Machine Learning based Load-Pull Algorithm for Fast Reflection Factor Synthesis

Load-pull is one of the most used techniques for estimating and optimizing the performance of device operating in the non-linear region. In order to solve the problem of instability and slow tuning speed of incident wave at the load port in active load-pull, a behavioral model based on Bayesian inference is proposed to predict the tuned incident wave. Compared with the traditional numerical method, this method can effectively reduce the number of iterations required for load impedance fitting and improve the accuracy of impedance synthesis. The efficiency of proposed method was validated by carrying out load-pull measurements on a fundamental frequency of a 10×60um GaN HEMT transistor.