This article summarizes the main results and contributions of the MagNet Challenge 2023, an open-source research initiative for data-driven modeling of power magnetic materials. The MagNet Challenge has (1) advanced the state-of-the-art in power magnetics modeling; (2) set up examples for fostering an open-source and transparent research community; (3) developed useful guidelines and practical rules for conducting data-driven research in power electronics; and (4) provided a fair performance benchmark leading to insights on the most promising future research directions. The competition yielded a collection of publicly disclosed software algorithms and tools designed to capture the distinct loss characteristics of power magnetic materials, which are mostly open-sourced. We have attempted to bridge power electronics domain knowledge with state-of-the-art advancements in artificial intelligence, machine learning, pattern recognition, and signal processing. The MagNet Challenge has greatly improved the accuracy and reduced the size of data-driven power magnetic material models. The models and tools created for various materials were meticulously documented and shared within the broader power electronics community.

The ability to provide galvanic isolation is a key requirement in power electronics applications such as data center power delivery network, telecom power system, Electrical Vehicles on board chargers or floating auxiliary supplies. Typically, isolation is provided by means of magnetic transformers, which share a noteworthy portion of the total volume and losses of the system. This work presents a novel, capacitive based isolated, fixed gain resonant switched capacitor (ReSC) DC/DC converter as a higher power density, lower volume, lower weight alternative to the transformer based isolated converters. Capacitors are used as the isolation coupling element as well as the differential mode power processing elements. The presented converter modulation makes it suitable for use in interleaved multi-cell architectures without the generation of additional circulating current between cells. A 100V, 200W proof of concept prototype is built to validate the topology. Higher than 97.5% efficiency in the whole operating range is reached and low frequency common mode isolation is demonstrated.

This letter presents a simplified dc-bias injection circuit utilizing a voltage mirror transformer configured in a unique way for improved equipment applicability and measurement accuracy. A mirror transformer can minimize the impact of the reflected actuation voltages on the dc source and improve the quality of the dc-bias injection. By combining the dc-bias and ac actuation current, the traditional three-winding dc-bias measurement setup was simplified with improved accuracy and reduced equipment complexity. The dc-bias injection circuit was implemented as a part of a highly automated data acquisition system designed for fast and accurate characterization of power magnetic materials under a wide range of operating conditions.

This paper presents a transformer-based encoder-projector-decoder neural network architecture for modeling power magnetics B-H hysteresis loops. The transformer-based encoder-decoder network architecture maps a flux density excitation waveform (B) into the corresponding magnetic field strength (H) waveform. The predicted B-H loop can be used to estimate the core loss and support magnetics-in-circuit simulations. A projector is added between the transformer encoder and decoder to capture the impact of other inputs such as frequency, temperature, and dc bias. An example transformer neural network is designed, trained, and tested to prove the effectiveness of the proposed architecture.

In this article, a setup to characterize the failure mechanisms and degradation indicators of gallium nitride (GaN) high-electron-mobility transistors (HEMTs) under short-circuit events is presented. Understanding how this technology fails is critical, especially for space applications where a common failure mechanism is the short-circuit event due to radiation. First, the previous literature in the field is reviewed. Both the behavior under single and repetitive short circuits and failure mechanisms are discussed. Then, a systematic method is proposed to measure the critical electrical parameters and analyze the behavior of the device under test under short-circuit failure to build a reliability model. A setup to characterize GaN HEMT devices is developed, and using it, multiple tests and short circuits have been performed to characterize the devices under test at different conditions and to identify critical parameters and aging indicators. The reliability challenge of GaN devices could be addressed by having on-board in-system prognostics and device health monitoring techniques to predict device failures well ahead of time.

This paper presents a simplified dc-bias injection circuit using a mirror transformer as a part of an automated data acquisition system for the characterization of soft magnetic materials for power electronics applications. Modified from the existing three-winding method for minimizing the reflected voltage on the dc-bias injection circuitry, the proposed two-winding method offers improved accuracy with reduced equipment complexity and capability. It enables rapid and fully automated material characterization across wide frequency, flux density, excitation waveform, temperature, and dc-bias ranges.

This paper presents the MagNet-AI platform as an online platform to demonstrate the “Neural Network as Datasheet” concept for B–H loop modeling and material recommendation of power magnetics across wide operation range. Instead of directly presenting the measured characteristics of magnetic core materials as time sequences, we employ a neural network to capture the B–H loop mapping relationships of magnetic materials under different excitation waveforms at different temperatures and dc-bias. Both LSTM and Transformer based neural network models are developed, verified, and compared. The training and inference process of the neural network are fully automated to minimize the impact of human error. The neural network can be used to rapidly predict B–H loops under different operating conditions, compare materials, and recommend materials for design. Neural networks are also effective in compressing the information contained in the raw database, and enable rapid material evaluation and comparison.

This paper motivates the development of sophisti- cated data-driven models for power magnetic material charac- teristics. Core losses and hysteresis loops are critical information in the design process of power magnetics, yet the physics behind them is not fully understood. Both losses and hysteresis loops change for each magnetic material, are highly nonlinear, and depend heavily on the electrical operating conditions (e.g., wave- form, frequency, amplitude, dc bias), the mechanical properties (e.g. pressure, vibration), as well as temperature and geometry of the magnetic components. Understanding the complexity of these factors is important for the development of accurate models and their applicability and limitations. Existing studies on power magnetics are usually developed based on a small amount of data and do not reveal the full magnetic behavior across a wide range of operating conditions. In this paper, based on a recently developed large-scale open-source database – MagNet – the core losses and hysteresis loops of Mn-Zn ferrites are analyzed over a wide range of amplitudes, frequencies, waveform shapes, dc bias levels, and temperatures, to quantify the complexity of modeling magnetic core losses, amplitude permeability, and hysteresis loops and provide guidelines for modeling power magnetics with data- driven methods.

This paper applies machine learning to power mag- netics modeling. We first introduce an open-source database – MagNet – which hosts a large amount of experimentally measured excitation data for many materials across a variety of operating conditions, consisting of more than 500,000 data points in its current state. The processes for data acquisition and data quality control are explained. We then demonstrate a few neural network-based power magnetics modeling tools for modeling the core losses and B–H loops. Machine learning allows multiple factors that may influence the magnetic characteristics to be mod- eled in a unified framework, while provides insights to quantify the complexity of magnetic characteristics and reduce the size of the measurement data required to build a precise model. Neural network models are found to be effective in compressing the measurement data and predicting the material characteristics. The behaviors of a typical power magnetic material (TDK N87) across a wide range of operating conditions (e.g., temperature, waveform, dc-bias) can be well described by a small-scale neural network (204 KB) which is 2,500 times smaller than the raw measured time-series data (512 MB), paving the way for “neural networks as datasheet” to assist power magnetics design.

In power electronics, galvanic isolation is required for safety reasons, and when large DC or AC voltages have to be withstood between the input and output of a power converter. Galvanic isolation in the form of a magnetic transformer prevents failures in one side of the converter to be transferred to the other side and provides voltage transformation. Transformers are used for power applications, but optocouplers or capacitive barriers can be used to transfer control signals. However, the main limitation of transformers is that they are heavy, bulky, and lossy, and, as a result, transformer-less topologies conventionally achieve better performance. In this paper, a topology is proposed where galvanic isolation is obtained through capacitors rated to the maximum common-mode voltage, rather than relying on transformers. First, the fundamental principles of the idea are discussed. The operation of the proposed 1:1 resonant switched capacitor converter is introduced, emphasizing the effects of varying common-mode voltages. Later, zero-voltage switching is covered, and finally, the operation is validated in simulation.

The ever-rising power inverter applications where high-frequency and high-power density are a must, demand solutions driven by new topologies, emergent semiconductors technologies like Enhancement-Mode Gallium Nitride Electron-Mobility transistors, and improved high-frequency magnetics materials like NiZn ferrites. Although Power Electronics tends towards high-switching frequencies looking for more compact designs with higher efficiency, some easily solvable problems at medium frequencies, such as common mode currents or delays effect, take an especial relevance at high-frequencies. Furthermore, some challenges related to new technologies like NiZN ferrites have shown up. In this paper, a Modular Cascaded H-Bridge converter working as an arbitrary waveform generator is detailed, where its modulation techniques are reviewed and its main design challenges are included. In addition, a modular multicell power inverter has been implemented with 5 Cascaded H-Bridges obtaining 20 MHz equivalent switched outputs with a variable frequency up to 1 MHz and 1.35 kW power peaks.

This paper describes the design and implementation of a compact and high-frequency LLC converter with wide input voltage range (from 220 V to 320 V), output power range (from 250 W to 1.5 kW), high efficiency (over 96 % at full power), high power-density (32 kW/dm 3 ) and narrow frequency variations (15 %). A complete and automated optimization process has been developed, including the comparison between different semiconductor technologies and the design of a Matrix Transformer that reduces the height of this magnetic component. A fully customized converter has been developed using an automated design tool.

In this paper, an analysis of the failure mecha-nisms and degradation indicators of GaN HEMTs is presented. Understanding how this technology fails is critical, especially for space applications. Due to radiation, a common failure mechanism in space applications for GaN devices is the short-circuit event. A systematic method is proposed to perform “on-board” measurement of the critical electrical parameters and analyze the behavior of DUTs under short-circuit failures to build a reliability model. A setup to characterize GaN HEMTs devices is developed, and multiple tests at different conditions have been performed. The reliability challenge of GaN devices could be addressed by having on-board, in-system prognostics and device health monitoring techniques to predict device failures well ahead of time.

This paper presents the concept of "Neural Network as Datasheet" for B-H loop modeling of power magnetics with sequence-to-sequence machine learning. Instead of directly presenting the measured characteristics of magnetic core materials, we employ a neural network to capture the B-H loop mapping relationships of magnetic materials under different excitation waveforms at different temperatures. The training and inference process of the neural network are fully automated to minimize the impact of human error. Neural networks are also effective in compressing the information contained in the raw database to avoid data search or interpolation. The neural network can be used to rapidly predict B-H loops under different operating conditions and support circuit simulations. Based on a recently developed large-scale magnetic core loss database - MagNet - we demonstrate that a neural network datasheet can effectively compress and release information about power magnetics and can play important roles in power electronics converter design.

This paper exposes the analysis, modeling and comparison of two different LLC converter configurations: the conventional LLC converter and the Three-Phase $\mathrm{Y}-\Delta$ LLC converter, which is firstly analyzed in this paper. Piecewise-defined analytical equations that accurately model the behavior of these converters have been developed and validated through simulation. Additionally, a comparison between these two power converters has been carried out, so the advantages and disadvantages of each of them are exposed. To validate the derived conclusions, a fully optimized Three-Phase $\mathrm{Y}-\Delta$ LLC converter is designed (including the design of the resonant tank, the three-phase transformer, the resonant inductor, and the selection of semiconductor devices) and compared to a 96% efficiency single-phase LLC converter.

This paper introduces an open-source database - MagNet - for data-driven magnetic core loss modeling. MagNet aims to support data-driven magnetics research by hosting a large amount of experimentally measured excitation waveform data for many materials across a variety of operating conditions. This database in its current state contains over 150,000 excitation waveforms for six ferrite materials - TDK{N27, N49, N87}, Ferroxcube{3C90, 3C94}, Fair-Rite{78} - in the 50 kHz to 500 kHz, 10 mT to 300 mT range for sinusoidal, triangle, and trapezoidal waveforms. This paper presents the purposes of building MagNet, introduces the data acquisition system and data format, discusses the data quality, and presents a few examples of using this database with data driven methods.

Generally, in non-isolated power converters, the highest temperature is reached in inductors and switching devices due to power losses. Extracting heat is always challenging, requiring a careful three dimensional (3-D) design and bulky heatsinks. For inductors, this challenge can also be addressed using custom magnetics; the shape of inductors could be designed aiming to decrease the reached temperature. In this work, a detailed study of new 3-D coils termed "Box inductors" has been carried out. The advantage of this new structure is that it has a large surface area for a given volume, which allows a better heat transfer and, thus, a decrease in temperature. This "Box inductor" is used in a single-phase single-stage transformerless inverter integrating the intermediate energy storage. Three Box inductor designs with the same PLT-3C95 core, but with different types of geometries are compared, which are also compared with a conventional RM14/I-N87 coil to establish which one leads to lower losses and temperature. Matlab and Maxwell are used to accurately estimate the losses. Prototypes of the three designs are built and tested, and the results of inductance value, losses, and temperature are compared.

This article shows the use of a high-frequency Full-Bridge inverter to design an alternating magnetic field generator for experimental studies on anti-cancer treatment with magnetic nanoparticles. In magnetic hyperthermia, nanoparticles are used to raise the pathological cancerous cell's temperature high enough to induce their death by apoptosis, but not as high as to destroy them by thermal ablation of the whole tissue volume, consequently leaving healthy cells alive. Conventionally, sinusoidal alternating magnetic fields are used to heat the nanoparticles, which can be more easily produced than other waveforms. However, there are no theoretical nor experimental reasons to choose the sinusoidal waveform. This work aims to develop an improved power system to study the effect of different waveforms of the magnetic field on heat production when exciting magnetic nanoparticles, aiming at demonstrating that other waveforms can be much more efficient in producing heat than conventional sinusoids. To prove our hypothesis, we designed an inverter able to generate four waveforms at high frequencies and a fifth sinusoidal signal derived by a resonant capacitor, in the range of 100 kHz to 1 MHz and up to 10 mT of peak intensity. Also, we used SiC devices to process high currents at high switching frequencies efficiently. Additionally, to enhance the system efficiency, Zero-Voltage Switching is used to reduce switching losses and minimize electromagnetic noise and interference. The experimental results obtained with non-sinusoidal waveforms have shown a remarkable performance improvement compared to classical sine wave excitation. The nanoparticles' heat dissipation depends on the applied alternating magnetic field's signal slope, signal frequency, and peak field intensity. We conclude that further work deserves to be done to find the optimum work conditions in function of the used particle and biological environment to test if this type of magnetic field generator could overcome the conventional system's performance.

This article presents a control linearization technique for a single-phase single-stage inverter with multiple modulation strategies. The power topology is based on an flying capacitor multi-level (FCML) inverter, and the control is based on the plant inversion. The proposed technique allows decoupling the control of the input current and the output voltage, which simplifies the control in two ways: first, the control of both variables is independent and second, the controllers are independent of the modulation applied, thanks to the plant inversion. This control is implemented in an field-programmable gate array. Since the converter operates at variable frequency, two different data acquisition alternatives are explored: sampling at variable frequency (once per switching cycle) or sampling at constant frequency (higher than the switching frequency).The proposed control is validated by simulation and experimental results with a 1-kVA prototype.