Ultra-fast curve fitting for pulses on FPGA

The hardware described in this work can process a pulse train out of a 1.5 GHz ADC and can summarize pulses with parameters such as amplitude, rise/fall times, and arrival time. It can handle back-to-back pulses with zero dead-time. The pulses can be as short as 9 samples. Such signals (and even many channels of it) can be found in high energy physics experiments, where particles are accelerated, collided, and detected. Similar physical setups are present in nuclear medical imaging, especially in positron emission tomography. We did the hardware implementation on FPGA. FPGAs offer tremendous parallelism and hence total compute power. In this article, we present our top-level architecture, submodule details, design flow, and implementation details. Our FPGA design offers space, power, deployment time, and operational cost reduction in the respective applications.