FPGA performance for signal reconstruction of the ATLAS Tile Calorimeter in the HL - LHC environment

The ATLAS Tile Calorimeter (TileCal) will face considerable challenges from increasing radiation and high data throughput in the High Luminosity Large Hadron Collider (HL - LHC) era. This study aims to find a balance between FPGA resource usage and latency with efficient signal reconstruction algorithms of the calorimeter pulses. The precise reconstruction of signal pulse from calorimeter pulses is crucial as it corresponds to the energy of the particle deposited in the calorimeter. Optimal Filtering (OF) is currently used for reconstructing the pulse amplitude from digitised samples. While efficient, the performance degrades considerably with increasing pile up. A solution to combat this challenge is to work with Neural Networks. This study shows the performance of OF contrasted with a Single Layer Perceptron (SLP) with data trained on simulated ATLAS TileCal pulses. The results show that the SLP outperforms the OF algorithm in the pile-up case, which is expected in the HL - LHC era.