A Support Vector Machine Method for Electrothermal Modeling of Power FETs

Author

Guo, Yunchuan ; Xu, Yuehang ; Wang, Lei ; Xu, Ruimin

Author_Institution

Univ. of Electron. Sci. & Technol., Chengdu

fYear

2007

fDate

16-17 Aug. 2007

Firstpage

1387

Lastpage

1389

Abstract

An accurate electrothermal modeling method for power FETs is presented. The thermal models are setup by using Support Vector Machine Regression (SVR) approach, which is like artificial neural network (ANN) method leading a knowledge-based model. Unlike traditional ANNs, Support Vector Machine (SVM )method requires fewer samples in statistical learning and is free of local minima in optimization. A comparison among the SVM model, the empirical model and the measurement data of a GaAs power pHEMT are given out to validate the proposed approach.

Keywords

III-V semiconductors; gallium arsenide; neural nets; power HEMT; power field effect transistors; regression analysis; semiconductor device models; support vector machines; GaAs; GaAs - Interface; artificial neural network; electrothermal modeling; knowledge-based model; power FET; power pHEMT; statistical learning; support vector machine regression; Artificial neural networks; Electrothermal effects; FETs; Gallium arsenide; Integrated circuit modeling; Microwave technology; Microwave theory and techniques; PHEMTs; Statistical learning; Support vector machines; Electrothermal model; field effect transistor (FET); support vector machine (SVM);

fLanguage

English

Publisher

ieee

Conference_Titel

Microwave, Antenna, Propagation and EMC Technologies for Wireless Communications, 2007 International Symposium on

Conference_Location

Hangzhou

Print_ISBN

978-1-4244-1045-3

Electronic_ISBN

978-1-4244-1045-3

Type

conf

DOI

10.1109/MAPE.2007.4393537

Filename

4393537