Shared Pixel Weighting for In-Pixel Convolution Sensors

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing camera sensors face energy inefficiency and throughput bottlenecks due to high data transmission volumes and require excessive circuit area, unusual circuit elements, and lack configurability in in-pixel processing.

Innovation Solution

An in-pixel processing system with a switch network that shares weighting values among pixels, allowing weight circuitry to be positioned outside light-sensitive regions, reducing circuit area and enabling flexible convolution stride without hardware changes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If in-pixel processing is implemented with local weighting circuitry in each pixel, then computation capability is improved, but circuit area and wiring complexity increase excessively

Engineering Contradiction:

Improvein-pixel processing capabilityVSAvoidcircuit area

Core Design Contradiction:

Extent of automationVSArea of stationary object

Solution Approach 1:

The weighting circuitry is extracted from individual pixels and placed in a separate weight generator located outside the pixel array. This allows each pixel to perform computation without containing full weighting circuitry, reducing pixel area while maintaining in-pixel processing capability through shared resources.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

A single weight generator serves multiple pixels simultaneously through a switching network, making the weighting circuitry universal rather than dedicated to each pixel. This multi-functional approach reduces total circuit area while preserving computation capability across all pixels.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Adaptability or versatility

If weighting circuitry is integrated within each pixel, then configurability is improved, but device complexity and wiring increase

Engineering Contradiction:

ImproveconfigurabilityVSAvoidcircuit complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system transitions from a two-dimensional pixel array to a three-dimensional architecture by placing the weight generator outside the pixel plane and connecting it through vertical interconnects. This spatial reorganization reduces in-plane wiring complexity while maintaining configurability through the switching network.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

A switching network acts as an intermediary between the weight generator and pixel array, managing the complex connections without requiring each pixel to contain full weighting circuitry. This mediator reduces device complexity by centralizing control while preserving configurability.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Productivity

If weight values are stored locally in each pixel, then processing speed is improved, but circuit area increases

Engineering Contradiction:

Improveprocessing speedVSAvoidpixel area

Core Design Contradiction:

ProductivityVSArea of stationary object

Solution Approach 1:

Multiple pixels share common weight values through the switching network, merging the storage function across the array rather than duplicating weights in each pixel. This reduces total circuit area while maintaining fast access to weight values for processing.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS20250350821A1Camera Sensor with Shared Pixel Weighting

Publication Date: 2025.11.13 WISCONSIN ALUMNI RES FOUND

AI summary

A camera sensor architecture provides in-pixel processing using weights that may be generated remotely from the pixels and applied to the pixels using a network switch for rapid in-pixel convolution useful as a first stage in a neural network.