20. Februar 2020, 14:00 Uhr | Ralf Higgelke
DESIGN&ELEKTRONIK editor Ralf Higgelke met Alex Lidow (right), CEO of GaN pioneer Efficient Power Conversion (EPC), in Munich in December 2019.
Alex Lidow is the CEO of Efficient Power Conversion, probably the most prominent advocate for gallium nitride, delivering the first GaN transistor in 2009. After a decade of selling products, DESIGN&ELEKTRONIK editor Ralf Higgelke met him to discuss some of the latest advances in that area.
DESIGN&ELEKTRONIK: Is the sixth generation of your eGaN devices already out on the market?
Alex Lidow: No, it isn’t. And by the way, we are not calling these Gen6 anymore. It’s too confusing to the people. But we have developed a series of products with a higher current density that will be released in the next few months (see Fig. 1).
How did you achieve this higher current density?
We shrunk the devices for the same current by increasing the metal layers, in particular thick copper layers on top. Furthermore, we’ve changed the connection technology from a ball grid array to a land grid array (see Fig. 2).
A big topic with GaN HEMTs is driving these devices. I learned that p-doping is very hard to do. So how do you create the driver and protection circuitry and in your eGaN devices?
Actually, gallium nitride is a very good at conducting electrons by this two-dimensional electron gas, or 2DEG. Without an efficient hole conduction we cannot make CMOS-like circuitry, that’s obvious. So in order to drive a monolithic half-bridge, for instance, we need to boost the supply voltage for the gate driver beyond the DC-link voltage. We integrate that boost circuitry into our ICs.
Fig. 1: Instead of a land grid array (LGA; top) EPC is using for the latest products a ball grid array (BGA, botton).
Well, two-dimensional hole gas is not really new at all. This concept first appeared at least some 15 years ago. The Cornell people claim that they have reached a higher hole mobility than before. But I could not find any figures on how high that mobility really was. If you are not able to get a hole mobility of more than 200 cm²/(V∙s), such a p-channel device was not efficient enough to drive a n-channel device or to complement n-channel devices as we find it today in silicon with CMOS. I personally believe that they did not achieve a hole mobility that is high enough.
The GaN landscape is changing dramatically. In July 2019, Power Integrations disclosed that they were already shipping GaN in volumes since a while, and in November Nexperia announced the release of their first GaN product. But all of these are high-voltage products for 600 or 650 Volt. In the low-voltage area EPC is still the only vendor, isn’t it?
As far as I know, GaN Systems has low-voltage GaN in their portfolio. Infineon has announced low-voltage GaN, but they don’t have it yet. The same applies to several other manufacturers. So I’m enjoying the fact of being more or less alone. There are a lot of good applications for low-voltage GaN, but there are also a lot of good applications for 600 Volts, too.
EPC reduced the distance from the gate to the drain electrode when transitioning from Gen 4 (left) to Gen 5 (right). This enables tighter packaging of transistor cells and reduces the on-resistance of each cell.
Power Integrations is now manufacturing GaN in volumes, using these devices in some of their products. They are not integrating it monolithically, but co-packaging with a silicon-based driver and their proprietary FluxLink technology. As reported, the company is not using silicon as a substrate, but sapphire. What do you think about using a different substrate?
I think Power Integrations was under pressure in the AC adapter business. They had a big market share in that area, but they were experiencing pressure by design wins from companies like Navitas. So they had to do something about it. As silicon is reaching its theoretical limit, the only way to enhance their products was to make use of gallium nitride.
They incorporated in the same package a depletion mode GaN device, the most primitive one you can find on the marketplace. I think it is of limited use, but by using their special control techniques Power Integrations was able to enhance the performance of their products surgically. But you won’t see discrete GaN devices on sapphire, because this material is such a bad heat conductor. Therefore, it is not suitable for high power applications. But for an AC adapter it’s sufficient.
We are talking with each other every now and when on new applications for gallium nitride. In the talk you gave today you stressed the topic 48 V in automotive applications. What have been the main points of your presentation?
I’ve been talking about two different applications for GaN in automotive. One is already there today which is Lidar. Lidar is famous to be used in autonomous cars, but it is not limited to passenger cars. It is also used in unusual autonomous vehicles like delivery pods and autonomous drones. The other big market for GaN in automotive is going to be the emerging 48 Volt to 14 Volt bidirectional DC-DC conversion in mild-hybrid cars. This is going to ramp up in 2022/23.
In Lidar applications gallium nitride transistors are not the best components, they are the only ones that are capable to fire the laser with the accuracy needed. In the 48 Volt to 14 Volt bidirectional DC-DC conversion there are also silicon solutions available. So this will be a head-to-head competition for cost effectiveness.
Alex, many thanks for taking the time.