A classification optimization method for broadband Doherty power amplifier designs using Support Vector Machine

Wang, Wenhao

doi:10.1016/j.aeue.2026.156586

A classification optimization method for broadband Doherty power amplifier designs using Support Vector Machine

2026-09-01

0

PRE

AI

DOI：10.1016/j.aeue.2026.156586

Original

Original request for help

Save

Abstract

En 中文

This paper proposes a classification-based optimization method for broadband, high-efficiency Doherty power amplifier (DPA) design, leveraging the Support Vector Machine (SVM) algorithm. Unlike conventional regression-based methods, the proposed method first identifies the feasible impedance region satisfying the design requirements, and then constructs binary classification models to distinguish viable from non-viable design candidates. The proposed method enables efficient navigation of high-dimensional impedance space without reliance on parametric curve-fitting. To verify the proposed method, two broadband DPAs operating within the Sub-6 GHz spectrum are designed: one covering 1.5-2.4 GHz, and the other covering 3.1-3.6 GHz. Measured results demonstrate that the 1.5-2.4 GHz DPA achieves a saturated drain efficiency (DE) higher than 61.0% and a 6-dB back-off DE higher than 48.0%, while the 3.1-3.6 GHz DPA achieves a saturated DE over 56.4% and a 6-dB back-off DE over 43.3%. It is to highlight that the proposed method inherently circumvents the overfitting issue endemic to regression-based methods, making it particularly robust for high-efficiency, wideband PA designs.

Keywords:

Broadband

Classification optimization method

Doherty power amplifier

High-efficiency

Support Vector Machine

A classification optimization method for broadband Doherty power amplifier designs using Support Vector Machine

Abstract

Journal

Organization

Cited Papers