When it comes to previous research and applications, using high voltage normally-off Gallium Nitride (GaN) devices, has paved the way for high power density and efficiency in bidirectional battery chargers. However, the topology for such chargers demands careful consideration. While widely adopted topologies like LLC resonant converters and phase shift full bridge converters have several advantages, they fall short in the context of bidirectional chargers employing GaN devices. There was hence the requirement for an optimized, high-frequency-efficiency dual active bridge converter (DAB) for battery charging applications.

GaN Multi-chip Module and its Characterization

To meet high current requirements, multiple GaN devices connected in parallel, with 6 to 8 devices typically paralleled were used as a test circuit. As shown in Figure 1, a half-bridge configuration served as the core topology, aided by ceramic capacitors for decoupling, significantly reducing loop inductance to 3.6 nH. This layout ensured an even distribution of power loops which was crucial for the optimal performance of the system.

Figure 1: Bidirectional battery charger topology using GaN

Figure 2 shows how on-resistance varies with junction temperature, while the absence of a body diode in normally-off GaN devices leads to a significant reverse conduction voltage drop. Dynamic Ron is seen as initial instability post-activation, stabilizing over time, and how the output capacitance measurements are essential for predicting zero-voltage switching (ZVS) transition times. Despite challenges posed by limited datasheet insights, direct characterization methods offered reliable solutions during this test. The integration of parallel-connected GaN devices, accurate layout design, and precise gate signal management ensured optimal performance in high-current applications. By ensuring thorough semiconductor characterization, engineers can gain valuable insights into device behavior, enabling informed decision-making for system optimization.

Figure 2: Voltage drop observed when reverse conduction takes place at a different current

Optimizing Parameters for DAB and its Selection for Efficiency

Initially, a DAB converter was designed to operate at 250V input and output, delivering 1kW power at 500 kHz. In order to match the energy transfer needs, an inductance model was used, resulting in a 2.7 µH inductor due to practical constraints, along with transformer leakage, creating a total commutation inductance of 3.36 µH. In this setup, there was an efficiency of 97.2% achieved wherein the loss breakdown revealed GaN device and transformer/inductor losses each accounted for 42%, with additional losses from PCB traces and Litz wire termination. Moreover, it was also noted that GaN device losses increase during hard-switching. Figure 3 illustrates the loss breakdown.

Figure 3: Maximum output current of the DAB

Optimizing the performance of a Plug-in Hybrid Electric Vehicle (PHEV) battery charger takes into consideration several critical parameters like the transformer turns ratio and commutation inductance (Llk). The efficiency of the converter varies with load conditions, particularly concerning the charger's ability to accommodate voltage fluctuations ranging from 270V to 430V. To ensure optimal efficiency within the battery voltage range of 310V to 390V, where the charger delivers a steady 3.3kW output, precise design choices are crucial for its operation. Selecting a DC link voltage of 350V simplifies the determination of a 1:1 transformer turns ratio. This decision aligns the charger's highest efficiency point with the mid-range battery voltage of 350V, enhancing overall performance. Moreover, adopting a sinusoidal charging methodology minimizes the reliance on DC link capacitance. In this approach, the output current regulation is achieved through a precise control of phase shift (φ) across the line cycle.

Figure 4: Conduction losses at 310V, 350V and 390V with the commutation inductance value up to 4 µH

This Sinusoidal charging could confirm that the peak output current was double the average, necessitating a commutation inductance lower than 4 µH to facilitate full 3.3kW power delivery at 310V. Phase shift modulation enabled a soft-switching turn-on for all eight switches within the charger architecture and the Zero Voltage Switching (ZVS) boundary, influenced by GaN device output capacitance, played a pivotal role in determining optimal operation conditions. Moreover, further analysis revealed that switching loss decreases with higher inductance, owing to the extended ZVS range. Notably, the lowest loss occurs at a battery voltage of 350V, attributed to the widest ZVS range at this voltage level. Conduction loss follows a similar trend, with the lowest observed again at 350V battery voltage. Despite this, selecting the optimal commutation inductance to minimize the total losses presents a nuanced challenge, necessitating consideration of total energy loss over the charging duration. Figure 4 illustrates the total semiconductor loss trends for different battery voltages, aiding in the comprehensive analysis of charger performance.