Genetic algorithm initialized artificial neural network based temperature dependent small‐signal modeling technique for GaN high electron mobility transistors

Abstract

This paper explores and develops efficient temperature‐dependent small‐signal modeling approaches for GaN high electron mobility transistors (HEMTs). The multilayer perceptron (MLP) architecture and cascaded MLP architecture of artificial neural network are employed to model temperature dependence of 2‐mm GaN‐on‐silicon device. It is identified that both architectures face problem of dependence on initials values of weights and biases. To overcome this issue, the genetic algorithm (GA) is incorporated in both MLP and cascaded MLP architectures. The models are trained on a large set of operating conditions (bias voltages and ambient temperatures) over a frequency range of 0.1 to 20 GHz and then tested for both temperature interpolation and extrapolation cases to assess their accuracy and robustness. An excellent agreement between the measured and the modeled S‐parameters over the entire frequency range demonstrate the quality and robustness of the proposed technique. It is also shown that the cascaded MLP with GA exhibits better performance but with increased complexity.

Demonstration of CAD Deployability for GPR Based Small-Signal Modelling of GaN HEMT

“…The solution of this problem could come in many folds. One of the solutions could be the incorporation of the global optimization techniques in the standard BP based algorithms [15]. The report discussed genetic algorithm (GA) initialized ANN based models for MLP and Cascaded MLP architectures.…”

“…This can therefore facilitate the easy implementation of the device models in circuit design tools and the PAs can be produced on mass-scale easily. Therefore, recently a number of papers have been reported to describe the smallsignal behavior of GaN HEMT using Artificial Neural Network [11]- [14], global optimization-oriented ANN [15] and Particle swarm optimization (PSO)-based Support Vector Regression (SVR) [16]- [17]. But due to parametric nature of ANN and SVR algorithms, the performance of developed models heavily depend upon the nature of the dataset.…”

Equivalent‐circuit extraction for gallium nitride electron devices: Direct versus optimization‐empowered approaches

“…Over the last half century or so, many efforts have been devoted to the equivalent‐circuit extraction for modeling the scattering S‐parameter measurements of active electron devices 1–21 . Although the frequency‐dependent behavior of the scattering parameters can be easily and quickly reproduced by using alternative representations (e.g., artificial neural networks 22–26 ), the calculation of a small‐signal equivalent‐circuit model is essential as this circuit can be utilized as a foundation for developing large‐signal 27–32 and noise 33–38 models. Typically for the field‐effect transistors (FETs), the challenging task of extracting the small‐signal equivalent‐circuit model is accomplished by using direct extraction techniques based on the well‐known “cold” approach 4,5,10,12,13 .…”

Comprehensive Investigation and Comparative Analysis of Machine Learning-Based Small-Signal Modelling Techniques for GaN HEMTs

“…furthermore, selection of suitable topology of ECMs, physically inconsistent values for complex topologies are major concerns for the designers [9], [10], [23]. Therefore, the alternative machine learning (ML) based techniques to develop SSMs are getting traction [24]- [28], [30]- [32], [34], [35]. These techniques have shown promise as they can emulate complex behaviors, manifest better prediction capability and generally are computationally efficient, nevertheless, requisite large sample size of the measurements [23], [24].…”