Full Speed GaN for High-Power Applications

Author:
Ally Winning, European Editor, PSD

Date
11/07/2023

 PDF
A Cambridge, UK, company has taken a new approach to bring out the best in GaN technology for high-power applications

QTP

Rob Gwynne, CEO and CTO of QPT

­

One phrase I never expected to hear in the power industry was, “sorry I was delayed, I was arranging delivery of a vector network analyzer”, but that is what happened when I called Rob Gwynne, CEO and CTO of QPT. VNAs are usually pieces of equipment used to test high frequency communications systems rather than power ones. However, once the discussion began, the it was clear how the VNA will be a core tool for QPT’s design.

 

GaN transistors can switch really quickly, at around 2 MHz, or at least they can in low power applications. As the power increases, EMI starts to be a problem for GaN systems. For many applications this is not a huge issue, but for heavy duty applications, such as electrical motors, designers usually temper the switching speed so they do not have to deal with EMI issues. Current high-power GaN systems typically only switch in the tens and hundreds of KHz, rather than at the 2 MHz level where the full benefits of the wide-bandgap material can be leveraged. By definition, high power applications have higher losses, even when operating at the same efficiency as lower power ones, but they are currently not even meeting the same efficiency level because of the slower switching speeds. Efficiency is important, especially for electric motors, where it is estimated that they use around 45% of all electricity generated globally.

 

QPT has taken a different approach to the problem. Instead of throttling the speed of the GaN transistors to prevent EMI problems, the company allows the transistors to work at full speed and then deals with the EMI. Gwynne’s background includes a period in RF engineering, where large amounts of power are often switched even more quickly than in power systems. He describes the current situation in the industry as, “GaN manufacturers design their circuits to slow down the transistor to make it easier to pass EMI testing. Unfortunately, by slowing down the GaN transistor, you can only really use it in resonant or quasi resonant systems. With zero-voltage switching it doesn't really matter if the transistor switches a bit slower as that doesn't affect efficiency.”

However, for higher-power applications, he says, “The motor industry uses a topology where a half bridge drives an LC filter, and to do that you are hard switching a DC bus. It's not the only way to do it, but it's the way everybody does it. Motors can be driven with resonance supplies, but it's incredibly complex and difficult. It would require the whole design process to be changed - the controller, control algorithms and topology, as well as retraining of engineers. If we are to persuade the power industry to change from SiC or silicon to GaN for high-power applications, then we have to have hard switching using today’s topologies, and that requires switching as quickly as possible”.

What that means in practice is a system built around GaN System's 650V GaN transistor. The transistor was the only one that QPT had found that had no additional circuitry to slow the transistor, so it was possible to run it at full speed. However, the rest of the task was much harder, as Gwynne explains, “The RF power was huge. It was switching very fast and producing very fast edges. We couldn’t use existing components or we’d have failed every EMC tests on the planet. The only way to fix it was to broadband match the output of the half bridge to the filter, and then the filter must have a self-resonant frequency up into the multiple gigahertz. These are the reasons that we required a VNA. We also had to design active clamping systems which broadband terminate the output and special output inductors. On top of that, we designed a new driver, as existing drivers would have blown up at the speed we were operating. All these components were then packaged in a module. Those weren’t the only hurdles we faced in the design – SPICE models were inaccurate at that frequency, so we had to make our own, and we are also also currently working with a major T&M provider to develop probes, as no probe available could measure 600 volt peak-to-peak with one nanosecond transitions”.

The company’s first qGaN module (Q650V15A-M01) will handle 15A RMS current driving 380V three phase motors. Its roadmap will have qGaN modules to handle various different power loads to suit different application area requirements.The QPT GaN modules are almost ready for launch to the market.

 

At the moment, the company is refining the production model by matching filters and setting everything else up correctly. The company has also taken on Geoff Haynes and Tony Astley from GaN Systems on advisor roles. The two bring both technical knowledge and experience of the market to QPT.

 

 

 

https://www.q-p-t.com/

 

 

RELATED