Fiber test day: OTDR traces, ghosts, gainers and dirty connectors

The splicers closed the last can on the pole line at dusk, and the crew lead's text arrived before the coffee did: "Test day. Shoot everything both directions, scope every connector, and nobody calls it done until the numbers are in the report." Test day is when a fiber job stops being about cable and starts being about light, and the vocabulary changes with it. Out on the pull, people talk about tension, rodders and mule tape. On test day they talk in decibels, kilometers and a few odd nicknames for things on a screen.

Shooting the fiber

Nobody on a crew says they are going to perform an optical time domain reflectometer measurement. They say they are going to shoot it. The OTDR fires short pulses of light down the fiber and listens to what comes back. What returns is mostly backscatter, light scattered by the glass itself, and the instrument plots it against distance as a trace: a line that slopes gently downhill as the fiber loses a little power every kilometer. Every bump, spike or step in that line is an event.

First the tech clips on a launch cable, a box of reference fiber a few hundred meters long, so the first connector on the link does not sit in the instrument's blind spot. That blind spot has a proper name, the event dead zone: after a strong reflection the OTDR needs a stretch of fiber to recover before it can see the next thing. A receive cable on the far end does the same job for the last connector.

"Launch on, receive on, shooting 1310 and 1550," the tech calls over the radio. Two wavelengths, because a tight bend barely registers at 1310 nm and jumps right out at 1550 nm.

Reading the trace

A fusion splice usually shows as a small step down with no spike. A connector shows as a spike, because the joint between two mated tips reflects a little light, followed by a step. The height of that spike is the reflectance, read as a negative number: minus 55 is a clean, healthy singlemode connector, and minus 30 is a problem.

Then there are things on the trace that are not really there. A ghost is a false event, a copy of a strong reflection that bounced back and forth inside the fiber and arrived late. The tell is spacing. "That event at 4.2 km is exactly twice the distance to the panel, and there's no loss after it. Ghost." A real event costs light. A ghost never does.

The other famous phantom is the gainer: a splice that appears to add light. Glass cannot amplify itself: the fiber after the splice simply scatters more light back than the fiber before it, usually because the fiber after the splice has a smaller mode field diameter than the fiber before it. Shoot it from the other end and it shows as a loss bigger than the real one, which is why the crew lead wanted both directions. Bidirectional averaging adds the two readings and halves them, and the gainer and its evil twin cancel out to the true splice loss.

"Splice 14 reads a gain of 0.08 from the A end and a loss of 0.22 from Z," the tech says. "Average is 0.07. Good burn." That last word is splicer slang for making a splice, after the electric arc that welds the glass.

The dirty connector is always guilty

Ask a fiber tech what causes most failed tests and the answer is instant: dirt. One speck of dust on the end face of a singlemode connector can block much of the core, which is only about nine microns across. Dust also rides on the ferrule, the ceramic cylinder that holds the fiber, and gets crushed into the glass the moment two connectors are mated.

Hence the rule every new hire hears in the first week: inspect before you connect. Every end face goes under the scope before it goes into a port. If it is dirty, out comes the click cleaner, a pen-shaped tool that drags fresh cleaning tape across the tip with one push. Then it gets scoped again, because cleaning sometimes just moves dirt around.

Polish type matters too. A UPC connector has a domed tip with no angle and, on singlemode, a blue body. An APC connector is polished at 8 degrees and colored green, so its reflection heads off into the cladding instead of back toward the laser. Mate a green connector to a blue one and an angled face meets a domed one across an air gap. The loss jumps, the return loss collapses, and both end faces can be damaged. "Green to green, blue to blue" gets said out loud on every job.

Loss tests and budgets

The OTDR maps a fiber. The OLTS settles arguments. It is a stable light source on one end and a calibrated power meter on the other, and it measures insertion loss, the total light the link swallows from end to end. Before the first reading the tech sets a reference, usually a one-jumper reference, so that the connectors at both ends of the link count along with the glass and splices in between.

Every result gets compared to the loss budget, the number worked out on paper from fiber length, splice count and connector count. "Budget's 2.1, it read 1.6, pass." Whatever the transceivers could tolerate beyond the real loss is margin, and network engineers like plenty of it.

Once equipment goes in and the fiber is lit, the talk moves to light levels: transmit and receive power from the optics, in dBm. "What's your receive on port 3?" "Minus 31." "Should be around minus 12." That gap means something between the two ends is losing 19 dB it has no business losing, and someone is heading back out with a scope.

Hunting the fault

For a break or bad bend near the panel, the fastest tool on the truck is the red light. Its proper name is VFL, for visual fault locator, and it pushes bright red laser light into the fiber. Wherever the glass is cracked or bent too tight, the jacket glows. A macrobend in a splice tray, a fiber pinched under a cover: each one lights up like a birthday candle.

Farther out, the OTDR earns its keep. A clean break shows as a big reflective spike followed by nothing at all, and the distance on the screen tells the crew which vault to open. If that distance lands on a road crossing where a contractor was digging, the cause has a name too: backhoe fade, the trade's dry joke for an outage delivered by an excavator bucket.

Not every failure is in the glass. If both transceivers show light going out, nothing coming in, and the link stays down, someone usually has to roll the pair, swapping the two fibers so transmit meets receive. The proper word for getting that right is polarity, and getting it wrong is the classic reason a perfectly good link looks dead.

Five words to learn first

1. Shoot: to test with the OTDR, ideally from both ends.
2. Ghost: a false event, a reflection of a reflection. Real events cost light, ghosts do not.
3. Gainer: a splice that seems to add light. Average both directions and it disappears.
4. Inspect before you connect: scope, clean, scope again, then mate.
5. Red light: the visual fault locator, the quickest way to find a break within reach.