Image Sensor Substrate Heat Management via Segmented Signal Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing photoelectric conversion apparatuses face issues with heat generation from the digital signal processor on the second substrate, leading to increased noise and decreased image quality due to heat propagation to the pixel array on the first substrate, and hindered processing speed from neural network computation models.
Innovation Solution
The apparatus includes a first substrate with a pixel array and a second substrate with A/D converters and signal processing sections, where the A/D conversion circuits and AI processing sections are arranged to dissipate heat effectively and perform parallel processing, reducing the impact of heat on the pixel array and enhancing processing speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a digital signal processor based on neural network computation model is disposed on the second substrate, then signal processing capability is improved, but heat generation increases and propagates to the pixel array causing noise increase
Solution Approach 1:
The second substrate is divided into multiple regions with different functions: a first region for A/D conversion circuits and a second region for signal processing sections. This spatial segmentation separates heat-generating components from the pixel array, reducing heat propagation while maintaining processing capability.
Solution Approach 2:
Different regions of the second substrate are assigned different functional qualities: the first region is optimized for A/D conversion with appropriate heat dissipation characteristics, while the second region is optimized for neural network computation. This local differentiation allows each region to perform its function efficiently while minimizing overall heat impact on the pixel array.
2Measurement precision
If multiple processes are performed sequentially based on neural network computation model, then processing accuracy is improved, but signal processing speed is hindered
Solution Approach 1:
The signal processing function is segmented into multiple independent signal processing sections, each capable of performing neural network computation. These sections are arranged in parallel on the second substrate, enabling simultaneous processing of multiple signals while maintaining the accuracy of neural network-based processing.
Solution Approach 2:
Multiple signal processing sections are combined on a single second substrate, integrating what would traditionally require multiple separate processing units. This merging enables parallel processing operations to occur simultaneously on one substrate, improving speed while preserving processing accuracy through the neural network computation model.
Data Source
AI summary
A photoelectric conversion apparatus includes a first substrate including a pixel array having a plurality of pixels including a pixel configured to output a focus detection signal and a second substrate layered on the first substrate, the second substrate including an analog-to-digital (A/D) converter including a plurality of A/D conversion circuits that converts a signal output from the first substrate to a digital signal, wherein the second substrate further includes a plurality of signal processing sections configured to perform a mechanical learning process, the second substrate including a first signal processing section and a second signal processing section, wherein the plurality of A/D conversion circuits includes a plurality of different sets, wherein the first signal processing section is disposed in correspondence with one of the plurality of sets, and wherein the second signal processing section is disposed in correspondence with another of the plurality of sets.