OTDR Testing Basics: Measuring Fiber Links and Locating Faults

When a fiber link goes down, the OTDR (Optical Time-Domain Reflectometer) is the instrument that tells you exactly where and what the problem is. Unlike a simple power meter, which only confirms that light is or is not arriving, an OTDR sends pulses into the fiber and measures the backscattered light to build a trace of the whole link - showing every splice, connector, bend and break with its distance. This guide covers the physics in plain terms, the settings you must get right (wavelength, pulse width, averaging, refractive index), and how to interpret a trace to find faults fast.

How an OTDR Works

The OTDR launches a short light pulse into the fiber and measures the returned light over time. Two effects produce the return signal:

• Rayleigh backscattering: microscopic density variations in the glass scatter a tiny fraction of light back toward the source - this creates the continuous trace slope.
• Fresnel reflections: at discrete events (connectors, mechanical splices, the far end, a break), a strong reflection occurs - visible as spikes on the trace.

Since the speed of light in fiber is known, time converts to distance: distance = (c × time) / (2 × group index).

That factor of two is worth pausing on. The pulse travels out to the event and the scattered light travels back, so the round trip is twice the physical distance. The "group index" (about 1.468 for standard single-mode at 1310 nm, about 1.468-1.470 at 1550 nm) is the effective speed of the light in the glass, not the vacuum speed of light. Every number the OTDR prints - event distance, total length, span loss - is derived from that one constant, which is why a wrong index corrupts the whole report rather than a single reading.

Two more physical behaviours shape what you see. First, backscatter level depends on the fiber type: a 50/125 multimode fiber scatters more strongly than a 9/125 single-mode fiber, so traces of different fiber types sit at different heights and cannot be compared by eye. Second, the trace slope itself is the attenuation coefficient in dB/km. On single-mode fiber at 1550 nm, a healthy slope is roughly 0.18-0.22 dB/km; at 1310 nm it is closer to 0.32-0.35 dB/km. The difference is why long-haul routes prefer 1550 nm and short campus links happily use 1310 nm - see our guide to WDM and expanding fiber capacity if you want to push more wavelengths over the same glass.

Key OTDR Settings

Pulse width deserves more explanation than a table cell allows, because it is the setting that most often ruins a measurement in either direction. A narrow pulse (say 10 ns) resolves events that are a metre apart but injects little energy, so the backscatter signal fades quickly and a long span never returns a usable trace. A wide pulse (say 1 µs) carries enough energy to trace tens of kilometres but smears every event across roughly 100 m of fiber, merging adjacent splices into one unrecognisable step. The practical technique is to shoot the link at two widths: a long pulse to prove continuity and find the far end, then a short pulse to characterise the events near the OTDR. Some instruments label these "auto" modes, but they still need a sane range and averaging value underneath.

Choosing the Right Test Mode and Tier

Before you press run, decide what kind of test you are performing. This decision drives every setting that follows.

• Tier 1 (basic / certification): an optical loss measurement confirming the link meets a loss budget - typically a light source and power meter, a length measurement, and optionally an OTDR trace. Tier 1 answers "does this link pass?"
• Tier 2 (extended): an OTDR trace on every fiber, analysed event by event, to prove the quality of each splice and connector. Tier 2 answers "why will this link be a problem in five years?"

For a new data-centre or campus build, Tier 2 traces are the ones you will be glad you kept when a splice degrades years later and nobody can remember what the installer left behind. Store them as the commissioning baseline, keyed by cable ID and fiber number, and never overwrite them with later tests.

Reading a Trace: Events

• Connector: sharp loss step + reflectance spike (e.g., 0.3 dB loss, -40 dB reflectance).
• Splice: small loss step, no reflectance (fusion splices typically 0.02-0.1 dB).
• Macrobend: gradual loss with no reflectance - usually a tight bend, check with a visual fault locator.
• Break: large reflectance followed by the trace dropping to noise floor - the fault is at that distance.
• End of fiber: final reflection (or flat end) - the trace ends.

Distance accuracy matters for dispatch: a 100 m error can send a technician to the wrong manhole. Verify the RI setting before trusting distances.

Two subtleties trip up people reading traces for the first time. First, a gainer is a splice that appears to increase the signal - it is an artefact, caused by the reflectance of an upstream connector artificially lifting the backscatter level, and it usually means the event before it has a bad polish. Second, a flat, black trace running to the bottom of the screen at some distance is not always a break: a tight macrobend can also extinguish the trace while the fiber remains physically intact. A visual fault locator resolves both cases in seconds because it shows where light actually leaks out.

Interpreting Loss, Reflectance and Slope

Numbers on the trace mean something specific, and acceptance criteria usually quote them. Splicing and connector standards differ, but the following figures are widely used as acceptance thresholds for single-mode fiber:

Reflectance matters because it is a leading indicator. A reflection spike that grows over successive tests means the connector end-face is dirty, scratched or loose, even though the loss figure still looks acceptable. Clean before you measure, and note the cleaning in the report - a large share of "splice loss" readings turn out to be a dirty connector that a swab would have fixed. For the transmit-side mirror image of this problem, see optical module fault diagnosis with power readings.

Dead Zone and the First Event

Every OTDR has a dead zone after the launch: the first few meters are saturated and events close to the OTDR cannot be resolved. Use a launch fiber (pigtail, 100-500 m) to push the first event beyond the dead zone - this is mandatory for premises/jumper testing where the first connector is often the one failing.

The dead zone scales with pulse width and with the instrument's recovery time. A typical specification quotes an event dead zone (how close two reflective events can be and still be told apart, often 0.8-3 m) and an attenuation dead zone (how soon the trace recovers enough to measure a non-reflective event accurately, often 4-20 m). Both grow dramatically with wider pulses. The fix is always the same: attach a launch cable long enough that the first real event sits past the dead zone, and for full bidirectional accuracy attach a receive cable at the far end too, so the final connector is also measured from a clean baseline.

Step-by-Step Fault Location Workflow

1. Connect the OTDR with a launch fiber; set wavelength, range, pulse width and RI.
2. Run the trace and identify the event list (distance, loss, reflectance).
3. Compare against the baseline trace from commissioning - differences are the faults.
4. Confirm a suspected break with a visual fault locator (VFL) on short links.
5. Re-test after repair and store the new baseline for future comparison.

A slightly more detailed version of the same workflow, including the checks that keep you honest, looks like this:

# 1. Record what you are testing before touching the fiber
#    - cable ID, fiber number, wavelength, RI, launch cable length

# 2. Set up and shoot a long pulse first (continuity + far end)
#    wavelength 1550 nm, pulse 1 us, range 2x link, averaging 30 s

# 3. Re-shoot with a short pulse to resolve near events
#    pulse 10-30 ns, range just past the link, averaging 60-180 s

# 4. Export the event table and compare against the baseline trace
#    look for: new events, loss growth > 0.1 dB, reflectance drift

# 5. Every event should be reconciled to the as-built splice/connector
#    schedule - any unexplained event is the fault

Bidirectional Testing and Why It Matters

Single-direction OTDR testing has a well-known bias. A splice that is slightly mismatched in mode-field diameter will read loss in one direction and apparent gain in the other, because the backscatter coefficients on each side of the joint differ. The only way to get the true splice loss is to average the two directional readings:

true_splice_loss = (loss_forward + loss_reverse) / 2

Standards bodies and acceptance documents generally require bidirectional measurement for splices above a certain threshold, and many test-report tools compute the average automatically from two trace files. If you only ever test one way, you will occasionally reject a perfectly good splice because of backscatter bias, and occasionally accept a bad one for the same reason. It costs one more orientation swap and it removes an entire class of argument with the installers.

Common Mistakes

• Wrong RI → distances off by several percent; on a 10 km link that is hundreds of meters.
• No launch fiber → the most common failure (the first connector) is hidden in the dead zone.
• Testing 1310 nm but the link carries 1550/1625 nm (DWDM) → loss and events differ.

A few more that show up repeatedly in real reports:

• Dirty end-faces. Measuring without cleaning first produces phantom loss and random reflection spikes. Clean and re-measure before you escalate.
• Range set too short. If the configured range is shorter than the link, the OTDR folds or truncates the trace and the far event is invisible - or worse, appears at the wrong distance.
• Not saving and naming traces. An unlabelled trace file is worthless six months later. Include cable ID, fiber number, test date, wavelength and direction in the filename.
• Over-driving a live link. On an in-service link, use a filter or verify the receiving optics can tolerate the test pulse before you inject light into an active system.
• Trusting auto-analysis blindly. Event-finding algorithms miss closely spaced events and misplace the end-of-fiber marker. Always eye-check the trace against the event table.

OTDR Test Report Checklist

When a contractor hands over a fiber plant, the trace files alone are not the deliverable - the structured report is. Before you accept it, confirm each of these is present for every fiber:

1. Cable ID, fiber number, and the two endpoints of the measured span.
2. Wavelength tested, pulse width used, and the refractive index setting.
3. Launch and receive cable lengths, so dead-zone effects can be judged.
4. Full event table with distance, loss and reflectance per event.
5. Total span loss and total length, matching the as-built schedule.
6. Bidirectional results (or a stated reason why they are not required).
7. Signed baseline traces stored in a format the tool can reopen.

Get that package at commissioning and the next fault becomes a comparison exercise instead of an investigation. For long spans where absolute loss matters more than event-by-event detail, pair the OTDR trace with an optical power budget calculation for the link, and for coherent routes where the underlying constraint is optical signal quality rather than raw loss, read our notes on pre-FEC BER and OSNR checks.

Related: WDM: expanding fiber capacity and optical module fault diagnosis.

原文链接：https://community.fs.com/article/otdr-basics.html