Speaker
Description
High-rate position-sensitive silicon trackers increasingly require real-time reconstruction close to the detector readout in order to reduce data volumes before storage. The MUonE experiment is a proposed fixed-target experiment at the CERN M2 beamline designed to independently measure the hadronic leading-order corrections to the muon anomalous magnetic moment ($g-2$). It consists of a high-intensity 160 GeV muon beam impinging on a sequence of thin targets instrumented with a high-granularity silicon tracking system and accompanied by a calorimeter. We present a real-time FPGA reconstruction chain for high-rate silicon tracking, demonstrated in the context of elastic muon-electron scattering events in MUonE at a 40 MHz input rate. Low-latency vertex reconstruction is required for the MUonE experiment to reduce the final design data rate from $\mathcal{O}(1 \ \mathrm{TB/s})$ to a manageable level for storage and offline analysis. The vertex reconstruction algorithm employs a hardware-oriented minimization method. The full reconstruction chain is implemented in C++ and synthesized to RTL using High-Level Synthesis (HLS). We demonstrate a live test on a physical FPGA using real detector data, and show agreement with the software vertex reconstruction within the detector resolution. We present latency, FPGA resource-utilization, and reconstruction-performance studies showing that the track-fitting block meets the MUonE real-time processing requirements, while the vertex reconstruction block is progressing toward the same target.