Glossary entry

Cutback method

A measurement technique for waveguide propagation loss in which the transmitted optical power is measured at multiple known waveguide lengths and the loss extracted from the slope of transmission versus length.

The cutback method extracts the propagation loss $\alpha$ of a waveguide by measuring transmitted optical power for a set of waveguides of different known lengths. Power transmitted through length $L$ obeys

P(L) \;=\; P_0 \cdot \eta_c^2 \cdot 10^{-\alpha L / 10},

where $P_0$ is the source power and $\eta_c$ is the per-end coupling efficiency. Taking the logarithm:

10 \log_{10} P(L) \;=\; \text{constant} - \alpha L.

A least-squares linear fit of $\log P$ versus $L$ yields $\alpha$ directly, independent of absolute source power and coupling efficiency.

Two variants are in standard use.

Destructive cutback measures the same waveguide at successively shorter lengths by cleaving or polishing between measurements. Each cut reuses the same input fiber and coupling geometry, eliminating per-waveguide coupling variation. Used primarily for optical fiber loss measurements over meter-scale lengths.

Paired-device cutback uses a set of nominally identical waveguides of different lengths fabricated on the same chip, measured through identical coupling geometry (typically grating couplers at standard fiber-array pitch). This variant is the dominant method in modern integrated photonics. Foundry process control modules typically include cutback waveguide sets spanning 0.1–10 cm in spiral or boustrophedon layouts.

For low-loss platforms (silicon nitride at $<0.1$ dB/cm), individual cutback waveguides are routed as long spirals to accumulate measurable loss; the longest path may be tens of centimeters to meters of total propagation length.

The full procedure including fit methodology and uncertainty analysis is in Waveguide Propagation Loss by the Cutback Method.