A real-time peak-detection approach for nuclear detection and its implementation on an FPGA

Detecting a pulse correctly is a key process in nuclear detection. Because the radiation emission is a random process, it is hard to design a suitable peak-detection approach in FPGA. The error detection will influence the final energy spectrum and flood histogram. In order to improve the result of nuclear detection, this paper proposes a novel method for nuclear signal peak-detection, which can improve both the effective counting rate and the quality of pulses in real-time. The main method is to establish a normalized reference pulse regardless of waveform through the least squares method. By calculating the loss between the incoming data stream and normalized reference pulse, this algorithm retains the pulses whose loss is below the threshold. We select the threshold based on statistical methods. The algorithm is implemented on field programmable gate array (FPGA) successfully, and this process is able to work in real-time. The result shows that the effective counting rate can improve about 19.8% and more than 99% pile-up and error pulses will be suppressed. By analyzing reserved pulses, the energy spectrum and flood histogram could be well rectified. The energy resolution increases 11% compared with traditional algorithm. Furthermore, due to this new algorithm, the low-energy threshold can be lower.