With high-voltage GaN devices close to commercialisation, manufacturers can, at last, look forward massive market growth. Compound Semiconductor talks to Philippe Roussel from Yole Développement to find out more. 

At last the GaN power industry expects to see market growth.

After several false-starts, the GaN power electronics industry finally looks set to take-off. Latest reports from France-based semiconductor analysis business, Yole Developpement, predict a hefty 80% CAGR for the market from 2016-2020.

As author, Philippe Roussel tells Compound Semiconductor: "We've lost two years. 600V devices were supposed to be with us two years ago, but these were not ready. These devices do now exist, but most are under qualification for another year to a year and a half."

In the meantime, 200 V GaN-on-silicon devices are seeing take-up in point of loadapplications, from servers and routers to switches and general-purpose point of load DC-DC converters. And as the likes of EPC and International Rectifier reap the first revenues from these devices, electric vehicle, motor drive and photovoltaic inverter manufacturers, to name but a few, are still holding out for the much-trumpeted 600 V GaN HEMT.

"There's been a lot of venture capital put into this technology, so maybe it had been a little over-promoted," says Roussel. "Today you cannot buy a 600 V GaN HEMT on Digi-Key or Mouser. You can get the devices if you are collaborating with the big manufacturers, but you and I cannot access them at the moment."

The analyst reckons the power supply and power factor correction sector will be the first major market for high voltage GaN devices, representing some 50% of device sales from 2015 to 2018. From here, electric vehicle makers are expected to start snapping up the devices, with manufacturers of PV inverters, wind turbines and rail traction systems close behind.

"Reverse costing simulations comparing silicon IGBT-based inverters with the same in GaN indicate come 2016, you can really make some savings with GaN at the system level," says Roussel. "Overall, 2020 could see an estimated device market size of almost $600 million, leading to approximately 580,000 six inch wafers to be processed."

Cost reductions come from the expected price erosion of GaN devices as market take-up increases, with savings on passive cooling, thanks to GaN's faster switching, also playing a key role. Factor in incremental improvements in conversion efficiency, and Roussel expects to see payback times of less than a year.

"So, from 2016, it seems all the necessary technical and economic parameters are in place," he adds.

Rival technologies
But 600 V GaN-on-silicon devices are not the only viable high performance technology for power electronics applications. The PV inverter sector has already adopted the SiC MOSFET with many products commercially available. According to Roussel, this qualified flavour of transistor, represents GaN-on-silicon's biggest threat.

"The main interest for the SiC MOSFET has been its chip to chip replacement. It's a normally-off device, like the silicon IGBT, so you just remove your IGBT, recompute your driver a little, and put in your SiC MOSFET," he says.

In contrast, the HEMT is traditionally normally-on device that can be turned off by applying a negative gate bias voltage, but this is a challenge to apply in typical power electronic circuit topologies, demanding different drivers and a system re-design. So, to rival the manufacturers of normally-off SiC MOSFETs, GaN players have developed and are now promoting the enhancement-mode HEMT.

Here, the design of the GaN HEMT is altered to shift the gate threshold voltage from negative to positive, yielding a normally-off E-mode devices. These HEMTs are already available up to 200 V, and 600 V will follow.

"The price also needs to be lower to give the same overall [system] cost as SiC MOSFETs," says Roussel. "But to my knowledge, PV inverter manufacturers are technology-agnostic; if they can find the same cheaper, they will buy it."

The analyst reckons GaN device manufacturers will also have to work on their supply chains. SiC devices are qualified and available from multiple sources of supply, while GaN supply is less established. "We see a lot of announcements from International Rectifier, Transphorm and so on, claiming to have 600 V normally-off GaN, but will these manufacturers all be ready at the same time, supplying roughly the same product with the same specification so you can easily switch from one to the other?" he asks. "We don't know yet, it's too early to say."

But what isn't too early to predict, is the inevitable industry consolidation. Late last year, Japan-based Fujitsu −the world's third-largest IT services provider −scooped up GaN power device manufacturer, Transphorm, so together the companies could drive high-volume, high-performance GaN device production forward.

Roussel describes this as a 'smart approach' for both and expects more of the same. Highlighting start-ups such as China-based CorEnergy, ExaGaN, France, Dynax Semiconductor and Avogy, both in the US, he says: "These have totally different business models and are fabless or fab-lite. [Such companies] have to find a larger manufacturer."

"Licensing technologies is the natural evolution of this business," he adds. "There are tonnes of GaN HEMTs to be produced and this isn't compatible with the small start-up. I am not sure all these companies will remain in the next two or three years; the assumption is they will be absorbed."