SPAD-Based Time-of-Flight (ToF) Depth Sensor for LiDAR Applications

This thesis investigates advanced techniques for noise and interference suppression in single-photon avalanche diode (SPAD)-based direct time-of-flight (d-ToF) light detection and ranging (LiDAR) sensors, targeting automotive urban-driving applications. In complex urban scenarios, LiDAR performance is critically degraded by strong ambient illumination, dominated by sunlight, and optical interference from other nearby LiDAR systems. To overcome these challenges, this work proposes three complementary techniques: time-coded time-correlated single-photon counting (TCSPC), adaptive coincidence detection, and pixel-wise range gating. Time-coded TCSPC introduces pseudo-random temporal modulation to the laser repetition rate, encoding each pulse to coherently reinforce genuine signal returns while dispersing interference across the histogram. Adaptive coincidence detection dynamically adjusts spatial coincidence thresholds among neighboring SPADs based on the supply voltage, compensating for voltage-dependent dark count variations and ensuring reliable photon detection and stable signal-to-noise ratio (SNR) under varying illumination. Pixel-wise range gating implements locally adaptive detection windows for each pixel, determined from coarse ToF estimates in a fast, pre-imaging phase. By excluding out-of-range and pile-up-induced events, it enhances SNR and supports high-frame-rate, high-resolution depth imaging with minimal memory overhead. The proposed techniques are evaluated using a SPAD-based sensor chip fabricated in 180-nm CMOS technology. The fabricated sensor features a 68×40 pixel array and eight time-to-digital converter (TDC) blocks capable of simultaneous coarse (10 ns) and fine (156.25 ps) ToF measurements at 200-MS/s conversion rate, enabling high-throughput, high-resolution photon timing. Indoor and outdoor experimental results of the LiDAR system in varying conditions show 20-fps depth and intensity imaging up to 34 m under 79-klux background illumination, as well as accurate single-point target detection up to 80 m under 103-klux illumination, matching the requirements of automotive urban-driving applications. These contributions establish a comprehensive framework for noise-tolerant, interference-resilient, and high-precision SPAD-based LiDAR, supporting scalable and reliable depth sensing for autonomous and assisted driving in dynamic urban environments.