Pile-up Pulse Recognition Method Based on Ballistic Deficit Shape Feature

In X-ray fluorescence spectroscopy analysis, high intensity X-ray is usually used to irradiate the sample for obtaining stable energy spectrum quickly, which also leads to serious pulse pile-up, count rate loss and energy resolution degradation. It is effective for improving energy spectrum performance to recognize and reject pile-up pulses. Fast shaping discrimination technology (FSDT) and pulse feature time discrimination technology (FTDT) are popular methods for identifying pile-up pulses. It is impossible for FSDT to identify pile-up pules with intervals shorter than the fast pulse resolution time. The measured feature time is subject to pulse-height variation and ADC sampling frequency, which limits FTDT detection capability. In this paper, a pulse shape discrimination method was proposed. An original digital pulse was shaped to a triangular pulse and a trapezoidal pulse. The amplitude ratio, which is also known as time-invariant pulse-shape signature (TIPS), is a pulse shape feature for recognizing pile-up pulse. It is not necessary to measure the pulse arrival time. TIPS is not dependent on ADC sampling frequency and pulse amplitude, but only on ballistic deficit and noise fluctuations. The triangular pulse shaper is a bandpass filter that can suppress high-frequency noise, which can reduce TIPS broadening. The trapezoidal pulse shaping algorithm can be immune to ballistic deficit, while the triangular pulse shaping algorithm cannot. Generally, there are higher TIPS value in pile-up pules because of ballistic deficit. A threshold can be used to determine whether the pulse is pile-up. The TIPS value of some piled pulses is less than that of non-piled pulses. If the threshold is set relatively low, many non-piled pulses will be misjudged. Conversely, a lot of pile-up pulses will be missed. The loss rate of the non-piled pulses can be determined by its TIPS distribution. A copper sample is used for the test experiment. When the tube current is set to 3.9 μA, the measured count rate is only 6 500 s-1 in the fast shaper so that the pile-up can be almost ignored. The measured TIPS distribution can be regarded as the TIPS distribution of non-piled pulses. When the tube current is set to 1 mA, the measured count rate reaches 1.22×106 s-1 in the fast shaper. When α is the same, the accuracy, recall rate and F1 score are all improved by using the proposed method to identify pile-up pulses. For example, when α is 15%, the accuracy, recall rate and F1 score are increased to 73.55%, 78.75% and 76.06% respectively. The peak-to-total ratio of the Kα peak of Cu is increased to 76.6%. The proposed method can effectively identify pile-up pulses.