Research On Quasi-Physical Large-Signal Model For Milimeter-Wave GaN HEMT Power Devices

Due to the high output power performance,power efficiency and high operating frequency of the GaN HMET transistors,they have been widely used in radar,satellite,and 5G communication system.An accurate,simple,scalable,physical meaning device model is very important for circuit design and device optimization.For the GaN HEMT devices working at W-band(75-110 GHz),as the parasitic effect becomes complex,the parameter extraction becomes hard,the accuracy and scalability at high frequency also need to improve.The emerging model of Quasi-Physical Zone Division(QPZD)with the advantages of less parameters,high accuracy,and physical meaning,but still have problems on the model equations and parameter extraction,also the significant short channel effects should be added in the model.Therefore,as the high frequency application of GaN HEMT,and the demand for accurate,concise,and clear physical meaning of large-signal model,it is significant to establish a quasi-physical large-signal model that can be applied to W-band and contains the main short-channel effects.This dissertation is aimed to investigate the quasi-physical model of millimeter-wave GaN HEMT power transistor based on 0.1?m domestic processing technology.1.Millimeter-wave small-signal model investigation.For the complex parasitic effect at W-band,a new equivalent-circuit model corresponding parameter extraction method is proposed.The taps of manifold are introduced in the equivalent-circuit,the layout of extrinsic part is divided into manifolds,taps and gate/drain fingers,the new model with clear physical meaning and easy to be scalable.For the issue of parameter increasing and hard to extract,for the extraction of capacitances,where a low frequency full-wave electromagnetic(EM)method has been used.For the inductance extraction,the approach combines the measurement data and analytical equations.The parameter values in this dissertation are more reasonable,and more accurate with the error 5%?8%in the frequency range from 1?110 GHz.2.The Quasi-Physical Zone Division(QPZD)large-signal model for GaN HEMT investigation.The QPZD model,suggested so far,has many fitting parameters,which are hard to extract.To circumvent this problem,in dissertation,a new set of model equations and the corresponding parameter extraction flow have been proposed.The proposed modified QPZD model also accounts for the effect of Drain Induced Barrier Lowing(DIBL),while it was not contained in the original QPZD model.The model shows good accuracy,and the number of parameters for Ids without DIBL effect deceases from 16 to 14 compared with the original QPZD model.The extraction flow proposed here is much easier,the number of parameters has been reduced up to 50%compared with empiricism-based Angelov model and physics-based surface-potential model.The whole large-signal model has been verified by measurement data,it can model the S-parameter,DC,output power,gain and power added efficiency accurately.3.The millimeter-wave scalable large-signal model for GaN HEMT investigation.For the issue of inaccurate for the traditional scaling rules,a new set of parasitic scaling rules are proposed.The scaling rule of tap part is introduced,and the scaling rules of manifold,gate/drain fingers are corrected.The verification results of transistors with different sizes,show that the proposed W-band QPZD large-signal model has high precision for S-parameter,DC,output power,gain and power added efficiency.