Comparison of HEMT non‐linear model extraction approaches based on small signal and on large signal measurements

An accurate analytical electrothermal drain current (I ds ) model for AlGaN/GaN HEMTs is presented in this letter. The model is implemented into our recently proposed quasi-physical zone division (QPZD) model for demonstration purpose. Compared with the original QPZD model, its electrothermal characteristics are enhanced by involving more fundamental temperature dependent elements. In addition, these elements are derived analytically based on physical mechanisms instead of former pure empirical fitting method. Thus, the extracted parameter values are more close to intrinsic values. An in house 0.15-μm GaN HEMT is used for validation. Compared with the measured data at a wide ambient temperature range (245-390 K), this electrothermal model demonstrates good accuracy to predict the DC characteristics and RF performances of GaN HEMTs.relatively simple while still provides a reasonable prediction on the device performance under different operating conditions. This large-signal modeling approach is usually based on the current and charge sources obtained by integration. 9,10 C. Wang et al 11 presented a temperature dependent large signal model for GaN HEMTs, which includes thermal, trapping, and RF dispersion effects. However, the method proposed in the present paper was to modify the Chalmers model that was originally proposed by I. Angelov 12,13 and is basically an empirical model. Empirical models for GaN HEMTs 14-18 usually contain dozens of fitting parameters, thus hardly offer an opportunity to understand device operation from the physical point of view.Compared with empirical models, physics-based analytical models for GaN HEMTs have been widely caught the attention in recent years 19-25 because they originate from device operation principle and contain most of the device physical parameters and much less empirical ones. Previously, we proposed an analytical model named quasi-physical zone division (QPZD) model 26,27 for GaN HEMTs. Compared with other fully physics-based analytical models, the QPZD model shows some advantages, such as concise formulas with superior accuracy, fewer parameters, better convergence, etc. However, empirical fitting functions are introduced to the electron sheet density n s and the critical electric field E c , which lack of clear physical meanings. Besides, only a few of the temperature dependent elements, such as the pinch off voltage V off and the maximal electron saturated velocity υ sat are included in the model. These two factors may result in unreasonable parameter values. For example, the obtained variation of the low-field mobility μ 0 with temperature is unreasonably large in our previous QPZD model. 27 The aim of this work is to improve the electrothermal characteristics of the QPZD model to incorporate the temperature effects on device performance. The effects of temperature variation on the fundamental elements (ie, ϕ B , E g , μ 0 , υ sat , and k) and also on the deduced ones (ie V off , n s , the thermal resistance R th , the maximal drain current I max , E c ) are inc...

Section: Introductionmentioning

confidence: 99%

Improved quasi‐physical zone division model with analytical electrothermal I_ds model for AlGaN/GaN heterojunction high electron mobility transistors

Yong-bo

Wang³

et al. 2019

“…Typically, microwave transistor nonlinear models are described in terms of nonlinear currents and charges, which account for carrier transport and charge variations within the semiconductor materials. A great deal of works aimed at extraction and identification of nonlinear models of microwave transistors can be found in literature, including . A largely used FET model topology is reported in Figure .…”

Section: Introductionmentioning

confidence: 99%

A procedure for the extraction of a nonlinear microwave GaN FET model

Avolio

Vadalà

Angelov

et al. 2016

In this paper, we describe an extraction procedure of nonlinear models for microwave field-effect transistor (FET). We use a nonlinear model available in commercial CAD tools, and we extract the parameters by combining direct extraction and numerical optimization. We determine a first estimate of the model parameters by few DC and S-parameter measurements. Next, we refine the parameters by optimization against low-frequency and highfrequency vector-calibrated large-signal measurements gathered with a Large-Signal-Network Analyzer (LSNA). As case study we consider a 0.25 × 200 μm 2 GaN FET on SiC for power amplifier applications. Ultimately, we want to show that a good accuracy level can be achieved while minimizing the extraction effort and that an accurate model can be built and suitably tailored depending on the final application.

“…It is based on high-frequency small-signal measurements for the identification of both the parasitic network and the strictly nonlinear dynamic effects, whereas low-frequency large-signal measurements are exploited for the description of the dynamic I/V characteristics of the device above the cut-off of low-frequency dispersion. A validation of the proposed model is provided by using time-domain nonlinear measurements carried out at 40 MHz and 100 MHz.in look-up tables [19][20][21] that are dynamically read during the simulation. To complete the tables with the missing data outside the measurement domain, standard extrapolation techniques have been used.It is worth noting that the transconductance delay has been considered through the transcapacitance C m (i.e., g m τ), which is typically adopted in large-signal models [11].…”

mentioning

confidence: 99%

“…in look-up tables [19][20][21] that are dynamically read during the simulation. To complete the tables with the missing data outside the measurement domain, standard extrapolation techniques have been used.…”

mentioning

confidence: 99%

Nonlinear modeling of LDMOS transistors for high‐power FM transmitters

Bosi

Crupi

Vadalí

et al. 2013

In this paper, a nonlinear model of a commercial 10-W laterally diffused metal oxide semiconductor (LDMOS) transistor for FM transmitters is identified. It is based on high-frequency small-signal measurements for the identification of both the parasitic network and the strictly nonlinear dynamic effects, whereas low-frequency large-signal measurements are exploited for the description of the dynamic I/V characteristics of the device above the cut-off of low-frequency dispersion. A validation of the proposed model is provided by using time-domain nonlinear measurements carried out at 40MHz and 100 MHz