The detection of a small number of large particles whose size lies outside the required size distribution is important in the processing of many materials when those particles result in the deterioration of the product quality. Laser diffraction, which is a very popular technique for on-line monitoring and process control, has, however, limited sensitivity to those few large particles. This paper deals with the improvement of this sensitivity. The fluctuations of the signal received by the detectors were analyzed by both experiment and simulation. Two statistical approaches, crosscorrelation and principal component analysis (PCA), were applied. A detection procedure is proposed which is based on the combination of sweep selection through PCA and an appropriate deconvolution of the selected sweeps. The preliminary experiments with Al2O3 powder showed an improved sensitivity to the large particles.