臺灣博碩士論文加值系統

無線通訊中，射頻積體電路(RFIC)設計已成為重要的專業技術，能否提供完整並準確的元件高頻模型，修關於增進電路設計之精確性與設計產能效益，因此高頻元件參數之萃取與元件模型建置已成為相關RFIC設計領域之關鍵環節。

本論文擬首先進行元件之高頻特性量測，以 “Cold-FET”理論為基礎，藉由Microwave Office®電腦輔助設計軟體，進行初期之高頻參數萃取，再利用Matlab之強大的數值運算與程式語言功能為平台，以期藉由數値分析之系統化趨近方式，取代現行之手動微調步驟，以期萃取出能同時精準描述4組S參數實驗特性之高頻元件參數組合，進而建立高效益、高精準、且具物理意義之高頻元件模型。

本論文之數值分析方法包含倒傳遞類神經網路與基因演算法。在倒傳遞類神經網路裡，輸入正確訊息並將其與輸出信號相比較。當兩者之誤差產生，以倒傳遞類神經網路學習演算法則完成神經網路的訓練。在量測S參數時，由於提供太少之頻率點予傳遞類神經網路當作學習之樣本數，導致模擬結果參數值不具有物理意義，且與所萃取的初始值有著非常大的關係。基因演算法源自於達爾文的”適者生存，不適者淘汰”的理論。在基因演算法的結果中顯示：S11、S21、S22的實部與虛部都非常吻合，然而，雖然S12的實虛部不完全吻合，但其差異極小，且在高頻的區域非常接近。在基因演算法中，模擬的S參數值趨近量測的S參數值，得到本計畫所期望之結果。

本論文提供一系統化精準萃取小訊號模型之方法，在電路設計方面提供精準之高頻元件模型，以期望提高電路效能。

In the wireless communication, radio-frequency integrated circuit (RFIC) design has already become one of the important professional techniques. Establishing a complete and accurate high-frequency model can improve the CAD application validity, benefiting the RFIC design. Therefore, precisely extracting the parameters of the high-frequency model for the RFIC design applications is the main motive for this thesis.

The first step is to measure the high-frequency S-parameters. By using the CAD tool of the Microwave Office®, we will be able to extract the initial parameter values of the equivalent circuit based on the "Cold-FET" theory. A systematic procedure of extraction is performed by utilizing the powerful numerical and matrix operations and embedded function of the MATLAB program. Four S-parameter sets can then be precisely described at the same time to establish the highly accurate and physically meaningful high-frequency model.

The numerical analysis to improve the extraction accuracy in this thesis includes two methods: (1) the back-propagation neural networks and (2) the genetic algorithms. In back-propagation neural networks, we assume the trial input information and compare with the output signal. Observing the encountered error with respect to the experimental results, the learning algorithm of the back-propagation neural network is used to complete training phase. The accuracy by using this method is not satisfied, mainly due to the limited sample data to provide more calculating iteration. Consequently, the parametric values are not physically meaningful and show sensitive dependence on the assumed initial values. On the other hand, the genetic algorithms, one of the stochastic searching techniques based on the “survival-of-the-fittest” of the Darwinism principles to emulate the natural genetic operators, has been used to improve the extraction accuracy. The results show that all of the extracted real/imaginary parts of the S11, S12, S21, and S22 have precisely and simultaneously matched with the measured characteristics.

This thesis has successfully provided a systematic method to extract the high-frequency small-signal models for the HEMT device. High extraction accuracy and efficiency have been accomplished by integrating the generic algorithm. The proposed methodology can be easily applied to extracting different high-speed devices and is promisingly useful for the RFIC design technologies.

Contents
Acknowledgement i
Abstract ii
Contents v
Figure Captions vii
Table Captions ix
1. Introduction 1
1.1 Why HEMT Modeling 1
1.2 Discussion of Research Work 2
1.3 Thesis Organization 3
2. HEMT Physics and Modeling 4
2.1 Physical Structure 4
2.2 Simple Principles of the HEMT Operation 6
2.3 HEMT Small-Signal Models 6
2.3.1 Parasitic Inductances Ls, Lg, Ld 8
2.3.2 Parasitic Resistances Rs, Rg, Rd 8
2.3.3 Pad Capacitances Cpd, Cpg 9
2.3.4 Intrinsic Capacitances Cgs, Cgd, Cds 9
2.3.5 Tranconductances gm 10
2.3.6 Output Resistance Rds 11
2.3.7 Tranconductance Delay τ 11
2.3.8 Charging Resistance Ri 12
3. Extraction Methodology 13
3.1 Introduction 13
3.2 Model Formulation at Bias Conditions 13
3.3 Intrinsic Circuit Extraction 14
3.4 Extrinsic Circuit Extraction 20
3.4.1 A Simplified Model of the HEMT at Pinch-off Voltage 20
3.4.2 Pad Capacitances 21
3.4.3 Extrinsic Inductances and Resistances 22
4. Back-Propagation Neural Network 24
4.1 Neural Networks 24
4.2 Perceptrons 24
4.3 Back-Propagation Neural Network Structure 28
4.4 Back-propagation neural network learning algorithm 29
4.5 The Procedure of Back-Propagation Neural Network 30
4.6 The Flow Chart of the Procedure 32
4.7 Results 33
5. Genetic Algorithm 37
5.1 Introduction of Genetic Algorithm 37
5.2 Characters of Genetic algorithm 38
5.3 The Structure of Genetic Algorithm 39
5.4 The Flow Chart of the Procedure 46
5.5 Results 47
6. Conclusion 55
7. References 57





Figure Captions
Figure 2.1 Cross Section of an AlGaAs/GaAs HEMT
Figure 2.2 HEMT Small-signal Model Including Parasitic Elements
Figure 2.3 HEMT Small-signal Model Showing Physical Origin of Elements
Figure 3.1 15-element small-signal HEMT model
Figure 3.2 Y-Parameter Network
Figure 3.3 Circuit Used to Determine Y12
Figure 3.4 Circuit Used to Determine Y22
Figure 3.5 Circuit Used to Determine Y21
Figure 3.6 Circuit Used to Determine Y11
Figure 3.7 Method for Extracting the Device Intrinsic Y Matrix
Figure 3.8 A Simplified Model of the HEMT at Pinch-off Voltage with Vds=0
Figure 3.9 A Reduced Model at Gate Voltage Equal to the Pinch-off Voltage.
Figure 4.1 The Single-Layer Perceptron Architecture
Figure 4.2 Activation Functions
Figure 4.3 An Example of Three-layer Networks
Figure 4.4 The Flow Chart of the Procedure of back-propagation neural network
Figure 4.5 The Smith Charts from the Procedure of Back-propagation Neural Network
Figure 5.1 Many-Peaked Curve Graph
Figure 5.2 Flow Chart of Genetic Algorithm
Figure 5.3 The chromosome is composed by binary string.
Figure 5.4 Coding Space and Solution Space
Figure 5.5 The Schematic of the “Roulette Wheel” for Proportional Selection
Figure 5.6 One-point Crossover
Figure 5.7 Two-point Crossover
Figure 5.8 Uniform Crossover
Figure 5.9 One-point Mutation
Figure 5.10 Bit Mutation
Figure 5.11 Uniform Mutation
Figure 5.12 The Flow Chart of the Procedure
Figure 5.13 The Real-part of Each S-parameter.
Figure 5.14 The Imaginary-part of Each S-parameter.
Figure 5.15 The Smith Charts from the Procedure of Genetic Algorithm
Figure 5.16 S11 Extracted by Different Theory.
Figure 5.17 S12 Extracted by Different Theory.
Figure 5.18 S21 Extracted by Different Theory.
Figure 5.19 S22 Extracted by Different Theory.

Reference
[1]W. R. Curtice and R.L. Camisa, “Self-Consistent GaAs FET Models for Amplifier Design and Device Diagnostics,” IEEE Trans. Microwave Theory Tech., vol. MTT-32, no.12, pp.1573-1578, Dec. 1984.
[2]G. Kompa and M. Novotny, “Highly Consistent FET Model Parameter Extraction Based on Broadband S-Parameter Measurements,” in IEEE MTT-S Dig., 1992, pp.293-296.
[3]M. Berroth and R. Bosch, “Broad-Band Determination of the FET Small-Signal Equivalent Cirucit,” IEEE Trans. Microwave Theory & Tech., vol.38, no.7, pp.891-895, July 1990.
[4]M. Berroth and R. Bosch, “High-Frequency Equivalent Circuit of GaAs FET’s for Large-Signal Applications,” IEEE Trans. Microwave Theory Tech., vol.39, no.2, pp.224-229, Feb. 1991.
[5]K. Shirakawa, H. Oikawa, T. Shimura, Y. Kawasaki, Y. Ohashi, T. Saito, and Y. Daido, “An Approach for Determining the GaAs MESFET Small-Signal Equivalent-Circuit for HEMTs,” IEEE Trans. Microwave Theory & Tech.,vol.43, no.3, pp.499-503, March 1995.
[6]B.-L. Ooi, M.-S. Leong, and P.-S. Kooi, “A Novel Approach for Determining the GaAs MESFET Small-Signal Equivalent-Circuit Elements,” IEEE Trans. Microwave Theory & Tech., vol.45, no.12, pp.2084-2088, Dec.1997.
[7]S. Lee, H.K. Yu, C.S. Kim, J.G. Koo, and F.S. Nam, “A Novel Approach to Extracting Small-Signal Model Parameters of Silicon MOSFET’s,” IEEE Microwave Guided Wave Lett., vol.7, no.3, pp.75-77, March 1997.
[8]J. Michael Golio, “Microwave MESFETs and HEMTs,”
[9]K. Shirakawa, H. Oikawa, T. Shimura, Y. Kawasaki, Y. Ohashi, T. Saito, and Y. Daido, “An Approach for Determining the GaAs MESFET Small-Signal Equivalent-Circuit for HEMTs,” IEEE Trans. Microwave Theory & Tech.,vol.43, no.3, pp.499-503, March 1995.
[10]B.-L. Ooi, M.-S. Leong, and P.-S. Kooi, “A Novel Approach for Determining the GaAs MESFET Small-Signal Equivalent-Circuit Elements,” IEEE Trans. Microwave Theory & Tech., vol.45, no.12, pp.2084-2088, Dec.1997.
[11]Dambrine et al. propose R. Anholt and S. Swirhun, “Equivalent-Circuit Parameter Extraction for Cold GaAs MESFET’s,” IEEE Trans. Microwave Theory Tech., vol.39, no.7, pp.1243-1247, July 1991.
[12]B.-S. Kim, S, Nam, and K.-S.Seo, “Analytic Intrinsic Model Based Parasitic Extraction Method for HEMTs,” Electron. Lett., vol.30, no.12, pp1005-1006, June 1994.
[13]Gilles Dambrine, Alain Cappy, Frederic Heliodore and Edouard Playez, “A New Method for Determining the FET Small-Signal Equivalent Cirucit,” IEEE Trans. Microwave Theory Tech., vol. 36, no. 7, pp. 1151-1159, July 1988.
[14]H. Kondoh, “An Accurate FET Modelling from Measured S-Parameters,” IEEE MTT-S Dig., pp.377-380, 1986.
[15]S.Yanagawa, H. Ishihara, and M. Ohtomo, “Analytical Method for Determining Equivalent Circuit Parameters of GaAs FET’s,” IEEE Trans. Microwave Theory & Tech., vol.44, no.10, pp.1637-1641, Oct.1996.
[16]H.O. Vickes, “Determination of Intrinsic FET Parameters Using Cirucit Partitioning Approach,” IEEE Trans. Microwave Theory Tech., vol.39, no.2, pp.363-366, Feb. 1991.
[17]P.M. White and R.M. Healy, “Improved Equivalent Cirucit for Determination of MESFET and HEMT Parasitic Capacitances from “Coldfet” Measurements,” IEEE Microwave Guided Wave Lett., vol.3, no.12, pp.453-454, Dec. 1993.
[18]R. Anholt and S. Swirhum, “Measurement and Analysis of GaAs MESFET Parasitic Capacitances,” IEEE Trans. Microwave Theory Tech., vol.39, no.7, pp.1243-1247, July 1991.
[19]S. Chaudhuri and M.B. Das, “On the Determination of Source and Drain Series Resistances of MESFE’s,” IEEE Electron Device Lett., vol. EDL-5, no.7, pp244-246, July 1984
[20]S.-M.J. Liu, S.-T. Fu, M. Thurairaj, and M.B. Das, “Determination of Source and Drain Series Resistances of Ultra-Short Gate-Length MODFET’s,” IEEE Electron Device Lett., vol.EDL-10, no.2, pp.85-87, Feb. 1989.
[21]P. Pouvil, B. Zemour, D. Pasquet, and J. Gaubert, “Determination of Source and Drain Parasitic Resistances of HEMTs”, Electron. Lett., vol.28, no.7, pp.618-620, March 1992.
[22]F. Rosenblatt, “The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain”, Psychological Review 65, pp.386-408, 1958.
[23]D.E. Rumelhart, G.E. Hinton, and P.J. Williams, “Learning Internal Representation by Error Propagation,” Parallel Distributed Processing, D.E. Rumelhart and J.L. McClelland, Eds. Cambridge, MA: MIT Press
[24]周鵬程，類神經網路入門-活用Matlab，全華科技圖書，2005。
[25]陳堃峰，應用模糊類神經網路於壓電致動器之精密定位控制研究，碩士論文，國立東華大學，2004。
[26]C.T. Lin, and C.S. Lee, Neural Fuzzy Systems – a Neuro-Fuzzy Synergism to Intelligent Systems, Singapore: Prentice-Hall, 1999.
[27]M.T. Hagan, and M. Menhaj, “Training Feedforward Networks with the Marquardt algorithm,”IEEE Trans. Neural Networks, vol. 5, no. 6, pp. 989-993,1994
[28]M. Mitchell, An Introduction to Genetic Algorithm, MIT Press, Cambrige, MA, 1996.
[29]J.C. Potts, T.D. Giddens and S.B. Yadav, “The Development and Evaluation of an Improved Genetic Algorithm Based on Migration and Artificial Selection,” IEEE Trans. System, Man, and Cybernetics, Vol. 24, No. 1, pp. 73-86, 1994.
[30]P. Scheunders, “A Genetic Lloyd-Max Image Quantization Algorithm,” Pattern Recognition Lett., Vol. 17, pp. 547-556, 1996
[31]Gen. Mitsou, Genetic Algorithms and Engineering Design, New York, Wiley, 1997.
[32]David A Coley, (1999). “An Introduction to Genetic Algorithms for Scientists and Engineers,” World Scientific Publishing Co. Plc. Ltd..
[33]Goldberg, D.E., (1989). Genetic Algorithms in search, Optimization, and Machine Learning, Addison Wesley.
[34]Cheng. T.M., Wu, H.T., Ke, Y.F., and Feng, C.W. (2002). “Applying Genetic Algorithms on the Computer Simulation for Improving Construction Productivity,” The Ninth International Conference on Computing in Civil and Building Engineering, April 3-5, Taipei, Taiwan.