Recent work has shown the importance of the fast Fourier transform (FFT) in the automatic analysis of fringe patterns. Three problem areas are encountered when the FFT is used as an analysis tool in this fashion, namely, aliasing, the picket fence effect and leakage. The authors consider leakage to be by far the biggest problem in this application. The paper defines what causes leakage and suggests how leakage may be controlled using data weighting functions. Comparative results obtained using a variety of functions are presented.