Técnicas de Deconvolução Aplicadas à Estimação de Energia Online em Calorimetria de Altas Energias em Condições de Alta Taxa de Eventos

Abstract

This work presents a study of deconvolution techniques that can be applied to online energy reconstruction in the ATLAS hadronic calorimeter (TileCal) first-level trigger system. The high-rate event environment expected in future LHC operations increases the likelihood of adjacent collisions, leading to signal superposition. The current energy reconstruction algorithm is not robust against this pile-up effect. To address this issue, several deconvolution methods are investigated, including two direct approaches that result in FIR filters and three iterative techniques based on Gradient Descent, which allow the incorporation of prior knowledge. As expected, the iterative methods achieve lower estimation errors compared to the direct FIR approaches. However, in terms of implementation, the iterative techniques are more complex and require greater hardware resources.