Measurement uncertainty of fiber optic measurements
This domain tests how confidently a fiber measurement can be compared with a limit. The exam expects you to identify uncertainty sources, combine them correctly, report an expanded uncertainty, and apply a guardband when a result is close to a pass/fail threshold.
Sources of uncertainty
For insertion loss, contributors include source power stability, power-meter accuracy and linearity, wavelength uncertainty, connector repeatability, reference-cord condition, temperature, and operator technique. For OTDR work, add index-of-refraction uncertainty, distance calibration, pulse-width effects, and event-reading resolution. Resolution is not accuracy: a display that reads 0.01 dB does not mean the measurement is good to 0.01 dB.
Combining uncertainty
Independent standard uncertainties are combined by root-sum-square (RSS), not by simple addition. The combined standard uncertainty is the square root of the sum of the squares. Expanded uncertainty usually uses a coverage factor k = 2, giving roughly 95% confidence, and is reported as ±U. When a measured value is within U of a limit, use a guardband or report the result as indeterminate rather than a clean pass.
Worked example
A loss measurement has standard uncertainty contributions of 0.10 dB from the meter, 0.05 dB from source stability, 0.12 dB from connector repeatability, and 0.08 dB from wavelength/reference effects. Squares: 0.0100 + 0.0025 + 0.0144 + 0.0064 = 0.0333. The square root is about 0.18 dB. Expanded uncertainty at k = 2 is about 0.37 dB. A result of 2.8 dB against a 3.0 dB limit should not be called a confident pass; the true value could exceed the limit.
Traps
- Adding uncertainty contributions linearly and overstating the uncertainty.
- Using k = 1 when the specification calls for k = 2.
- Treating display resolution as measurement accuracy.
- Ignoring connector repeatability, which often dominates field loss tests.
- Reporting a pass without a guardband when the result is inside the uncertainty band.