Pulse selection algorithm for NM64 neutron detector

Abstract We propose an algorithm to sepearate pile-up pulses in neutron detectors by utilizing a standard pulse. For this method to be effective, the data must consist mostly of isolated pulses. First, we define a reference pulse by averaging a sample of clearly isolated pulses. Then, for an arbitrary signal, we calculate a shape deviation by summing up the squared residuals between it and the reference pulse. The pulse with the greatest shape deviation is removed from the process. We then recalculate the reference pulse and repeat until the remaining pulses have shape deviation within a threshold. These remaining pulses exhibit a very good linear trend between area and height, allowing us to screen those suspected as pile-ups. A pulse much higher than the final reference pulse, despite being in the area-height-trend and having low shape deviation, is considered a pile-up of two identical pulses. The final reference pulse is fitted to a function defined by two pieces of Gaussian-multiplied polynomial, normalized, and called the standard pulse. We attempted to fit the pile-up pulse with one standard pulse to separate pile-up pulses. If the sum of squared normalized residuals is higher than some threshold, we add one more pulse, try fitting again, and repeat up to three pulses. We apply this algorithm to the pulses collected from one counter at the Princess Sirindhorn Neutron Monitor station at the summit of Doi Inthanon Mountain, Chiang Mai, Thailand, measured by an oscilloscope. The algorithm correctly separates obvious pile-up cases, allowing to record individual pulse timing with improved accuracy. However, we found small pulses, usually belong to gamma rays, that blend with neutron pulse and pass 100 mV output filtration. After removing those under 100 mV, the pulse area distribution of the separated pile-up is consistent with that of single pulses except at very small pulse sizes.