Application Of Large Signal Model Of GaN Device In Yield Design Of MMIC

Wide bandgap semiconductor transistors—GaN high electron mobility transistors?GaN HEMTs?have a wide range of applications in the field of high frequency,high efficiency,and high power,and have become a research focus in microwave semiconductor devices at home and abroad.The large signal transistor model based on GaN HEMT is the hub of transistors and circuits,which plays a guiding role in optimizing circuit design and improving transistor performance.At present,the more mature transistor models are mostly empiricism-based models,which represent the self-heating effects and dispersion characteristics of transistors by a large number of fitting parameters.However,with the continuous improvement of process and the development of devices,physical-based large signal models that can guide the design of transistor technology and structure are urgently needed.On the other hand,the yield optimization design of microwave compound semiconductor chips has always been a difficulty in integrated circuit design.The main reason is that it is difficult to precisely control the process parameters in transistor fabrication,and the circuit design lacks an accurate large signal process statistical model.Therefore,how to establish the physical-based statistical model of transistor microwave characteristics related to physical parameters has become the key to the design and application of microwave gallium nitride circuit.In this paper,the quasi-physical-based large signal model and physical-based statistical model were studied for GaN HEMT,and a method to improve the yield of MMIC was proposed based on the statistical model.The main research contents are as follows:1.Research on GaN HEMT quasi-physical-based large-signal model based on zone division theory.Based on the GaN HEMT device process with a grid length of 0.15?m,the dc I-V,multi-bias S-parameters and large-signal characteristics of the transistor were first tested,and then based on the zone division theory,a quasi-physical-based large signal model was established.Finally,the verification results of MMW amplifier chip show that the output power error of the large signal model is 1.4%,the efficiency error is 1.8%and the gain error is 2.2%in the 23dBm input and the 35GHz frequency point.2.Research on the microwave GaN HEMT physics-based large signal statistical model.Using four batches of 25 GaN HEMT tra nsistors,based on the self-built regional partitioning model,starting from I_ds data sets,using principal component analysis,factor analysis method to deal with data sets,is used to establish the statistical model which contains many physical parameters statistical properties such as barrier layer thickness,the largest electronic saturated velocity,maximum electron sheet density,electron sheet density at pinch-off,Saturated electron mobility.The large signal validation results show that the statistical model average accuracy higher than 90%at36GHz with different input power.3.Optimization design of yield based on large signal statistical model.Based on the statistical analysis of parameter sensitivity in the large signal statistical model,an optimization design method of yield based on large signal statistical model was proposed.This method first obtains the impedance distribution diagram through loadpull simulation under multi-parameter fluctuation,then establishes the relation between the output characteristics corresponding to the impedance points and the yield,and finally finds out the impedance points with the highest yield for circuit design.The results show that the new chip yield is increased by nearly 10%compared to the original chip in 32GHz-38GHz bandwidth.