Deep-Learning-Based Pixelated Microwave Filter Design and Characterization using Electro-Optical Electric-Field Measurements

2026-06-16

AI Summary

This work addresses the limitations of conventional microwave filter design, which relies on predefined topologies and manual parameter tuning, thereby constraining performance and efficiency. The authors propose an automated design framework that integrates convolutional neural networks with a genetic algorithm to directly synthesize pixelated microwave filter structures. For the first time, electro-optic near-field measurement techniques are employed to uncover the underlying physical mechanisms—such as emergent equivalent coupled transmission lines or stubs—in AI-generated devices. The resulting low-pass filter exhibits a flat passband up to 7 GHz and achieves better than 20 dB suppression beyond 9.5 GHz. Strong agreement between simulation and experimental measurements validates both the efficacy and interpretability of the proposed approach.

Abstract

Traditional microwave filter design typically relies on iterative parameter tuning and predefined topologies, which limits design space and increases development time. This study uses a deep learning approach combining convolutional neural networks with genetic algorithms to automate pixelated microwave filter synthesis. To validate the approach experimentally, both S-parameter and spatial electric-field measurements were analyzed. The synthesized low-pass filter demonstrated excellent agreement between simulated and measured performance, achieving a 7 GHz passband with over 20 dB suppression beyond 9.5 GHz. Electro-optical measurements, for the first time, revealed electric field patterns that resemble coupled transmission-lines or stub structures, providing insight into the emergent characteristics of AI-generated designs.

Problem

• microwave filter design
• design space limitation
• iterative parameter tuning
• predefined topologies
• development time

Innovation

• deep learning
• pixelated microwave filter
• electro-optical electric-field measurement
• genetic algorithm
• automated design