Bidirectional topologies for fast charging of EVs: options to optimize for size, power, cost or simplicity

Dc fast chargers for electric vehicles (EVs) under development now have to meet a more demanding set of specifications than today’s installed base of chargers. This high requirement stems from two market pressures: first, to provide for faster charging of the higher-capacity batteries embedded in the latest EVs. And second, to enable bidirectional power flow, supporting new vehicle-to-grid (V2G) and vehicle-to-building (V2B) applications, technology that will help to balance the power grid as more energy is generated from fluctuating sources such as wind and solar power. 

Charger manufacturers can build flexibility into their designs by implementing dc fast chargers as modular building blocks: multiple modules can be connected in parallel to scale up the power output to as high as 300 kW. This means that when multiple modules are stacked in a charger’s enclosure, the power density of the module becomes a key issue, as does thermal management. 

In answer to these challenges, EV charger manufacturers have turned to silicon carbide (SiC) power switches, which can switch faster than equivalent silicon MOSFETs or IGBTs with much lower switching losses: this enables the use of smaller magnetic components, resulting in higher system power density. SiC technology also provides an answer to the thermal problems of high-density power designs, because of its higher efficiency and toleration of higher operating temperatures than silicon devices, enabling the use of a smaller heat-sink while reducing thermal stress on the system’s components. 

There are today few semiconductor manufacturers with the technical capability and manufacturing capacity to supply SiC devices in volume to charger manufacturers: leading the group are Infineon, onsemi and STMicroelectronics. Many EV charger designers will therefore look to these companies to provide the reference design boards with which they can evaluate the performance of various SiC power switches, and assess their suitability for their application. 

These reference designs provide implementations of some of the most appropriate topologies for bidirectional dc fast chargers, both at the PFC stage and the dc-dc converter stage. An understanding of these topologies, and their benefits and drawbacks, will help the designer to work out which best fits the requirements of their application. Let’s look first at the choice of topologies for an EV charger’s active front-end PFC stage.  

 

Active front-end PFC stage of a dc charger module

The ac-dc stage converts a normally three-phase input in the range 380 V to 415 V ac into a stable dc link voltage of around 800 V. All the topologies examined in this article are for bidirectional systems, so the conversion will also go the other way, from dc to ac.

SiC devices are particularly well suited to bidirectional converters, which are based on a half-bridge configuration. Usually bidirectional systems perform repetitive hard commutation: in this case, a silicon power switch’s long reverse-recovery time at the device’s body diode will lead to high power dissipation and low efficiency, as well as higher thermal stress and lower system reliability. So a bidirectional converter requires low or even zero reverse-recovery time, a feature of SiC MOSFETs, as shown in  Figure 1.

Fig. 1: Low body diode reverse-recovery time is essential in bidirectional converters, which have a half-bridge configuration and are exposed to repetitive hard commutation

 

There are three topologies worth considering for the active front-end PFC stage of three-phase bidirectional dc chargers: 

• Two-level PFC
• Three-level neutral point clamped (NPC)/active NPC (ANPC) 
• Three-level T-type NPC
  

Two-level PFC topology

The six-switch boost-type rectifier of the two-level PFC topology is a very simple circuit that is easy to control, and shown in Figure 2. It facilitates bidirectional power flow and can achieve a high power factor with reasonable efficiency. Compared to a three-level topology, it has a lower component count and an easier PCB layout.

Fig. 2: The two-level PFC topology

 

On the other hand, it requires switches with a high voltage rating to block the entire dc link voltage. For example, in an 800 V dc application, a SiC MOSFET with a 1,200 V blocking capacity is required. 

An additional drawback of this topology is the bulky filter inductor, which is required to minimize the total harmonic distortion (THD) at the input current. Three-level topologies, which do not require so much inductance, enable lower power density. Another factor to consider is the high peak-voltage stress, which compromises the lifetime of the semiconductor and other passive devices. 

Finally, the converter’s EMI performance is substantially lower than that of the multi-level PFC topologies described below.

 

Three-level NPC/ANPC PFC topology

In the three-level NPC/ANPC topology, each switch only needs to be capable of blocking half the bus voltage, so MOSFETs with a lower voltage rating can be used, and the voltage stress on devices is much lower, as shown in Figure 3. This means that this topology can be easily scaled across multiple platforms for implementation with SiC, GaN and silicon power switches to meet the needs of applications with different power, cost and efficiency requirements.

In an 800 V application, 600 V-rated MOSFETs may be used. As well as offering lower switching losses than 1,200 V MOSFETs, 600 V MOSFETs can support much faster switching frequencies. 

In the NPC topology, lower ripple is observed in the output current, and the output-voltage transient is 50% lower. This reduces the requirement for filtering and isolation, and allows for the use of a smaller filter inductor. With less inductance needed to regulate THD in the current waveform, the design can achieve higher power density. This multi-level converter topology’s output voltages also suffer very low disturbance, which minimizes the dv/dt stresses across devices, and improves EMI performance. 

While the NPC topology offers lower switching losses and higher efficiency at switching frequencies above 50 kHz than the two-level PFC, it does require more switches, and each switch needs its own gate-drive circuit. This means that control is more complex, and the bill-of-materials (BoM) cost is higher. 

In addition, this topology uses both active semiconductor switches and diodes, and the resulting asymmetrical distribution of losses across the power stage can make thermal management difficult. Some designers prefer a more symmetrical distribution of losses, so replace the diodes of the NPC topology with active switches in an ANPC converter, as shown in Figure 4. 

It is useful in both NPC and ANPC topologies that the reduced blocking voltage across all switches means that high-efficiency gallium nitride (GaN) switches can be used to improve efficiency and power density.

Fig. 3: Three-level NPC PFC topology

Fig. 4: Three-level ANPC PFC topology

 

Three-level T-type NPC PFC topology

In three-level T-type NPC PFC converters, the conventional two-level voltage source converter topology is extended with an active, bidirectional switch to the dc link midpoint, shown in Figure 5. For 800 V dc link voltages, 1,200 V IGBTs/diodes would normally operate on the high and low sides on each phase, as the full voltage has to be blocked. But in the T-type configuration, the bidirectional switch to the dc link midpoint only needs to block half of the voltage. This means it can be implemented with lower-voltage devices, for example two
600 V IGBTs that include antiparallel diodes. 

Due to the reduced blocking voltage, the middle switch generates very low switching losses and acceptable conduction loss. Unlike the three-level NPC topology, the T-type NPC topology does not connect devices in series that must block the whole dc link voltage. 

In the NPC topology, switching transitions made directly from a positive to a negative dc link voltage, and vice versa, are usually avoided, as they might cause an uneven share of the voltage to be blocked momentarily when both FETs connected in series turn off at the same time. This undesirable effect cannot occur in the T-type topology. Therefore it is not necessary to implement low-level routines that prevent such transitions, or to provide for transient voltage balancing across series-connected IGBTs. 

An additional benefit of using single 1,200 V devices to block the full dc link voltage is reduced conduction losses. Whenever the output is connected to positive or negative, the circuit is exposed to the forward voltage drop of only one device; in the NPC topology, two devices are always connected in series. This considerably reduces conduction losses, making the T-type topology valuable in applications that switch at a low frequency.

Overall, conduction losses are significantly lower than those of the NPC topology, but switching losses are high because of the devices that block the full dc link voltage. So the T-type rectifier is best suited for applications switching at up to 50 kHz, while the NPC topology performs better at frequencies higher than 50 kHz. 

Fig. 5: Three-level T-type NPC PFC topology

 

Table 1 shows a high-level comparison of the advantages and drawbacks of the three PFC stage topologies described above.

	Two-level PFC	Three-Level NPC	Three-level ANPC	Three-level T-type NPC
Power Density	Low	Higher	Higher	Highest
Efficiency	Low	Very high at high frequencies	Highest	High
Conduction Loss	Low	High	High	Mid
Switching Loss	High	Low	Low	Mid
Peak Voltage Stress	High	Low	Lowest	Low/(high blocking voltage)
Cost	Low	High	Highest	Mid
Control	Easy	Mid	Mid	Mid
Input Inductor	Large	Small	Small	Small
Number of Active Switches	6	12	18	12
Number of SiC Diodes	0	6	0	0

Table 1: Summary of features of PFC topologies 

 

Dc-dc converter stage of a dc charger module

The dc-dc stage is the galvanic isolated converter which converts the incoming dc link voltage of 800 V to a regulated dc output voltage for charging the EV’s battery, bypassing the vehicle’s on-board charger, which is used only when connecting to an ac charger. 

The topology for the bidirectional dc-dc stage can be implemented in one of two ways:

• Dual active bridge 
• Dual active bridge in CLLC mode

 

The dual active bridge topology

The dual active bridge (DAB) converter consists of a full bridge with active switches on both the primary and secondary sides, connected via a high-frequency transformer, shown in Figure 6. Because of the inherently lagging current in one of the bridges, the current discharges the output capacitance of the switches of one bridge at a time. While the secondary side switches are discharged, some switches on the primary side enable zero-voltage switching turn-on. Loss-less capacitive snubbers can also be used across the switches to reduce turn-off losses. 

The main advantages of this converter topology are: 

• Bidirectional capability, which is achieved by controlling the phase angle between the two bridges 
• Modularity, which allows for it to be scaled to higher power levels
  

  Fig. 6: Dual active bridge dc-dc converter topology

  
  In single-phase shift modulation, the DAB topology is simple to control. For extended, dual- or triple-phase shift modulation, however, the control scheme becomes complex. This topology may be used to cover a wide range of battery voltages with single-phase shift modulation, but circulating currents in the transformer rise to elevated levels, which greatly reduces efficiency.

  With advanced modulation schemes such as triple-phase shift modulation, however, the converter can theoretically perform zero-voltage switching over the entire operating range. The ratio of output power to the transformer’s KVA rating is high in this topology. The output capacitance required to handle ripple currents is also low. 

  Overall, this converter is ideal for applications in which power density, cost, weight, isolation and reliability are critical factors, because of its range of attractive features: 

  • Relatively low component count 
  • Soft-switching commutation
  • Low cost
  • High efficiency

  It is worth noting, however, that the DAB converter often requires an additional shim inductor to support zero-voltage switching: this increases the size of the circuit and reduces power density.

   

  DAB topology in CLLC mode

  The CLLC circuit configuration performs all the functions of the classic LLC, but has the added advantage that the use of active switches across the secondary side enables bidirectional power transfer, which is shown in Figure 7.

  This converter operates in zero-voltage/zero-current switching mode, resulting in high efficiency. When there is room to vary the bus voltage by a margin of 10%, this converter can cater for a widely varying battery voltage while maintaining good efficiency. With a fixed bus voltage, however, this topology has a very limited range of operation. 

  The presence of capacitors on both the primary and secondary sides eliminates the risk of saturation of the transformer’s core. 

  The DAB converter in CLLC mode is actually best suited to ac-dc onboard chargers. It can though be used at power levels higher than onboard chargers handle, up to 15 kW. But scaling to higher power levels and paralleling can be difficult, as it requires a highly symmetrical tank structure and synchronization of multiple modules, which is a difficult task.
  

  Fig. 7: DAB topology in CLLC mode

  Both the DAB and DAB in CLLC mode topologies are commonly used in 800 V isolated dc-dc converters. The voltage conversion ratio controls the connection for the converter terminals, which affects the breakdown-voltage rating required for the switches: a high-voltage converter could be connected in series or in parallel in one terminal, while another remains connected in parallel. This means that there are four possible configurations for the topology’s connections.

  Two examples of the DAB topology in CLLC mode are shown: Figure 8 shows a series input configuration, and Figure 9 a parallel input configuration for an 800 V bus voltage with a wide output-voltage range of 200 V to 1 kV.

  The advantage of the series-input full-bridge CLLC is its narrow resonant-frequency range over a wide output-voltage range, resulting in lower switching losses; here, a 650 V device could be used. But this calls for more complex control of the dc bus capacitor voltage in series on the dc bus side. In addition, to achieve a given efficiency, a device with lower on-resistance is required than in a single full bridge using 1,200 V devices.

  The advantage of the parallel-input full-bridge DAB converter in CLLC mode is that, for a given efficiency, the circuit can use devices with a higher on-resistance, while the control scheme is easier. A wide resonant-frequency range is required to support a wide output-voltage range.

  Fig. 8: DAB converter in CLLC mode with series input

  Fig. 9: DAB converter in CLLC mode with parallel input

  
  The advantages and drawbacks of the two dc-dc stage topologies are shown in Table 2.

  	DAB	DAB in CLLC mode
  Peak Device Stress	Low	High
  Switching Frequency	High	Very High
  Control	PWM (simple)	Frequency (moderate)
  Wide Battery Voltage Range, Fixed Bus Voltage	Yes (with reduced efficiency)	Limited range
  Input RMS Currents	Low	High
  Conduction Losses	Low	Medium
  Turn-on Switching Loss	ZVS	ZVS
  Turn-off Switching Loss	High (device turn-off at peak leakage inductor current value)	Low (primary-side turn-off decided by magnetizing inductor current, secondary-side turn-off is zero to ZCS)
  Total Losses	Medium	Low
  Efficiency	High	Very High
  Paralleling Modules	Easy	Difficult
  Number of Active Switches	8	8

  Table 2: Summary of the features of two dc-dc stage topologies 

   

  Reference designs accelerate implementation of new dc charger circuits

  The leading SiC MOSFET manufacturers supply reference designs that provide a blueprint for new high-power dc charger designs, either in part or in whole. 

   

  The STMicroelectronics STDES-PFCBIDIR 15 kW bidirectional PFC stage converter implements the T-type NPC topology, shown in Figure 10. Digitally controlled, it converts between 400 V ac and 800 V dc. Efficiency is almost 99%. ST has optimized the passive components for both size and cost, and the converter offers high power density.

Fig. 10: The STDES-PFCBIDIR PFC reference design from STMicroelectronics is notable for its high efficiency and power density

 

A pairing of the STDES-PFCBIDIR with ST’s 25 kW STDES-DABBIDIR provides a complete solution for a bidirectional EV charger. The
STDES-DABBIDIR implements the DAB topology switching at 100 kHz with a SiC MOSFET power module in an ACEPACK 2 package, shown in Figure 11. Digital control is performed by an STM32G474RE MCU. Soft-switching operation is managed by adaptive modulation techniques, which respond to variations in the load and voltage. 

Fig. 11: The STDES-DABBIDIR dc-dc converter reference design board has a power rating of 25 kW

The Infineon REF-DAB11KIZSICSYS is a bidirectional dc-dc converter stage that implements the DAB topology in CLLC mode, shown in Figure 12. It provides an output of up to 11 kW at 800 V, at higher than 97% efficiency. 

Fig. 12: The Infineon REF-DAB11KIZSICSYS reference design board has an 11 kW output rating

Based on IMZ120R030M1H CoolSiC™ MOSFETs driven by the Infineon 1EDC20I12AH gate driver, the board combines high power density and reliability at low cost.

Infineon has also developed a complete dc-dc charger reference design, the REF-EV50KW2SICKIT, due for release in 2023. This 50 kW dc charger sub-unit is intended for use as a module in stacked high-power charging systems, and shown in Figure 13. 

The design achieves a power factor of higher than 0.95, and maximum efficiency of 96%.

Fig. 13a, 13b: The REF-EV50KW2SICKIT from Infineon implements a complete dc-dc fast charger

On release, Infineon will offer fully assembled boards that fit into a 19″ 4U rack. A power control card plus software with GUI will also be available. 

Another complete dc-dc charger design is supplied by onsemi. The SEC-25KW-SIC-PIM-GEVK is a 25 kW charger which implements the two-level PFC and DAB topologies, shown in Figure 14. 

Fig. 14: SEC-25KW-SIC-PIM-GEVK reference design from onsemi is a 25 kW dc-dc charger

The SEC-25KW-SIC-PIM-GEVK features multiple NXH010P120MNF1 half-bridge SiC modules which have a breakdown-voltage rating of 1,200 V. Notable for their very low on-resistance of 10 mΩ and low parasitic inductance, these SiC modules substantially reduce conduction and switching loss. Conversion operations are controlled by a powerful universal controller board based on a Zynq^®-7000 SoC FPGA. The output-voltage range is 200 V to 1,000 V, and efficiency is up to 96%. 

Multiple SEC-25KW-SIC-PIM-GEVK boards can be stacked together in a single cabinet to supply the output power required by the application. 

 

Conclusion

The choice of topologies in the PFC and dc-dc converter stages of bidirectional EV fast chargers gives the designer the choice of optimizing for size, cost, efficiency, output power, component count, and ease of control. 

The availability of high-performance reference designs from the leading SiC device manufacturers gives designers a head start in implementing some of these topologies.