OTDR settings that change the answer: pulse width, averaging, wavelength, and dynamic range

Two techs can test the same fiber span and come back with different loss numbers — and both can be right, because the OTDR is not measuring the fiber alone. It is measuring the fiber through four settings: pulse width, averaging, wavelength, and dynamic range. Change any one and the trace changes: events merge or split, the far end appears or vanishes, the loss of the same splice reads differently. The settings are not a detail to leave on "auto" and forget; they are part of the measurement, and a loss number without them is not evidence — it is a number nobody can reproduce.

The four settings, and what each one does

Setting	What it controls	Turn it up and…	Turn it down and…
Pulse width	How long each light pulse is — sets event resolution and dead zone	More range, but nearby events smear into one and the dead zone grows	Finer resolution and a shorter dead zone, but less range — the far end may vanish
Averaging	How long the OTDR averages the returning signal — sets trace noise	Cleaner trace, small events (a 0.05 dB splice) become readable; costs minutes	Faster trace, but noise hides small events and loss readings wobble
Wavelength	The light the OTDR sends — must match the span and the optic	1550 nm shows bending loss more clearly; 1310 nm is the usual single-mode working wavelength	850 nm for multimode; testing at the wrong wavelength measures a different span
Dynamic range	How far the OTDR can still see a signal above the noise floor	Longer pulse and more averaging extend it; the far end and far events appear	Short pulse, short averaging — the trace dies before the end and looks like a break

Pulse width: the resolution knob

Pulse width is the single most misunderstood setting on the OTDR. A short pulse — say 10 to 30 ns — puts a short burst of light into the fiber, so two events a few metres apart reflect back as two separate blips. That is what you want inside a building, where connectors and splices sit close together. The price is range: a short pulse carries less energy, so on a long span the signal dies into noise before the far end, and the trace ends mid-fiber. A long pulse — 1 µs or more — reaches much further, but it smears: two events inside the pulse length merge into one wide event, and the dead zone after a reflective connector grows, hiding whatever sits behind it.

The rule that works in practice: use the shortest pulse that still reaches the far end. If the far end is missing, step the pulse up one notch and re-test. If two events merge into one, step it down. Never compare a 30 ns trace against a 1 µs trace and argue about which loss number is right — they measured the span at different resolutions.

Averaging: the noise knob

Averaging is the OTDR re-shooting the same span and combining the results. More averaging — 30 seconds to 3 minutes is typical for a handover trace — pulls the trace out of the noise floor, so a 0.05 dB fusion splice that was invisible at 5 seconds of averaging shows up clearly at 60. The cost is time, and the temptation is to cut it: a quick 10-second trace is fine for "is the far end there," but it is not a handover measurement. If a loss reading wobbles between re-tests, or a small event appears and disappears, the averaging is too short, not the fiber. Set the averaging before you start, write it down, and use the same averaging when you re-test a span later — otherwise the comparison is meaningless.

Wavelength: test at the wavelength the span runs at

Fiber loss depends on wavelength, so the wavelength is not a preference — it is part of the test. Single-mode spans run at 1310 nm or 1550 nm; multimode at 850 nm or 1300 nm. A span that passes at 1550 nm can fail at 1310 nm, and that is expected: the fiber itself loses more at 1310, and a bend that barely registers at 1310 shows up clearly at 1550, which is exactly why comparing the two wavelengths is a standard bend test. The mistake is testing at one wavelength and judging against a budget written for another, or against the optic's wavelength. Before every test, check the wavelength on the OTDR matches the span and the optic that will use it, and write the wavelength in the ticket next to the loss — "1.1 dB" means nothing until you know whether it was 1310 or 1550.

Dynamic range: when the trace dies before the end

Dynamic range is how far the OTDR can still see a signal above its own noise floor — it is the result of pulse width plus averaging plus the instrument itself, not a knob you turn directly. When the far end is beyond the dynamic range, the trace simply fades into noise partway along the span, and that looks exactly like a fiber break. Before you call it a break, check the settings: step the pulse width up, lengthen the averaging, and re-test. If the far end appears, the fiber was never broken — the OTDR just could not see that far. On long spans, that re-test is the difference between a needless emergency splice call and a passing span.

And the reverse trap: a trace that ends cleanly at a connector with no far-end reflection may be a genuine break or a range limit — the settings tell you which one to suspect first.

A settings check before every test

1. Wavelength first. Set it to the span's working wavelength — 1310 or 1550 nm single-mode, 850 nm multimode — and confirm it on screen before the first shot.
2. Shortest pulse that reaches. Start short (10–30 ns for in-building spans). If the far end is missing, step up one notch and re-test.
3. Averaging for the job. Quick check: 10–30 s. Handover or dispute measurement: 60 s or more, until the trace stops changing between shots.
4. Launch cable on. The first connector sits in the dead zone without one — use a launch cable long enough to clear it, and record its length.
5. Same settings on re-test. Comparing today's trace to last month's only works if the pulse width, wavelength, and averaging match. If they do not, re-test at the old settings instead of arguing about the numbers.

Recording the settings in the ticket

A loss number without its settings cannot be reproduced or challenged — the next tech has to re-test from scratch instead of comparing against your trace. The minimum useful line records all four settings plus the launch cable:

ODF-B 12 ↔ ODF-D 07, SM, 1.2 km.
OTDR, 1310 nm, 100 ns pulse, 60 s averaging,
1 km launch cable.
Splice 412 m: 0.08 dB. End connector: 0.35 dB.
Total 1.1 dB vs budget 1.5 dB → PASS.

The shape to copy: span, wavelength, pulse width, averaging, launch cable length, each event with its loss, total vs budget, pass/fail. Six fields, and the next person on the ticket can re-shoot the same span at the same settings and get the same answer — which is the whole point of writing it down.