High-speed link consisting of drivers and interconnects is essential for high-speed data transmission. In this article, a surrogate modeling technique based on graph neural network (GNN) and recurrent neural network (RNN) is proposed for signal integrity (SI) analysis of high-speed links with variable physical parameters and variable topologies. First, GNN extracts global features that can fully characterize components of the high-speed link from their topologies and physical parameters. Second, RNN takes the extracted global features and the excitation waveforms as inputs to predict the response waveforms. Finally, the well-trained GNN-RNN surrogate models of components of the high-speed link are cascaded as the entire surrogate model of the high-speed link. Numerical examples of the dri-ver model, the interconnect model, and the entire high-speed link model are provided for validation. It is shown that the proposed GNN-RNN surrogate models achieve low mean squared errors (MSEs), mean absolute errors (MAEs), and high efficiency.

In this article, a folded ridge half-mode substrate integrated waveguide (FRHMSIW) combined with the novel spoof surface plasmon polaritons (SSPP) is proposed to reduce both the lateral size and the longitudinal size. The SSPP structure consisting of sub-wavelength U-shaped grooves, metal patches and metal blind holes is loaded into the FRHMSIW. Firstly, the SSPP increases the effective permeability and permittivity in the waveguide, which can decrease the cutoff frequency and hence reduce the lateral size. Secondly, the SSPP has slow-wave effect, which can reduce the longitudinal size of the waveguide. Analytical formulas of the effective permeability and permittivity are derived based on anisotropic media theory, which are verified by comparing the calculation results with full wave simulation results. According to the derived analytical formulas, a prototype of the proposed FRHMSIW-SSPP waveguide is designed and fabricated using the PCB process. The measurement results show that the proposed waveguide reduces lateral width by 78.7% and the longitudinal dimension by more than 74% compared with the classical SIW with the same cutoff frequency. Subsequently, the FRHMSIW-SSPP structure is applied to design a 3-dB coupler. Results show that the coupler based on the FRHMSIW-SSPP can reduce the surface area by 92.5% compared with the classical SIW coupler.

In this article, an enhanced artificial neural network (ANN) is proposed for equivalent radiation source reconstruction in the electromagnetic interference (EMI) analysis. The half-space Green's function of the dipole array is taken as the input and the radiation field obtained by near-field scanning is taken as the output of the enhanced ANN. The loss function of the enhanced ANN consists of the magnitude loss and the gradient loss. Compared with the conventional ANN only using the magnitude loss, this article introduces the gradient information of the radiation field into the loss function in the enhanced ANN training, which can accurately analyze EMI problems with higher efficiency. Numerical and measurement examples show that the proposed method can accurately reconstruct the radiation source with the error less than 5%, and the number of the scanning points in the enhanced ANN is reduced to 60% compared with that in the conventional ANN.

Pixel antennas expressed in the form of discrete variable’s structure (DVS) demonstrate immense potential in electromagnetic (EM) performance, but the design process relies on a huge full-wave simulation dataset. To efficiently design pixel antennas with high degrees of freedom (DoFs), in this study, a multi-head discrete action calibration proximal policy optimization (MDAC-PPO) reinforcement learning (RL) method is proposed. Specifically, a sophisticated multi-head discrete mechanism is developed to generate the placement status of all metal patches in a probabilistic form. Furthermore, a coupling calibration mechanism that considers local metal density is proposed to break down the intricate optimization target, expediting the iterative process. Additionally, to delve deeper into potential designs that might otherwise be neglected, an exclusive OR (XOR) operation state update manner is integrated into the PPO agent. Three design examples on a pixel antenna with a 2 ²⁵⁶ -sized solutions space, including optimization targets on impedance matching, gain, efficiency, and related bandwidth, are presented to show the optimization efficiency and performance of the MDAC-PPO method. The measured |S11| of the fabricated prototypes aligns well with simulations, further validating the reliability of the MDAC-PPO method in practical applications. Meanwhile, this method also provides a new perspective for designing other EM devices with DVS.

This article proposes a microstrip-to-microstrip (MS-to-MS) vertical transition with pixel structures and then proposes a knowledge-assisted proximal policy optimization (PPO), which is a reinforcement learning (RL) for the design of this transition. The transition requires fully connected structures and a novel mechanism to generate the pixel structures with fully connected shape is proposed and incorporated into PPO. The proposed method is compared with the particle swarm optimization (PSO) and the genetic algorithm (GA) and demonstrates benefits in improving design efficiency. The designed MS-to-MS transition is fabricated using the PCB process. Measurement results indicate that the designed MS-to-MS vertical transition operates in the band from 3.4 to 14.8 GHz with low insertion loss.

This article proposes a wideband microstrip-to-microstrip vertical transition with multi-layer pixel structures, alongside a multi-branch knowledge-assisted proximal policy optimization (MB-KPPO) method for its automatic design. The proposed transition consists of the three-layer pixel structures with high design degrees of freedom to realize a wide bandwidth. The MB-KPPO adopts a multi-branch policy network instead of a single-branch policy network in the PPO to improve design efficiency. In addition, the MB-KPPO integrates a fully connected shape generation mechanism to incorporate physical requirements. An MS-to-MS vertical multi-layer pixel transition is designed and fabricated by PCB technology. Measurement results show that the multi-layer transition has a frequency range from 3.5 to 17.8 GHz, with a bandwidth that is 25% higher than the single-layer pixel transition towards higher frequencies.