Facilitating the interface between the power grid and the high-voltage battery inside an EV requires an isolated DC/DC power stage: two popular choices are the CLLLC and DAB topologies.
In order to optimize power in electric vehicles (EVs), the onboard charger (OBC) must be highly-efficient, light in weight, and small in size. A lighter EV also requires less power to move the vehicle, which increases overall efficiency.
The OBC needs to support an appropriate grid-to-vehicle (G2V) voltage and current battery-charging algorithm; as such, it functions as the power-conditioning interface between the power grid and the EV (Figure 1). In addition, it must be able to provide power from the vehicle-to-grid (V2G) so that the EV can supplement renewable energy sources that might have a fluctuating peak capacity.
Figure 1 The OBC needs to support an appropriate grid-to-vehicle (G2V) voltage and provide power from the vehicle-to-grid (V2G).
Facilitating the interface between the power grid and the high-voltage battery inside the EV requires an electromagnetic interference (EMI) filter, power factor correction (PFC), and an isolated DC/DC power stage. Figure 2 illustrates this architecture.
The scope of this discussion is limited to the DC/DC stage. As of this writing, two popular choices for the DC/DC stage are the capacitor-inductor-inductor-inductor-capacitor (CLLLC) and the dual-active-bridge (DAB) topologies (figures 3 and 4). Both options can achieve a small solution size and provide the necessary G2V and V2G power demands.
Figure 3 This schematic show the basic topology of the CLLLC.
Figure 4 The DAB topology is shown in this schematic.
Maximize OBC performance and minimize its size
To understand how these two topology options can affect the size and performance of the OBC, let’s further limit the scope to the battery-charging phase of operation, or G2V, considering how to minimize the charging time by providing the maximum battery power that the switches can tolerate. For example, consider a switch under the following operating conditions:
The switch will have a TJ = 125°C, according to equation 1:
The switches in this design cannot tolerate a temperature above 125°C; therefore, this condition represents the highest power level that the OBC can provide to the battery without compromising the switch. The goal is to minimize the power dissipation in the switch and charge the battery as quickly as possible.
Two main factors drive the majority of power losses in the switches: root-mean-square (RMS) current and the switch’s ability to maintain zero-voltage-switching (ZVS).
Given their low capacitance and fast turn-on and turn-off characteristics, Texas Instruments’ GaN switches enable the converter to operate at a higher switching frequency than what would be possible with silicon. Higher-frequency operation directly affects the size of the reactive components and results in a smaller transformer, inductors, and capacitors. Let’s start by establishing a baseline design for both the DAB and CLLLC, and then explore a circuit enhancement to extend the ZVS range of the converters.
Baseline DAB and CLLLC performance comparisons
Table 1 outlines the basic requirements for the OBC.
Table 1 OBC power requirements
Creating a detailed design for both the DAB and CLLLC helps determine the most viable tank designs. The procedures for doing this are beyond the scope of this discussion; however, circuit simulation is best for adequately approximating the losses in the switches and verifying compliance with the overall functionality. I configured the simulator to run in batch mode over different power levels and input and output voltages and tested different DAB and CLLLC inductor, capacitor, and turns-ratio values. In each simulation run, I collected data on parameters such as VIN, VOUT, switch power, RMS current, and switch ZVS conditions. Table 2 summarizes the two optimized topology designs.
Table 2 DAB and CLLLC optimized designs
Figure 5 illustrates the salient simulation results. While there are eight switches in each topology, the graphs plot only the switch with the highest power loss. For each switch, there are three plots. The first is the total losses in the switch. The second is the RMS current through that switch. The third plot, on the far right, shows the worst-case drain-to-source voltage that a given GaN switch experiences at turn-on. This is a figure of merit of how much ZVS has been lost; the higher this voltage, the larger the losses will be in that switch. Thus, the switch’s RMS current coupled with its ability to maintain ZVS represents the greatest portion of power losses in the device.
Equipped with these facts and a careful examination of the data, it’s clear that the CLLLC is able to maintain ZVS over a wider range of operation. Enhanced ZVS is therefore responsible for the lower power losses seen in the CLLLC switches. Having said that, at 6.6 kW of operation, the DAB has superior performance, which comes from good ZVS and reduced RMS current over most of the range. These observations suggest looking for a way to improve ZVS without adversely impacting the RMS current.
Improving ZVS with commutation inductors
Figures 6 and 7 show the same CLLLC and DAB circuits as figures 3 and 4 with extra inductors (highlighted in yellow) added to the topologies to provide the additional current required to maintain ZVS over a wider range of operation. For now, consider a case when these extra inductors are operational all the time.
Figure 7 This schematic shows the DAB with commutation inductors.
Table 3 lists the values for the new inductors, and repeats the other tank parameters for convenience.
Table 3 DAB and CLLLC designs with commutation inductor (LC) values
Figure 8 shows the results after repeating the simulations in Figure 5.
In this case, notice that the DAB is able to achieve full ZVS over the entire range of operating conditions. This is clearly illustrated by the fact that the VDS of the GaN switch is always at 0V at turn-on. The CLLLC, while not achieving full ZVS, is able to achieve significantly-improved ZVS. Also notice, however, that the ZVS improvement comes at a significant expense to the RMS currents in both topologies. Looking at the power losses alone, it appears that the DAB converter has the advantage over most of the range.
Before going too far, go back and compare Figure 8 to Figure 5, and you’ll notice that under some conditions the commutation inductor actually makes the losses worse. This begs the question – is it possible to create a hybrid approach where you can achieve the lowest of the losses shown in Figure 5 and Figure 8?
Minimizing total losses: have your cake and eat it too
The addition of a commutation inductor creates a broader range of operating conditions where the converter maintains ZVS. This is of tremendous benefit when the converter can’t maintain ZVS. The problem with a commutation inductor is that it only improves the losses when ZVS would otherwise be lost. If the converter is already in ZVS, the commutation inductor hurts operation by increasing the current, which results in more ohmic loss in the switches.
This thought process leads to the testing of a hybrid approach where the commutation inductors are left off at the heavier loads and turned on at lighter loads. Figure 9 shows the results after repeating the simulations with this approach, which enables the design to harness each topology’s lower RMS currents and natural ZVS ability at heavy loads.
I was careful to add only enough commutation inductance and operation time to fit within the thermal envelope of the switches, in order to prevent unneeded RMS current to the switches or an unnecessary solution size. Note that the DAB converter does not achieve full ZVS over the operating range. The ZVS conditions are much improved, but only as much as they are needed to stay within in the 20-W switch target previously discussed.
In order to better visualize the trade-offs, Figure 10 summarizes the power losses for each case. You can see that the DAB converter has a clear advantage in terms of power losses in the switch.
To better illustrate the performance capabilities between these two converters, Figure 11 reformats and plots the data shown in Figure 10. The graph shows the maximum power that each converter can supply, assuming that the switch cannot safely dissipate more than 20 W of power. Remember, 20 W represents the largest loss the switch can tolerate and still keep the junction temperature below 125°C.
Figure 11 This plot shows the maximum power that each converter can supply.
Is CLLLC or DAB better?
As evidenced by the fact that the blue line is above the red in Figure 11, the DAB converter can provide more power over the entire range than the CLLLC. This makes it tempting to assume that the DAB is the clear winner. However, remember that minimal size and weight are central requirements of an OBC. The DAB converter needs two extra inductors, but the CLLLC only needs one. In my opinion, that gives the win to the CLLLC.
Like most things in engineering, what’s best is almost always a matter of trade-offs against the requirements. It’s pretty rare that big advantages come for free, and in this case it’s no different. To me, the CLLLC edges out the DAB because it appears to have a clear size advantage. Please comment if you have any thoughts on which option is better, especially if you disagree, and why.
This article was originally published on EDN.
Brent McDonald is a system and applications manager for the Texas Instruments Power Supply Design Services team.