Scholarlink

<p>This paper presents the design, implementation, and modeling of an ultra-wideband (UWB) GaN HEMT power amplifier (PA) using Class-F mode theory and machine learning-based predictive analysis. The PA operates from 1.6 to 4.8 GHz, achieving high efficiency through a broadband impedance environment and harmonic tuning network. A Cree CG2H40010F GaN HEMT transistor is employed with a harmonic control network that enables smooth transitioning between F waveforms. To support efficient design validation, an XGBoost regression model is developed to predict gain and power-added efficiency (PAE) across varying input powers and frequencies. Experimental results show a peak PAE of 58–70%, a gain range of 9–14 dB, and output power exceeding 39–40 dBm. The XGBoost model exhibited superior regression performance, attaining an RMSE of 0.05397 for gain and 0.0119 for PAE, along with an <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(R^2\)</EquationSource> </InlineEquation> score of 0.99 for both parameters. These results validate the model’s robustness in capturing complex PA behavior across a broad frequency span. The integration of ML-based predictive modeling with PA design significantly reduces simulation overhead while maintaining high accuracy, establishing a scalable framework for RF/microwave component optimization. This methodology provides a foundation for data-driven design strategies, enabling the rapid prototyping of intelligent, high-efficiency amplifiers for wideband 6G applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning based design of ultra-wideband GaN HEMT power amplifiers for next-generation 6G systems

  • 期刊论文

  • 发表日期：

    2025年8月23日

查看全文
  • Hemant Kumari,
  • Amartya Paul,
  • Shubhankar Majumdar

摘要

This paper presents the design, implementation, and modeling of an ultra-wideband (UWB) GaN HEMT power amplifier (PA) using Class-F mode theory and machine learning-based predictive analysis. The PA operates from 1.6 to 4.8 GHz, achieving high efficiency through a broadband impedance environment and harmonic tuning network. A Cree CG2H40010F GaN HEMT transistor is employed with a harmonic control network that enables smooth transitioning between F waveforms. To support efficient design validation, an XGBoost regression model is developed to predict gain and power-added efficiency (PAE) across varying input powers and frequencies. Experimental results show a peak PAE of 58–70%, a gain range of 9–14 dB, and output power exceeding 39–40 dBm. The XGBoost model exhibited superior regression performance, attaining an RMSE of 0.05397 for gain and 0.0119 for PAE, along with an \(R^2\) score of 0.99 for both parameters. These results validate the model’s robustness in capturing complex PA behavior across a broad frequency span. The integration of ML-based predictive modeling with PA design significantly reduces simulation overhead while maintaining high accuracy, establishing a scalable framework for RF/microwave component optimization. This methodology provides a foundation for data-driven design strategies, enabling the rapid prototyping of intelligent, high-efficiency amplifiers for wideband 6G applications.