Reading an OTDR trace

THE SHORT VERSION

An OTDR (optical time-domain reflectometer) fires short pulses of light into one end of a fiber and listens to what comes back. Glass scatters a tiny, steady fraction of light backward everywhere along its length, and anything abrupt — a connector, a break, the far end — throws back a brighter flash. Time the echoes and you know where things are; measure how faint they've grown and you know how much light was lost getting there. It's sonar, with photons.

THE AXES

Horizontal: distance. The instrument measures round-trip time and converts it with the fiber's group index of refraction (IOR): distance = (speed of light × time) ÷ (2 × IOR). The ÷2 is the round trip. A wrong IOR setting stretches or shrinks every distance on the trace by the same proportion, so set it from the cable manufacturer's data before you trust a locate.

Vertical: relative power, in dB. It's a logarithmic scale of backscattered power, and the OTDR scales it so a drop on screen reads as one-way loss — the light actually went out and back, and the instrument halves that for you. The absolute height means little; slopes and steps are what you read.

A healthy trace reads left to right: a tall spike at the front panel, a long straight line sloping gently down, small steps where events are, and a final spike (or a plain drop) where the fiber ends, after which only the instrument's noise floor remains.

EVENTS

Reflective events

Where light crosses a glass-air-glass boundary, some of it bounces straight back (Fresnel reflection) — a spike above the backscatter line, followed by a step down for the loss. Connectors and mechanical splices do this, and so do a cleaved or cracked end. A spike at an unexpected distance on a link with no connector there is worth a hard look: it may be a crack.

Non-reflective events

A fusion splice melts two fibers into continuous glass, so there's no boundary to reflect from — just a step down. A macrobend (a fiber bent tighter than it tolerates) leaks light out of the core and also shows as a plain step.

Telling a splice from a bend: bend loss grows with wavelength, so a bend shows a noticeably bigger step at 1550 nm than at 1310 nm, while a fusion splice reads about the same at both. Shooting two wavelengths is the cheapest bend detector you own.

DEAD ZONES AND THE LAUNCH CABLE

A bright reflection briefly blinds the OTDR's receiver, the way a camera flash leaves you seeing spots. The distance it takes to recover is a dead zone, and there are two kinds:

• Attenuation dead zone — the distance after a reflective event before the trace settles back onto the backscatter line well enough to measure the loss of the next event. Always the longer of the two.
• Event dead zone — the shortest distance after a reflective event at which a second reflective event shows up as a separate event at all.

Both grow with pulse width: a wider pulse carries more energy and sees farther, but blurs more of the trace. That's the basic OTDR trade — range versus resolution.

The worst flash of all is the instrument's own front-panel connector, which means the first connector of the link you actually care about sits inside a dead zone. A launch cable (also called a pulse suppressor) — a known length of fiber between the OTDR and the link — moves that first connector out where it can be measured. A receive (tail) cable at the far end does the same for the last connector, which otherwise has no backscatter after it to measure against. How long? It depends on pulse width and the instrument; for outside plant, a few hundred meters to a kilometer or more is common. That's a rule of thumb, not a standard — check your OTDR's documentation.

SLOPE

The straight stretches between events slope down at the fiber's attenuation rate, in dB/km. Read it as the drop between two points on a clean stretch divided by the distance between them. Typical values for cabled singlemode outside-plant fiber:

Typical, not guaranteed: the cable's datasheet is the authority for your cable. Design limits in the TIA-568 family are looser than these typical figures. A stretch that slopes noticeably steeper than its neighbors is a clue — a stressed section, or a different fiber type spliced in.

THE GAINER

Sometimes a splice reads as a gain — the trace steps up. Glass doesn't amplify light; this is an artifact. The OTDR assumes both sides of the splice scatter light back equally well. If the downstream fiber has a different mode field diameter or backscatter coefficient, it returns more light per meter, and the trace jumps up even though light was lost.

Shoot the same splice from the other end and the error flips sign: it reads as extra loss. Average the two directions and the backscatter difference cancels, leaving the true splice loss. That's why bidirectional testing is the reference method for splice loss: a one-way reading of any splice between dissimilar fibers is biased, and a gainer is just the bias made obvious.

WORKED EXAMPLE

A described trace at 1310 nm, shot from end A through a 500 m launch cable. The numbers are illustrative, chosen to show each kind of event:

• 0 m: big spike — the front panel. Ignore it; that's what the launch cable is for.
• 500 m: spike, then a 0.30 dB step. Reflective, at the end of the launch cable: the first connector of the link, measured cleanly. 0.30 dB is inside the 0.75 dB TIA allowance for a mated pair.
• 500 m to 4,500 m: straight line falling from −3.10 dB to −4.42 dB on the vertical scale. Slope = 1.32 dB ÷ 4.0 km = 0.33 dB/km — right where typical singlemode sits at 1310 nm.
• 4,500 m: the trace steps up 0.08 dB with no spike. Non-reflective, so a fusion splice — and the step up says gainer. Shot from end B, the same splice reads 0.20 dB of loss. True loss = (−0.08 + 0.20) ÷ 2 = 0.06 dB. A good splice that only looked strange.
• 7,200 m: a 0.15 dB step, no spike — at 1310 nm. The 1550 nm shot shows 0.60 dB at the same spot. A splice wouldn't care about wavelength that much; a bend would. Go look at that location for a tight slack loop or a crushed section.
• 9,000 m: a spike, a step, then a short stretch of backscatter and a final spike into noise: the far-end connector, then the receive cable and its open end. Link length: 9,000 − 500 = 8,500 m.

Verdict: the link is sound except one location. The gainer is fine once averaged; the event at 7,200 m is the call to make.

SOURCES AND LIMITS

Mechanisms are general OTDR physics. The mated-pair allowance is from the TIA-568 family; attenuation figures are typical values, not limits. The worked trace is illustrative. Your instrument's manual and your cable's datasheet outrank this page.

← All field notes