Real-time pattern recognition with FPGA at LHCb, an O(n) complexity architecture

¹ Università di Pisa, Pisa, Italy
² INFN sezione di Pisa, Pisa, Italy
³ INFN sezione di Cagliari, Cagliari, Italy
⁴ University of Chinese Academy of Sciences, Beijing, China
⁵ Rutherford Appleton Laboratory, Chilton, United Kingdom
⁶ Università degli Studi di Milano-Bicocca, Milano, Italy
⁷ INFN sezione di Milano, Milano, Italy
⁸ Scuola Normale Superiore, Pisa, Italy
⁹ Universitat de València, València, Spain
¹⁰ Consejo Superior de Investigaciones Científicas, Madrid, Spain
¹¹ Università degli Studi di Siena, Siena, Italy
¹² Ruprecht-Karls-Universität Heidelberg, Heidelberg, Germany

^* e-mail: federico.lazzari@cern.ch

Published online: 7 October 2025

Abstract

The LHCb collaboration is planning an upgrade (LHCb “Upgrade- II”) to collect data during Run 5 at an instantaneous luminosity an order of magnitude larger than the current one (Run 3). LHCb relies on a complete realtime reconstruction of all collision events at LHC-Point 8, which will have to cope with both the luminosity increase and the introduction of correspondingly more granular and complex detectors. After an intensive R&D programme, LHCb approved a FPGA-based system to pre-reconstruct tracks in the SciFi detector at readout level during Run 4, as an intermediate step towards a system that could be extended to other tracking detectors in the future. It is based on the “artificial retina”, an extremely parallel architecture. Using simulated data, the performance of a hardware demonstrator of this architecture has been tested as a function of instantaneous luminosity and system size, and was found to have O(n) complexity, which is a crucial feature for high luminosity applications.