3D-Stacked Image Sensor Architecture for Low-Latency Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image sensors require extensive external processing, leading to high bandwidth demands and latency in object detection tasks.
Innovation Solution
An image sensor architecture with multiple stacked integrated circuit layers, including pixel sensor groups, image processing circuitry groups, and neural network circuitry groups, allowing for distributed parallel processing and reduced data transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If image sensors use traditional external processing architecture, then device complexity is reduced, but bandwidth requirements increase and processing latency increases
Solution Approach 1:
The patent transitions from a planar 2D sensor architecture to a 3D stacked architecture with multiple layers (sensor layer, processing layer, neural network layer). This vertical dimensionality change enables simultaneous processing operations within the sensor module, reducing data transmission distance and processing latency while maintaining compact form factor.
Solution Approach 2:
The patent combines sensing elements, image processing circuitry, and neural network processing units into a single integrated stacked module. This merging of previously separate components (sensor, processor, memory) into one unified system eliminates external data transmission interfaces, reducing both bandwidth requirements and processing latency.
2Productivity
If image sensors perform extensive external processing, then processing capability is reduced, but bandwidth requirements decrease
Solution Approach 1:
The patent implements preliminary processing operations (pixel-level processing, feature extraction, neural network inference) directly within the sensor module before data leaves the sensor. This preliminary action reduces the volume of data that needs to be transmitted externally, as only processed results or selectively filtered data are output rather than complete raw image streams.
Solution Approach 2:
The patent segments the processing pipeline into multiple stages distributed across different layers: pixel-level processing in the sensor layer, intermediate processing in the processing layer, and neural network processing in the top layer. This segmentation enables parallel processing of different image regions and features, increasing overall processing speed while reducing the amount of data requiring external transmission.
3Productivity
If image sensors use stacked layered architecture, then processing efficiency is improved, but manufacturing complexity increases
Solution Approach 1:
The patent addresses manufacturing challenges by moving complex processing functions to the vertical dimension through stacked layers, rather than increasing horizontal integration density. This approach allows each layer to be manufactured using standard planar processes, then bonded together using established 3D integration techniques, reducing overall manufacturing complexity compared to monolithic 3D integration.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A image sensor includes a first integrated circuit layer including pixel sensors that are grouped based on position into pixel sensor groups, a second integrated circuit layer in electrical communication with the first integrated circuit layer, the second integrated circuit layer including image processing circuitry groups that are configured to each receive pixel information from a corresponding pixel sensor group, the image processing circuitry groups further configured to perform image processing operations on the pixel information to provide processed pixel information during operation of the image sensor, a third integrated circuit layer in electrical communication with the second integrated circuit layer, and the third integrated circuit layer including neural network circuitry groups that are configured to each receive the processed pixel information from a corresponding image processing circuitry group and perform analysis for object detection on the processed pixel information during operation of the image sensor.