Cascaded H-bridge Converters (CHBs) are a promising solution in converting power from a three-phase medium voltage of 6.6 kV...30 kV to a lower DC-voltage in the range of 100 V...1 kV to provide pure DC power to applications such as electrolyzers for hydrogen generation, data centers with a DC power distribution and DC microgrids. CHBs can be interpreted as modular multilevel converters with an isolated DC-DC output stage per module, require a large DC-link capacitor for each module to handle the second harmonic voltage ripple caused by the fluctuating input power within a fundamental grid period. Without a zero-sequence voltage injection, star-connected CHBs are operated with approximately sinusoidal arm voltages and currents. The floating star point potential enables to utilize different zero-sequence voltage injection techniques such as a third-harmonic injection with 1/6 of the grid voltage amplitude or a Min-Max voltage injection. Both well-known methods have the advantage to reduce the peak arm voltage and thereby the number of required modules by 13.4 % (to √ 3 2). This paper proves analytically that the third-harmonic injection with 1/6 of the grid voltage amplitude reduces the second harmonic voltage ripple by only 15.1 % compared to no-voltage injection for unity power factor operation and balanced grid voltages. Then it is shown, that the Min-Max injection has the often overlooked advantage of reducing the second harmonic voltage ripple by even 18.8 %. By applying the here proposed zero-sequence voltage injection in saturation modulation, the second harmonic voltage ripple of the DC-link capacitors is reduced by even 24.3 %, while still requiring the same number of modules as the Min-Max injection. For a realistic number of reserve modules, the overall energy ripple in the DC-link capacitors is reduced by 40 %.

LLC resonant converters are typically unsuitable to be applied for wide voltage-transfer ratio applications. With a full-bridge inverter, however, they can be operated in a variety of different modulations. Most notably, by permanently turning on one MOSFET and turning off the other MOSFET of the same bridge leg, the LLC can be operated in half-bridge configuration reducing the gain by a factor of two. The resonant capacitor is hereby charged to an average voltage of half the input voltage. In this modulation, however, the switch that is permanently turned on is stressed by the complete resonant current while exhibiting no switching losses. This paper proves that the frequency-doubler modulation can better balance the losses among all MOSFETs and should be the preferred mode of operation favored over the conventional half-bridge modulation. This paper analyzes the beneficial loss distribution, proposes an on-the-fly morphing modulation and discusses potential operating strategies to further reduce the junction temperature. Furthermore, it is shown that this modulation can also be altered to achieve the asymmetrical LLC operation. Experimental measurement results show that the modulation results in a substantial decrease of the maximum MOSFET temperature and shows that the converter can be smoothly transitioned during operation from full-bridge modulation to the frequency-doubler half-bridge operation and back.

DC-DC converters for on-board chargers (OBC) of electrical vehicles are usually galvanically isolated allowing modular single-phase PFC front-end solutions, but require transformers which are more bulky, costly and lossy than inductors of non-isolated DC-DCs. Furthermore, for vehicle-to-grid applications, bidirectional converters with transformers are generally more complex and have a higher count on semiconductor switches than transformerless solutions. However, when using non-isolated DC-DC converters within an OBC, the large common-mode (CM) capacitance comprising capacitive parasitics of the traction battery as well as explicit Y-capacitors connecting the high-voltage DC-system (HV-system) within specific HV-loads to ground has to be considered. For the PFC front-end stage, when supplied from the three-phase mains this means that generation of high-frequency and high-amplitude CM voltages, as it is common e.g. with the conventional six-switch full-bridge converter, has to be strictly avoided. For this reason, a modified topology is suggested leading to a different mode of operation and to a very low common-mode noise behaviour: The three-phase four-wire full-bridge PFC with split DC-link, whose midpoint is connected to the mains neutral provides very stable potentials at the DC-link rails and therefore it can be classified as Zero-CM-topology.For dedicated single-phase operation, as required for most OBC, an additional balancing leg may be added to the topology to reduce the required DC-link capacitance and allow non-electrolytic capacitors.The function of the bidirectional Zero-CM three-phase four-wire full-bridge PFC was verified by simulation and on an 11 kW-laboratory sample. The power factor is above 0.999 and an efficiency of 98 % is measured.

In electric vehicles (EV) the large common-mode (CM) capacitance comprising capacitive parasitics of the traction battery as well as explicit Y-capacitors connecting within specific loads the high-voltage DC-system (HV-system) to ground, can cause issues when using non-isolated EV Chargers. One solution for a power factor correction (PFC) rectifier that is capable to operate with a non-isolated DC-DC converter, is the three-phase four-wire full-bridge PFC, with split DC-link, whose midpoint is connected to the mains neutral. Therefore, it provides very stable potentials at the DC-link rails and accordingly can be classified as Zero-CM topology, which facilitates a common-mode-free operation. When to be operated at a single-phase supply, which is a common requirement for On-board chargers (OBCs) this topology results in the voltage-doubler PFC (V2-PFC) being characterised by a comparably large DC-link voltage ripple at mains frequency. If the DC-link capacitance shall be minimized, for instance to avoid lifetime-limited electrolytic capacitors, two more circuits in addition to the original V2-PFC are proposed for keeping the common-mode-free operation: A balancing circuit (BC), that balances the voltages over the split capacitors and a ripple port (RP), that buffers the 100 Hz power pulsation of the mains. For both circuits the available two bridge legs of the three-phase topology in single-phase operation may be utilized. A 3.7 kW laboratory sample verifies the functionality of the additional circuits in conjunction with the V2-PFC and achieves an efficiency of 95 %.

LLC resonant converters typically employ power MOSFETs in their inverter stage. The generally weak reverse recovery behaviour of the intrinsic body diodes of those MOSFETs causes significant turn-on losses when being forced to hard commutations. Continuous operation in this way will lead to self-destruction of the transistors. Consequently, zero-voltage switching (ZVS) is essential in a MOSFET-based inverter stage. To ensure ZVS, the LLC converter is operated in the inductive region. On the contrary, IGBTs show dominant turn-off losses and are therefore conventionally not applied in LLC converters typically requiring high switching frequencies to achieve low output voltages. However, if the LLC converter is intentionally designed for capacitive operation, zero-current switching (ZCS) is enabled and thus robust and cost-efficient IGBTs can be applied in the inverter stage. The aim of this work is to investigate the use IGBTs in the inverter of an LLC converter. The theory behind the capacitive operated LLC is derived using a switched simulation model and compared with the fundamental harmonic approximation (FHA). The results prove FHA to be useless for practical converter design. Instead, a stress value analysis based on switched model simulations is proposed to the design a capacitive operated LLC utilizing ZCS. A 2 kW prototype for on-board EV applications was built to verify the theory and design approach. The prototype confirms the derived theory and thus the deployment of IGBTs in the inverter stage of LLC resonant converters. Synchronous rectification turns out to require a specific control solution, but if given the resulting efficiency in the most critical operation point exceeds the value of a MOSFET-based (inductive operated) LLC-design of an identical application. Therefore, this concept should be further developed.

Modular solid-state transformers (SSTs) are a promising technology in converting power from a 10kV three-phase medium voltage to a lower DC-voltage in the range of 100…400V to provide pure DC power to applications such as electrolyzers for hydrogen generation, data centers with a DC power distribution and DC micro grids. Modular SSTs which can be interpreted as modular multilevel converters with an isolated DC-DC output stage per module, are designed with redundant modules to increase reliability. Usually, each of the three arms operates independently, and therefore, only a fixed number of faulty modules can be compensated in each arm, even if all modules are operational in the remaining two arms. With the proposed zero-sequence voltage injection, up to 100% more faulty modules can be compensated in an arm by employing the same hardware. In addition, module power imbalances are nearly eliminated by utilizing a fundamental frequency zero-sequence voltage. A dominant 3rd harmonic zero-sequence voltage injection in combination with the 5th, 7th and several higher order harmonics with adaptive (small) amplitudes minimize the required arm voltages at steady-state. For nominal operation or symmetrical faults, the proposed technique is equivalent to the well known Min-Max voltage injection, which already reduces the peak arm voltage by 13.4% compared to a constant star point potential. A statistical analysis proves, that the expected number of tolerable faulty modules of the 1MW SST increases by 12% without the need for additional hardware.

In this paper, a full-bridge modular multilevel converter (MMC) and two half-bridge-based MMCs are evaluated for high-current low-voltage e.g. 100 - 400V DC-applications such as electrolysis, arc welding or datacenters with DC-power distribution. Usually, modular multilevel converters are used in high-voltage DC-applications (HVDC) in the multiple kV-range, but to meet the needs of a high-current demand at low output voltage levels, the modular converter concept requires adaptations. In the proposed concept, the MMC is used to step-down the three-phase medium-voltage of 10kV, and provide up to 1 MW to the load. Therefore, each module is extended by an LLC resonant converter to adapt to the specific electrolyzers DC-voltage range of 142 - 220V and to provide galvanic isolation. The six-arm MMC converter with half-bridge modules can be simplified and optimized by removing three arms, and thus halving the number of modules. In addition, the module voltage ripple and capacitor losses are decreased by 22% and 30% respectively. By rearranging the components of the half-bridge MMC to build a MMC consisting of grid-side full-bridge modules, the voltage ripple is further reduced by 78% and capacitor losses by 64%, while ensuring identical costs and volume for all MMCs. Finally, the LLC resonant converter is designed for the most efficient full-bridge MMC. The LLC can not operate at resonance with a fixed nominal module voltage of 770V because the output voltage is varying between 142 - 220V. By decreasing the module voltage down to 600V, additional points of operation can be operated in resonance, and the remaining are closer to resonance. The option to decrease the module voltage down to 600V, increases the number of required modules per arm from 12 to 15, which requires to balance the losses of the LLCs and the grid-side stages.

A full-bridge modular multilevel converter (MMC) is compared to a half-bridge-based MMC for high-current low-voltage DC-applications such as electrolysis, arc welding or datacenters with DC-power distribution. Usually, modular multilevel converters are used in high-voltage DC-applications (HVDC) in the multiple kV-range, but to meet the needs of a high-current demand at low output voltage levels, the modular converter concept requires adaptations. In the proposed concept, the MMC is used to step-down the three-phase medium-voltage of 10 kV. Therefore, each module is extended by an LLC resonant converter to adapt to the specific electrolyzers DC-voltage range of 142-220V and to provide galvanic isolation. The proposed MMC converter with full-bridge modules uses half the number of modules compared to a half-bridge-based MMC while reducing the voltage ripple by 78% and capacitor losses by 64% by rearranging the same components to ensure identical costs and volume. For additional reliability, a new robust algorithm for balancing conduction losses during the bypass phase is presented.

Although there are numerous design methodologies for the LLC resonant converter, they often do not consider the possibility of input voltage adjustment. In the proposed concept, a modular multi-level converter (MMC) is used to step-down the three-phase medium voltage of 10 kV, and provide up to 1 MW of pure DC power to the load consisting of electrolyzers for hydrogen generation. Therefore, each module is extended by an LLC resonant converter to adapt to the specific electrolyzers DC voltage range of 142...220 V and to provide galvanic isolation. In order to achieve a high efficiency for a wide range of load conditions, the input voltage of the LLC converter is adjusted between 600 V and 770 V while operating at resonance or close to resonance. The parameters of the 11kW LLC resonant converter with an integrated leakage inductance are systematically optimized to maximize the efficiency for all loads while achieving zero-voltage switching. For a fast estimation of eddy current losses, a new method is proposed, which uses a single FEM simulation to fit newly developed loss equations. The calculated average efficiency is 97.8%. The prototype of the LLC converter reaches a peak efficiency of over 98% at resonance at half load which is similar to the precalculated value.

Heat dissipation is a limiting factor in the performance of many power electronic components. Especially in the TO-263-7 package, which is used for several SiC-MOSFETs, the heat transfer must take place through the cross section of the printed circuit board (PCB) to the heatsink at the bottom side. Most commonly, thermal vias are used to form this path in a perpendicular direction through all PCB-layers. In a given soft- and hard switched example applications with the use of C3M0065090J SiC-MOSFETs, this conventional approach limited the component’s maximum heat dissipation to approx. 13 W. A recent alternative approach are massive copper blocks (”pedestals”) being integrated in PCBs and reaching from their top- to the bottom-side in relevant footprint areas under SMD-housed power semiconductors. Pedestals allowing to increase the heat dissipation in the given case to even 36 W. This step is achieved due to the clearly superior heat spreading capability of that massive thermal connection between SiC-MOSFET and heatsink. For the hard switched example application the number of switch-elements can be halved to one, by using the pedestal instead of thermal vias. Independently of optimizing the heat transfer path, the up-front avoidance of losses helps to stay within existing heat dissipation limits, of course. The dominant conduction losses of the mentioned soft-switched example application could be halved by changing to SiC-MOSFET types with significant lowered RDSon. By using pedestals and changing to SiC-MOSFETs with lowered RDSon, the number of switch-elements can also be halved for the soft switched application.

Automotive DC-DC converters linking the traction battery to the auxiliary battery are characterized by the wide input and output voltage ranges resulting from the varying state-of-charge of the traction and auxiliary battery. The wide voltage transfer ratio needs to be covered for the entire load range conventionally requiring two-stage converter architectures. Considering a less complex single-stage solution potentially enabling cost and weight advantages, traditional LLC converters are unsuitable topologies since it results in a too wide operating frequency range. Most alternative topology candidates show comparable difficulties. To overcome this issue, the gain range of the LLC with full-bridge inverter can be extended by operation in half-bridge mode for low voltage transfer ratios. Phase-shift operation is utilized for intermediate gains and low loads. This paper describes a detailed design methodology for the resonant tank. The experimental results with a peak efficiency of 96.5 % and a power density of 2.1 kW/l prove the proposed concept.

An onboard DC-DC converter connects the high voltage traction battery to the low voltage auxiliary battery of an EV. It has to provide power across a wide range of input and output voltages. This paper presents the design and evaluation of an economical two-stage converter concept consisting of a first-stage boost converter and a second-stage LLC converter. While for low input voltages, the boost converter can supply the second-stage LLC with the optimum bulk voltage, for high input voltages, the boost converter is turned off and the LLC regulates the output voltage on its own. Whereas this is unproblematic for high output currents, for low loads high switching frequencies become necessary. For this purpose, the LLC needs to be designed for a wide gain range. Traditionally, this is achieved through a small magnetizing inductance resulting in increased conduction losses. If an asymmetric duty cycle operation is used to cover the low gains at low output current, the LLC can be optimized for a better efficiency. A prototype design proves that the asymmetric duty cycle operation is feasible to achieve a wide gain range at a high efficiency whereas the conventional design achieves very poor efficiencies.

LLC resonant converters generally employ MOSFETs in the inverter stage, which can be of half-bridge (HB) or full-bridge (FB) type. The generally weak intrinsic (body) diodes of the MOSFETs cause turn-on losses when being forced to hard current commutations finally leading to the components self-destruction when operated constantly in this way. Consequently, zero-voltage switching (ZVS) operation is more or less essential in a silicon (Si) MOSFET-based HB or FB. To ensure ZVS, the LLC is operated in the inductive region, i.e. with lagging resonant current. On the contrary, IGBTs show dominant turn-off losses and therefore are conventionally not applied in LLCs typically requiring high switching frequencies to achieve low output voltages. Yet, if the LLC is intentionally designed for the capacitive region, i. e. operation with leading current, zero-current switching (ZCS) enabling IGBTs in the inverter stage can be ensured. This paper explores in detail the LLC in the capacitive operating region and gives design considerations for a capacitive LLC utilizing both robust and cost-efficient IGBTs for an exemplary 2.2 kW automotive on-board DC-DC converter application. The results of a loss analysis show that the LLC resonant converter can be operated well in the capacitive region. In the given case, significantly lower overall and 30 % lower inverter stage losses are achieved in the thermally relevant worst-case comparison with an inductive LLC based on Si MOSFETs.

For future Vehicle-to-Grid (V2G) applications, the six-switch full-bridge is often used as AC-DC front-end converter of a three-phase EV-charger. In many publications, the common mode (CM) noise is not taken into account. However, this must not be neglected considering the large effective capacitance, of up to 3 muF, as allowed by new standards. In this paper, different modulation techniques are investigated, related to their CM-noise. Based on electric circuit simulations, CM-filters are estimated, and the CM-currents are investigated. Accordingly, the conventional six-switch full-bridge is practically difficult to use in non-isolated chargers, because the resulting CM-currents and/or the required EMI-filter become too large, even if CM-Voltage optimized modulation techniques are used.

Utilisation of SiC semiconductors' fast switching speeds and high switching frequencies as a consequent are often in discussion. But which switching frequency is really optimal in terms of converter volume, losses and costs? Based on the example of a buck converter, this question is investigated, and a tool for loss calculation and design is described in this paper. A Pareto optimization of the converter is performed where the switching frequency is one of several design parameters. The buck converter can be realized by multiple rails interleaved, and several switches can be placed in parallel. Considered converter modes are continuous conduction mode, that allows hard-switching, ZVS, and incomplete ZVS depending on the switching frequency. Based on Pareto optimizations, a design is selected, and a laboratory sample of 5.5 kW for application in an EV battery charger is built up. Efficiencies of 99.5 % are achieved with switching frequencies of around 100 kHz.

Due to recent developments in MMCs, they are used in many medium-voltage and high-power applications today, but efficient and modular solutions for high-power at low-voltage such as for electrolysis are still required. The proposed Y-MMC converts the grid AC-voltage into a DC-voltages, and an LLC converter is connected to each submodule capacitor to provide the required current to the DC-load. This paper proposes a topology that uses only half the number of submodules and moreover to reduce the effective switching frequency by a third, while preserving the same THD und output power of an YY-MMC.