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.
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).
| Property | Typical value (illustrative) | What it means for you |
|---|---|---|
| Gain | ~15–25 dB | How much it boosts every channel — must match the span loss it's covering |
| Noise figure (NF) | ~4–6 dB | How much ASE noise it adds — lower NF = better OSNR downstream |
| Band | C-band ~1530–1565 nm | Amplifies the whole band in one shot |
| Total output power | Limited (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:
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).
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.
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).
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 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.
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/ALS — Automatic 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.
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:
| Class | Roughly means | In fiber terms |
|---|---|---|
| Class 1 | Safe under normal use | Low-power / enclosed; safe as long as the enclosure is intact |
| Class 1M | Safe to the naked eye, hazardous with optics | Do NOT view with a loupe/microscope/magnifier — collecting optics concentrate it |
| Class 3R | Low risk, but exceeds Class 1 limits | Direct viewing is a hazard; treat with care |
| Class 3B | Hazardous to the eye on direct exposure | Amplified 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.
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.
1. Why is 1550 nm light especially dangerous to work near compared to visible laser light?
2. A technician pulls a connector on a live amplified span. What is APR/ALS supposed to do, and under which standard?
3. A link is noise-limited (poor OSNR). Is cranking up launch power the right fix? Why or why not?
4. Why does the pre-amp get the lowest noise figure of the amplifiers on a span?
5. Name the three Kerr nonlinear effects and what raises them.
6. What does Raman amplification buy you, and why is it a heightened safety hazard?