The Only Cheap Moment to Measure
At handover, everything is open, the crew is on site, and the equipment is already out. Measuring one more span costs minutes.
Three years later, the same measurement means a truck roll, a customer outage window, and an engineer who has never seen this plant before trying to work out what should be there.
Acceptance testing is not paperwork. It is the baseline that every future fault diagnosis is measured against. That framing changes what you choose to record.
The Minimum Acceptance Pack
For each fiber path:
Bidirectional OTDR traces, both wavelengths. Saved as native trace files, not screenshots. A screenshot cannot be re-analysed with different markers; a trace file can.
End-to-end insertion loss, both wavelengths. With the reference method recorded, as covered in the insertion loss lesson.
Optical return loss where the specification calls for it — particularly on PON plant with video overlay.
Connector end-face images for the terminations that were made or mated during the work. This one is skipped almost universally and is disproportionately valuable, because connector degradation is the most common later fault.
Fiber length measured, alongside the design length. A significant discrepancy at handover is a finding worth chasing immediately.
The Metadata That Makes It Usable
The measurements are only half of it. What turns a folder of files into a usable baseline:
- Which fiber, in which cable, on which route. Named the same way the drawings name it.
- Which direction each trace was taken in.
- Reference method used for loss measurement.
- Instrument model and serial, and pulse width and range settings.
- Launch cable used, and its length.
- Date and technician.
The instrument settings matter more than people expect. When a future engineer compares their trace against yours and gets a different answer, the first question is whether the settings match. Without that record, the comparison is worthless.
Pass Criteria Have to Be Agreed Before Testing
This sounds obvious and is routinely violated. If the criteria are settled after the numbers are in, the discussion becomes a negotiation instead of a test.
What needs to be agreed in writing beforehand:
- Maximum end-to-end loss per path, derived from the design budget rather than just the equipment limit.
- Maximum individual splice loss, and whether it is judged on the bidirectional average.
- Maximum connector loss and minimum return loss.
- Reference method for insertion loss.
- What happens on a marginal result — retest, rework, or accept with a note.
That last item prevents the most common dispute. A splice at 0.28 dB against a 0.3 dB limit passes, but if three of them appear on the same span, something systematic is happening and "it passed" is the wrong conclusion.
Read the Results as a Set
Individual pass marks hide patterns. Before signing off, look across the whole job:
Are splice losses clustered high? As covered in the splicing lesson, that usually means tooling — a worn cleaver blade — not bad luck.
Is one crew's work consistently different? Worth knowing, and worth addressing as training rather than rework.
Do measured lengths track the design? Systematic overlength suggests slack storage that was not in the plan, which will matter when someone converts fiber distance to dig location.
How much margin is left? As covered in the power budget lesson, a link that passes with 1 dB spare is a link that will fail on its second repair. Recording the remaining margin at handover gives the operator a way to prioritise.
What Handover Should Include
The acceptance pack is one part. The full handover also needs:
- As-built drawings reflecting what was actually installed, not what was designed.
- Splice schedules — which fiber connects to which, in which tray, in which closure.
- Slack locations and lengths, which is what makes future fault localisation accurate.
- Closure and hand-hole locations, ideally with coordinates.
A network with perfect test results and missing as-builts behaves during a fault like an undocumented network, because that is what it is.
Summary
- Handover is the only cheap moment to measure. Treat the results as the baseline for the plant's whole life.
- Save native trace files, not screenshots. Record instrument settings, direction, reference method and launch cable.
- Agree pass criteria — including what happens on marginal results — before testing starts.
- Read results as a set. Clustered marginal splices are a tooling signal, not luck.
- Record remaining margin. It tells the operator how much future repair the link can absorb.
- As-builts, splice schedules and slack locations are part of the deliverable, not an afterthought.