Data-driven Optimization for High-efficiency Power Amplifier Designs

The load-pull technique is a primary routine for designing and optimizing power amplifiers (PAs). However, most load-pull techniques only focus on the fundamental impedance while fixing the harmonic terminations, which limits the achievable high efficiency of PA designs. In this paper, a data-driven searching technique is proposed for high-efficiency PA design, which achieves high efficiency by proposing simultaneous novel fundamental and harmonic impedances. Use of practical circuit simulation results as training data, allows surrogate models to be constructed using least-square support vector regression. After that, a second-order ascending strategy is adopted to optimize the surrogate model to find optimum terminations. To verify the effectiveness of the proposed algorithm, a high-efficiency PA is designed. Simulation results show that the PA achieves a maximum power-added efficiency of 80.5 % at 2.6 GHz with an output power of 41.2 dBm and a gain of 13.2 dB.