In the WP4 context in which INFN-Milano is active, a real-time tracking device is being implemented using commercial FPGA technology (Field Programmable Gate Array). An highly parallel algorithm that makes use of both precise timing ans space information of the particle hits in the detector has been studied and developed. In fact, on one hand the TIMESPOT collaboration has already demonstrated the timing capabilities of socalled 3D sensors proving a time resolution down to 30 ps, on the other hand the inclusion of timing information is identified as crucial for improving the capabilities of future high luminosity experiments.

The algorithm is based on early identification of doublets of particle hits in adjacent detector planes and the precise timing information is used to suppress fake hit combinations. Candidate tracks are identified in parallel from groups of stubs with similar parameters.

The algorithm has been for a possible application to the VELO Upgrade II detector and implemented on Xilinx Virtex UltraScale devices. The parallel nature of the algorithm has mostly driven the choice of the FPGA technology that allows for high performances in specific tasks, low latency and highly pipelined implementations, typical of hardware technologies, while keeping the possibility of the logic to be reprogrammable, typical of software approaches. Moreover FPGA technology allow for very high input/output data rates (hundreds of Gbps).

The highly parallized and pipelined architecture makes the system modular, suitable for implementation on multiple FPGA boards and scalable to large detectors.
A prototype demonstrator has been tested with simulated data from a sector (∼ 1/64) of an LHCb VELO-like detector and proved to be capable to process events at a rate of 40 MHz.