Their results suggest that the topology with the best current quality for one module need not offer the strongest overall package for an interleaved modular charger, because harmonic cancellation can offset some of the simpler design’s structural disadvantages. The study examined the grid-side AC–DC stage of high-power off-board fast chargers rather than the downstream DC–DC stage that connects to a vehicle. The modular architecture allows charging capacity to be increased by adding identical AC–DC modules in parallel instead of replacing an entire charging installation.

In the computational analysis, each 100 kW module was treated as a building block, and systems containing one through six modules represented aggregate ratings from 100 to 600 kW. The authors were not attempting to reproduce the complete behavior of a megawatt charging station; the models were designed to isolate how parallel current aggregation and carrier interleaving affect the current delivered to the grid. Practical systems would also require protection, disconnection, pre-charge, synchronization and other components not fully represented in the idealized comparison.

For all three simulated topologies, the researchers used 500 μH coupling inductors per phase, a split DC link with 800 μF in each half, and a 50 kHz switching frequency. They compared parallel modules operating with identical carrier phase against modules whose pulse-width-modulation carriers were displaced by 360° divided by the number of modules. Without phase shifting, grid-side current THD remained practically constant as the number of modules increased.

Although absolute ripple and the fundamental current rose with aggregate power, harmonic and fundamental components increased at approximately the same rate, leaving their ratio nearly unchanged. With carrier interleaving, switching-related components partially canceled at the grid connection. For two full-bridge modules, the 180° displacement produced a local minimum in THD; with more modules, the relevant cancellation was distributed across multiple spectral terms and THD converged more smoothly.

The simulated single-module results separated the three converters clearly. With carrier interleaving specified as 360°/n, the t-type converter produced the lowest grid-side current THD, approximately 0.75%. The full-bridge followed at about 1.13%, while the Vienna rectifier registered approximately 1.64%.

From two modules onward, the interleaved full-bridge and t-type systems converged to similar THD values of roughly 0.50%, making their modeled grid-current quality nearly independent of which of those two topologies was selected. The Vienna design did not share that convergence: its THD remained around 1.53% and changed little with module count. The authors associated this limitation with passive diode-conduction intervals, during which the controller has less influence over the voltage applied to the coupling inductor and therefore cannot fully regulate the current waveform.

These converter-side results should not be read as direct demonstrations of compliance with a universal grid-code threshold. The paper notes that current-distortion requirements at a point of common coupling depend on installation-specific information, including available short-circuit current and maximum demand load current, while voltage-quality limits refer to a different measured quantity. The simulations nevertheless show that lower converter-side distortion reduces the harmonic contribution of a charging system at the common coupling point.

They also reveal an important structural contrast. The full bridge uses six controlled silicon-carbide MOSFETs and synthesizes two phase-to-neutral voltage levels. The Vienna rectifier contains six MOSFETs and six passive diodes and provides three phase-to-neutral voltage levels, but it supports only unidirectional power flow.

During each AC half-cycle, one diode provides a passive conduction path while the actively controlled cell regulates the remaining interval. This arrangement can reduce the number of active commutations and gives the converter a comparatively low switching ripple, yet the passive intervals constrain current control and contributed to its higher THD in the experiments. The t-type converter uses 12 active MOSFETs and also provides three phase-to-neutral levels, but all voltage-level transitions are produced through controlled switching in both half-cycles.

That control supports bidirectional operation and low ripple and THD in a single module, at the expense of more devices, gate signals and modulation complexity. It should not be characterized as a simple midpoint between the other converters: it delivered the best single-module current quality, whereas the full bridge proved more competitive with it under multi-module interleaving because of its lower hardware complexity. To isolate topology effects as far as possible, the investigators applied the same continuous-control-set model predictive current controller to all three converters.

Described as a deadbeat predictive law, the controller uses a discrete model of the converter and coupling inductor to calculate the continuous-valued converter voltage needed for the next inductor current to match its current reference. That voltage reference is translated into switching signals by a topology-specific modulation stage. A proportional-integral controller regulates the DC-link voltage and generates the phase current references.

The computational and experimental systems used different tuned gains, but all three converters shared the appropriate gains within each campaign. The computational model also assumed symmetrical modules, matched electrical parameters and common reference signals; it did not include a dedicated low-frequency circulating-current controller. Experimental validation used one 25 kW prototype that could implement each topology while retaining the same power printed circuit board, sensing architecture, control hardware, DC link and coupling inductors.

The converter contained 24 1,200 V silicon-carbide MOSFETs arranged in parallel pairs at the switching positions to support its rated current. A digital signal processor performed the control functions, with current sensors on the three AC phases and voltage sensors for the grid and split DC link. The laboratory platform operated at the same 50 kHz carrier frequency used in the simulations.

Thermal management, grid-side filtering and electromagnetic interference required careful design at the prototype’s current levels and switching rate, but complete thermal, EMI and long-term reliability qualification was outside the scope. Quantitative conversion efficiency was not measured, so the study cannot establish an efficiency ranking among the three designs. In the representative steady-state laboratory tests, the full-bridge converter produced 231 mA of grid-side current ripple and 6.48% THD.

The Vienna topology generated slightly less ripple, 212 mA, but higher THD of 7.42%, demonstrating that a lower switching ripple did not necessarily produce lower total distortion. The t-type converter recorded the lowest measured values: 183 mA of ripple and 3.46% THD. These measurements were not accompanied by a repeatability or uncertainty analysis, and the absolute THD values were higher than those predicted computationally.

The authors attributed the difference to practical non-idealities including semiconductor dead time, digital control delay, measurement noise, hardware-layout parasitics and small distortions in the laboratory grid voltage. Despite the numerical gap, the ordering of the topology results was consistent with the simulations. For a charger built from one module, or from parallel modules without carrier interleaving, the evidence favors the t-type converter when grid-current waveform quality is the priority.

The full-bridge remains attractive when fewer active devices and simpler implementation are more important. In the modeled multi-module arrangement with a 360°/n carrier phase shift, however, the full bridge achieved THD comparable to the t-type design while retaining lower hardware complexity. That multi-module finding was computational, not experimental, and the authors cautioned that module parameter tolerances, circulating currents, synchronization errors, differing grid impedance, thermal constraints, electromagnetic interference and protection coordination could reduce the harmonic cancellation seen in the idealized model.

Testing several physical modules in parallel would therefore be needed to determine how much of the predicted improvement survives in a practical high-power charging installation.

Subject of Research: Comparative study of SiC AC–DC converter topologies and carrier interleaving for grid-side power quality in modular high-power EV charging

Article Title: Module-level grid-side power-quality comparison of SiC-based three-phase AC-DC converter topologies for modular high-power EV charging

Article References: Pereira, P., Soares, T., Coelho, S., Afonso, J. L., & Monteiro, V. (2026). Module-level grid-side power-quality comparison of SiC-based three-phase AC-DC converter topologies for modular high-power EV charging. Frontiers in Future Transportation, 7, Article 1904835. https://doi.org/10.3389/ffutr.2026.1904835

Image Credits: AI Generated

DOI: 10.3389/ffutr.2026.1904835

Keywords: electric vehicle charging, silicon carbide converters, power quality, current harmonic distortion, carrier interleaving, modular converters, full-bridge converter, t-type converter, Vienna rectifier, predictive control, Module-level, grid-side