Fiber Optic Attenuation

Two intrinsic mechanisms and several extrinsic ones. Intrinsic Rayleigh scattering arises from microscopic density and composition fluctuations frozen into the glass during drawing; it scales as λ⁻⁴, so shorter wavelengths scatter far more. Intrinsic infrared absorption arises from the vibrational overtones of the SiO₂ tetrahedral lattice; it grows steeply beyond 1600 nm. The famous 1550 nm minimum is where the falling Rayleigh curve and the rising infrared curve cross. Extrinsic effects include hydroxyl (OH⁻) absorption at 1383 nm from residual water in the preform, macrobending losses at tight bend radii, microbending from cabling stress, and connector/splice losses at every junction.

Because in multimode fiber the guided light travels in many spatial modes, some of which extend further into the cladding than others. Modes near the core-cladding boundary experience more scattering and more absorption than modes near the core centre, and any bend in the fiber preferentially strips the outer modes. The result is that the effective attenuation coefficient is higher than for single-mode fiber, where only the fundamental mode is guided and it is more tightly confined. Typical values are 2.5–3.5 dB/km for OM1/OM2 at 850 nm, versus 0.35 dB/km for SMF-28 at 1310 nm and 0.20 dB/km at 1550 nm.

Standard single-mode fiber has a small absorption peak around 1383 nm caused by residual hydroxyl (OH⁻) ions in the glass. The peak adds 0.5–1.0 dB/km to the intrinsic loss and effectively blocks the 1370–1420 nm region from use in WDM systems. This splits the low-loss spectrum into two windows — the O-band (around 1310 nm) and the C/L bands (around 1550–1625 nm). "Low-water-peak" or "full-spectrum" fiber, made with a modified preform process, reduces OH⁻ content by two orders of magnitude, opening the E-band (1360–1460 nm) between them. Full-spectrum fiber adds roughly 50% to the usable bandwidth of a single span and is important for CWDM systems that want to use many channels across the full 1260–1625 nm range.

A complete link budget includes: (1) fiber attenuation (α × L); (2) connector losses, typically 0.3–0.5 dB per mated pair, with maybe 2 connectors at each end; (3) fusion splice losses, typically 0.05–0.1 dB each, with one splice every 2–5 km in a cabled span; (4) macrobend losses at patch panels and cable bends, typically a few tenths of a dB per occurrence if routing is good; (5) aging margin, historically 3 dB over the 20–25 year design life of a long-haul link; (6) repair margin, typically 0.1 dB/km for buried cable to cover future splice repairs; and (7) any passive components in the path — WDMs, isolators, splitters, circulators — each contributing their insertion loss. A 100 km span of standard SMF-28 at 1550 nm has 20 dB of pure fiber loss, and a realistic budget of 25–28 dB after all additions.

The standard technique is the cut-back method: measure the power at the far end of the fiber link, then cut the fiber a known short distance from the source and measure again. The difference in dB divided by the cut length gives α. This is destructive but precise, and is the primary reference method in standards such as IEC 60793-1-40. In the field, technicians use an OTDR (Optical Time-Domain Reflectometer), which sends a short pulse down the fiber and measures the backscattered light as a function of time. The slope of the OTDR trace gives the local attenuation coefficient, and any sudden drop indicates a splice, connector, bend or break. OTDR is non-destructive and can survey an entire span from one end, but its absolute accuracy is limited by the launch conditions and backscatter capture fraction, so it is usually cross-checked against a power-meter based insertion-loss measurement.