There’s been a lot of attention and excitement lately over semiconductor materials. In the power-semiconductor arena, silicon has begun to hit the wall in terms of the operating performance it is able to deliver to device manufacturers. Power-semi consumers want more efficient devices that won’t lose as much power to switching losses.
Thus, power-semi vendors are beginning to turn to alternate materials, namely silicon carbide (SiC) and gallium nitride (GaN). There are a number of good reasons why, too. For one, their wide bandgap makes them very attractive for high-temperature applications.
For one, the Ford Motor Co. is very interested in GaN devices for power conversion in their hybrid vehicles. In the past, Ford has used silicon power MOSFETs in these applications, but would have to locate them far from the engine block where most of the power conversion needs to happen. Why? They just can’t stand the heat, so they had to be moved out of the kitchen. The result is long wiring runs and IR losses, which conspire to defeat the purpose of hybrids. In contrast, GaN devices can stand temperatures of up to 300°C and still operate within their specified parameters.
Of course, GaN power semiconductors are going to cost more, right? Well, not as much more as you’d think, especially since the industry has adopted the practice of growing GaN on top of a silicon substrate and an aluminum nitride buffer layer. This allows device makers to use their existing silicon process technology and to grow the actual PN junctions on top of the substrate. That’s brought GaN pricing down from 10X to 15X the price of traditional silicon to a level that’s much more palatable, if you absolutely need the performance boost.
Thus, we have the promise of high-performance materials that provide more efficiency, more robustness, and at a price point that makes them feasible for volume applications like hybrid vehicles, industrial motor drives, and the like. GaN comes a lot closer than silicon ever could to the attributes of the perfect power switch, which would block voltages of infinite magnitude, carry infinite current, switch instantaneously, and require zero drive power. Of course, it doesn’t approach perfection, but it gets closer than silicon: GaN delivers higher voltage blocking, a lower on-resistance, and a lot more speed. How much speed? Well, a GaN FET channel can demonstrably switch in a matter of nanoseconds, even while carrying as much as 10 A. We’re talking switching frequencies of maybe 80 MHz, well into RF.
Ah, but a gotcha comes along with this miraculous material technology. In a GaN switch, current peaks faster and voltage drops correspondingly faster. Not only do GaN devices switch faster, but their turn-on threshold is lower as is their drain-to-source resistance.
Designers will need to take an oscilloscope to these devices to characterize their behavior and measure losses during transitions. Voltage swings in these devices, which might be 600 V or more, must be measured much faster. The scope that’s going to keep up must have enough bandwidth to respond fast enough to the transition and also provide a good deal of resolution to get to the lower voltage levels at which these transitions occur.
One solution to this problem comes from the probing side of the equation. “We’ve heard from engineers using power semiconductors made with GaN,” says Randy White, technical marketing manager at Tektronix. “They want kilovolt-range measurements with bandwidth of more than a gigahertz.” Probes are a limiting factor here, as the best high-voltage differential probes only provide about 100 MHz of bandwidth. The top-shelf single-ended HV probes will get you up to about 800 MHz, which might be workable for that 600-V swing.
Another probing issue is safety at these high voltages. Tektronix and other scope makers put a lot of thought and effort into the construction of their high-voltage probes, ensuring that there’s enough insulation and clearance between the tips. But Tek’s customers balk at the safety features. Yes, they make the probes safer to use but they limit performance.
What about adding longer leads? No good; too much lead inductance that results in circuit loading and ringing. It can be pretty tough to figure out if that ringing is coming from instability in the circuit or from the test setup. Tek tries to address this requirement for high-frequency measurements with high fidelity at high voltages by adding damping resistors to the probes to snuff out oscillation.
On the scope side, there’s the requirement of more resolution. Most of today’s scopes today deliver 8 bits of resolution. The way around this limitation, says White, is post-processing of the measurements with averaging. “We do remind people that averaging works, but there are tradeoffs,” says White. For averaging to work, the signal under measurement should ideally be of a repetitive nature. What about a glitch coming from a motor drive, or crosstalk in the circuit? That’s a single-shot measurement but you still need to average it to increase the resolution. The answer there is a boxcar-averaging technique in which the signal is oversampled; the effective result is a reduction in random noise and a higher-resolution measurement.
GaN power semiconductors are in production in limited quantities. International Rectifier is targeting full production within the next 12 to 18 months. So the bottom line is that measurement of these devices could become an issue for many engineers. Look for lots of news and developments in this area in the next year or two, both on the components side and on the test side.