The traction designs are a little different and are currently dominated by silicon IGBTs and SiC MOSFETs. “Traction inverters are sized for power and that situation is dominated by conduction losses but there’s little differentiation between Gan and Silicon and SiC,” he said. “But when you look at the load profile more of the time is spent at low load, around the city, and GaN dominates with the switching losses at low load so the way the market is bifurcating is the high voltage and for medium and low voltage systems they are moving to GaN.”
Another issue is the qualification of parts. Several suppliers have now announced AEC qualification of GaN devices. “We have detailed test programmes with all the automotive manufacturers for AEC Q+, they have all the test results,” said Witham.
However GaN is an enabling technology and that needs a broad portfolio, from wireless chargers in consumer designs to highly efficient converters in data centres. “The approach that we take is these are the basic building blocks for any power system from 25W to 25kW,” he said.
“For the data centre we have power supplies from Bel Power that provide twice as many watts into the same space that frees up extra space for servers and memory in the rack, and in industrial designs the smaller power supplies are a third the size for PLCs.”
That can bring challenges for a growing company.
“The winners in this GaN business will have a full product line, but you do have to focus as application support becomes critical,” he said. “We probably have ten different semiconductor IC companies with reference designs with our chips, we have half a dozen micromodule makers, and a dozen module companies producing power modules for high power systems so we have a very active power partner programme as you have to build the system around the transistor and we need partners.”
The sweet spot for GaN is 100 to 1200V and from a few amps to hundreds of A, and then its taking these basic building blocks and coupling them with other things, he says. However there’s a long path to integration and that may not be the right direction.
“Integration is a three step process starting from discrete and getting customers to buy them,” he said, “Then you integrate separate die into a micromodule to get the best of breed components and then in some cases where it makes sense from a cost perspective, in low power, high density applications, it makes sense to integrate onto a single chip. That last step only occurs in really low power density applications and it’s not clear that chargers are the right place.”