基于Neuro-SM的新型HBT晶体管直流特性建模  

DC characteristic modeling of new HBT transistor based on Neuro-SM

作　　者：闫淑霞[1,2] 张爽 YAN Shu-xia;ZHANG Shuang(School of Electronics and Information Engineering,Tiangong University,Tianjin 300387,China;Tianjin Key Laboratory of Optoelectronic Detection Technology and System,Tiangong University,Tianjin 300387,China)

机构地区：[1]天津工业大学电子与信息工程学院,天津300387 [2]天津工业大学天津市光电检测技术与系统重点实验室,天津300387

出　　处：《天津工业大学学报》2021年第6期66-70,共5页Journal of Tiangong University

基　　金：天津市自然科学基金项目(19JCQNJC03300);天津市教委科研项目(2017KJ088)。

摘　　要：为在不降低建模设计要求的条件下改进建模方法,提高建模精度,对神经网络空间映射(neuro-space mapping,Neuro-SM)直流特性建模方法进行优化,提出了一种适用于异质结双极型晶体管(heterojunction bipolar transistor,HBT)的新型Neuro-SM模型,该方法在输入映射基础上增加了输出映射神经网络,并将该模型在HBT晶体管实例中进行验证。结果表明:输入输出映射网络可以共同作用将粗模型的电流电压信号映射到细模型,新型Neuro-SM模型可以自动调整输入信号以准确地匹配设备数据;通过HBT晶体管的实验证明,新型Neuro-SM模型在直流特性建模中可以达到优化的效果,且建模误差由传统Neuro-SM模型的0.744%降低到0.477%,比传统Neuro-SM模型提高了精度。In order to improve the modeling method and improve the modeling accuracy without reducing the requirements of the modeling design, the Neuro-SM DC characteristic modeling method was optimized. This paper proposes a new Neuro-SM model suitable for HBT transistors. This method adds an output mapping neural network to the input mapping, and verify the new Neuro-SM model in the HBT transistors. The results shows that the input and output mapping neural networks can work together to map the current and voltage signals of the coarse model to the fine model. The proposed new Neuro-SM model can automatically adjust the input signals to accurately match the device data. Experiments with HBT transistors prove that the proposed new Neuro-SM model can achieve optimization effects in DC characteristic modeling, the modeling error is reduced from 0.744% of the traditional Neuro-SM model to 0.477%, and the accuracy is improved.

关 键 词：Neuro-SM 输出映射神经网络 HBT晶体管 直流特性 建模 

分 类 号：TN710.2[电子电信—电路与系统]