GaN-based high-power EV inverters
Article By : Alex Q. Huang
The roadmap to increased power density starts with reducing the conduction dynamic losses.
The roadmap to increased power density starts with reducing the conduction dynamic losses. Even more than silicon carbide, gallium nitride can substantially reduce dynamic loss and therefore can reduce overall loss. So, this is one approach to get to high power density in the future.
A second parameter is the thickness of the overall inverter stack; a design with a flat and thin inverter housing is important. The third possibility is to increase the temperature of operation. The high-temperature roadmap calls for operating the device at 175°C and, even more, at 200°C in the future.
For the first aspect, reduced losses, we need to move to silicon carbide- and gallium nitride-based power devices. For EV applications, we’ll talk about devices in the range of 600 to 1,200 V. Here, silicon carbide and gallium nitride devices can outperform even silicon bipolar devices such as IGBTs, reducing conduction loss and switching loss substantially.
This is true for a couple of reasons. First, they are majority carrier devices, meaning they do not have any minority carrier storage effects. GaN’s lateral construction gives it a slight advantage over silicon carbide in terms of dynamic or switching losses. Compared with a SiC equivalent (vertical device, inversion channel), a 600-V GaN device’s high-mobility 2D electron gas channel enables a lower channel resistance and therefore a smaller chip size, and the lateral structure also allows a lower capacitance.
Looking at on-resistance times input capacitance, output capacitance, and reverse-recovery characteristics, we see that 600-V GaN and SiC wide-bandgap devices outperform a silicon superjunction equivalent. Ron × Ciss is an indication of how fast you can drive your gate loop. SiC and GaN yield a faster gate loop, reducing switching loss, but GaN is substantially better than SiC, partially because of the lateral structure and partially because the design rule used is deep-submicron CMOS. The second figure of merit is Ron × Coss. This again is a reduction of the turnaround losses, because the capacitance will store energy that will be dissipated when you turn on the devices. Third is the reverse-recovery characteristic (Ron × Qrr); here, silicon carbide is substantially better than silicon and GaN slightly better than SiC. So together, we can see that the reverse recovery in GaN and SiC are basically eliminated. At 20 kHz, which is typical for an EV inverter, the switching loss is almost negligible, so conduction loss will start to dominate the overall losses. These I2R losses can be reduced proportionally by using more and more devices in parallel as long as there is sufficient space to put the device in your AV inverter, which of course is directly related to the power density.
Of course, one important parameter for increasing power density is the packaging form factor, or the thickness of the device. Here again, we think GaN has the advantage over silicon carbide.
This article was originally published on EEWeb.
Alex Q. Huang, professor, University of Texas at Austin
