Download this article in .PDF format
This file type includes high resolution graphics and schematics when applicable.

It is impossible to deny the significant performance capabilities that gallium-nitride (GaN) high-electron mobility transistors (HEMTs) have bestowed upon all but the most niche technologies. In terms of power density, frequency capability, and bandwidth, there are few other technologies that compare. These factors have created a gulf between past modeling and characterization technologies and what is needed today.

But GaN HEMTs most likely represent just the first round of next-generation high-performance technologies that will revolutionize the RF industry. To pave the way for the next round of technologies, computer-aided-design (CAD) techniques must rapidly advance (Fig. 1).

Individual segment benefits

The complexity and performance of GaN HEMT devices is currently limited by two main factors: the ability to effectively design circuits that make full use of the technology’s capability, and the ability to flow enough thermal energy from GaN transistors to operate these devices at peak performance. Sophisticated modeling and characterization techniques can aid in the design process of HEMT devices, potentially mitigating the effects of these limitations.

To accomplish the level of detail, efficiency, and accuracy required to perform this type of value-added simulations, the latest test and measurement and modeling techniques must be employed (Fig. 2). Power-amplifier (PA) test techniques are commonly employed to evaluate GaN transistors, as GaN technology lends itself particularly well to this application.

“Accurate active device models form the heart of basic power-amplifier design,” notes Timothy Boles, technology fellow/director of strategy for Macom. “These models need to go further than just predicting IV characteristics and output-power levels. In order to optimize overall amplifier efficiency, the models need to provide simulations that accurately reflect the active-device input and output impedances.”

Given the high-end performance of GaN HEMT technology, even minute inaccuracies in a design can significantly impact the overall performance of the final devices. Boles continues, “This enables the optimal matching topology to be presented for the amplifier input/output and—most importantly—all interstage terminals in order to achieve the lowest loss with the minimum number of transformation sections.”

Model-extraction details

GaN, as a material, is a class-III/V, compound wide-bandgap semiconductor with many incredible mechanical and electrical properties. GaN technology has been employed in applications ranging from violet laser diodes to high-frequency and high-power electronics. Thanks to GaN’s rugged mechanical properties, GaN transistors can operate in extreme environments as solar cells in satellites and in other hazardous aerospace applications. GaN also is a mechanically strong material. As a result, it can be deposited as a thin film on silicon, silicon carbide, GaN, sapphire, diamond, and even more insulating materials.

Many aspects of GaN technology are exciting for researchers, commercial companies, and military technologists alike. The challenge lies in effectively developing simulation models that are precise enough to unleash those aspects. Several techniques have been adapted from the last generation of high-performance RF substrates. In addition, several new techniques are emerging to bridge the gap.

“One high-performance technique used to characterize GaN transistors is based on the pulsed-IV characterization system,” explains Dr. Christophe Charbonniaud, deputy director and compact modeling leader for AMCAD. “The pulsed-IV method was originally developed for gallium-arsenide (GaAs) devices and has been extended to GaN devices. Principal evolutions are concerned with the increase of voltage, current swing, and commutation times when dealing with high voltages and currents.”

Using pulsed-IV characterizations demands the generation of certain measurements, which could enable the extraction of a transistor’s isodynamic equations. The trapping characteristics and thermal parameters that function as a product of the quiescent bias conditions are isodynamic parameters. If those parameters are characterized properly, they will enable the modeling of GaN transistors closer to real-world bias conditions and behavior.

For example, thermal failure based on electron trapping issues is a significant concern. It also is difficult to model using DC IV testing. As a result, pulsed-based measurements have become invaluable for accurate model development.

Beyond IV characteristics, pulsed-based S-parameter measurements have recently been used to further detail the self-heating effects and thermal dependencies within these models. The process for deriving the pulsed S-parameters and IV characteristics involves the development of a device’s early model data, which is manipulated to include the device’s scale. These parameters are then analyzed based on their accuracy toward physically measured device parameters. Fits and adjustments are performed on these parameters as a function of iterative testing.

Note that the devices commonly used in PAs have very low-impedance input/output parameters as a result of the very large device size. The parameters are therefore extrapolated to account for the increase in size.  “Either of these modeling techniques needs to be validated by designing, producing, measuring, and iterating against actual device performance,” adds Boles. “Adding a thermal node to these models is more complex. It requires a set of equations that describe the basic material properties as a function of temperature and the resulting effect on the device terminal characteristics.”

Thermal limitations are a significant barrier to optimal GaN PA-device development. As a result, a significant amount of design effort is invested in ensuring desired thermal performance prior to fabricating runs on these devices. Boles states, “Thermal management and modeling are key to producing high-reliability/high-efficiency power amplifiers.”

This process reveals two areas where the detail of the data would limit the ability to effectively simulate device behavior: the quality of the measurement information on the device parameters and the model error that scales with the need to extrapolate on the fitted model data. In response to these issues, many companies have developed tools that enable a single-environment system.

Among those companies are National Instruments/AWR, Keysight, AMCAD, and Maury Microwave. Their single-environment systems promise to bridge the gap between test and measurement and model development. Aspects of these software suites aid in taking fundamental test data and creating models within a single environment. As a result, it is possible to more accurately process the evaluated device behavior. In addition, iterative testing can be repeated in a manner that far exceeds the accuracy of a human-controlled process.

Download this article in .PDF format
This file type includes high resolution graphics and schematics when applicable.