Four digital methods are investigated for neutron-gamma
discrimination in organic scintillation detectors. A fast-digital
pulse processing system has been used to optimize the PSD methods
and to investigate a mixed radiation field with light output in the
range of 100 keV to 5 MeV from an Am-Be source. The methods used
for pulse shape discrimination are the charge comparison, the
simplified digital charge collection, the frequency gradient
analysis, and the discrete wavelet transform method. The performance
of each of these methods has been assessed by calculating the
figure-of-merit and the probability of misclassification. It has
been found that the frequency-domain algorithms perform more
effectively for low-energy pulses and require a smaller processing
gate width than the time-domain methods. They are therefore more
suited to environments with high count rates where pulse pileup is a
common attribute of data. In addition, frequency-domain methods are
more resistant to random noise present in the dataset compared to
time-domain methods.