Measurement Uncertainty of Fiber Optic Measurements
This domain covers how much you should trust a number. The exam tests which error sources dominate a fiber measurement, how independent contributions combine, and how to handle a result that sits close to a limit.
Where the error comes from
- Power meter: accuracy, linearity and wavelength calibration, typically ±0.2 dB and up to ±0.5 dB at extremes.
- Source: output stability, warm-up drift, and wavelength tolerance - a few nanometres of drift changes multimode loss slightly.
- Connector repeatability: ±0.2 to 0.5 dB every time a pair is mated, which is often the largest single term.
- Reference uncertainty: the reference method itself, adapter quality, and whether the reference was disturbed during the run.
- Launch condition: multimode mode distribution; uncontrolled launch biases loss low.
- Environment and handling: temperature, bend radius, fiber movement, contamination.
- Operator technique: cleaning quality and consistency matter more than most people expect.
Combining contributions
Independent contributions combine by root-sum-square, not by addition: u = √(u₁² + u₂² + …). The expanded uncertainty is U = k·u with k = 2, giving roughly 95% confidence. Adding contributions arithmetically overstates the uncertainty and can wrongly fail good links.
Reporting and guard bands
Report the result with its uncertainty. If a reading falls inside the uncertainty band around a limit, the honest conclusion is indeterminate - investigate, re-clean, re-measure, or agree an acceptance rule with the customer. Do not declare a pass on a number you cannot distinguish from a fail.
Worked example
Contributions for a 1310 nm single-mode insertion loss measurement: meter accuracy 0.20 dB, source stability 0.10 dB, connector repeatability 0.30 dB, reference uncertainty 0.20 dB. Combined: √(0.04 + 0.01 + 0.09 + 0.04) = √0.18 = 0.42 dB. Expanded uncertainty U = 2 × 0.42 = 0.85 dB. With a 2.0 dB limit, a measured 2.4 dB is a clear fail; a measured 2.2 dB is indeterminate; a measured 1.0 dB is a clear pass.
Traps that cost marks
Adding uncertainty terms linearly; quoting a meter specification of ±5% of reading as though it were a dB value (convert it at the actual power level); ignoring connector repeatability because it is not printed on a datasheet; treating a single measurement as exact when the reference was set once and never re-verified.
Key points
- Independent uncertainty contributions combine by root-sum-square.
- Expanded uncertainty uses k = 2 for roughly 95% confidence.
- Connector repeatability of ±0.2 to 0.5 dB is often the dominant term.
- A result inside the uncertainty band at the limit is indeterminate, not a pass.
- Percent-of-reading specifications must be converted before use in a dB budget.