An installer hands over a fiber link with a short note: tested, working. Six months later you upgrade to 10 Gbit/s, insert new transceivers, and one link in two either will not come up or keeps accumulating errors. The old transceivers had more margin and tolerated the poor splice or the dirty connector. The problem was always there — nobody had measured it.
A fiber link with a light on is not a tested fiber link. This article explains what happens during splicing, what each test measures, how to read an OTDR trace and what the handover documentation should contain. It is useful when you accept a network infrastructure project or assess the condition of an existing fiber network.
Splices, connectors and where the light goes
Every joint in a fiber link loses some light. In fusion splicing, the machine aligns the two fiber ends precisely and melts them together with an electric arc into a single continuous fiber. A good fusion splice loses very little — well under a tenth of a decibel. Mechanical splices are quicker but lose more and are less stable, so they are rarely used in permanent infrastructure.
A mated connector pair (for example, an LC in the patch panel and one on the patch cord) loses much more than a splice and, more importantly, is sensitive to dirt. In a typical building, fiber is terminated in one of two ways: fusion-spliced to factory pigtails in the patch panel, or using factory-terminated cables. Both approaches are sound — what matters is that the result is measured and documented.
Two tests, two different questions
Fiber testing comes in two tiers, and they complement rather than replace each other:
- Loss testing with a light source and power meter (OLTS, often called Tier 1) answers: how much light does the link lose in total? This is the main acceptance criterion, because it is exactly how a transceiver sees the link.
- Reflectometry (OTDR, Tier 2) answers: where is the light lost? The instrument sends pulses into the fiber and measures the light scattered back, drawing a map of the link — the trace — with every splice, connector and bend.
- Both tests should be run at the wavelengths the link will use: typically 850 and 1300 nm on multimode, 1310 and 1550 nm on single-mode.
The loss budget: how much loss is normal
A measured result only means something when compared with a budget. A simple formula: loss budget = fiber length (km) × fiber attenuation (dB/km) + number of mated connector pairs × 0.75 dB + number of splices × 0.3 dB. The 0.75 and 0.3 dB figures are the maximums permitted by the standards — good workmanship usually achieves considerably less. Multimode fiber attenuation at 850 nm is about 3.5 dB/km; single-mode is well under 1 dB/km.
For example, a 300 m multimode link with two connector pairs and two splices: 0.3 × 3.5 = 1.05 dB, plus 2 × 0.75 = 1.5 dB, plus 2 × 0.3 = 0.6 dB — about 3.15 dB in total. If the meter reads 1.8 dB, the link is in good shape; 4 dB means there is a defect somewhere, and the OTDR should show where. Also check the limit of the specific application — 10 Gbit/s over multimode, for example, tolerates less total loss than 1 Gbit/s.
How to read an OTDR trace
A trace is a graph with distance on the horizontal axis and backscattered signal level on the vertical. The line slopes down gradually — that is the fiber's natural attenuation. Along it you see events:
- A sharp spike followed by a drop — a reflective event, usually a connector or a mechanical splice.
- A small step down with no spike — a fusion splice or a bend.
- A drop that is much larger at 1550 nm than at 1310 nm — almost always a macrobend: an over-tightened cable tie, a sharp corner in the tray, or a cramped splice cassette.
- A gainer at a splice — a physically impossible gain, meaning the joined fibers have different backscatter characteristics; an accurate value needs a bidirectional measurement.
- A large spike followed by noise — the fiber end or, if it appears earlier than expected, a break.
What goes wrong most often, and why
Most fiber problems are not about the quality of the glass but about installation and handling. At the top of the list there is usually the same cause — a dirty connector.
- Dirty or damaged connector end-faces: a single dust particle on the core causes noticeable loss and reflection, and mating a dirty connector can scratch the surface permanently.
- Tight bends: over-tightened cable ties, fiber slack coiled too tightly in the cassette, a patch cord crushed by a cabinet door.
- Poor cleaves before splicing: an angled or uneven fiber end gives high loss even when the splicer rates the result as acceptable.
- Mixed fiber types: OM3 and OM4, or multimode and single-mode, in one link.
- Polarity errors: transmit and receive are swapped — the link is technically sound, but no connection is established.
That is why every connector should be inspected with a microscope before mating and cleaned if needed — including new patch cords straight out of the bag. This matters especially in industrial environments, where dust and vibration are constant factors.
What the handover package should contain
- A consistent identifier for every link and fiber, matching the labels in the patch panels.
- Loss test results for every fiber, at both wavelengths, compared with the calculated budget and with a clear pass or fail.
- OTDR traces as native electronic files, not just screenshots — they are the baseline for future comparison.
- A splice tray map: which fiber is presented on which port.
- Connector inspection results, if the project specifies them.
- The actual cable route and length on as-built drawings.
This package is not a formality. When a link causes trouble two years later, comparing a new trace with the baseline shows within minutes what changed and where. It is exactly the information a technical support team relies on during diagnosis.
If you are about to accept a fiber installation or need an assessment of existing links, write to us — we will help you define the test scope and acceptance criteria.