The FTK First-Stage Tracking Board: A High Bandwidth Track Fitter for the ATLAS Trigger*

The FastTracKer (FTK) is a real-time pattern-recognition hardware tracker for the ATLAS experiment. FTK receives particle hit coordinates on 12 detector layers and provides best-fit helical tracks with a design latency of 100 microseconds. Fitting all possible combinations of hits becomes exponentially more complex with increasing accelerator collision rates and is not possible in real-time on CPUs given the dense environment at the LHC. To reduce the required number of fits, FTK first uses custom ASICs to perform coarse pattern recognition, then uses FPGAs to fit all combinations of hits within each pattern. This talk will cover the design and performance of the First-Stage Tracking Board (AUX), which uses 8 out of 12 layers to compute track seeds at a peak rate of 4 billion fits per second per board. The AUX computes the coarse resolution hits used for pattern matching, arranges them to allow quick lookup of full-resolution hits given a pattern, and performs linearized chi-squared fits for all combinations of hits in each pattern. The full functionality of the AUX was demonstrated with maximum-rate ATLAS data at the end of 2018, and FTK is currently under commissioning to scale to the full 128-AUX system for use in 2021.