Deep Neural Network-Based Load-Pull Measurement for Linearity Prediction in Mobile Front-End Impedance Matching Application

Abstract

In this article, a simple deep neural network (DNN) is proposed to predict the linearity of power amplifier modules (PAMs) in load-pull measurement for mobile front-end impedance matching, not for power amplifier design by transistors. PAM is a crucial and fully matched packaged product in the transmitter for amplification in mobile products, which contains digital control circuits, passive components, RF switches, and multiband power amplifiers (PAs). For the 3GPP standard with low current consumption to be met, load-pull measurement of the PAM is essential for the mobile front-end impedance matching application to optimize the final product. However, traditional measurement using all impedance points for plotting load-pull contours is time-consuming. Compared with the traditional measurement method, the proposed method can minimize the measurement time by more than half. The impedance points used for the load-pull measurement are randomly split into two datasets with different ratios for verification. A set of impedance points is used for DNN model training. Another set of impedance points is used for linearity prediction. Experiments have been conducted, and the results highlight that the proposed DNN approach has high accuracy in linearity prediction and significantly minimizes the load-pull data measurement time, almost by half compared with the traditional measurement method. This study demonstrates the effectiveness of DNN with simple MLP structure in load-pull contour exploration in mobile front-end impedance matching applications.