How to Read an OTDR Trace: Loss Events, Ghosts, and the Ticket Line

A light meter tells you whether light arrives. An OTDR tells you where it was lost — and that difference is the whole point. The trace is a map of the span: distance along the bottom, loss down the side, and every step down is an event at a known location. Read it event by event, match each event to something real on the span, and you can tell a bad splice from a dirty connector from a bend without opening a single tray. Read it as a pass/fail verdict and it will mislead you at least as often as it helps.

What the trace is actually showing

The OTDR fires pulses down the fiber and measures the light scattered back. The steady downward slope is the fiber's own attenuation (about 0.2 dB/km at 1550 nm on single-mode, about 0.35 dB/km at 1310 nm). Anything steeper than that slope is an event. Two things matter about each event: how much loss it shows, and whether it carries a reflective spike. The spike is the fingerprint — connectors reflect, splices (good ones) do not.

Those loss figures are rules of thumb for judging whether an event is worth chasing, not acceptance limits for your plant — your site's spec is the limit that matters. A 0.4 dB connector is unremarkable; a 0.4 dB fusion splice is worth re-doing. The signature tells you which one you are looking at.

Reading the trace, in order

1. Set the launch right. Use a launch cable (a few hundred meters of fiber before the span under test) so the first connector is not buried in the OTDR's dead zone — the blind stretch right after the pulse fires, where the instrument cannot resolve events. Without a launch cable, a fault in the first panel is invisible, and the trace starts mid-span.
2. Walk the events against the span record. Panel at 0.3 km, splice tray at 1.2 km, bulkhead at 2.0 km — each known point should produce exactly one event, at the right distance. An event with no matching physical point is the one to investigate; a physical point with no event is usually fine (a clean fusion splice can be nearly invisible).
3. Compare wavelengths on a suspect bend. A bend loses more at 1550 nm than at 1310 nm. If the same event measures 0.3 dB at 1310 nm and 1.2 dB at 1550 nm, it is a bend, not a splice — re-route or relieve the cable instead of re-splicing it. A splice or connector loses roughly the same at both wavelengths.
4. Shoot from both ends before touching anything. A one-direction trace cannot separate three traps: a gainer (two different fiber types spliced together read as a loss from one end and a gain from the other — the true splice loss is the average of the two directions), a ghost (a false event from reflections bouncing between two connectors, which moves or vanishes from the other end), and a real fault (which appears at the same location from both ends). Bidirectional traces are the difference between a diagnosis and a guess.
5. Check the end of fiber before declaring a break. A trace that ends well short of the far end with a large reflective drop is a break or a disconnected far end; a trace that ends at the expected distance with a normal end reflection is a complete span, even if the total loss looks high — in that case the loss is distributed across the events, not concentrated in one fault.

The traps that waste the most time

Chasing ghosts. A ghost event sits at a distance equal to the sum of two real reflections, often beyond the end of the fiber. It looks exactly like a fault and there is nothing at that location to fix. The other-direction shot is the only reliable test — do it before opening any tray.

Reading a gainer as a repair. A splice that reads as a gain from one direction has not improved the fiber; it joins two fiber types with different backscatter levels. Average both directions for the true loss, and write both numbers in the ticket, or the next tech will read the single-direction gain and conclude the splice is "better than perfect."

Testing at the wrong wavelength. A span qualified at 1310 nm can fail at 1550 nm (bend sensitivity) and vice versa. Test at the wavelength the optics actually use, and say which wavelength you used — a trace without its wavelength is unverifiable.

Recording an OTDR result in the ticket

A trace summary that nobody can re-check is barely better than no trace. The minimum useful note:

Span: SW-11 ↔ SW-14 via ODF-C (expected length 3.4 km).
OTDR 1550 nm from ODF-C, launch cable fitted.
Splice tray 1.24 km: 0.08 dB (normal).
Bulkhead 2.01 km: 0.35 dB (connector, within spec).
End of fiber 3.42 km (matches record — span complete).
Total loss 1.9 dB. Trace file attached: otdr-sw11-sw14-1550.pdf.
No fault found on this span; fault lies elsewhere.

The shape to copy: span and expected length, wavelength and direction, each event with distance and loss, total loss, the trace file itself, and the conclusion the trace supports. "OTDR shows no fault" without wavelength, direction, and distances is an assertion; with them, it is evidence the next tech can stand behind.

See RackReady   Free sample   Fiber no-link checklist

Support: support@hasherventures.com

Free tool: turn your trace findings into a structured ticket note with the RackReady Ticket Builder.