Ponente
Descripción
Pulse pile-up in detection systems can degrade energy and time resolution, especially in high-rate environments. While traditional rejection methods discard overlapping events, some of these signals may still contain valuable physical information. We present a novel reconstruction method based on a one-dimensional convolutional autoencoder (1D-CAE), trained on data acquired for our first experience, from the Neutron Detector Array (NEDA). This method reconstructs pile-up events, allowing further analysis such as Pulse Shape Analysis (PSA) neutron-gamma discrimination using the Charge Comparison method (CC). As part of ongoing work, we have executed the model on a 32-bit microprocessor and are currently implementing the solution on FPGA hardware. Using the HLS4ML library, we have generated a hardware IP core, although successful simulation of the IP remains an open task. These developments aim to enable real-time, on-detector pulse reconstruction in future high-throughput experiments.
