W7020E PathWave IC-CAP DynaFET Model Generation

The W7020E PathWave IC-CAP DynaFET Model Generation includes:

• One-stop package includes advanced DC and S-parameters measurement control, neural network training, and verification
• Works in conjunction with the NVNA S94522B Arbitrary Load Control software
• Models dynamic self-heating and trapping effects of GaN devices
• Improved broadband performance versus frequency across the entire device operating range versus NeuroFET
• Improved prediction of large-signal nonlinearities (distortion, load-pull, PAE) and long-term memory effects versus NeuroFET
• PathWave Advanced Design System (ADS) and other simulator support via Verilog-A

The Keysight DynaFET model represents another significant innovation in device modeling. It models dynamic self-heating and trapping effects attributed to critical III-V FET dynamic phenomena such as drain lag, current collapse, and gate lag.

DynaFET uses a novel model identification procedure to identify (extract) trap states and junction temperature from large-signal waveform data and derive the detailed coupling of these dynamical variables to the device's electrical behavior. Like NeuroFET, DynaFET also uses Artificial Neural Networks (ANNs) to describe device charges and currents, but its formulation includes device temperature and charge trapping states as additional inputs to the neural networks. A sophisticated machine learning algorithm trains the networks, using DC, S-parameter, and large-signal waveform measurement data at different ambient temperatures. Nonlinear waveform data are measured with the Keysight Nonlinear Vector Analyzer (NVNA) at a low fundamental frequency (100 MHz) for different temperatures, DC operating points, power levels, and load impedances.

The DynaFET model applies to high-efficiency power amplifiers, radar applications, and low noise amplifiers of GaN and GaAs FETs / HEMTs. It does not cover switching or passive mixer applications.