Design of a Ku-Band Single Stage MMIC PA Based on Cluster Guided Bayesian Optimization

Abstract

In this paper, a single stage gallium arsenide (GaAs) based monolithic microwave integrated circuit (MMIC) power amplifier (PA) is designed. The Bayesian optimization (BO) algorithm with clustering guided Gaussian process upper confidence bound (CG-GPUCB) method, is proposed for the optimization of the MMIC PA. To validate the developed optimization strategy, a 14GHz-16GHz KU-band PA has been designed using 0.25-μm GaAs pseudomorphic high-electron-mobility transistor (pHEMT) process from Lion. The circuit is initially designed by commercial design tool with standard input-output matching. Both the proposed Bayesian method and the tool embedded optimization method are employed for circuit optimization. Simulation results show that the performance based on the proposed technique can be greatly improved compared to both the initial design and the tool embedded optimization method, the PA can achieve the output power over 28.5 dBm with an input of 21dBm, and the power-added efficiency (PAE) over 47.5% and a gain over 7.5dB.