Deep Learning-Driven Inverse Design of Doherty Power Amplifiers Using Pixelated Combiners and Dual-State Impedance Synthesis

2026-06-16 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Hardware Architecture, Signal Processing · Depth: Expert, quick

Summary

A novel methodology for the inverse design of Doherty power amplifier (PA) output combiners integrates deep convolutional neural networks (CNNs), pixelated layout representations, and genetic algorithms (GA) with dual-state impedance synthesis. This approach addresses the complex challenge of designing combiners that manage load modulation, impedance matching, and phase compensation for both peak and back-off power conditions. As a proof of concept, two GaN HEMT Doherty PA prototypes were designed and fabricated, incorporating three-port pixelated combiners. These prototypes achieved a measured saturated output power exceeding 44.2 dBm and a peak drain efficiency above 71.2% within the 2.6-2.8 GHz band. Furthermore, a drain efficiency of 64% was measured at the 6-dB back-off level, and after digital predistortion, each prototype demonstrated an adjacent channel leakage ratio (ACLR) better than -51.3 dBc.

Key takeaway

For RF Design Engineers developing high-efficiency power amplifiers, this deep learning-driven inverse design approach offers a powerful alternative to traditional methods. You should consider integrating CNNs and genetic algorithms with pixelated layouts to rapidly synthesize complex Doherty PA combiners. This can significantly streamline the design process for multi-objective optimization, enabling you to achieve superior performance metrics like >71.2% peak drain efficiency and excellent ACLR in the 2.6-2.8 GHz range.

Key insights

Deep learning and genetic algorithms enable inverse design of complex Doherty PA combiners for optimal performance.

Principles

• Combining CNNs and GAs optimizes multi-objective RF design.
• Pixelated layouts facilitate AI-driven physical design synthesis.
• Dual-state impedance synthesis addresses varied power conditions.

Method

The methodology combines deep CNNs, pixelated layout representations, genetic algorithms, and dual-state impedance synthesis for three-port Doherty combiner design.

In practice

• Design GaN HEMT Doherty PAs for 2.6-2.8 GHz band.
• Achieve >44.2 dBm output power with >71.2% peak efficiency.
• Apply digital predistortion for ACLR < -51.3 dBc.

Topics

• Deep Learning
• Doherty Power Amplifiers
• Inverse Design
• GaN HEMT
• RF Design
• Genetic Algorithms

Best for: AI Scientist, AI Hardware Engineer, Research Scientist

Related on AIssential

• Deep-Learning-Based Pixelated Microwave Filter Design and Characterization using Electro-Optical Electric-Field Measurements — Deep learning combined with genetic algorithms automates pixelated microwave filter synthesis, validated by novel electro-optical measurements.
• Dual-Domain Equivariant Generative Adversarial Network for Multimodal CT-PET Synthesis — Combining dual-domain learning with geometric equivariance significantly enhances multimodal medical image synthesis accuracy and robustness.
• AttentionCap: Transformer Based Capacitance Matrix Learning Toward Full-Chip Extraction — AttentionCap uses a customized Transformer and novel loss for accurate, fast, and transferable full-chip capacitance extraction.
• From SGD to Muon: Adaptive Optimization via Schatten-p Norms — Dynamically adapting LMO geometry from runtime data improves optimizer performance and design.
• Machine learning enables roughness-driven inverse design of milling processes — An ML-based framework integrates DNN/RF models with Bayesian optimization for inverse design of milling processes, optimizing surface roughness with <5% error.
• Constructive Circuit Amplification: Improving Math Reasoning in LLMs via Targeted Sub-Network Updates — Targeted sub-network updates can significantly improve specific LLM capabilities with minimal collateral impact.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.