本篇論文欲針對InGaP/GaAs HBT建立準確的大訊號模型，希望模型能充分的反應元件的特性，降低電路設計者的成本。對VBIC大訊號模型與SPICE Gummel-Poon模型進行分析討論，比較優劣，並介紹VBIC模型與SPICE Gummel-Poon模型的參數轉換方法。 我們利用GCTC的InGaP/GaAs的HBT進行量測，建立VBIC大訊號模型；我們詳細的介紹VBIC模型的直流與交流參數的萃取方法；在直流參數部分包含寄生電阻參數萃取，電流相關參數萃取，溫度相關參數萃取的方法介紹，並利用比對量測與模擬電流-電壓輸出特性曲線作為直流參數的驗證；而在交流參數方面包含寄生電容參數的萃取，及傳輸時間參數的萃取流程，並利用比對模擬與量測的S參數作為交流參數的確認。 為了驗證元件非線性的特性，則利用負載-推移(load-pull)量測元件操作在2.4GHz及5.2GHz的輸出功率特性，並與元件的負載-推移模擬進行比對，驗證了VBIC模型對描述功率特性的準確度。 VBIC模型在多射級(multi-finger emitter)結構元件的應用，我們針對多射極結構元件的熱耦合(thermal coupling)效應，及因為layout所導致的拉線寄生電容效應進行修正；並對多射極結構元件負載-推移的模擬與量測進行比對，證實了加入了以上兩種效應的修正，可使得模擬的結果更接近量測的結果，使得VBIC模型在多射級結構元件的使用上更為準確。 We try to establish an accurate VBIC model for InGaP/GaAs HBT, and the VBIC model can react characteristics of device fully, then we can decrease the cost of the circuit designers. We discuss the advantage and shortcomings to compare with SPICE Gummel-Poon(SGP) model, and introduce the method how to convert between VBIC model parameters and SGP model parameters. We measure the InGaP/GaAs HBT of GCTC and established the VBIC large signal model；We expand the ways to extract the DC and AC parameters of VBIC model. In DC parameters part, we introduce how to extract the resistance dependent parameters, the current dependent parameters and the temperature dependent parameters, and used the I-V curve to check the accuracy of DC parameters. In AC parameters part, we introduce how to extract the parasitic capacitance dependent parameter and transient time dependent parameters, and used the S parameter to check the accuracy of AC parameters. In order to verify the nonlinear characteristics of the device, we used load-pull measurement to measure the output power characteristics of the device which operated in 2.4 GHz and 5.2 GHz, and compare with the load-pull simulations. We prove the accuracy of the VBIC model in power characteristics. To apply VBIC model to multi-finger emitter device, we aimed the thermal coupling effect and the parasitic capacitance from layout to correct VBIC model. And we used load-pull measurement in multi-finger device compare with load-pull simulation, and we proved that we improve the accuracy of VBIC model and measured data when we take the two effects into account.
