Coherence scanning interferometry (CSI) is an optical profilometry technique that uses the scanning of white light
interference fringes over the depth of the surface of a sample to measure the surface roughness. Many different types of
algorithms have been proposed to determine the fringe envelope, such as peak fringe intensity detection, demodulation,
centroid detection, FFT, wavelets and signal correlation. In this paper we present a very compact and efficient algorithm
based on the measurement of the signal modulation using a second-order nonlinear filter derived from Teager-Kaiser
methods and known as the five-sample adaptive (FSA) algorithm. We describe its implementation in a measuring system
for static surface roughness measurement. Two envelope peak detection techniques are demonstrated. The first one,
using second order spline fitting results in an axial sensitivity of 25 nm and is better adapted to rough samples. The
second one, using local phase correction, gives nanometric axial sensitivity and is more appropriate for smooth samples.
The choice of technique is important to minimize artifacts. Surface measurement results are given on a silicon wafer and
a metallic contact on poly-Si and the results are compared with those from a commercial interferometer and AFM,
demonstrating the robustness of the FSA algorithm.