In-Pixel Computing Weights Using Embedded ROM for Sensor Arrays
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional pixel sensors are separate from memory and processing elements, leading to size, weight, and power bottlenecks, as well as data transfer inefficiencies and security concerns due to the segregation of sensing and processing functions.
Innovation Solution
Integration of sensors with amplifiers, weights, and memory elements within pixels, enabling in-pixel computing through multi-bit, multi-kernel, and multi-channel ROM embedded architectures, allowing for local data processing and reduced data transfer needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If sensors are physically separated from processing elements, then device complexity is reduced and manufacturing is easier, but data transfer bandwidth requirements increase and power consumption rises
Solution Approach 1:
The patent merges sensors, memory elements, and processing elements into integrated pixel units. Each pixel contains photodetectors, ROM-based weight storage, and amplifiers that perform local signal processing. This integration eliminates the need for extensive data transfer between separate components, reducing power consumption while maintaining manufacturing feasibility through standardized integrated circuit fabrication processes.
Solution Approach 2:
The patent segments the processing function by distributing computational capabilities directly to individual pixels rather than concentrating all processing in a central unit. Each pixel independently performs signal weighting and preliminary processing using locally stored weights in ROM, dividing the overall processing task across numerous independent units that operate in parallel.
2Ease of manufacture
If sensors are physically separated from processing elements, then device layout is simpler, but data transfer speed and processing latency increase
Solution Approach 1:
The patent combines sensors, weight storage (ROM), and processing amplifiers into single integrated pixel units. This merging eliminates data transfer delays between separate components by enabling immediate local processing of sensor signals. The amplifier within each pixel directly accesses weights stored in the pixel's ROM, performing computations without external data transfer.
3Productivity
If multiple data transfer channels are used for sensor data, then processing capability increases, but device complexity and connectivity requirements increase
Solution Approach 1:
The patent segments processing capability by distributing computational functions to individual pixels rather than using a centralized processing architecture. Each pixel independently performs signal weighting and processing operations, enabling massive parallel processing without requiring complex inter-connectivity between dedicated processing units. This segmentation achieves high productivity through parallelism while maintaining relatively simple device structure.
4Device complexity
If sensor data is transmitted to remote computing entities, then centralized processing is simplified, but security vulnerabilities and bandwidth requirements increase
Solution Approach 1:
The patent implements self-service processing where each pixel independently performs signal weighting and preliminary computation using locally stored weights in ROM. This distributed self-processing reduces the volume of data that must be transmitted to external computing entities, minimizing security exposure during data transfer. Critical processing functions are performed autonomously at the sensor level, reducing reliance on external systems.
Data Source
AI summary
Provided is an integrated circuit comprising: an amplifier; a sensor electrically connected to a first input of the amplifier; and a set of weights electrically connected to a second input of the amplifier, wherein the amplifier is configured weight an output of the sensor according to an output of the set of weights. Provided is an array of integrated circuits comprising an amplifier; a sensor electrically connected to a first input of the amplifier; and a set of weights electrically connected to a second input of the amplifier, wherein the amplifier is configured weight an output of the sensor according to an output of the set of weights.