Understand practical GaN and SiC differences for EV onboard chargers
The debate continues around choosing wide bandgap (WBG) semiconductors over standard silicon, particularly in power conversion applications. The main advantage put forward is the higher switching efficiency of gallium nitride (GaN) and silicon carbide (SiC) devices.
However, WBG devices are also more expensive than silicon. Moving to WBG pivots on the value you gain from lower losses. GaN-based high electron mobility transistor (HEMT) cells have been traditionally expensive to make, with low yield from high lattice defect rates in the wafers. However, this has improved tremendously. The simplicity of HEMT cells should eventually mean devices are cheaper to produce than devices with multiple epitaxy layers, which includes SiC MOSFETs.
While this transition progresses, choosing between GaN and SiC for use in new designs is not easy. As average selling prices approach parity, the main consideration may become the practical implementation. Is it viable to design power converters using both device types, and what would that do to the overall system cost?
GaN and SiC material comparisons
If efficiency is the goal, a relevant headline material characteristic of GaN and SiC is electron mobility for switching speed and consequent low dynamic losses. GaN is around 1700 cm2/V-sec, which is similar to Si, at 1450. Electron mobility for SiC is only around 900 cm2/V-sec.
Thermal conductivity for GaN as a material is relatively poor, 3.5 times lower than SiC, and even a little worse than silicon. This means that while GaN on-resistance per unit die area might theoretically be 100 times lower than SiC, removing the heat for acceptable junction temperature rise is more difficult than with SiC.
The maximum junction temperature for SiC is higher than GaN, but both can stand more than 450°C, so in practice, packages set the limit to a much lower common value. The picture is a little mixed so far, but another significant difference is the critical breakdown voltage of the materials – GaN is 3.5 x 106 V/cm while SiC is 3 and Si is 0.3. This means that WBG devices can be much smaller for the same bulk breakdown rating than silicon and therefore device capacitances can be much lower, which reads across to faster switching speeds.
However, lateral GaN HEMT cells are limited by surface breakdown effects to a rating of about 650V, while vertical trench SiC MOSFETs are available commonly at 1200V and sometimes higher. SiC devices also have an avalanche rating for over-voltage which GaN cells do not have.
Material differences
A summary of some major GaN and SiC material differences
Gan and SiC device comparisons
There are some characteristics of GaN and SiC devices that affect their application directly. In addition to the-already mentioned available voltage ratings, the presence of a parasitic body diode in SiC affects dynamic losses, particularly in hard-switched converters. GaN has a body diode effect so it can conduct in reverse but with no recovery losses.
When either device conducts in the third quadrant, they drop a relatively high voltage, around 4V for SiC and 1.3 to 6V for GaN, depending on its gate drive arrangement. This causes dissipation if the dead time when they might self-conduct in reverse. This is a significant proportion of the switching period and increasingly so as the switching frequency goes higher.
Gate drive complexity is also a difference between SiC and GaN. A SiC MOSFET has a relatively low threshold voltage of around 2.5V but needs around 18V for full enhancement. This is worryingly close to its absolute maximum, which might be 22V. GaN starts to conduct at around 1.3V on its gate so has a small noise margin. This creates the risk of spurious turn-on and potential shoot-through and failure if the gate drive is not carefully controlled. Its absolute maximum is also only around 7V
The WBG choice in EV on-board chargers
Having discussed some of the practical differences between SiC and GaN switches, let's look at how each might fit into an EV onboard charger (OBC). These are currently rated between about 1.2kW and 7kW for single-phase types and more rarely up to around 22kW for three-phase supplies. The OBC application is an ideal candidate for WBG switches, as power density and weight are prime considerations. A modern high-efficiency OBC design feature a ‘Totem-Pole’ PFC stage, followed by an isolated ‘CLLC’ DC-DC conversion stage.
OBC topology
This possible EV OBC topology is based on a totem-pole PFC stage followed by an isolated DC-DC conversion stage
In this example, the totem-pole PFC stage is operated in hard-switched continuous conduction mode (CCM), where the high frequency switches Q1 and Q2 are forced to conduct in reverse during part of the switching cycle. CCM is chosen because it keeps peak currents lower and L1 can be smaller but at the expense of potentially higher EMI. GaN devices are a good choice here as there are no reverse recovery losses, which would be significant with SiC MOSFETs. Voltage stress with the single-phase supply shown is limited to the AC peak and is well within the range of 650V GaN devices. A secure input EMI filter is needed however to avoid AC mains transients stressing the GaN switches. Q3 and Q4 could be simply silicon MOSFETs or even diodes as they only switch at line frequency and dynamic losses are negligible.
The CLLC stage comprising Q5-8 is resonant and the body diodes of the devices do not conduct in normal operation. Because of this, SiC MOSFETs could be chosen to have similar dynamic and static losses to GaN but at a lower cost and with less critical gate drive, especially for the two high-side devices.
The output synchronous rectifiers, Q9-12 also switch resonantly and can be SiC devices for batteries up to 800V or even a little higher. At lower voltages, say 100-400V, GaN cells will give fewer dynamic losses from charge and discharge of lower device capacitances along with potentially lower conduction losses, for ultimate efficiency, if premium parts are specified. GaN could be used for the output switches at 800V by stacking lower voltage windings from the transformer with separate rectification stages. This increases the device count and cost, however.
The higher voltage from a three-phase AC supply would suggest using 1200V SiC MOSFETs in the PFC stage, as the voltage stress in a basic three-phase totem-pole PFC stage would approach the 650V limit for GaN. This would mean suffering significant reverse recovery losses, but GaN could still be used in a multi-level input topology such as the three-phase Vienna rectifier where switch voltage stress is halved.
Conclusion
The costs of GaN and SiC switches have become closer, and while still higher than silicon for comparable headline volt and amp ratings, the efficiency gains in an EV OBC application can deliver overall system cost savings. These appear from lower energy bills, reduced heatsinking, and smaller magnetics and capacitors from higher switching frequencies. At the same time, battery charge time is shortened and the reduced weight and size of the charger help to maximize cabin space and driving range. The choice of GaN or SiC still depends on the detail of the application but both have benefits over silicon.
Resources
This article focuses on onboard chargers, but the technologies discussed are also applicable to power conversion in charging infrastructure. For more:
About Author
Avnet Staff
We use Avnet Staff as a collective byline when our team of editors and writers collaborate on the co...
silica content library/articles/understand-practical-gan-and-sic-differences-for-ev-onboard-chargers
Understand practical GaN and SiC differences for EV onboard chargers | Avnet Silica
