Crosstalk Correction in Multispectral Image Sensors

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Multispectral imaging sensors face challenges with crosstalk between elementary pixels and limitations in manufacturing yield due to the complexity of integrating microlenses and filters, leading to reduced sensitivity and spectral response modifications.

Innovation Solution

A method to limit crosstalk in imaging sensors by using a CMOS detector with Fabry-Pérot type interference filters and microlenses, where the filtering module is attached to the detector using a glue border, and employing a Gaussian function to determine ideal spectral responses and estimate coefficients for correcting crosstalk, utilizing the generalized reduced gradient method for optimization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Use of energy by moving object

If microlenses are integrated with filters and detectors in a stacked configuration, then sensitivity is improved by focusing light onto photosensitive areas, but manufacturing yield is reduced due to the complexity of integrating multiple components

Engineering Contradiction:
ImprovesensitivityVSAvoidmanufacturing yield
Core Design Contradiction:
Use of energy by moving objectVSProductivity

Solution Approach 1:

The system is divided into separate functional modules: detectors on a first substrate, filters on a second substrate, and microlenses on a third substrate. These modules are aligned and integrated using alignment marks and adhesive layers, allowing independent manufacturing of each component while maintaining the light-focusing function to improve sensitivity without compromising manufacturing yield.

Inventive Principle:
Principle #1Segmentation

2Reliability

If filters are placed under microlenses with large angular aperture, then the complete optical path is maintained, but spectral response is modified due to large angle of incidence on filters

Engineering Contradiction:
Improveoptical path completenessVSAvoidspectral response accuracy
Core Design Contradiction:
ReliabilityVSManufacturing precision

Solution Approach 1:

The filter substrate is designed with a curved surface that matches the curvature of the microlens array. This local adaptation of the filter surface geometry ensures that light rays passing through different zones of the microlens array strike the filters at appropriate angles, maintaining spectral response accuracy while preserving the complete optical path for improved reliability.

Inventive Principle:
Principle #3Local quality

3Adaptability or versatility

If multiple manufacturing operations are performed for detectors, filters, and microlenses, then functional completeness is achieved, but overall manufacturing yield is reduced by multiplying the individual yields

Engineering Contradiction:
Improvefunctional completenessVSAvoidoverall manufacturing yield
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

Each component (detectors, filters, microlenses) is manufactured separately on its own substrate with dedicated process optimization. The separate substrates include alignment marks that enable precise integration without requiring complex multi-step manufacturing sequences, thereby maintaining functional completeness while improving overall manufacturing yield through independent production of each module.

Inventive Principle:
Principle #1Segmentation

4Measurement precision

If crosstalk correction coefficients are estimated and applied to macropixels, then crosstalk is reduced improving image quality, but processing time and computational complexity increase

Engineering Contradiction:
Improveimage qualityVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

Crosstalk correction coefficients are pre-calculated and stored in memory before image acquisition. During operation, these pre-computed coefficients are directly applied to the captured images without requiring real-time computation, thereby achieving crosstalk reduction and improved image quality while minimizing additional processing time.

Inventive Principle:
Principle #10Preliminary action

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

Substantially reduces crosstalk in imaging sensors while maintaining sensitivity, improving manufacturing yield by simplifying the integration process and enhancing spectral response accuracy.

Implementation Method 1

a filtering (MF) formed of interference filters (FP1, FP2, FP3) of the Fabry-Pérot type

Methodology Applied
Scientific EffectInterference: Interference

Implementation Method 2

an array of microlenses (ML1, ML2, ML3) for focusing incident light on the photosensitive areas of the pixels

Methodology Applied
Scientific EffectRefraction: Refraction

Data Source

PatentEP3465112B1Method for limiting crosstalk in an image sensor
Publication Date: 2022.08.03 SILIOS TECH
  • EP3465112B1 patent drawingFigure 1a~2c
  • EP3465112B1 patent drawingFigure 3a~3f
  • EP3465112B1 patent drawingFigure 4~5

AI summary

The invention relates to a method for limiting crosstalk in an image sensor, said sensor being an array of macropixels defining an image, each macropixel being formed by a matrix of elementary pixels each dedicated to a distinct spectral band, all of the elementary pixels dedicated to the same spectral band forming a sub-image, this image being topologically broken down into at least one plot, the method comprising the following steps: measuring 700 the spectral response of each elementary pixel λ1, λ.2 , λ3,..., λ9, - calculating 701 the average spectral response of each sub-image in a plot, - targeting 702 to establish the ideal response of each sub-image in this plot, - calculating 703 a series of coefficients to minimise the crosstalk in this plot, - applying 704 said coefficients to the macropixels to correct the sub-images in the plot. The method is characterised in that the ideal response is a Gaussian function.