SPAD Image Sensor Readout Architecture for Low-Power LiDAR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
LiDAR systems face challenges in providing high-quality depth information for mid-ranges due to ambient light interference and high power consumption, particularly in mobile applications, and existing on-chip histogram and peak detection methods require large memory and circuit area, which is contrary to low-power operation.
Innovation Solution
A LiDAR system with a reduced number of TDCs and histogram bins, using a readout architecture with RG channels for timestamp generation and hardwired addressing, along with a dynamic correlation filter to minimize circuitry area and power consumption, and a reduced bin size for efficient detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a relatively greater optical power is used to compensate for strong ambient light, then the depth information quality is improved, but the power consumption increases which is contrary to low-power operation
Solution Approach 1:
The pixel array is divided into multiple read groups, each with its own TDC and histogram circuit. This segmentation allows parallel processing of depth information from different regions, improving measurement precision without requiring increased optical power, thereby maintaining low-power operation.
Solution Approach 2:
The system performs preliminary ambient light measurement and compensation before final depth calculation. By pre-processing the ambient light effects and applying compensation algorithms, the system maintains high depth information quality without needing to increase optical power, thus avoiding increased power consumption.
2Measurement precision
If on-chip histogram and peak detection circuitry is implemented, then detection quality is improved, but the memory and circuit area increase which is contrary to low-power operation
Solution Approach 1:
The histogram circuit is divided into multiple bins, with each bin handling a specific time range. This segmentation allows the circuit to process depth information efficiently without requiring a large monolithic memory structure, reducing overall circuit area while maintaining detection quality.
Solution Approach 2:
The system implements peak detection only for the most relevant histogram bins rather than processing all bins. By focusing computational resources on the most significant peaks that contain actual depth information, the system maintains high detection quality while minimizing the required circuit area and power consumption.
3Speed
If an entire pixel array is read to reduce latency, then the speed is improved, but the number of TDCs and memory required increase which uses large chip area and power
Solution Approach 1:
The pixel array is divided into multiple read groups that can be read out in parallel. Each read group has its own TDC and histogram circuit, allowing simultaneous processing of multiple regions. This segmentation enables high-speed reading without requiring a single large TDC array, thus reducing device complexity and chip area.
Solution Approach 2:
Multiple read groups are merged into a unified processing architecture where timestamp information from different groups is combined in the histogram circuits. This merging approach allows the system to achieve high reading speed through parallel processing while using a manageable number of TDCs per group, reducing overall device complexity.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system provides low-power, high-speed depth information with reduced ambient light effects, increasing detection range and minimizing delay, while maintaining detection quality for both short and long distances.
Implementation Method 1
The SPAD array may include at least one pixel read group that outputs a detection event signal in response to a light pulse that is incident on the at least one pixel read group
Implementation Method 2
The RG channel may include a TDC configured to generate timestamp information corresponding to the detection event signal
Data Source
AI summary
A LiDAR system is disclosed that includes a SPAD unit array and J read group (RG) channels. The SPAD unit array is arranged in M rows and N columns of pixel read groups. Each row includes K pixel read groups. Each pixel read group outputs a detection signal in response to a light pulse that is incident on the pixel read group. Each RG channel corresponds to at least one row of pixel read groups and includes L time-to-digital converters that respectively generate timestamp information corresponding to detection event signals of each of L pixel read groups in the at least one row of pixel read groups in which J<M and L≤K. Each RG channel stores the timestamp information in an accumulator bin of a histogram circuit corresponding to a value of the timestamp information using hardwired addressing based on the value of the timestamp information.