A circuit‐based neuro‐space mapping technique for small‐signal modeling of multi‐gate GaN HEMT switches. Issue 11 (13th September 2022)

• Record Type: Journal Article
• Authors: Hu, Zhifu; Na, Weicong; He, Meilin; He, Ruicong; Li, Jiahui; Zhang, Qijun; Ma, Kaixue
• Is Part Of: International journal of RF and microwave computer-aided engineering. Volume 32:Issue 11(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2022-09-13
• Subjects: gallium nitride (GaN) -- high‐electron‐mobility transistor (HEMT) -- modeling -- neuro‐space mapping (neuro‐SM) -- switch; Microwave devices -- Computer-aided design -- Periodicals; Computer-aided engineering -- Periodicals; 621.3813
• DOI: 10.1002/mmce.23396
• Languages: English
• ISSNs: 1096-4290

Abstract

Abstract: GaN HEMT switches has become more and more important in RF front‐end. This article proposes a circuit‐based neuro‐SM technique for the small‐signal modeling of multi‐gate GaN HEMT switches for the first time. A general equivalent circuit model for multi‐gate switch HEMT is proposed using the cascaded single‐gate switch HEMT model to represent the tendency rather than the exact behaviors of the small‐signal responses. We utilize this proposed equivalent circuit model as the coarse model and propose a novel circuit‐based neuro‐SM modeling technique. Neural networks are incorporated to learn the difference between the coarse model and the fine device data, improving the neuro‐SM model accuracy. After being trained, the obtained neuro‐SM model can be used for high‐level circuit design to increase the speed and accuracy of circuit design. Compared to the existing modeling techniques for multi‐gate switches, the proposed neuro‐SM achieves better model accuracy. Examples of a dual‐gate GaN HEMT switch and a triple‐gate GaN HEMT switch have been examined.