X-parameter modeling investigation for microwave power devices. (September 2022)

• Record Type: Journal Article
• Authors: Lin, Qian; Wang, Xiao-Zheng; Wu, Hai-Feng
• Is Part Of: Microelectronics journal. Volume 127(2022)
• Publication Date: 2022-09
• Subjects: X-parameter -- GRNN -- ELM -- Microwave power device -- Large signal; Microelectronics -- Periodicals; Microélectronique -- Périodiques; Microelectronics; Electronic journals; Journals - contents and abstracts; Periodicals; 621.3805
• DOI: 10.1016/j.mejo.2022.105523
• Languages: English
• ISSNs: 0959-8324

Abstract

Abstract: In order to address the complex problem of large-signal parameter extraction for microwave power devices, in this paper an X-parameter modeling approach for microwave power devices based on artificial neural network (ANN) is proposed. Since extreme learning machine (ELM) and general regression neural network (GRNN) have outstanding performance in expressing nonlinearity. Thus, ELM and GRNN are used to model X-parameters of transistors respectively. Next, for the sake of verify the effect for these two models, harmonic balance experiment is carried out, after that the third harmonic and modulus value of predicted data and expected data are obtained. After comparisons, it is proved that the three harmonic errors of ELM model are 3.577dBm, 1.092dBm and 2.511dBm, respectively. And the three harmonic errors of GRNN model are 0.130 dBm, 0.001 dBm, 1.235 dBm, respectively. Otherwise, the three harmonic modulus errors of ELM are 0.002, 0.321e, 0.685e, respectively, and the errors of GRNN model are 0.001, 0.235e, 0.304e, respectively. Therefore, the established GRNN model can accuracy can accurately characterize large signals more than for GaN high electron mobility transistor (HEMT).