Peak search and analysis of gamma-ray spectra with very low statistics

A study has been made of the efficiency of various techniques for use in the analysis of peaks in gamma spectra with very poor statistics. A comparison was made of the efficacy of variously shaped transforms used for smoothing and of four different zero-area transforms when used to search for very weak peaks. Results are given for computer generated data consisting of a peak on a flat background with added ''random'' noise. A related problem is the accurate analysis for area and position of peaks with only a few counts per channel. The standard least squares method failed to give a satisfactory fit. A simple technique is given for correcting this difficulty