Uncertainty quantification in compact GaN-HEMT transistor models

The shift to 6G demands the use and integration of advanced technology such as GaN-on-Si as it stands out for its high efficiency, power density, and performance in both high-frequency and switching applications. To improve the development of technology, it is crucial to be able to determine the physics-based parameters to better understand the high frequency, nonlinear, thermal, and trapping effects as a function of the transistor’s architecture.

The goal is to extend the physics-based compact modeling techniques, based on e.g. the ASM-HEMT or the Angelov model, using statistics-based analysis methods. This enables the extraction of not only the parameter values of the transistor model, but also the uncertainties and correlations/dependences of the various parameters onto each other. Technology engineers can use the knowledge of these uncertainties/dependencies to better understand the importance of the extracted parameters.

The aim of the PhD is therefore to

• extend the classical HEMT GaN modeling for both RF and switching applications with statistical-based system identification techniques to enable the computation of the uncertainties on the parameters of the extracted models,
• implement the models within Verilog-A to enable the uncertainty quantification using commercially available tools such as ADS or Cadence,
• perform TCAD simulations of GaN-on-Si transistors to determine the validity of the models in the absence of measurement errors and noise,
• design the optimal experiments and measurements to determine the model parameters in a more efficient way, and
• validate the technique using measurements of GaN devices (DC, linear S-parameter, pulsed-measurements, nonlinear load-pull measurements, modulated load-pull measurements).

The results of this PhD will enable the circuit designers to determine the variability on the key transistor design parameters and will provide feedback to GaN technology’s engineers.