ABSTRACT This paper presents a compact 3 dB branch‐line coupler leveraging pixelated layouts and the binary particle swarm optimization (BPSO) algorithm. Conventional branch‐line couplers are constrained by quarter‐wavelength transmission line structures, which result in bulky circuit footprints and pose tradeoffs between operational bandwidth and coupling balance. To tackle these limitations, the critical regions of the branch‐line coupler are discretized via pixelation, where microstrip conductors are partitioned into discrete grid units parameterized by binary 0–1 matrices. The BPSO algorithm is adopted to globally optimize circuit topologies and electromagnetic responses, enabling impedance matching optimization under fixed dimensional constraints. A pixelated 3 dB branch‐line coupler operating across 1.9–2.1 GHz is designed, simulated, and physically fabricated. Measured results reveal that the proposed coupler achieves favorable 3 dB power splitting performance within the target frequency band, featuring input return loss superior to 16 dB and output phase deviation near 90°. Moreover, the overall circuit size is substantially reduced by 26.5% compared with traditional branch‐line coupler implementations.
ABSTRACT Platform structures can significantly alter antenna radiation characteristics and input impedance, limiting the accuracy of conventional free space receiving models. This paper proposes a hybrid numerical method for rapid prediction of the transient coupling response of a platform mounted GPS antenna under ultra‐wideband high‐power microwave (UWB‐HPM) excitation. Broadband antenna characteristics obtained from a one‐time full‐wave simulation are incorporated into a reciprocity based receiving model to calculate the voltage response across a 50 Ω load. Validation using a GPS microstrip patch antenna mounted on a metallic platform shows a peak voltage error of 3.9% and an approximately 4.3 fold improvement in computational efficiency compared with direct full‐wave simulation. The extracted antenna parameters can be reused to rapidly evaluate different incident directions, polarization states, and pulse widths. A 200 case Latin hypercube study demonstrates the capability of the proposed method for large‐sample analysis under uncertain incident conditions. The peak voltage exhibits a right skewed distribution, with median and 99th percentile values of 38.9 and 188.6 V, respectively. The proposed method enables efficient prediction of transient coupling responses over a wide range of incident conditions.
ABSTRACT This paper introduces a fast numerical method for computing magnetostatic fields in cylindrical structures with arbitrary and nonhomogeneous cross‐sections. The efficiency of the approach results from the combination of curvilinear coordinate systems, specifically adapted to each interface, with a pseudo‐spectral Fourier method used to enforce boundary conditions analytically and to couple multiple subdomains in a S‐matrix formalism. Unlike modal techniques, which are restricted to piecewise homogeneous layers, the proposed formulation is able to handle nonhomogeneous media within each layer without increasing the size of the global system. This is achieved through a domain decomposition strategy based on a scattering‐matrix (S‐matrix) formalism, which allows multilayered configurations to be treated recursively while avoiding the construction and inversion of large monolithic matrices. A key contribution of this work is therefore the generalization of the S‐matrix algorithm to nonhomogeneous media. In contrast to previous developments focused on eddy‐current problems, the present formulation is extended here to the magnetostatic regime for the first time, with the objective of addressing magnetic flux leakage (MFL) field computations (in the linear case). The resulting numerical model is validated through numerical experiments and compared with finite element simulations, demonstrating both accuracy and computational efficiency.
ABSTRACT Loss minimization in interior permanent magnet synchronous motors (IPMSMs) leads to a nonlinear fourth‐order equation when both copper and core losses are simultaneously taken into account. This paper presents a novel, noniterative numerical method based on a sigmoid function transformation to solve this equation online, eliminating the need for model simplification and avoiding conventional iterative optimization procedures. The proposed approach identifies the desired root with accuracy comparable to that of the Newton–Raphson method while offering improved numerical robustness and reliable convergence. Extensive MATLAB/Simulink simulation results under diverse operating conditions validate the stable dynamic operation and superior loss minimization performance of the proposed strategy over conventional control schemes.
ABSTRACT Wave Digital (WD) models are a well‐established framework for the emulation of linear and nonlinear circuits. In practice, however, WD models are still implemented largely manually or through limited automated toolchains. Manual programming of WD models is time‐consuming and error‐prone for larger systems with pronounced algebraic dependencies, and existing frameworks lack a unified semantic representation from which parameterization, evaluation, and executable code can be derived and generated automatically in a backend‐agnostic manner. The aim of this work is to enable fully automatic generation of executable WD models from a declarative specification, by introducing a semantic framework that makes parameterization and timestep‐level evaluation dependencies explicit and directly compilable. We introduce a typed declarative graph as a modeling interface, specifying topology, parameters, and algebraic dependencies. From this graph, each element is mapped to two complementary local Petri nets: one representing parameter dependencies for parameterization, and one representing signal dependencies for timestep‐level evaluation. These nets are composed into global Petri nets that serve as a semantic intermediate representation whose firing semantics directly induce admissible evaluation orders and can be lowered systematically into backend‐specific executable code. The Petri net representation naturally enables future use of established analysis techniques, dynamic execution, model checking, and the development of loop‐breaking heuristics through systematic Petri net exploration.
