研究目的
To develop a differential evolution based method for small signal modeling of GaN HEMT to obtain optimized values of intrinsic and extrinsic elements for a compact small signal model, overcoming limitations of previous methods such as local minima and nonphysical values.
研究成果
The DE-based method effectively optimizes both intrinsic and extrinsic elements for a compact small signal model of GaN HEMT, producing physical and accurate results with low error percentages. It outperforms other methods in terms of computational efficiency and accuracy, as shown by comparative studies, and is robust with a small standard deviation over multiple runs.
研究不足
The method is applied to a specific GaN HEMT device at a particular bias condition (Vgs = -2 V, Vds = 20 V) and frequency range (1-30 GHz), which may limit generalizability. It relies on initial parameter extraction that could be affected by measurement inaccuracies, and the optimization may still be computationally intensive compared to some hybrid methods.
1：Experimental Design and Method Selection:
The method uses a differential evolution (DE) algorithm for optimization, incorporating a unique search space exploration strategy to handle nonphysical aspects and small values of circuit elements. It involves mutation, crossover, and selection operations to optimize 16 circuit elements (intrinsic and extrinsic) based on S-parameter data.
2：Sample Selection and Data Sources:
Measured S-parameter data of a 4 × 0.1 × 75 μm2 GaN/SiC HEMT in the frequency range of 1 to 30 GHz at a bias point of Vgs = -2 V and Vds = 20 V is used.
3：1 × 75 μm2 GaN/SiC HEMT in the frequency range of 1 to 30 GHz at a bias point of Vgs = -2 V and Vds = 20 V is used.
List of Experimental Equipment and Materials:
3. List of Experimental Equipment and Materials: Not specified in the paper.
4：Experimental Procedures and Operational Workflow:
Initial estimates of circuit elements are obtained using extraction methods (e.g., open structure and short structure methods). The DE algorithm is applied with a population size of 35, and fitness is evaluated using a weighted function based on S-parameter differences. Steps include mutation, crossover, selection, and convergence checking.
5：Data Analysis Methods:
The fitness function is calculated using Equation (42), which compares measured and calculated S-parameters. Error percentages and convergence profiles are analyzed to validate the model.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
