Anomalies of convolutional smoothing and differentiation

The original Savitzky-Golay convolutional smoothing method for simultaneous data smoothing and differentiation has been a useful tool for processing experimental data for nearly 30 years, However, unlike splining techniques, it leads to truncation of the end sections due to the center-point moving window approach, For a long time this was considered to be a necessary evil that could not be overcome. In 1990, Gorry addressed the truncation problem for uniformly spaced data in a mathematically elegant way. We implemented and tested this algorithm in our laboratory and observed some anomalies in the results produced. We offer explanations for these anomalies and suggestions of modifications so that the observed undesirable effects can be avoided.