Fiber Characterization
This domain covers the measurements beyond insertion loss that decide whether a fiber can carry a given bit rate: attenuation versus wavelength, chromatic dispersion, polarization mode dispersion, return loss, and multimode bandwidth. The exam tests units, limits, and the difference between linear and square-root accumulation.
Attenuation versus wavelength
Characterization starts with a bidirectional OTDR or a spectral attenuation set at 1310, 1550 and 1625 nm. Legacy fiber shows a water peak near 1383 nm from hydroxyl absorption; modern low-water-peak fiber does not. Splice losses are averaged from both directions.
Chromatic dispersion (CD)
Different wavelengths travel at different speeds, so a modulated signal spreads. Units are ps/(nm·km) and accumulated dispersion adds linearly with length. Standard single-mode fiber is about 17 ps/(nm·km) at 1550 nm and near zero at 1310 nm; non-zero dispersion-shifted fiber is around 4.5 ps/(nm·km) at 1550 nm. Compare accumulated CD with the transceiver's dispersion tolerance; coherent receivers compensate electronically, older 10G NRZ optics do not.
Polarization mode dispersion (PMD)
Random birefringence spreads the two polarization states. The coefficient is in ps/√km and accumulates as the square root of length: PMD_link = coefficient × √L. New fiber is typically 0.05–0.2 ps/√km. Rough tolerances are about 10 ps for 10G NRZ, 2.5 ps for 40G and around 1 ps for 100G before coherent compensation.
Return loss and multimode bandwidth
Optical return loss matters for analog and RF-over-glass links and for some coherent optics; a dirty connector can degrade ORL by many dB. Multimode bandwidth (OM3 ≈ 2000 MHz·km, OM4 ≈ 4700 MHz·km at 850 nm) must be measured with an encircled-flux launch, because an overfilled launch understates the modal bandwidth.
Worked example
A 40 km standard single-mode span: CD = 17 × 40 = 680 ps/nm; PMD = 0.1 × √40 = 0.63 ps. Reasoning: 680 ps/nm is inside the roughly ±800 ps/nm tolerance of many 10G NRZ interfaces, and 0.63 ps is far below the 10 ps PMD tolerance, so the span is fine for 10G. At 40G the same dispersion is well beyond tolerance and needs compensation, even though PMD is still acceptable.
Traps that cost marks
Adding PMD linearly instead of as √L; extrapolating PMD from a short spool as if it were deterministic (it is statistical); quoting CD in ps/nm instead of ps/(nm·km); characterizing a link without cleaning connectors first, which corrupts ORL; treating characterization as a synonym for a loss test.
Key points
- CD in ps/(nm·km) adds linearly with length; about 17 ps/(nm·km) at 1550 nm for standard single-mode.
- PMD in ps/√km adds as the square root of length.
- Rough PMD tolerances: 10 ps for 10G, 2.5 ps for 40G, about 1 ps for 100G.
- Multimode bandwidth must be measured with an encircled-flux launch.
- Characterization is required for high bit rates and long spans, not for every premises link.