An Asynchronous Flash LiDAR Sensor Based on Pixel-wise Time-of-Flight Validation via Background Light Adaptive Threshold

다국어 초록 (Multilingual Abstract)

The demand for 3D imaging technologies is rapidly increasing across various fields, including autonomous vehicles, robotics, augmented reality, and industrial automation. These applications rely on accurate and efficient depth sensing systems capable of capturing high-speed and high-resolution data in real time. Such systems must handle a wide range of environmental conditions, including variable lighting, diverse target reflectivity, and dynamic scenes, while maintaining robustness and reliability. The key challenges include managing wide input dynamic ranges, mitigating background light interference, and optimizing resource utilization such as memory and power. These constraints drive the need for innovative solutions to advance the performance of depth sensing technologies. Flash LiDAR sensors have emerged as essential tools to meet these demands, offering unparalleled speed and resolution. However, existing systems face significant limitations in their ability to efficiently process wide dynamic ranges and cope with environmental noise. This research addresses these challenges through two key contributions: the development of a novel Delta Intensity Quaternary Search (DIQS) histogramming Time-to-Digital Converter (hTDC) and its integration into an asynchronous flash LiDAR sensor. Together, these innovations enable precise and robust depth measurements while minimizing resource consumption. The DIQS hTDC represents a substantial advancement over traditional zooming histogramming TDCs by introducing a coarse-to-fine histogramming mechanism combined with adaptive bin allocation. This approach enhances memory efficiency and extends the dynamic range, ensuring accurate time-of- flight (ToF) measurements even under challenging conditions. The DIQS hTDC iteratively narrows the region of interest using a quaternary search method, concentrating computational resources on the most relevant photon arrival times. This technique inherently suppresses background noise, reduces memory usage, and enables scalability for resource-constrained ToF systems, addressing critical bottlenecks in previous designs. Building on the DIQS hTDC, an asynchronous flash LiDAR sensor is proposed to autonomously validate pixel-level ToF measurements. Each pixel dynamically adjusts its integration time based on the intensity of background light, leveraging a compact bitwise arithmetic unit to compute thresholds in real time. This pixel-level autonomy eliminates the need for complex global synchronization while selectively reading out validated pixels. Non-validated pixels continue collecting photons, improving data efficiency and robustness. The asynchronous architecture simplifies system complexity and enhances adaptability across diverse environmental and operational conditions. A prototype asynchronous flash LiDAR sensor incorporating the DIQS hTDC was fabricated in a 90- nm CMOS process and extensively tested. The sensor achieved a precision of 6 cm and an accuracy of 8 cm over ranges from 1.5 to 22.5 meters indoors. Under outdoor conditions with ambient illumination up to 30 klux, the sensor maintained a true detection rate exceeding 95% at a distance of 21 meters. Its ability to dynamically adjust frame rates from 5 to 250 fps demonstrates versatility and adaptability across varying signal-to-noise ratio conditions, catering to applications ranging from high-speed operation to high-precision imaging. This research makes two significant contributions to the advancement of SPAD-based flash LiDAR systems. First, the DIQS hTDC introduces a scalable and efficient histogramming solution that overcomes memory and noise challenges. Second, the asynchronous flash LiDAR sensor demonstrates the practical integration of this technology into a robust and compact system capable of addressing the complex demands of real-world environments. Together, these innovations pave the way for the next generation of 3D imaging systems, offering a practical, high-performance, and scalable solution for diverse applications requiring precise and efficient depth sensing. Keywords: 3D imager, CMOS image sensor, depth sensor, LiDAR, time-of-flight (ToF), direct ToF, histogramming, in-pixel histogramming, single-photon avalanche diode (SPAD), time-to-digital converter (TDC), delta intensity quaternary search (DIQS), asynchronous.

목차 (Table of Contents)

• Chapter 1: Introduction1
• 1.1. Light detection and ranging applications 1
• 1.2. Time-of-flight technique 2
• 1.3. Thesis organization 4
• Chapter 2: SPAD-based ToF sensor 5
• 2.1. Single Photon Avalanche Diode (SPAD) 5
• 2.1.1. Principle of SPAD operation 5
• 2.1.2. SPAD performance parameters 6
• 2.1.2.1 Dark count rate (DCR) 7
• 2.1.2.2 Photon detection probability (PDP) 7
• 2.1.2.3 Afterpulse 8
• 2.1.2.4 Crosstalk 8
• 2.1.2.5 SPAD jitter 8
• 2.1.3. SPAD architecture 9
• 2.1.3.1 Electrical structure 9
• 2.1.3.2 Optical structure 10
• 2.2. SPAD front-end 11
• 2.2.1. Passive quenching circuit. 11
• 2.2.2. Active quenching circuit 13
• 2.3. SPAD-based ToF sensors 13
• 2.3.1. SPAD-based dToF sensors 13
• 2.3.2. SPAD-based iToF sensors 18
• 2.4. Chapter Summary 23
• Chapter 3: Delta intensity quaternary search histogramming TDC24
• 3.1. Histogramming techniques. 24
• 3.1.1. Full-histogramming. 24
• 3.1.2. Zooming-histogramming 26
• 3.1.3. Sliding-histogramming 30
• 3.1.4. Tracking-histogramming 31
• 3.2. Delta intensity quaternary search (DIQS) hTDC 31
• 3.2.1. Operation principles of DIQS hTDC 32
• 3.2.2. Dynamic range extension from delta-intensity histogramming 34
• 3.2.3. Overall architecture. 36
• 3.3. Circuit implementation. 36
• 3.3.1. SPAD with front-end 36
• 3.3.2. Hybrid asynchronous-synchronous up-down counter 38
• 3.3.3. Global clock generator 40
• 3.3.4. In-pixel timing generator 41
• 3.4. Measurement results 44
• 3.4.1. Measurement setup 44
• 3.4.2. One-point measurement 45
• 3.4.3. Precision analysis. 46
• 3.4.4. Captured 3D image 50
• 3.4.5. Prototype LiDAR sensor performance 50
• 3.5. Chapter Summary 54
• Chapter 4: Asynchronous flash LiDAR sensor.55
• 4.1. Input dynamic range of LiDAR sensor 55
• 4.2. Proposed Asynchronous LiDAR with DIQS hTDC 56
• 4.2.1. ToF detection probability. 56
• 4.2.2. ToF validation 64
• 4.2.3. Approximation function 64
• 4.2.4. Operating principle of asynchronous flash LiDAR sensor 66
• 4.2.5. Overall architecture. 71
• 4.3. Circuit implementation. 73
• 4.3.1. SPAD 73
• 4.3.2. Timing generator 73
• 4.3.3. ToF validation circuit. 75
• 4.4. Measurement results 78
• 4.4.1. Measurement setup 79
• 4.4.2. One-point measurements 81
• 4.4.3. Captured 3D image 85
• 4.4.3. Prototype asynchronous flash LiDAR sensor performance 88
• 4.5. Chapter summary. 90
• Chapter 5: Conclusion & Discussion91
• REFERENCES.93
• CURRICULUM VITAE.99
• ACKNOWLEDGEDMENT102