Event-Driven Frequency Decomposition with Spiking Phase-Locked Loops

Event-driven sensors are highly effective at capturing stimuli in real-world scenarios where signals are sparse in both time and space. Unlike clock-based approaches, these sensors capture data from the environment only upon perceiving a variation in the signal, translating temporal sequences into trains of unary events, often referred to as spikes. However, while the sensor functions in an event-driven manner, subsequent data processing often still relies on clock-based methods. This lack of event-driven processing units undermines the latency, power efficiency and responsiveness benefits of the event-driven approach. In this work, we present a circuit implementation of a spiking phase-locked loop (sPLL), which decomposes event data into frequency components using fully event-driven CMOS circuits. Through experimental measurements on a CMOS chip, we demonstrate that the fabricated sPLL circuit can accurately synchronize with input frequencies and perform frequency decomposition. This approach demonstrates the feasibility of event-driven CMOS circuits for processing and decoding complex event data, offering a promising solution for low-power and low-latency applications that require spectral analysis.