Artificial neural network (ANN)-based surrogate models have been widely used to reduce the computational cost of electromagnetic (EM) optimization for microwave filters. However, repeated surrogate reconstruction in iterative optimization may still introduce considerable overhead as the design point moves. To address this issue, an adaptive neuro-coupling matrix (neuro-CM)-based surrogate optimization framework is developed in this article. In the proposed framework, coupling matrix (CM)-related intermediate variables are extracted from full-wave EM responses, predicted by ANNs, and then used for physics-based response reconstruction. The neuro-CM surrogate model is then embedded into a trust-region-based optimization procedure with full-wave EM validation. To further improve the efficiency of multiround surrogate reconstruction, a CM-related sensitivity-guided modeling region update strategy is introduced. Specifically, the design-error sensitivity is transferred from the intermediate-variable space to the geometrical-parameter space, and the modeling radii are allocated anisotropically according to parameter activity. In this way, subsequent surrogate reconstruction can focus on the parameter directions that are more relevant to performance improvement. Two microwave filter examples demonstrate that the proposed method reduces the number of surrogate reconstruction rounds and the total optimization time compared with the considered baseline methods.

In modern MIMO radar systems, the length of the transmit waveform directly affects both energy consumption and vulnerability to interception. Short-duration waveforms are highly desirable for power-constrained platforms, as they reduce per-pulse energy consumption and extend battery life; moreover, shorter waveforms inherently offer improved low-probability-of-intercept (LPI) performance by reducing the radar’s on-air exposure time. In this paper, we consider the joint design of transmit waveform with minimum- length and receive filters for a collocated multiple- input multiple- output (MIMO) radar system under the existence of signal- dependent interference sources. The problem of waveform design is formulated as minimization of the actual length of transmit waveform under the signal-to-interference-plus noise ratio (SINR) and constant modulus constraints. To the best of our knowledge, this problem is barely investigated in the current literature. The length of transmit waveform is a formal variable and cannot be optimized directly. In the light of this, we introduced a code-choosing vector with several constraints to make it solvable. The formulated problem possesses complicated fractional constraint, and we introduce several auxiliary variables to simplify it. Then we utilize the alternating direction method of multipliers (ADMM) framework to solve it iteratively. Numerical examples demonstrate the effectiveness of our model and proposed algorithm to design minimum- length waveform under the SINR and constant modulus requirement.

For microwave filters with non-resonant nodes, while routine network synthesis-based methods often require significant design expertise and a complicated design process, existing machine learning-assisted optimization methods often cannot handle filters with non-resonant nodes. Hence, there is a demand for an effective and easy-to-use design approach for filters with non-resonant nodes. To address this challenge, a hybrid approach utilizing filter design knowledge and a machine learning-assisted evolutionary algorithm is proposed. The key innovation is a systematic design procedure combining the advantages of both kinds of methods. Its effectiveness and efficiency are verified by a 3-2 and 6-2 waveguide filters with a single dangling node, and a third-order waveguide filter with multiple dangling nodes. Moreover, due to the avoidance of equivalent circuit models, the proposed hybrid approach is fully programmable.

This study proposes an equivalent circuit-fitting-based neural transfer function modeling method (ECFB neuro-TF) to enhance the robustness, interpretability, and capability of handling nonideal factors in parametric modeling of coupled resonator microwave bandpass filters. Unlike previous approaches that rely on vector fitting (VF), this method introduces physical constraints through equivalent circuit (EC) mapping. Compared with the previous systematic neuro-transfer function method based on a compact embedded format (SCEF neuro-TF), this work includes the following advances: First, constraining pole/residue extraction via predefined circuit topology, effectively overcoming VF’s inherent issues of order inconsistency and parameter discontinuity. Second, establishing a complex exponential phase shift function through theoretical and experimental validation, providing accurate quantification of port effects. Furthermore, demonstrating the pole uniqueness and parameter continuity under perturbation from filter synthesis theory. The developed neuro-TF model incorporates these circuit-based features while employing neural networks to correct residual nonideal effects. Experimental results demonstrate that our method achieves superior accuracy over broader parameter ranges compared to both VF-based and multilayer perceptron (MLP) approaches, offering an efficient and reliable solution for complex electromagnetic (EM) modeling scenarios.

Microwave filters play an essential role in wireless communication systems. Manufacturing errors are inevitable during the fabrication of microwave filters, and reducing their impact on filter design is of crucial importance. This problem is typically defined as a yield optimization problem. Thus, the question emerges: How can the yield optimization algorithm enhance the robust design of microwave filters? This article will provide an answer to this question.

This paper presents a WR-2.2 band (330-500 GHz) waveguide with E-plane bends and a waveguide bandpass filter. A micro-metal additive manufacturing technology employs copper to process these devices. This technology includes several micromachining processes, which can produce integrated, miniaturized, and complex three-dimensional metal microstructures when meeting the process constraints. To successfully apply this technology to the fabrication of terahertz waveguide devices, many design details are considered. First, the E-plane bent transition structures in the waveguide overcome size limitations imposed by the manufacturing process, allowing for accurate measurement using standard waveguide interfaces. Second, the rectangular resonators compose the fifth-order waveguide bandpass filter spanning 20 GHz with a center frequency of 380 GHz. Given that the excess photoresist needs to be removed, some release holes are designed in the proper positions of the filter. Third, to increase the uniformity of the device layout and improve the machining quality in the electroforming process, a lot of rectangular holes are arranged around the device. By adopting a meticulous and novel design method in accordance with the process advantages and limitations, the fabrication accuracy of the device size is greatly increased. The integrated filter with exceptional performance is obtained. The measured average insertion loss within the passband is better than 1.2 dB, and the return loss exceeds 15 dB. Additionally, the filter demonstrates a center frequency offset of approximately 0.5%. Good consistency between simulation and measurement shows the validity of design and fabrication.

Abstract Three‐dimensional (3‐D) printing technology has become a promising manufacturing method to fabricate microwave components in recent years. This article focuses on waveguide components fabricated using 3‐D printing technology. The development of passive waveguide devices based on different 3‐D printing technologies is introduced and compared. Depending on the material and fabrication techniques, two fabrication modes are normally adopted for waveguide components. Featured waveguide components based on these two modes are illustrated with examples. Then, state‐of‐the‐art waveguide filters based on spherical resonators (including single‐mode and dual‐mode cavity resonators) are introduced. These filters are suitable for 3‐D printing fabrication techniques, demonstrating superior performance over filters based on conventional waveguide filters.

Dataflow architectures can achieve much better performance and higher efficiency than general-purpose core, approaching the performance of a specialized design while retaining programmability. However, advanced application scenarios place higher demands on the hardware in terms of cross-domain and multi-batch processing. In this article, we propose a unified scale-vector architecture that can work in multiple modes and adapt to diverse algorithms and requirements efficiently. First, a novel reconfigurable interconnection structure is proposed, which can organize execution units into different cluster typologies as a way to accommodate different data-level parallelism. Second, we decouple threads within each DFG node into consecutive pipeline stages and provide architectural support. By time-multiplexing during these stages, dataflow hardware can achieve much higher utilization and performance. In addition, the task-based program model can also exploit multi-level parallelism and deploy applications efficiently. Evaluated in a wide range of benchmarks, including digital signal processing algorithms, CNNs, and scientific computing algorithms, our design attains up to 11.95× energy efficiency (performance-per-watt) improvement over GPU (V100), and 2.01× energy efficiency improvement over state-of-the-art dataflow architectures.

This paper presents a WR-2.2 band waveguide orthogonal mode transducer (OMT) for an ice cloud remote sensing radiometer. A holistic approach is proposed to the design and manufacturing. Given the difficulties of using computer numerically controlled (CNC) milling techniques to fabricate the OMT and its assembly, both the design and fabrication processes are optimized to reduce the possibility of error, which could affect OMT performance. In the design, the OMT utilizes a side-arm structure, with a compact 90° waveguide twist designed in the side-arm path to reduce insertion loss and achieve two symmetric and opposing output ports in the radiometer. Regarding the challenges in CNC milling technology for such high frequencies, a comprehensive sensitivity analysis is performed following the OMT design to identify the design parameters with high sensitivity. Then, an acceptable tolerance range for CNC milling is obtained. To control fabrication uncertainty and minimize tolerance, an effective fabrication refinement strategy is proposed for accessible CNC machining, reducing the range of fabrication errors to − 3 to 2 µm within 3 iterations. Comprehensive measurements and analyses are performed on the fabricated OMT as well as the one with an antenna and a receiver. The measured insertion losses of the H and V copolarization of the OMT are better than 1.4 dB and 2 dB, respectively. The isolation is better than 40 dB. The measured H and V cross-polarizations are better than 25 dB. They demonstrate the OMT with a good performance for radiometer system.

This article presents a waveguide with a pair of ${E}$ -plane bends and two waveguide bandpass filters (BPFs) which operate in WR-3.4 band (220–330 GHz). These devices are fabricated by a micro metal additive manufacturing (M-MAM) technology on copper in one piece. Both filters are fifth order and operate at a center frequency of 280 GHz with a bandwidth of 8 GHz. Cylindrical resonators are employed for filter design. The standard feed waveguides are designed in bend structures, facilitating the measurements. The M-MAM technology is a layer-by-layer additive manufacturing process constructing the waveguide devices in five layers. Since the layout of the devices has a strong impact on the electrostatic field distributions and hence the electroforming quality, both full-wave electromagnetic (EM) simulations and electrostatic analysis are simultaneously considered in the device design. It eases the subsequent electroforming and polishing processes, effectively improving the performance of the devices. The insertion losses and return losses of the measured filters are lower than 1.65 dB and better than 15 dB. The center frequency shifts are less than 0.3%. The excellent results attribute to the delicate co-design method between EM and electrostatic analyses encountered both in the filter design and in the additive manufacturing process.

Surrogate models are widely used in filter yield optimization methods to improve efficiency, which can be divided into online and offline. State-of-the-art offline surrogate model-based filter yield optimization methods are shown to be effective for filter cases with more than ten sensitive design variables. In these methods, a keystone is the appropriate definition of the space for building the surrogate model, deciding success/failure, or at least the efficiency of the yield optimization. However, there is a lack of systematic methods to achieve it. To address this challenge, a new method, called p attern s earch o ptimization-based surrogate m odeling s pace d efinition method (PSOMSD), is proposed. The performance of PSOMSD is demonstrated by a real-world filter case with 14 sensitive design variables. Analysis shows the appropriateness of the defined surrogate modeling space and advantages compared to empirical methods.

This article presents a wideband coaxial D-band frequency tripler fabricated by copper additive manufacturing (AM). The central part of the tripler consists of two waveguide–coaxial transitions, a coaxial low-pass filter, an antiparallel Schottky diode chip, and an output impedance matching network. Most constitutions, except for the diode chip, are monolithically fabricated in one piece. The diode chip with additional thermal paths is specially designed to obtain a good thermal performance and better power handling. To facilitate the measurement, a waveguide fixture is also designed to convert the ports to standard waveguide ports (WR-19 and WR-6.5). The measurements show that the maximum output power of the tripler is 9.44 mW with about 182 mW of input power at 147 GHz and its 3-dB bandwidth is 45.2% from 108 to 171 GHz. Additionally, the tripler delivers $>$ 1.87 mW of output power and better than 10 dB of return loss over the 100–180-GHz range. To the best of authors' knowledge, this is the first AM coaxial ultra-wideband frequency multiplier reported in the open literature.

This letter presents the design of a third-order filtering waveguide aperture antenna based on coupled cavity resonators. Three offset-coupled rectangular waveguide cavities are employed in the design realizing two nested loaded-stubs without costing extra structure and size. The loaded-stubs introduce two controllable transmission zeroes and enhance the out-of-band realized gain selectivity. To validate the predicted results, a prototype operating at the X- band frequencies has been fabricated monolithically using the three-dimensional selective laser melting printing technique. The measured results are in very good agreement with the simulated results, showing a flat gain response of 7.0 ± 0.2 dBi from 9.5 to 10.5 GHz with very good out-of-band selectivity. The fractional bandwidth is about 10% at 10 GHz when S 11 = −20 dB. Compared to the previously designed filtering antennas, the proposed design has the advantages of stronger out-of-band gain selectivity and low profile.

In resonator-coupled bandpass filter 3D design, it is a routine that the filter optimization methods are guided/supervised by designers' experience to carry out an iterative design optimization process. To realize automated or unsupervised filter 3D design optimization, a new method, called hybrid surrogate model-assisted evolutionary algorithm for filter optimization (H-SMEAFO), is proposed. H-SMEAFO aims to automatically obtain a highly optimal filter 3D design without designers' interaction (i.e., unsupervised) and is also not restricted to certain kinds of filter structures. In H-SMEAFO, the key innovations include a hybrid response feature-based objective function and a hybrid surrogate model-assisted global optimization algorithm; both are designed bespoke for filter design landscape characteristics. The performance of H-SMEAFO is demonstrated by an 8th-order dual-band waveguide filter with four transmission zeros and a 6th-order waveguide filter with two transmission zeros, for which, unsupervised design optimization does not appear to be possible using existing methods. Numerical results show the effectiveness and advantages of H-SMEAFO.

This article presents a D-band waveguide diplexer fabricated using micro metal additive manufacturing (M-MAM) technology. The diplexer features two sixth-order filters with passbands of 130–134 GHz and 141–148.5 GHz, respectively, for a crucial RF application band. Cylindrical resonators were employed in the filter design, and a nine-layer process was developed specifically for this frequency band, allowing the diplexer to be manufactured as a single piece. The process was optimized to minimize photoresist deformation during repeated thermal cycles, so a larger waveguide can be fabricated. To ensure accurate measurements, a fixture equipped with a straight-through calibration channel was also designed and manufactured. The measured insertion losses and return losses of both channels are less than 2 dB and greater than 15 dB, respectively, and the center frequency shifts are negligible at less than 0.15%. The excellent correlation between simulations and measurements is a result of the carefully designed process and its high level of accuracy.

We report a $D$ -band waveguide diplexer, with two passbands of 130–134 and 151.5–155.5 GHz, fabricated using micro laser sintering (MLS) additive manufacturing with stainless steel. This is the first demonstration of metal 3-D printing technology for multiport filtering devices at a sub-terahertz (THz) frequency. For comparison, the same diplexer design has also been implemented using computer numerical controlled (CNC) milling. The diplexer, designed using coupling matrix theory, employs an all-resonator and $E$ -plane split-block structure. The two channels are folded for compactness. A staircase coupled structure is used in one channel to increase the isolation performance. The printed waveguide flanges are modified to adapt to the limited printing volume from the MLS. Effects of fabrication tolerance on the diplexer are investigated. An effective and unconventional electroless plating process is developed. The measured average insertion losses of the gold-coated diplexer are 1.31 and 1.37 dB, respectively. The respective frequency shifts from design values are 0.92% and 1.1%, and bandwidth variations are 4% and 15%. From a comprehensive treatment of the end-to-end manufacturing process, the work demonstrates MLS to be a promising fabrication technique for complex waveguide devices at a sub-THz frequency range.

Microwave filters are indispensable passive devices for modern wireless communication systems. Nowadays, electromagnetic (EM) simulation-based design process is a norm for filter designs. Many EM-based design methodologies for microwave filter design have emerged in recent years to achieve efficiency, automation, and customizability. The majority of EM-based design methods exploit low-cost models (i.e., surrogates) in various forms, and artificial intelligence techniques assist the surrogate modeling and optimization processes. Focusing on surrogate-assisted microwave filter designs, this article first analyzes the characteristic of filter design based on different design objective functions. Then, the state-of-the-art filter design methodologies are reviewed, including surrogate modeling (machine learning) methods and advanced optimization algorithms. Three essential techniques in filter designs are included: 1) smart data sampling techniques; 2) advanced surrogate modeling techniques; and 3) advanced optimization methods and frameworks. To achieve success and stability, they have to be tailored or combined together to achieve the specific characteristics of the microwave filters. Finally, new emerging design applications and future trends in the filter design are discussed.

RF/Microwave devices are mission-critical components for communication systems. Using three-dimensional full-wave electromagnetic (EM) simulators to design those devices is considered a must for most microwave/RF engineers today. Since today's desktop workstations outperform supercomputers of years ago, widely available commercial EM simulators solve those practical problems on low-cost desktop computers. At the same time, the communication industry faces increasing design challenges in function and frequency integration, miniaturization, and multiple tight performance requirements for microwave/RF components. This contributes to extreme time costs in EM-based design. The key to efficient design is surrogate-based approaches for now and future. In this paper, recent advances in surrogate related techniques including sampling, modelling, and optimization are reviewed with applications in complex microwave/RF structure design, post-production tuning, and yield design. The focus of these methods is to achieve fast design/tuning/modelling without sacriflcing accuracy,

In this work, an efficient framework is designed to obtain the optimal design solutions with a high yield. In the proposed framework, all objectives are modeled using the Polynomial chaos method. The optimal design with the yield-constrained is obtained using a global optimizer. A four-pole waveguide microwave filter example is used to verify the performance of the proposed method. The proposed framework does not require an initial design. The electromagnetics simulated verification results show that the proposed framework obtains a better design with a higher yield using lower computational cost than a Monte Carlo-based method and a state-of-the-art yield optimization method.’

Inverse time overcurrent protection can be applied to distribution networks with distributed power sources because of its superior protection characteristics, but its time limit coordination and value setting are complicated, which limits its large-scale engineering application. Therefore, this paper proposes an inverse time overcurrent protection setting strategy for distribution network(DN) based on improved gray wolf optimizer (GWO) algorithm. Firstly, the current setting optimization model and the protection operation characteristics are established considering the reliability, rapidity and selectivity. Secondly, the GWO algorithm is improved by introducing the population initialization strategy based on elite backward learning, adaptive weights and variation strategy based on the Cauchy operator for the characteristics of the traditional GWO algorithm which is easy to trap in local optimum and low convergence accuracy. The improved algorithm does not introduce new parameters and achieves a balance between global and local. Finally, the case study results represent that the improved GWO algorithm has high accuracy and stability in both two-phase short-circuit and three-phase short-circuit scenarios, and has good practical applicability.