Reading an OTDR Trace: What the Steps, Spikes and Tails Are Telling You

An OTDR trace is a picture of one thing: how much light comes back, against how far along the fibre it came from. Everything you read off it — splices, connectors, bends, breaks, the end of the link — is a change in that returning light. Learning to read one is mostly learning which shapes mean a real event and which are artefacts of the instrument itself.

What the trace is actually showing

The instrument sends a pulse of light down the fibre and listens to what comes back. Two things return: a faint, continuous scatter from the glass itself along the entire length, and sharp reflections from anything that presents a change in refractive index — a connector, a mechanical joint, a break, the far end.

Plot the returning power against elapsed time, convert time to distance, and you have the trace. The gentle downward slope across the whole thing is the fibre's own attenuation. Anything that interrupts that slope is an event.

The horizontal axis is fibre distance, not ground distance. Cable is longer than the route it follows: it coils in joint enclosures, loops through manholes, rises and falls in risers. A trace reading 1,340 metres does not put the event 1,340 metres along the pavement, and converting one to the other needs the route record.

The shapes, and what each one means

A step down, with no spike. The trace drops to a lower level and carries on at the same slope. That is a fusion splice, or a bend. No reflection, just light lost. Fusion splices should be small steps; a large one is a bad splice or a tight bend.

A spike, then a step down. A sharp reflection followed by a drop. That is a connector, a mechanical splice, or anything with an air gap or an index change. Connectors reflect because glass meets something that is not identical glass.

A spike that falls to the noise floor. The trace shoots up and then collapses into noise and never recovers. That is the end of the link — either the far end connector, or a break. If it appears where you expect the far end, it is the end. If it appears at 1,340 metres on a 4-kilometre route, it is a break.

A gradual sag over a stretch. Not an event at a point, but excess loss over a length. Usually a bend along a section, a cable under strain, or water in a joint affecting a run.

The tail, after the end. Everything past the final reflection is noise. There is no information there, and reading anything into it is a mistake beginners make often.

Why a splice can appear to produce gain

This one catches people out and looks like an instrument fault. The trace steps up at a splice — as if the fibre produced light.

It did not. The OTDR infers loss from backscatter, and how much a fibre backscatters depends on its own properties. Splice a fibre with a lower backscatter coefficient to one with a higher one, and the returning signal rises at the joint. The instrument, which can only measure what comes back, reports a gain.

The consequence is practical: a one-directional measurement of that splice is wrong. Shoot from the other end and the same splice shows an exaggerated loss. The true figure is the average of the two, which is why bidirectional testing is the standard for anything that matters and why a single-direction trace should never be the only evidence for a splice loss figure.

If a report hands you a splice showing a gain and no second direction, the report is incomplete rather than remarkable.

Dead zones — where the instrument is blind

Every strong reflection saturates the detector briefly, and for a short distance afterwards the instrument cannot resolve anything. That is the dead zone, and it has two consequences on real jobs.

Events near the launch connector can be invisible. The reflection from the connector at the instrument blinds it for the first stretch of fibre. A launch lead — a spool of fibre between the OTDR and the link under test — moves that blind spot off the link and into the lead, which is why a proper test set includes one.

Two events close together merge into one. A splice a short distance after a connector may not appear separately. It is not absent; it is hidden.

The practical rule: anything you cannot see from one end, shoot from the other. Faults near either end are the classic case where one direction reports a clean link and the other finds the problem.

Pulse width — the trade you cannot avoid

The pulse width setting is the single biggest influence on what a trace can tell you, and it is a straight trade.

• A short pulse gives fine resolution and short dead zones — it separates events close together — but carries less energy, so it does not reach far and the trace gets noisy at distance.
• A long pulse carries further and produces a cleaner far end, but smears events together and lengthens dead zones.

There is no setting that does both. The working method on an unfamiliar link is to shoot long first to find the far end and confirm the total length, then shoot short to resolve the events near the section you care about. A trace taken at one setting and presented as the whole story is usually hiding something at the other.

What a test report must contain

A trace on its own is a picture. A report is evidence, and for it to be worth anything on handover it needs:

• The trace file itself, in a format that can be re-opened — not a screenshot
• Pulse width, wavelength and range settings used, so the result can be reproduced
• Bidirectional results for splice loss, or a stated reason why only one direction exists
• Event table: each event, its distance, its loss, and its type
• End-to-end loss for the link, against the acceptance figure it is being judged on
• Date, technician and the link identifier — which link, on which route

A summary letter saying "all links tested and passed" is not a test report. It cannot be checked, it cannot be compared against a later trace, and it proves nothing when a fault appears eighteen months on.

The habit that separates useful testing from box-ticking

Testing at commissioning is not really about proving the installation was good. It is about creating the baseline that makes every future fault cheap to find.

When a link degrades two years later, the question is always the same: has this changed, or was it always like this? With a commissioning trace, the answer takes minutes — overlay the new trace on the old and the changed event is obvious. Without one, the crew is diagnosing from scratch on a live fault, at whatever hour it happened.

That is the whole argument for insisting on proper traces at handover, and it has nothing to do with distrusting the installer. The value of the record is not in the day it is made. It is in the day something goes wrong.

— Samuel

Testing, Documented Properly

Fiber Tech Solutions is an IMDA-licensed contractor with FOA-certified splicing and termination teams, working across telecom exchanges, transport, utilities, data centres and commercial buildings in Singapore. Fibre testing with bidirectional traces, event tables and end-to-end loss figures, handed over as part of the job.

If you are accepting a handover and what you have been given is a summary rather than traces, that is worth a second look before you sign it.