What are you looking for?
RF Compound Semiconductors Modeling
A major challenge with III-V semiconductor modeling has been the difficulty to accurately describe effects such as trapping and non-linear thermal memory effects in terms of analytical equations. Unlike the silicon device modeling world where equations in compact models are mostly derived from the semiconductor physics, industry standard compound semiconductor models are mostly based on empirical equations (e.g. EEFET, EEHEMT, Curtice, Angelov).
Because of the challenge of using compact models to achieve accurate results over the entire range of bias conditions, measurement based models have been successfully used for the past 20 years. The first industry standard measurement based model was the Keysight Root model and was introduced by Dr. David E. Root more than 20 years ago. It constructed nonlinear model I-V and Q-V constitutive relations from mathematical transformations of DC and S-parameter data acquired adaptively over the entire device operating range. This data was stored in tables and interpolated, dynamically, by the model during simulation. More recently, Keysight has introduced the NeuroFET model. This model includes an improved data acquisition procedure for DC and S-parameters to minimize device degradation and uses artificial neural networks (ANNs) to compute current and charge constitutive relations, resulting in more uniform accuracy and better distortion simulations.
A fundamental irony of today’s active device modeling is that models are extracted using DC and linear S-parameter measurements while most applications are inherently non-linear, such as power amplifiers, mixers and switches. When electrical nonlinearities are further complicated by the presence of significant memory effects, caused by dynamic self-heating and trapping effects; for example, it becomes extremely difficult to predict nonlinear behavior from static and linear data. This is especially true for transistors fabricated in new material systems like gallium nitride (GaN). Fortunately, breakthrough instrumentation technology is now available to help. Keysight has recently released a Nonlinear Vector Network Analyzer (NVNA), a special version of the PNA-X series of vector network analyzers. The NVNA can stimulate the device under nonlinear conditions at various loads and measure phase and amplitudes of incident and reflected waves, including harmonics. A new modeling technology is currently being evaluated and will soon be available to extract empirical model parameters and even generate models directly from nonlinear measurements. Nonlinear model validation is a significant additional benefit of this approach.
For over two decades, Keysight IC-CAP has been the software of choice for RF device modeling. It supports all the major industry standard models (EEFET, EEHEMT, Angelov-GaN) as well as Keysight proprietary models like Keysight Root, Keysight HBT and Keysight NeuroFET. IC-CAP links to Keysight Advanced Design System (ADS), the industry standard simulator for RF circuit design, and provides a flexible environment that it is the key for successful RF modeling (e.g. custom de-embedding and extraction).
