With the rapid development of wireless communication, the requirements of high-integration and low-cost radio frequency (RF) front-end modules in mobile phones result in more usage of silicon-on-insulator field effect transistors (SOIFETs) for the manufacturing of RF switches. However, the nonlinear behavior of SOIFET switches in off-state always produces unwanted harmonics distortion interference when they are excited by a large signal. In this article, we simplify the well-known physics-based surface potential model to an interelectrode nonlinear capacitance (INC) model since it adequately describes the harmonic effects produced by the transistor. The INC model, referred to as the coarse model, cannot match the behavior of the real switch, since many parameters of the switch cannot be accurately determined. This article proposes a novel dynamic neuro-space mapping network model, referred to as the fine model, to optimize the INC model. The proposed model takes advantage of the high accuracy of the fine model and the fast speed of the coarse model. Ultimately, the proposed method can accurately predict the harmonics interference for SOIFET switches in the complex RF front-end environment and provides an intuitive guideline under the design stage to prevent EMI problems.

中文翻译：

---

SOIFET射频开关谐波干扰预测的神经空间映射方法


随着无线通信的快速发展，手机中对高集成度和低成本射频（RF）前端模块的需求导致更多地使用绝缘体上硅场效应晶体管（SOIFET）来制造射频开关。然而，当 SOIFET 开关受到大信号激励时，处于关断状态的 SOIFET 开关的非线性行为总是会产生不需要的谐波失真干扰。在本文中，我们将众所周知的基于物理的表面电势模型简化为电极​​间非线性电容（INC）模型，因为它充分描述了晶体管产生的谐波效应。 INC 模型（称为粗模型）无法匹配真实开关的行为，因为开关的许多参数无法准确确定。本文提出了一种新颖的动态神经空间映射网络模型（称为精细模型）来优化 INC 模型。该模型利用了精细模型的高精度和粗模型的快速速度。最终，所提出的方法可以准确预测复杂射频前端环境中 SOIFET 开关的谐波干扰，并在设计阶段提供直观的指导以防止 EMI 问题。