Deep Learning-Based ASM-HEMT I-V Parameter Extraction

TL;DR: For the 12 model parameters that dominate the 40-nm GaN HEMT RF characterization, the combined root-mean-squared (RMS) error of 2.5% between the ANN prediction and the training set is acceptable for most design tasks.

TL;DR: Compact model parameter extraction via derivative-free optimization efficiently extracts tens of parameters from complex device models, streamlining the process and reducing the time required for model fitting.

Abstract: ABSTRACT On the basis of the canonical section‐wise piecewise linear (CSWPL) technique, this work presents the novel modified drain and gate current models for advanced spice model (ASM) that take temperature effects into account. Compared with the classic ASM current model, the proposed approach incorporates an error‐correction module based on the CSWPL technique, providing enhanced prediction accuracy. In addition, temperature dependence is explicitly modeled through a polynomial function, enabling reliable performance across a wide temperature range. To experimentally validate the new model, multi‐temperature current measurement data obtained from a 0.25 × 440 μm 2 gallium‐nitride (GaN) high‐electron‐mobility transistor (HEMT) are used. The achieved results demonstrate that the modified model significantly improves current prediction accuracy compared to the classic current model.

TL;DR: In this paper , a deep learning technique was used to extract the set of Berkeley short-channel IGFET model-common multigate (BSIM-CMG) compact model parameters directly from experimental capacitance-voltage measurements.

TL;DR: In this article, a complete solution from parameter extraction to large-signal electrothermal model generation for gallium nitride (GaN) HEMTs is presented with the consideration of trap deduced gate and drain lag effects.

TL;DR: A comparative study on the application of the ANNs for modeling the scattering parameters of a variety of FET technologies versus bias point, ambient temperature, and geometrical dimensions is presented.

TL;DR: In this paper, an artificial neural network (ANN)-based large-signal model (LSM) of AlGaN/GaN high electron mobility transistors (HEMTs) with accurate buffer-related trapping effects characterization and modeling is proposed.

TL;DR: A new, frequency-domain, behavioral modeling methodology for gallium nitride (GaN) high-electron-mobility transistors (HEMTs), based on the Bayesian inference theory, is presented, with results showing that the proposed approach demonstrates improved accuracy, while at the same time, alleviating the well-known ANN overfitting issue.