Modeling Method for Accurate Electrothermal Behavior of GaN HEMT Based on Modeling Data and Optimization Algorithm
肖龙 1陈冬冬1
扫码查看
点击上方二维码区域，可以放大扫码查看
作者信息
1. 闽南理工学院电子与电气工程学院,泉州 362700
折叠
摘要
为了实现对氮化镓高电子迁移率晶体管GaN HEMT(gallium nitride high electron mobility transistor)高速开关带来的开通过压、误导通、开关振荡和EMI噪声等问题展开定量的仿真分析,提出了一种基于建模数据和最优化算法的门极增强型GaN HEMT电热行为模型建模方法.相比较于常规GaN HEMT行为模型,所提出的建模方法采用2个简单的建模公式实现了对GaN HEMT在第一和第三象限宽工作温度范围内的电热特性进行准确的建模.同时采用一个紧凑的建模公式实现对GaN HEMT非线性寄生电容的精确建模.此外,提出了一种遗传算法和Levenberg-Marquardt算法组合的优化算法,基于该优化算法和建模数据实现了对建模参数的快速提取,在较大程度上减小了建模时间和工作量.仿真表明,所提出的建模方法能够实现对不同公司多个型号的GaN HEMT器件展开精确的建模.最后通过吻合的动态仿真和实验数据验证了所提建模方法的正确性和有效性.
Abstract
A modeling data and optimization algorithm driven electrothermal behavior model of gallium nitride high electron mobility transistor(GaN HEMT)is proposed to facilitate the quantitative analysis of problems caused by high speed switching,such as turn-on overvoltage,false turn-on,oscillation and EMI noise.Compared with the traditional behavior models of GaN HEMT,the proposed model can precisely depict the electrothermal characteristics of GaN HEMT in a wide temperature range in both the first and third quadrants by only two compact equations.Meanwhile,the nonlinear parasitic capacitances of GaN HEMT can be accurately modeled by one compact equation.In addition,an optimization algorithm combing the genetic algorithm and Levenberg-Marquardt algorithm is put forward,and a one-step extraction of modeling parameters is realized based on this optimization algorithm and modeling data,which can reduce the modeling time and work load to a certain degree.Results show that the proposed modeling method can precisely model multiple types of GaN HEMT devices manufactured by different companies.Finally,the correctness and effectiveness of the proposed modeling method was verified by the well-matched simulated dynamic waveforms and experimental measurement data.
关键词
氮化镓高电子迁移率晶体管/电热行为模型/最优化算法/参数提取
Key words
Gallium nitride high electron mobility transis-tor(GaN HEMT)/electrothermal behavior model/optimization algorithm/parameter extraction
维普
万方数据