MODELING AND DESIGN OF SINGLE PHOTON AVALANCHE DIODES FOR LIDAR APPLICATIONS

Abstract

Single-photon avalanche diodes (SPADs) hold great potential as optical detectors in single-photon-counting (SPC) applications. With a primary focus on the fast-developing field of light detection and ranging (LiDAR) systems in automotive market, this thesis explores the important role of SPADs in achieving extended detection ranges, higher resolutions, and fast response times. As advanced complementary-metal-oxide-semiconductor (CMOS) technologies becomes more available, the cost-effective solutions of SPADs also become more achievable. First, a detailed review of recent SPAD applications within LiDAR systems is presented. This discussion encompasses both commercial products and research works based on various CMOS technologies. SPADs fabricated using different technologies exhibit significant variations in the performances. Therefore, SPAD models that are used to simulate critical performances prior to fabrication are very important. Subsequently, a comprehensive review introduces the evolution of SPAD models, from key fundamentals to the modeling process. Based on the 65 nm standard CMOS technology and information from literature review, an enhanced SPAD modeling process is introduced. This model, considering a 2-D distribution of electric field, improves the accuracy of dark count rate (DCR) predictions. To validate its effectiveness, SPADs are meticulously designed using the TSMC 65 nm standard CMOS technology for the calibration and comparison purposes. Measured results indicate negligible afterpulsing probability (~ 0%) and decent DCR level (~ 14 kHz at a 0.7 V excess voltage), thanks to the implementation of output buffers. The measured wavelength dependence of photon detection probability (PDP) also agrees with the simulations. Additional discussions are conducted to figure out some deviations between simulations and measurements. Finally, important research challenges are proposed based on the simulation and measurement results. Aiming to address these challenges, potential directions for optimizing SPAD models and design are presented, followed by conclusions.