OTDR dead zones: how to choose the pulse width and stop losing events

A patch panel, two connectors three metres apart, a report showing only one of them. Nobody cheated: the instrument was set to a 100 ns pulse, which in fibre occupies some ten metres, and everything happening inside those metres becomes a single event. Pulse width is the setting that weighs most on an OTDR measurement, and it is almost always the one the specification never mentions.

What the two dead zones really are.

The OTDR sends a light pulse into the fibre and listens to the fraction that comes back by backscattering. The pulse has a duration, so while it travels it illuminates a length of fibre. At a reflective event — a connector, a mechanical splice — part of the light returns all at once and the receiver saturates. Hence two dead zones, which are not the same number:

the event dead zone: the minimum distance after a reflective event within which a second event can no longer be distinguished;
the attenuation dead zone: the distance, always longer, after which the trace returns to the backscatter level and loss becomes measurable again.

They do not appear in the report: they are a characteristic of the instrument. Manufacturers declare them in line with IEC 61746-1 (2009), Calibration of optical time-domain reflectometers (OTDR) – Part 1: OTDR for single mode fibres, which ITU-T G.650.1 cites as the reference for calibrating backscattering equipment. But a dead zone without its conditions says nothing: read it together with the pulse width at which it is measured and the reflectance of the test event.

Pulse width is quoted in nanoseconds and read in metres.

In fibre, light travels at roughly 2×10⁸ m/s (a typical group index of around 1.47) and the OTDR measures a round trip: one microsecond of pulse is worth about 100 metres of fibre on the trace. The practical scale:

5 ns → ~0.5 m
10 ns → ~1 m
100 ns → ~10 m
1 µs → ~100 m
10 µs → ~1 km

The event dead zone cannot be shorter than that footprint, and the attenuation one is always longer. Translated: with a 1 µs pulse nobody can separate two splices fifty metres apart, and a thirty-metre jumper inside a rack simply does not exist. But with a few nanoseconds a twenty-kilometre backbone disappears into the noise before the far end.

The trade-off is written into the Recommendation.

ITU-T G.650.1, edition 01/2024 approved on 13 January 2024, states in clause 6.4.2.2.1 that “adjustment of the pulse width may be required to obtain a compromise between resolution and dynamic range”; clause 6.4.2.2.2 adds that pulse width and repetition rate should be consistent with the desired resolution and the length of the fibre. Two sentences that contain the whole craft:

short pulse = fine resolution, little energy in the fibre, a noisy trace, reduced dynamic range;
long pulse = more energy, you reach the end of long runs, but nearby events merge into one.

Rule of thumb: doubling the pulse buys about 1.5 dB of dynamic range, because the backscattered power doubles while the vertical scale of the trace is in one-way decibels. Averaging for longer reduces noise; it does not improve resolution.

So a single acquisition is not a test: you need at least two passes, a short pulse over the patching area and a long pulse over the run. Then there is the repetition rate, which G.650.1 names and almost nobody sets by hand: if the instrument fires again before the last echo has returned, ghost events appear on the trace that the fibre does not contain.

How long the launch fibre has to be.

ITU-T G.650.3, edition 08/2017 approved on 13 August 2017, is explicit in clause 6.1.3: OTDR testing should be carried out with a launch lead “long enough (typically 1 km to 2 km)” to measure the first connector in the link, with a similar tail lead at the far end for the last one. The same clause asks, when commissioning a new link, for measurement in both directions on every fibre and at least two wavelengths (1310 and 1550 nm on G.652 fibres).

The criterion is not traditional but physical: the lead has to exceed the attenuation dead zone at the pulse width in use, and give enough fibre for the analysis to sit on a stable backscatter slope. Change the pulse and you change the lead you need. Appendix I of the same Recommendation gives a real example at 1550 nm: a link of about 20 km, sections of about 2 km, launch and tail leads of about 3.8 km each. There, one connector measures 0.534 dB in one direction and 0.136 dB in the other, giving a bidirectional average of 0.335 dB; a splice reading 0.362 dB from one side appears as −0.062 dB from the other — an impossible gain — and the true value is the average, 0.150 dB. Without the two leads the end connectors would have stayed inside the dead zones. The rest of the report is covered in this note.

In a data centre the geometry flips — a few hundred metres, pulses of a few nanoseconds, launch cords of 100-150 m — but the lead is always sized on the pulse, not on habit.

What to write into the specification.

Seven lines that make bids comparable:

pulse width declared for every trace — G.650.1, clause 6.4.2.4, lists it among the data to be presented with the result;
at least two acquisitions with different pulse widths wherever events sit close together: patching, building entries, adjacent closures;
launch and tail leads with their length stated, consistent with the pulse width used (1-2 km as a guide in outside plant);
the instrument’s dead zones, event and attenuation, declared with their measurement conditions;
group index set and stated: if it is wrong, the distances do not match the as-built;
two wavelengths and bidirectional measurement with a per-splice average, as G.650.3 requires;
native files (.sor or equivalent) handed over with the PDF.

And one clause worth more than the rest: the pass/fail threshold has to be written together with the pulse width used to measure it. A splice at 0.1 dB with 1 µs and the same splice with 30 ns are not the same statement — and the difference lands on the optical budget of the run — the margin you will still have in five years.

The point.

A dead zone is not a fault of the instrument: it is the consequence of a setting, and it has to be declared. A report that does not say which pulse width it was measured with is not verifiable, today or at the next maintenance visit. That is why every run we certify comes out with the measurement parameters next to the values — on backbones as much as in data halls, where three metres between two connectors is the norm, not the exception.

Do you have a test to set up or a specification to write? Get in touch: defining pulse widths, launch fibres and thresholds before the measurement costs far less than repeating it.