Something I used to say is that if you or your customer doesn’t have an application where you have to switch more than thousand volts at greater than 20 kilohertz, you don’t need silicon carbide. The value proposition is there, but it’s not that compelling, yet. However, the advances in topologies and demands have driven many applications to 1,200 volts and above.
By Alix Paultre, Editor Power Electronics News (based on a technical paper presented by Ranbir Singh, GeneSiC)
In the case of electric vehicles, for example, there is a compelling case to hook directly to the electric grid at either 4160 or 13.8 kV, and that is necessitated by the need to fast-charge, or in data centers, where companies like Google are talking about direct conversion from full grid voltages down to even 48 volts or sometimes 12 volts, that’s quite a challenge, and again, requires the need for very high voltage devices. Traction systems, renewable energy, and other new and legacy application spaces are moving to higher voltages.
So what does an end-user want from want from a switch? Well, silicon carbide has an ability to achieve very high switching speeds. Turn on and turn off times are of the order of less than 50 nanoseconds. Ideally, it should have a positive temperature coefficient of on-resistance. This has been the beginning of silicon IGBTs for a long time, especially in high voltage regimes. Silicon carbide offers a positive temperature coefficient and a higher operating temperature. The amount of cooling required for high voltage devices is often overwhelming, and often consumes more energy than the losses in the circuits. High reliability is a given, and is pretty extensive in terms of what that means. Things like high avalanche energies are part of the robustness a device, as well as good short circuit capability and the like.
GeneSiC has been developing silicon carbide MOSFETs, initially with an emphasis on the high voltage side, and the figure below shows some of the results from our 4,600-volt MOSFET, with the reverse for blocking characteristics and forward characteristics. Features include a fairly low on-resistance for 4,600 volts.
This is near the theoretical that we are able to achieve. The figure below shows the RDS(on) increasing with temperature, almost exactly to the bulk temperature performance of such devices.
Threshold voltage changes with temperature and there are different definitions of threshold voltage, so it is important to look at it carefully. We define our threshold voltage at a drain current of 5 milliamps. It’s what a lot of people follow in silicon carbide, and it is a well-behaved reduction in threshold voltage from room temperature up to 175 degrees C, and follows near the theory of threshold voltage shifts. We experimentally demonstrated a positive temperature coefficient of breakdown voltage, so the blue curve is at 25 C and 100 C, you can see that the breakdown voltage increases.
The following figure shows the switching performance of these devices for various drive conditions, from 20 volts to -3.3 and 15 to -3.3, showing extremely fast switching speeds of less than 30 nanoseconds. The fall time was less than 30 nanoseconds, even at 1,800 volts, and this is not going to change much, even if you were to be close to 3,300 volts or so. For higher voltage devices, 4,600 volts in this case, but even higher voltage devices, the switching speed is not a limitation fundamentally from the device.
Driving at speed
The key is going to come from how slow you want to drive the MOSFETs, and that is a huge mindset shift. If you look at most of the silicon world, they are almost always limited by the switching speeds of the devices themselves, be it IGBTs, or Thyristors, or whatever. It’s many orders of magnitude bigger than what you can make changes with, meaning whatever’s in your control.
GeneSiC has measured the switching energies loss, the switching energies that were measured, and in the figure below, the energy is measured in microjoules versus different gate resistances. There’s a 22% decrease in Eon when gate driver voltage is increased from 15 volts to 18 volts, and further increase in dry voltage yielded only 9% improvement.
Threshold voltage stability is a new thing for silicon carbide, as it sometimes shifts a little bit as you stress the gate of the device over long periods of time. The next figure shows that under positive gate bias stress, we do see a 0.35 volt shift in threshold voltage, but then it stabilizes when a stress of 20 volts is applied. This is actually somewhat common in many silicon carbide MOSFETs. In the negative bias, however, we see no shift here.