Neural Network Distortion Filter for Optical Transmission

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Optical transmission systems face reliability issues due to inter-symbol interference (ISI) caused by filter narrowing, which distorts signals and introduces noise, making them less reliable.

Innovation Solution

A neural network-based system that receives distorted optical symbols, identifies and predicts the original symbols using a trained neural network, thereby mitigating filter narrowing-induced ISI without requiring additional system information like baud rate or OSNR, and decodes the predicted symbols to improve transmission reliability.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If narrower filters are used to increase spectral efficiency, then spectral efficiency is improved, but signal distortion and inter-symbol interference increase

Engineering Contradiction:

Improvespectral efficiencyVSAvoidsignal quality

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The neural network is trained in advance using known transmitted symbols and received distorted symbols to learn the filtering effects of the transmission system. This preliminary training enables the network to predict and compensate for filter narrowing effects before actual data transmission, allowing the use of narrower filters while maintaining signal quality through pre-learned compensation patterns.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

A neural network is introduced as an intermediary component between the distorted received signal and the decoding process. The network acts as a mediator that learns and compensates for the filtering effects, enabling the system to use narrower filters for spectral efficiency while the neural network restores signal quality by predicting original symbols from distorted ones.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Productivity

If filter narrowing is applied to improve spectral efficiency, then spectral efficiency is improved, but inter-symbol interference increases

Engineering Contradiction:

Improvespectral efficiencyVSAvoidinter-symbol interference

Core Design Contradiction:

ProductivityVSObject-generated harmful factors

Solution Approach 1:

The neural network is trained on the actual distorted symbols produced by filter narrowing, converting the harmful interference patterns into learnable features. By training with the distorted signals as input and original symbols as target, the network learns to recognize and reverse the specific interference patterns, transforming the harmful effect into a compensatable distortion that improves overall system performance.

Inventive Principle:

Principle #22Blessing in disguise (Convert harm into benefit)

3Device complexity

If traditional filtering methods are used, then system complexity is low, but compensation for filter narrowing effects is insufficient

Engineering Contradiction:

Improvesystem complexityVSAvoiddistortion compensation

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

Traditional mechanical or electronic equalization methods are replaced with a neural network-based approach. Instead of using complex adaptive filters or equalizers that require detailed knowledge of channel characteristics, the patent substitutes a data-driven neural network that automatically learns the compensation patterns from training data, achieving better distortion compensation with simpler system architecture and without requiring baud rate or OSNR information.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS10708094B2Transmission filtering using machine learning

Publication Date: 2020.07.07 IP WAVE PTE LTD

AI summary

Systems and methods for transmission filtering are provided. A receiver includes an input coupled to a transmission line to receive distorted optical symbols. A distortion filter is coupled to the input to replace the distorted optical symbols with predicted symbols using a trained neural network. A decoder is coupled to the distortion filter to decode the predicted symbols.