Fibre plant faults and how they show up

Most fibre "outages" are not broken fibres. They are a fingerprint on a connector, a cable tie pulled tight, a patch cord of the wrong type, a splice done in the rain, a connector left uncapped in a dusty room — small mechanical facts that cost decibels. This page lists the plant faults by how often they happen, gives for each the symptom at the transceiver, the signature on the instruments and the fix, and adds the seasonal and slow faults that monitoring catches before users do.

The fault list

#	Fault	How common	Loss caused	Where
1	Contaminated connector endface	most common by far (> 50 % of tickets)	0.5–3 dB, or intermittent	any mating; especially rarely-touched panel ports and module receptacles
2	Macrobend	very common	0.5–10 dB, wavelength-dependent	cable ties, panel doors, slack loops, bent patch cords behind racks
3	Wrong fibre or connector type	common at installation	2–20 dB, or no link	62.5 vs 50 µm, SMF vs MMF, APC vs UPC, MPO polarity
4	Damaged endface (scratch, chip)	common	0.3–2 dB + reflections	mated without cleaning, dirt ground in
5	Bad splice	occasional	0.3–1 dB, or high reflectance (mechanical)	field splices in bad conditions, mis-cleaved
6	High reflectance	occasional	0–1 dB but laser instability, BER	open or air-gap connectors, mechanical splices, damaged UPC
7	Crushed or kinked cable	occasional	1–20 dB or break	doors, heavy cables on top, rodents, construction
8	Water in closure / water peak	rare but slow	0.5–2 dB rising at 1383 nm and 1550+	flooded manholes, cracked closures; old fibre with hydrogen ageing
9	Fibre break	rare	total	digging, rodents, ice, vehicle strikes
10	Excess connector count / wrong design	design error	budget exhausted	too many panels, cassettes
11	Overload (not a plant fault)	occasional	errors with strong signal	long-reach module on short link (Link budget)

Symptoms at the transceiver

DDM / link observation	Likely plant fault
Rx power 1–3 dB below the link passport, one direction	dirty or scratched connector on the Tx side of that direction, or a bend in that fibre
Rx power down symmetrically both directions	bend or crush of the duplex cord, common panel, or fibre attenuation increase (water)
Rx fine at 1310 (LR) but poor at 1550 (ER) on the same fibre	macrobend — loss grows with wavelength (Attenuation)
Rx −40 dBm / no light one direction	break, disconnected, Tx/Rx swapped at one end, wrong panel port
Rx good, link flaps or errors	reflectance (laser instability), dispersion, or a connector making intermittent contact (Link flapping)
Loss worse in cold weather	microbending from cable contraction, gel stiffening; ice loads
Loss worse in heat / rain	closure ingress, thermal expansion pulling a splice tray
New link 5+ dB above design loss	wrong fibre type, mode mismatch, mis-referenced test
MPO link: some lanes down	polarity method mismatch, or one fibre of the trunk damaged (Breakout & MPO)

Bench-side first steps: No link, Physical mismatches.

Instrument signatures

Fault	OTDR	LSPM	Inspection scope	VFL
Dirty connector	reflective event with excess loss (0.5–3 dB), often high reflectance	total loss above design	contamination in zone A/B	—
Macrobend	non-reflective step; loss at 1550/1625 ≫ 1310	loss wavelength-dependent	—	red glow at the bend through the jacket
Wrong fibre type	step at the joint (2–4 dB for 62.5→50); different backscatter slope	large loss	different core size visible	—
Bad fusion splice	non-reflective step > 0.3 dB (average both directions)	—	—	—
Mechanical splice / air gap	reflective event, reflectance −35 … −45 dB	small loss	—	—
Damaged endface	reflective, moderate loss	—	scratches/chips in zone A	—
Crush	localized step, may be reflective; sometimes distributed loss	—	—	glow if severe
Water / hydrogen ageing	uniform slope increase, strongest at 1383 and > 1550 nm	loss at 1550 up, 1310 less	—	—
Break	end-of-fibre spike or noise at a distance shorter than the route	no light	—	glow at the break if within metres
Ghost	reflective spike beyond real events at multiples of a distance	—	—	—

Instrument use and trace reading: Testing and measurement.

Fix per fault

Fault	Fix	Prevent
Contamination	inspect, clean (dry click cleaner; wet-dry for oils), re-inspect both sides	caps on everything, clean before every mating, no touching ferrules
Macrobend	re-route, loosen ties, radius ≥ 30 mm; use G.657 cords in tight spaces	cable management with bend guides
Wrong type	replace cord/plant with the right type; mode-conditioning cords are a 1G-only workaround	colour discipline: yellow/blue = SMF UPC, green = APC, aqua = OM3/OM4
Damaged endface	replace the connector (re-terminate or new cord)	clean, inspect, don't force
Bad splice	re-splice; check cleaver, electrodes, fibre prep	qualified splicers, indoor or tented work
Reflectance	replace air-gap/mechanical parts with fusion or APC; cap open ports	APC on PON/DWDM/analog; never leave unmated UPC in path
Crush	replace the cable section; add protection	route away from doors and heavy cable
Water	dry and reseal closure; replace fibre with water damage	IP68 closures, drainage, inspection schedule
Break	OTDR locate → splice or replace section	route diversity, marking, permits
Design	reduce connector count, cassettes → direct splices, longer-reach optics	budget review before purchase

Slow faults and monitoring

Signal	Cause	Detection
Rx power drifting down 0.1 dB/month	connector contamination, closure ingress, fibre ageing	DDM trend with a baseline (Monitoring)
Seasonal ±1 dB swing	temperature effects on outdoor plant	acceptable if margin ≥ 3 dB; record it
Loss increases after works nearby	new bend or crush	compare with passport; OTDR against installation trace
Tx power falling at the far end	far module ageing, not plant (Tx bias & ageing)	Tx DDM and bias trends
FEC correctable errors rising with stable power	reflectance or dispersion growth (rare) or module degradation	FEC counters (VDM & FEC metrics)

The triage order

Compare with the passport (Rx now vs Rx at commissioning) at both ends.
Which direction? One direction bad → look at that direction's Tx end connectors and fibre; both → common elements.
Swap the module with a known-good one for 5 minutes — plant or module.
Inspect and clean every accessible connector in the path.
LSPM the link at two wavelengths; compare with the installation report.
OTDR from the end with the bad direction; find the event.
Fix, re-test, update the passport.

In CodingBox

Step 3 is where the bench comes in: a module read in CodingBox with normal Tx power, sane thresholds and a plausible Rx from a reference source is proven good, and the plant is the suspect. Recording the module's DDM at commissioning gives the passport that step 1 needs (DDM in the app, Code database).

Preventing these faults and recovering from the big ones — maintenance cycle, monitoring, spares, the restoration process: Maintenance and restoration; working on the plant safely — laser classes, live fibre, glass, site hazards: Safety and handling.