Author:
Ally Winning, European Editor, PSD
Date
07/26/2022
PDF
Dr. Shahin Sharifzadeh has over 30 years of semiconductor experience and was most recently the SVP of World-wide Operations at Atmel. Before Atmel, Shahin was the EVP of Technology and World-wide Manufacturing at Cypress.
GaN has already transformed the power electronics industry, but there may be much more to come. The technology has matured somewhat, and each new generation of GaN devices brings incremental benefits - switching gets faster, efficiency rises and power products get smaller. Those gains will continue into the future as better manufacturing techniques are implemented and new ways of working with the compound are found. However, there could be a greater leap forward in performance on the horizon, and according to Dr. Shahin Sharifzadeh, co-CEO of NexGen Power Systems it may be here sooner than you think. NexGen has developed a method of building GaN transistors on a GaN base, what the company terms vertical GaN, and the figures that it claims so far seem promising.
Competitors build GaN devices by growing it on silicon and other substrates, in a typical factory, using many of the same processes as they do for silicon. However, the difference in the crystal alignment between the two different materials means that GaN on an unmatched substrate is not as robust as it could be, and that limits the voltage that can be handled by the transistors. The vertical GaN technique does not have the same crystal mismatch and therefore will be able to handle much higher voltages, while switching much faster. The limit to voltage handling only being determined by the doping concentration in the channel. NexGen has also taken a different route with the technology’s manufacture, and uses its own dedicated factory to produce GaN materials. The process is not based on the well-understood methods other companies use for GaN on substrate technology, as silicon can contaminate NexGen’s process, and even act as a doping agent in the GaN semiconductor itself.
Sharifzadeh claims that the move to vertical GaN is just a natural progression and other semiconductor technologies have taken the same route. He clarified, “You start with surface channel devices, or surface devices, and then you have to move to vertical devices and the reasoning is pretty simple. You need to drop the voltage to reduce the electric field over a distance, and as the voltages become bigger and bigger, that distance becomes larger and larger. So if you have a surface device to air it just becomes too big and cost prohibitive. That’s why vertical devices are the most efficient type of transistors. Vertical devices also have the additional benefit that when you reduce the area, it has a direct impact to the capacitances and also the switching speeds.”
Of course it is not all upsides. Normal GaN transistors are more complicated to drive than silicon MOSFETs, and systems using vertical GaN transistors are more complex still. Sharifzadeh explains, “GaN on silicon switching peaks at around one megahertz, while GaN on GaN can go all the way up to 10 megahertz or higher. That makes the system a whole lot more efficient. However, designing at 10 megahertz is not something that has been done before in power systems. Designers are currently moving from 10 or 20 kilohertz to 100 kilohertz, and vertical GaN asks them to move to between 5 and 10 megahertz, which brings some whole system issues”.
The NexGen solution to those issues, at least in the short term is to take control of the whole process – from growing the GaN crystals to developing products using the resulting transistors. The first product using vertical GaN will be available soon for a high value application for the consumer industry, possibly in a power supply for a gaming laptop. Sharifzadeh thinks that the way to prove the technology will be in a large scale consumer application. After that he says, NexGen will look to partner with experts in other areas, possibly automotive. He explained, “There might be some markets that we seek partners to work with our chips. Those markets are very complex, such as the automotive market. We probably don’t have the infrastructure to do an inverter or a motor drive. People have invested billions of dollars in making those. In those cases, we are going to try sell modules or even smaller systems, but that is quite some distance away”.
The technology sounds very interesting, and NexGen has posted some impressive figures on the company’s website about handling SiC type voltages and switching 5 – 10X faster than GaN on Silicon technology. If the physical devices back those type of figures up, then it could be a really exciting time for the industry.
