Things are smaller in Belgium: Belgian team wins
Little Box Challenge for solar inverters

A team from CE+T Power in Belgium has won the Little Box Challenge, a competition to invent a 3x smaller inverter for interconnecting solar power systems to the power grid. The winning team received a $1 million prize.

Judging the competition were representatives from Google, the IEEE Power Electronics Society, and NREL. The winning inverter design had a power density of 143 W/in³-far greater than the minimum requirement of 50 W/in³ and 50% higher than the nearest competitor — and a volume of only 14 in.³, smaller in volume than a cube measuring 2.5 in. on each side. The winning inverter also performed better on measurements of electromagnetic emissions compliance.

Last October, finalists brought their inverters to the Energy Systems Integration Facility on the National Renewable Energy Laboratory (NREL) campus in Colorado for testing. There, contest personnel first verified the inverters met all critical safety-related specifications and then operated the inverters at a number of different operating points and to verify the meeting of key specifications throughout the three hours. These tests narrowed the field of eighteen finalists to three which then underwent a 100-hour simulation of real-life conditions. These included handling a direct-current source of electricity that emulated a solar power system, with rapid ramp-ups and ramp-downs in power typical of an intermittently cloudy day, as well as a realistic, changing load typical of a residence consuming power. And each inverter had to meet most of the same specifications required of commercially-available inverters.

A key factor in the winning inverters was the use of wide bandgap semiconductors. The winning inverter employed GS66508P gallium-nitride power transistors from GaN Systems. Said team member Olivier Bomboir, VP of Product Management and New Business at CE+T Power, “The reduced gate drive and switching losses of GaN Systems GS66508P were critical to our thermal and power density goals. Additionally, we were highly impressed at how reliably the devices performed over the months of rigorous, real-world testing by the NREL team.”

“Wide bandgaps offer a lot of advantages over traditional silicon that enabled teams to hit some of the miniaturization and efficiency targets that were needed to be successful in the competition,” said Blake Lundstrom, NREL project lead. “Not every single team used wide-bandgap devices, but the vast majority did.”

The winning team’s contest write-up sheds further light on its use of GaN transistors. The entry points out that though GaN devices have many advantages, their fast switching qualities also create challenges: They can be tough to drive and they require sensitive electromagnetic noise management. Another pitfall is the high voltage drop due to the reverse current when the GaN turns off. The team overcame these
difficulties by controlling all the GaN transistors using soft switching for the entire operation range. The design uses a five-leg topology to minimize energy transfer
within the inverter. Two half bridges (HB) generate the neutral voltage, two other half bridges generate the line voltage, and the last serves an active filter.

One complicating factor was that GaNs switch fast enough to generate high dv/dt across the control isolation (far beyond the allowed values for most drivers). And their the gate voltage threshold is low. So the Belguim team had to design a compact, inexpensive and robust driver circuit. GaN packaging was also a consideration. The group used an SMD model with two source accesses: one for the power, one for the command. The small package reduced the parasitic inductances and consequently the functional overvoltage. Team members also said the PCB layout and the positioning of the decoupling capacitors were crucial to operating the GaN properly.

To meet the ripple requirement on dc input, the team designed a parallel active filter that can compensate the ripple more efficiently than a large capacitor at the input. The team says this is more reliable than the use of a boost-based topology for which the working voltages could rise to the V_max limit of the GaN transistors.

The active filter works with higher voltage variations (~200 Vpk-pk) and stores the
corresponding energy in ceramic capacitors whose capacitance rises as the voltage drops. This has the benefit of reducing the size of the input tank capacitor (less than 15 µF), reducing the size of the filter capacitor to less than 150 µF, and making the inverter more bullet-proof thanks to the use of the GaNs below 450 Vdc.

Also playing a role in the compact inverter design were strategies aimed at using components that were as small as possible. In this regard, the team used multilayer ceramic capacitors for the energy storage. Magnetic components were mainly ferrite whose magnetic losses are low at high frequencies. The design used Litz wires (specifically designed to reduce skin effect and proximity effect losses) and wound them directly onto the ferrite, without a coil former. Their cooling was via a fan augmented by use of an
aluminum oxide foil placed in the middle of the ferrite to create an air gap and a
thermal drain.

The team minimized the size of the filter capacitors and inductors by allowing more ripple current. To minimize the size of the output current measurement device, they used an open-loop Hall sensor combined with an electromagnetic shield, offering galvanic decoupling and reducing the sensitivity to common-mode and parasitic inductance noise.

All other current measurements, such as inductor current, are made by state observers rather than by readings from current sensors. This reduces the overall inverter size.
The team also used a specific GaN control modulation which reduces the current within the filter inductors to keep their core size small without reaching saturation level.

Finally, the team used a custom heatsink made by EDM (Electrical Discharge Machining), an
ultra-thin PCB (0.012 inch thick), and a silicone foam to spread the GaN contact pressure on the heatsink. The team credits all these techniques with greatly helping to reduce the size of the inverter.