SPAD

# SPAD Single-Photon Avalanche Diode LIDAR Timing

A LIDAR system measures distance by timing exactly how long light takes to travel out to an object and reflect back, and a single-photon avalanche diode makes that timing possible at the extreme precision LIDAR needs by amplifying even a single detected photon into a large, sharply-timed electrical pulse through avalanche multiplication. An ordinary photodiode generates a signal proportional to how much light hits it, but a weak, distant reflection returning to a LIDAR sensor might deliver only a handful of photons — far too few to produce a clean, precisely-timed signal through ordinary photodetection. A SPAD instead operates biased above its breakdown voltage, so that even a single absorbed photon triggers a self-sustaining avalanche of charge carriers, producing a large, fast, sharply-timed electrical pulse from that one photon's arrival — exactly the kind of crisp timing signal needed to measure a light pulse's round-trip travel time to nanosecond or better precision.

Because the avalanche is self-sustaining once triggered, the SPAD has to be actively reset after every single detection event before it can detect the next photon, and that reset — called quenching — directly determines how quickly the detector can recover and how many photon arrivals per second it can actually register. An avalanche, once started, keeps drawing current until something actively intervenes; quenching circuitry detects the avalanche and rapidly drops the bias below breakdown to stop it, then restores the bias to ready the SPAD for the next photon. The speed of that quench-and-reset cycle sets a hard ceiling on detection rate — a SPAD that takes longer to reset simply can't register as many photon events per second, which matters directly for LIDAR systems scanning quickly across a scene and needing many timing measurements completed in a short window.

Because a single SPAD can only register one event per quench cycle, high-resolution LIDAR systems use large arrays of SPADs operating in parallel, each independently timing its own photon arrivals, letting the overall sensor build a complete timing map across many points simultaneously rather than waiting for one detector to cycle through every measurement sequentially. This is directly analogous to how a stacked CMOS image sensor's logic die handles readout for every pixel in parallel rather than one at a time — SPAD-array LIDAR sensors similarly depend on massive parallelism at the detector level to achieve the overall scan rate a working LIDAR system needs, with each individual SPAD's quench-reset speed setting the per-channel limit that the whole array's parallelism works around.

ReRAM forming voltage described a one-time bring-up challenge a memory cell has to pass through before normal operation; SPAD quenching is instead a continuous, every-single-detection-event constraint that directly caps a working sensor's real-time performance — two very different kinds of "the device has to recover before it can do its job again" problems, one happening once per cell's lifetime, the other happening potentially millions of times per second during active operation.