value (illustrative)What it means for you
Gain~15–25 dBHow much it boosts every channel — must match the span loss it's covering
Noise figure (NF)~4–6 dBHow much ASE noise it adds — lower NF = better OSNR downstream
BandC-band ~1530–1565 nmAmplifies the whole band in one shot
Total output powerLimited (a ceiling)Shared across all channels — more channels means less power each

Amplifier positions on a span

An amplifier's job depends on where it sits. Three classic positions:

AMPLIFIER POSITIONS (Tx site) .................... (Rx site) Tx ─►[ BOOSTER ]═══════════[ INLINE ]═══════════[ PRE-AMP ]─► Rx post-Tx: pushes mid-span: makes up pre-Rx: lifts the high launch power accumulated span weak arriving onto the fiber loss on long hauls signal for a clean (power out) (power in the middle) decode BOOSTER — high output, sets the launch power into the span. INLINE — sits at a hut/mid-span site; both amplifies what arrived and re-launches into the next span (often has its own NF budget). PRE-AMP — LOW noise figure is critical here; it sets the OSNR the receiver actually sees, so pre-amps are the lowest-NF units.
Why pre-amp NF matters most

The pre-amp is the last amplifier before the receiver, so its noise figure directly sets the OSNR the receiver decodes at. A noisy booster hurts; a noisy pre-amp hurts more. That's why designs spend the money on a low-NF pre-amp and can be more relaxed about the booster.

Gain tilt and flattening

An EDFA does not amplify every wavelength equally — its gain varies across the band. That unevenness is gain tilt: channels at one edge of the band come out hotter than channels at the other. Left uncorrected, tilt compounds span after span until edge channels are either buried in noise (too low) or driving nonlinearity (too high).

GAIN TILT across the C-band (before flattening) per-channel power out │ ______ │ ____/ \____ │ / \ channels at the edges land │ / \ higher/lower than the middle └───────────────────────► wavelength (1530 ......... 1565 nm) tilt = the slope/ripple across the band FIX: gain-flattening filters + per-channel power balancing (often at ROADMs) per-channel power out │ ___________________ every channel lands at the same │ / \ target power → uniform OSNR across └───────────────────────► the band, no edge channel starved.

The remedies are gain-flattening filters (built into the amplifier to level its natural shape) plus per-channel power balancing (often done at ROADMs / with variable attenuators) so every wavelength lands at its target power. Uniform power across the band means uniform OSNR — no channel starved and none overdriven.

The launch-power sweet spot

How hard should you launch each channel into the fiber? There is an optimum, and it sits between two failure modes:

Launch too LOW → poor OSNR

If per-channel launch power is too low, the signal sits too close to the amplifier noise floor. OSNR suffers, and the link runs out of noise margin early. This is the noise-limited side.

Launch too HIGH → nonlinear penalties

If launch power is too high, the intense light distorts the glass's own refractive index (the Kerr effect), producing signal-wrecking distortions. This is the nonlinear-limited side. More power now makes things worse.

OPTIMAL LAUNCH POWER — the "bathtub" (or smile) curve link penalty │\ / (worse ▲) │ \ noise-limited nonlinear-│ │ \ (too little power, limited / │ \ OSNR poor) (Kerr) / │ \_ _/ │ \__ __/ │ \____ OPTIMUM ___/ │ \____ ▼ ___/ │ \___/ ◄─ best OSNR-vs-nonlinear balance └──────────────────────────────────────► per-channel launch power too low just right too high The best link quality is NOT maximum power — it's the bottom of the curve. Push past the optimum and Kerr nonlinearity degrades the signal faster than the extra power helps.
The Kerr nonlinear effects, named

At high power the glass's refractive index shifts with the light's own intensity (the Kerr effect), producing three penalties worth knowing by name:

  • SPM — Self-Phase Modulation: a channel distorts its own phase as its intensity varies.
  • XPM — Cross-Phase Modulation: one channel's intensity distorts the phase of its neighbours.
  • FWM — Four-Wave Mixing: channels mix to generate new unwanted tones that land on other channels (worst when dispersion is near zero, which is one reason zero-dispersion fiber can be bad for dense WDM).
Common mistake — "just turn it up"

Cranking launch power to fix a marginal link. Past the optimum you trade a noise problem for a nonlinear one, and nonlinear penalties can't be filtered out downstream — they're baked into the signal. The fix for a noise-limited link is lower NF, Raman, or fewer/shorter spans — not simply more launch power.

Raman amplification

An EDFA is lumped gain in a box. Raman amplification is distributed gain: a high-power pump laser is injected into the transmission fiber itself (often from the far end, counter-propagating), and through a scattering effect it transfers energy to your signal along the fiber — amplifying it before it has fully decayed. Because the signal never gets as weak, Raman adds roughly 3–5 dB of effective OSNR and extends reach, and it's often paired with EDFAs on the very longest spans.

Raman is a safety hazard

Raman uses very high pump power launched onto the live transmission fiber — far more than an EDFA keeps inside its box. An open connector or fiber break on a Raman-pumped span is exceptionally dangerous. Raman spans get the strictest safety interlocks for exactly this reason.

LASER SAFETY

Everything above puts invisible, high-power infrared light onto glass that humans handle. This section is not optional reading. The light that carries your traffic can permanently damage an eye before you know it was there.

The invisible-light danger

1550 nm light is completely invisible — it is far outside what the human eye can see. Critically, there is no blink reflex: your eye will not flinch, water, or squint, because it never registers the light at all. On an amplified or Raman-pumped span the power can be high enough to injure the retina. You can be staring into a lit fiber and feel absolutely nothing until damage is done. Never assume a fiber is dark because you see nothing.

APR / ALS — automatic power shutdown (ITU-T G.664)

The primary engineered protection is APR/ALSAutomatic Power Reduction / Automatic Laser Shutdown, defined in ITU-T G.664. The idea: if the fiber breaks or the received signal disappears (which is exactly what happens when a connector is pulled or a cable is cut), the system assumes a human may now be exposed to the open fiber end and automatically reduces or shuts off the optical power. It then uses a periodic low-power restart handshake to test whether the fiber is whole again before ramping back to full power.

APR / ALS SEQUENCE (fiber break → shutdown → safe restart) [ITU-T G.664] 1. NORMAL Tx ═══full power═══► ................ ►═══► Rx (LOS? no) │ 2. BREAK / PULL Tx ═══full power═══► ✂ ✗ open end Rx sees LOSS OF │ (light escaping) SIGNAL (LOS) ▼ 3. SHUTDOWN Tx ─ power reduced/OFF ─X (within a defined short time) │ open fiber end is now SAFE to be near ▼ 4. RESTART PROBE Tx ─►low-power test pulse─►? sent periodically │ "is the fiber whole yet?" ▼ 5. RECOVERY fiber repaired → probe gets through → handshake completes Tx ═══ramps back to full power═══► NORMAL again Net effect: a cut or open connector makes the span go SAFE automatically, and it only re-lights once the path proves continuous again.
Operational note on ALS

ALS is a safety system, not a maintenance convenience — and it is why a freshly repaired span may sit dark for a moment before the restart handshake re-lights it. Never defeat or disable ALS to "make troubleshooting easier." If you must work with power present (e.g. testing), follow the local lockout and PPE procedure, not a workaround.

IEC 60825 — laser hazard classes and levels

IEC 60825-1 classifies laser products by how dangerous their emission is; IEC 60825-2 covers optical fiber communication systems specifically, defining hazard levels for the light present on the fiber. You do not need to memorize the physics, but you must recognize the classes:

ClassRoughly meansIn fiber terms
Class 1Safe under normal useLow-power / enclosed; safe as long as the enclosure is intact
Class 1MSafe to the naked eye, hazardous with opticsDo NOT view with a loupe/microscope/magnifier — collecting optics concentrate it
Class 3RLow risk, but exceeds Class 1 limitsDirect viewing is a hazard; treat with care
Class 3BHazardous to the eye on direct exposureAmplified outputs / Raman pumps can reach here — direct beam is dangerous

The higher the class, the more the system relies on interlocks, labeling, ALS, and procedure to keep people out of the beam. Amplified spans and Raman pumps sit at the dangerous end — treat any amplifier or Raman output connector as a live high-power source until a meter proves otherwise.

NEVER eyeball a fiber — meter it

The single rule that never bends: never look into a fiber, connector, or bulkhead that might be lit — check it with a power meter or VFL, not your eye. 1550 nm is invisible and triggers no blink reflex, so "it looks dark" tells you nothing. Point suspect connectors away from faces, cap unused ports, and confirm a fiber is dark with an optical power meter before you go near the end face. On amplified and Raman spans this is not caution — it is the difference between keeping your eyesight and losing it.

Field checklist — amplifiers & laser safety

1. Why is 1550 nm light especially dangerous to work near compared to visible laser light?

It is completely invisible and triggers no blink reflex — your eye never registers it, so it won't flinch or squint away. On amplified/Raman spans the power can injure the retina before you know any light was present. "It looks dark" proves nothing; meter it.

2. A technician pulls a connector on a live amplified span. What is APR/ALS supposed to do, and under which standard?

On loss of signal (which a break/pull causes), APR/ALS automatically reduces or shuts off the optical power to make the open end safe, then uses a periodic low-power restart handshake to re-light only once the fiber is whole again. It is defined in ITU-T G.664.

3. A link is noise-limited (poor OSNR). Is cranking up launch power the right fix? Why or why not?

No. There's an optimal launch power. Past the optimum, higher power triggers Kerr nonlinear penalties (SPM/XPM/FWM) that degrade the signal and can't be filtered out. Better fixes: lower-NF amplifiers, Raman gain, or fewer/shorter spans.

4. Why does the pre-amp get the lowest noise figure of the amplifiers on a span?

Because it's the last amplifier before the receiver, its NF directly sets the OSNR the receiver decodes at. A noisy pre-amp hurts final OSNR more than a noisy booster, so designs invest in a low-NF pre-amp.

5. Name the three Kerr nonlinear effects and what raises them.

SPM (self-phase modulation), XPM (cross-phase modulation), and FWM (four-wave mixing). They grow with excessive per-channel launch power (and FWM is worst near zero dispersion). They are the "too high" wall of the launch-power sweet spot.

6. What does Raman amplification buy you, and why is it a heightened safety hazard?

Distributed gain along the transmission fiber adds roughly 3–5 dB of effective OSNR and extends reach. It's hazardous because it launches very high pump power directly onto the live fiber, so an open connector or break on a Raman span is exceptionally dangerous — hence the strictest interlocks.