Novel discrimination parameters for neutron-gamma discrimination with liquid scintillation detectors using wavelet transform

Abstract

It has been observed that the discrimination performance of the wavelet transform method strongly depends on definition of discrimination parameters. These parameters are usually obtained from a combination of scaling functions at different scales, which represents the energy density of the wavelet coefficients. In this paper, the discrete wavelet transform (DWT) at minimum possible values of scale was investigated. Novel pulse shape discrimination parameters have been proposed for neutron and gamma discrimination in a mixed radiation field and tested with modeled pulses. The performance of these parameters was also validated in terms of quality of discrimination using experimental data of mixed events from an AmBe source collected with BC501 liquid scintillation detector. The quality of discrimination was evaluated by calculating a figure of merit (FOM) with all parameters under same experimental and simulation conditions. The FOM obtained with the proposed novel parameters was also compared with the charge comparison method. The proposed parameters exhibit better FOM as compared to the charge comparison method when high levels of noise are present in the data.

Optimization of digital neutron-gamma discrimination methods in time and frequency domain

“…The DWT method is also a frequency-domain method that calculates the wavelet coefficients 𝑊 𝑓 (𝑎, 𝑏) by convoluting the given discrete time pulse 𝑓 [𝑛], 𝑛𝜖Z with a suitable mother wavelet function 𝜓 𝑗,𝑘 (𝑡) sampled over a dyadic grid, at different values of shift (𝑏) and scale (𝑎) [19,23]. The wavelet coefficients are further used to plot a scaling function 𝑃(𝑎) corresponding to each pulse, which is then identified using various discrimination parameters [10]. There are several mother wavelet functions that can be used for PSD, however, the best results have been obtained with the wavelets from the Daubechies family [6].…”

“…TDMs are simple to use because they determine the particle type by directly extracting a discrimination feature from the time-domain digitized pulses. However, the number of pile-up events in the dataset and the level of noise present in the dataset have a significant impact on the performance of TDMs [6,10]. Additionally, TDMs always need a noise-filtering algorithm during the pre-processing stage, adding to the computational load.…”