Digitization Signal Correction Using Nonlinear Memory Modeling

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing digitization systems introduce unwanted frequency components, primarily non-linear distortions, which are difficult to correct due to complex mathematical models requiring knowledge of the ideal digital version of the analog input signal and causing significant computational overhead.

Innovation Solution

A method using a non-linear model with memory and discrete time to estimate model coefficients without knowing the ideal digital version, employing a programmable logic circuit to correct distortions by minimizing modeling errors through least squares optimization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If complex mathematical models (Volterra, Hammerstein, Wiener-Hammerstein) are used to correct non-linear distortions, then the correction accuracy is improved, but the computational complexity and overhead increase significantly

Engineering Contradiction:
Improvecorrection accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts only the essential non-linear distortion components from the complex mathematical models, focusing on the dominant terms that contribute most to distortion. This selective extraction maintains correction accuracy while significantly reducing computational complexity by eliminating less significant model terms.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent transforms the complex model parameters into a simplified parameter set that captures the essential non-linear behavior. By changing the parameter representation and focusing on key coefficients rather than full model complexity, the system achieves effective distortion correction with reduced computational overhead.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If classical least squares algorithms are used to estimate model coefficients, then the modeling error is minimized, but the requirement for knowledge of the ideal digital version of the analog input signal increases implementation difficulty

Engineering Contradiction:
Improvemodeling error minimizationVSAvoidimplementation ease
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The patent enables the system to estimate model coefficients using only the actual digital output signal from the ADC, without requiring external knowledge of the ideal digital version of the input signal. The algorithm uses the available output signal itself to self-determine the correction parameters, eliminating the need for additional reference signals or calibration inputs.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

Instead of using the ideal input signal to determine coefficients (as in classical least squares), the patent inverts the approach by using the distorted output signal to estimate the coefficients. This inversion allows coefficient estimation to proceed with readily available information rather than requiring unavailable reference data.

Inventive Principle:
Principle #13The other way round (Inversion)

3Measurement precision

If high-order non-linearities and memory depths are included in the mathematical models, then the correction precision is improved, but the real-time processing capability deteriorates due to significant computational overhead

Engineering Contradiction:
Improvecorrection precisionVSAvoidreal-time processing capability
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent applies partial action by including only the necessary order of non-linearities and memory depths required for adequate correction, rather than implementing all possible high-order terms. This selective inclusion achieves sufficient correction precision for practical applications while maintaining real-time processing capability by avoiding excessive computational requirements.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3503406B1Method for correcting faults introduced by a digitisation system and associated devices
Publication Date: 2026.02.18 THALES SA
  • EP3503406B1 patent drawingFigure 1~2
  • EP3503406B1 patent drawingFigure 3~4
  • EP3503406B1 patent drawingFigure 5~6

AI summary

The invention relates to a method for correcting defects introduced by a digitization system (20), the method comprising at least the steps of: - obtaining the output signal of the digitization system (20), - calculating the coefficients of a theoretical model ( ) by minimizing the error between the output signal calculated by the theoretical model ( ) and the output signal obtained, to obtain a calculated theoretical model, - applying the calculated theoretical model to the output signal obtained to determine the distortions introduced by the digitization system (20), and - correcting the output signal to obtain a corrected output signal by subtracting the determined distortions.