Parameter rankings at one bias point need not describe the full current response of a p-GaN gate high-electron-mobility transistor (HEMT). We analyze 27 combinations of AlGaN barrier thickness (t B), Al mole fraction (x B) and nominal Mg concentration (N Mg) using two-dimensional technology computer-aided design (TCAD) simulations in Sentaurus. The balanced three-level design separates deterministic main and interaction effects by bias and current scale under a common transient protocol. At gate voltage V G = 6 V and drain voltage V D = 40 V, x B contributes 55.5% of the linear drain-current (I D) variation and t B contributes 36.5%. Their interaction share falls from 6.39% on linear I D to 2.49% on log10 I D. Along the transfer path at V D = 10 V, t B leads the logarithmic drain-current decomposition over the sampled V G interval of 2.05-5.40 V. N Mg leads the total gate-current (I G) decomposition at 6 V on both scales and the complete-grid electron-plus-hole gate-current decomposition. Replacing the AlGaN ionization relation preserves the leading Mg term for logarithmic electron-plus-hole gate current over 2-6 V, but changes drain-current rankings and thresholds. Below-range threshold classification and reference-centered local responses complement the complete-grid analysis. These results show why parameter priorities must be tied to the bias, response scale and model parameterization used for a device comparison.
ABSTRACT In this paper, a 10 × 10 broadband planar antenna array is proposed for ground control stations of UAV systems operating across the C‐band. The planar array consists of 10 vertically polarized linear antenna arrays, each comprising 10 radiating dipoles. The dipoles in each linear array are excited through a broadband multistage feeding network. Unlike conventional binary 2 N ‐way feeding networks, the proposed feeding network is specifically tailored to provide 10 output excitations with a synthesized power distribution to reduce the side‐lobe level (SLL) of the antenna array across the desired frequency band. The 10 linear antenna arrays are, in turn, excited using a horizontal replica of the same 10‐way feeding network employed in the linear array configuration. The application of the multistage design technique enables broadband operation over the frequency range from 4.8 GHz to 6.4 GHz, corresponding to a fractional bandwidth of 29% with a VSWR ≤ 2. Furthermore, the stability of the synthesized power division ratios across the entire operating band maintains a low side‐lobe radiation pattern, making the proposed antenna array an excellent candidate for UAV ground control stations employing frequency hopping within the C‐band.
ABSTRACT Researchers have focused their attention toward meminductors after the development of memristors. Due to the difficulties of obtaining such elements as discrete components in the market, researchers intend to mimic them. This article introduces a novel Second‐Generation Current Conveyor (CCII) based grounded meminductor emulator. It consists of three CCIIs, two MOS capacitors, one NMOS, and two grounded resistors. The proposed circuit avoids analog multiplier circuits. The suggested meminductor can maintain pinched hysteresis curve between flux and current effectively within a 90 MHz frequency. The designed meminductor emulator circuit is capable of demonstrating both hard‐ and smooth‐switching behaviors without altering the circuit architecture. The functionalities are validated utilizing SPICE in conjunction with the TSMC 180 nm technology parameter. The power consumption of the proposed meminductor is 1.27 mW. Experimental validation is also performed using IC AD844 and IC HCF4007UBE. Further, the impact of nonidealities on the designed meminductor emulator is examined. Additionally, other critical attributes, including nonvolatility testing, temperature variations, supply voltage fluctuations, transistor size variations, and process corner simulations, have also been demonstrated. A comparison of the proposed meminductor with existing counterparts in the literature has been made, focusing on various performance parameters. Moreover, a meminductor‐based chaotic oscillator is designed as an application of the proposed meminductor.
ABSTRACT This work presents a comprehensive TCAD‐to‐RF modeling framework for composition‐engineered multifin AlGaN/GaN HEMTs incorporating fin‐dependent aluminum composition. The proposed device employs asymmetric barrier engineering (AlGaN with across three fins) to simultaneously tailor the threshold voltage and improve linearity while maintaining adequate RF performance. To accurately extract the intrinsic small‐signal parameters, a numerically stable Berroth–Bosch‐based extraction methodology is developed by integrating LOESS‐assisted smoothing, robust mid‐band slope estimation, and physically constrained parameter estimation within the de‐embedding procedure. Unlike conventional extraction techniques, which are susceptible to numerical oscillations and nonphysical parameter values, the proposed methodology consistently yields stable intrinsic parameters enabling reliable evaluation of the RF figures of merit and . Furthermore, robustness against realistic measurement uncertainty is demonstrated through standard‐deviation Gaussian perturbation of the TCAD‐generated S ‐parameters, where the proposed methodology maintains physically meaningful parameter extraction while conventional methods exhibit significant numerical instability. The extracted parameter set represents an equivalent compact small‐signal model of the complete multifin device without increasing circuit complexity as the number of fins increases. The proposed framework provides a reproducible and numerically robust foundation for compact modeling and RF characterization of composition‐engineered multifin GaN HEMTs for high‐frequency integrated circuit applications.
ABSTRACT Microwave radiation detection technology is extensively utilized for identifying targets across diverse terrains, with significant applications in remote sensing and security inspection. However, existing microwave radiometer measurement systems struggle to differentiate subtle brightness temperature variations across similar ground objects (e.g., turf or cement with varying roughness) that exhibit extremely close physical emissivities. To address this gap, this study proposes a feature‐enhanced inversion algorithm based on a fully polarized passive microwave radiometer. Six selected features from and are combined with the mean, maximum and standard deviation of each of and , giving 12 model inputs. Furthermore, a hybrid neural network combining Particle Swarm Optimization and Levenberg–Marquardt Backpropagation (PSO‐LM‐BP) integrates global initialization and local refinement to map these polarimetric features to reference brightness temperature. Comparative analyses employing multiple feature sets and optimization algorithms confirm the superior inversion accuracy of the proposed method, achieving a mean squared prediction error of 0.3243 K 2 relative to the reference labels under a record‐level split. In conclusion, this work provides an effective algorithmic framework and feature extraction approach for high‐precision microwave target identification under complex ground object backgrounds.
