A 32×32 SPAD LiDAR Sensor with In-pixel Spatial-temporal Coincidence and Per-SPAD Pulse Delay for Resolution-lossless Imaging

Single-photon Avalanche Diode (SPAD)-based LiDAR technology finds significant applications in advanced driver assistance systems, etc. Extracting distance information necessitates creating a histogram from enormous frames to filter out the overwhelming background light (BGL) data. As the formats of SPAD pixel arrays scaling up [1], the processing, storage, and transmission of extensive data result in ultra-high power consumption and increased layout area. In order to reduce BGL, previous studies [2][3] [4] have leveraged spatial coincidence and histograms in pixels, and [5] utilized external double pulses to enhance the signal-to-noise ratio. However, a significant amount of data still needs to be processed within the sensor. To address these challenges, this paper presents a $32 \times 32$ SPAD LiDAR sensor that incorporates in-pixel spatial-temporal coincidence to effectively compress random background light and a per-SPAD pulse delay method to realize resolution-lossless imaging. This approach achieves an impressive 98.3% compression rate of background light under 100 klux illumination while maintaining a low power consumption of $14.6 \mu \mathrm{~W}$ per pixel.