In a previous article we looked at a silicon carbide (SiC) FET from Wolfspeed. That component is one example of what seems to be an accelerating trend: there are actually real non-silicon semiconductor devices being manufactured and sold. However, it’s not quite as simple as that.
The thing is, alternative semiconductor technologies are not new. Some quick Internet searching indicates that materials such as gallium nitride, gallium arsenide, and silicon germanium have been used in various specialized applications for over twenty years. What does seem new, however, is that manufacturers are increasingly promoting alternative-material-based devices as replacements for ordinary components such as transistors. Presumably we’re reaching the point at which the costs and difficulties associated with alternative materials are more justifiable with respect to the performance improvements that engineers are looking for.
In this article we’re focusing on a gallium nitride (GaN) field-effect transistor. You might be familiar with GaN in the context of RF circuitry. It offers excellent high-frequency performance, and its high power density makes it very valuable as a material for RF amplifiers, since higher power density means that you can get the same amount of power from a smaller device. The FET we’re discussing in this article—the EPC2046 from Efficient Power Conversion (EPC)—also benefits from GaN’s higher power density.
A (Much) Smaller Solution
Miniaturization is big business these days, and thus it doesn’t surprise me to find that EPC is emphasizing the reduced board space required by the EPC2046:
Image courtesy of EPC.
The factor-of-twelve reduction in size is significant, to be sure, though some of that reduction is due to the fact that the comparison is between a chip-scale EPC2046 and a typically packaged silicon device. As EPC points out in the EPC2046 press release, the chip-scale package inevitably improves thermal performance since the heat goes straight to the environment instead of being held within the insulating plastic of a typical IC package. The difference in size between a silicon solution and a GaN solution would be less extreme if both devices were in chip-scale packaging.
But EPC doesn’t present reduced component size as the only significant benefit. They refer to a “fifth-generation” manufacturing process that is enabling GaN devices to offer higher performance and lower cost. My favorite excerpt from the press release is the assertion that this latest GaN technology “opens up entirely new applications beyond the reach of the aging silicon MOSFET.” I don’t disagree that silicon FETs have been around a long time, but I’ve not yet noticed anyone explicitly labeling them as “aging.” I guess it just strikes me as a mildly amusing adjective to use when describing a transistor.
Anyways, the specs do seem impressive, though it would be nice to see a direct comparison with equivalent silicon devices. The EPC2046’s maximum on-state resistance is 25 mΩ, and the part can survive a drain-to-source voltage of 200 V. The steady-state drain current rating is 11 A at room temperature, and it can handle 55 A of pulsed current—keep in mind that this part is really small:
Image taken from the datasheet.
Silicon carbide FETs are a bit difficult with respect to the threshold voltage. The article mentioned above points out that though the theoretical threshold voltage (VTH) of the device is ~2.5 V, you need a lot more voltage than that to achieve good on-state resistance. This app note from Fairchild (now part of On Semiconductor) confirms that a silicon carbide MOSFET with VTH ≈ 2.5 V will not be fully enhanced until the gate-to-source voltage is up around 16 V. The app note also mentions that SiC MOSFETs suffer from high threshold-voltage instability.
So do GaN FETs exhibit the same weaknesses? With respect to GaN technology in general, I don’t really know. A quick online search didn’t turn up any obvious information on the topic, and I’m allergic to the language used in academic papers, so I didn’t attempt to explore those resources. However, based on the information in the EPC2046 datasheet, I’m inclined to say that this particular part doesn’t have equivalent complications related to VTH .
The nominal VTH (1.6 V) is lower, which is generally a good thing, and the min-to-max range (0.8 V to 2.5 V) is reasonable. Furthermore, in the following plot the on-state resistance seems to approach its minimum value at a relatively low voltage (certainly much lower than 16 V).
Plot taken from the datasheet.
Do you have any experience with high-power GaN FETs? Let us know in the comments.