A design methodology for SPAD sensors in long-range LiDAR

Light Detection And Ranging (LiDAR) systems have the ability to generate high resolution 3D images for environmental mapping and situational awareness, which is a key requirement for self-driving cars in the near future. Direct Time-of-Flight (dToF) sensors using Single Photon Avalanche Diodes (SPAD) are promising for this application as they can achieve high sensitivity with remarkable timing precision. Additionally, these SPAD sensors can be manufactured efficiently in CMOS processes to provide compact and cost-effective solutions. The application of these sensors in long-range automotive LiDAR however poses many challenges, such as imaging a variety of objects including vehicles, pedestrians, vegetation, etc. having low surface reflective properties while on the other hand retroreflective objects such as number plates, lane markers, road signs, etc. have high reflectivity. These objects are to be detected at long distances (approximately 200 m), in a wide field of view (typically 120°x30°), at high frame rates (about 10 to 30 fps), and in unpredictable weather conditions (such as intense sunlight, fog, rain and snow). The aim of this research is to deliver a methodology for the design of LiDAR receivers suitable to perform in the highly dynamic automotive environment. An analysis model is presented to optimise the sensor parameters based on the system specifications and the optical power governed by laser safety requirements. The model provides analytical equations and simulates the statistical behaviour of SPADs to evaluate the impact on their performance parameters. The analysis further investigates the design implementations of combining circuits and Time to Digital Converters (TDC) on the imaging range of the system. A reconfigurable SPAD array, designed in STMicroelectronics 40nm CMOS technology, interfaced to a FPGA serves as a platform to validate the simulated sensor performance. The SPAD test chip consists of variations in pixel sizes with capability of 128 channels to be readout simultaneously at 100MHz. The FPGA provides the flexibility of implementing and comparing the different combining and TDC techniques, with maximum sampling rate of 4GS/s. A new approach, the Synchronous Summation Technique (SST), is proposed to optimise the photon processing throughput in order to achieve an extended range in the LiDAR system along with improved resistance to background conditions. The evaluation and characterisation of the sensor is demonstrated using two experimental setups. First, an emulated LiDAR system is built to evaluate the sensor response for the extent of expected signal and solar irradiance in the given application. Second, a scanning LiDAR system is developed to understand the practical implications and to further authenticate the simulation model. The optical response from these systems is integrated into the simulation framework leveraging synthetic data sets. This aids the assessment of the sensor performance at long-range in realistic scenarios of the automotive LiDAR application and can overcome the various challenges of prototyping in such an environment. The proposed design methodology comprises of the simulation framework with verified sensor response and utilisation of synthetic data sets. It offers a comprehensive guide to customise future sensor design solutions as per desired system requirements