The Erbium-Doped Fibre Amplifier

FINDING A FAULT IN AN AMPLIFIED FIBRE SPAN
Work top down. Each branch ends in a measurement, not a guess.

START: the link is down or degraded.

1. IS THERE ANY LIGHT AT THE RECEIVER? (optical power meter at the receive port)
   NO LIGHT AT ALL -> go to 2.
   LIGHT, BUT LOW  -> go to 3.
   LIGHT AT NORMAL LEVEL, STILL ERRORS -> go to 4.

2. NO LIGHT. The path is broken.
   - Shoot an OTDR from the receive end. The trace ends where the fibre ends.
     A near-vertical drop to the noise floor with a reflective spike = a clean BREAK
     or an unmated connector. Without a spike = a crushed or bent fibre.
   - Distance on the OTDR is optical, so divide by the group index (about 1.468)
     to get metres of cable, then apply the cable's own slack factor to get metres
     of duct. Skipping that step is how digging teams end up in the wrong field.
   - No break on the trace? Check the transmitter: laser bias current, and whether
     an amplifier upstream has shut down on its own loss-of-signal interlock.

3. LOW LIGHT. Something is eating the budget.
   - Compare against the commissioning record. You are looking for a CHANGE.
   - A step in the OTDR trace with no reflection = a splice or a bend that has
     degraded. A step WITH a reflection = a connector.
   - Loss that is much worse at 1550 nm than at 1310 nm = macrobending. Loss that
     is similar at both = a splice, a connector, or a break in the coating.
   - Dirty connectors are the single commonest cause. Inspect with a fibre scope
     BEFORE cleaning, so you learn whether it was dirty; then clean and re-measure.
   - Amplifier output down? Check pump laser current and case temperature first.

4. NORMAL POWER, STILL ERRORS. The power is fine; the SIGNAL is not.
   - Measure OSNR on a spectrum analyser, not power on a meter. An amplifier chain
     happily delivers full power made largely of its own spontaneous emission.
   - OSNR low -> an amplifier in the chain has lost pump power and the ones after
     it are amplifying noise to make up the level. Find the first one whose output
     spectrum has a raised noise floor; the fault is AT it, not after it.
   - OSNR fine -> suspect dispersion or nonlinearity. Has anyone changed the launch
     power? Too MUCH power is a real fault: above roughly +3 dBm per channel,
     four-wave mixing and self-phase modulation start corrupting the signal, and
     turning the amplifier DOWN fixes it.
   - Errors that come and go with the weather = a duct or a joint closure moving.

RULE THAT SAVES THE MOST TIME: always compare against the commissioning trace.
An absolute number tells you little; a CHANGE from the day it was installed tells
you where to dig.