SSTC or DRSSTC with GaN Transistors

[Zipdox]

Has anyone ever built a solid state coil with gallium nitride transistors? I would assume that using more efficient transistors would reduce the need for cooling and risk of damaging them.

I've looked at a company that makes them called Transphorm and their FETs are only about 3x as expensive as regular silicon FETs for the same current rating (compared to IRFP460), at a slightly higher voltage. What's interesting is that they can cram a lot of current capability into small packages because of the high efficiency.

Let me know what you think.

[Mads Barnkob]

I think I have seen around 3-4 designs that was show-cased on social media or forums, I can not find links right now.

But I think its people like Steve Ward, Weston and Loneoceans that have built with GaN or SiC switches.

[Weston]

I used some GaN parts from GaN Systems for a small Class E tesla coil a few years ago and use GaN parts a lot for research.



If you are going to use wide bandgap semiconductors, I would suggest SiC parts instead of GaN. GaN parts are typically pretty fragile and do not survive voltage spikes well. The also typically have low thermal mass with an on resistance that significantly increases with temperature. The main advantage with GaN is faster switching, but most tesla coil designs have soft switching, so switching loss should be pretty low.

For either SiC or GaN you probably will not be able to use a gate drive transformer due to the higher switching speed and different gate voltage requirements.

The Transphorm parts are a cascode where you drive a Si MOSFET gate, which solves the issue of the weird voltage requirements for the GaN gate, but the parts are kind of funky and have issues with oscillation. Steve Ward experimented with these parts in the past and had issues with this. The app notes for those parts talk about all sorts of snubber networks that are required to not have things blow up.

Consider looking at SiC parts. They are a lot more robust that GaN and have a better on resistance temperature coefficient. I use a SiC FET module in my QCW tesla coil, where it allows me to hard switch ~70A at 500V for a phase shifted full bridge https://highvoltageforum.net/index.php?topic=1024

[Zipdox]

Hmmm I already imagined they'd be more fragile than regular silicon. In that case I think I'll stick to old fashioned silicon unless I ever feel a hole burning in my wallet and have more spare time.

[Hydron]

Moderately specified SiC parts aren't crazy money. $12.50 for a 1.2kV 75mR MOSFET is one part I'm looking at - not cheap but performs a lot better than silicon if you need a MOSFET rather than an IGBT.

You can still use smaller silicon IGBTs up to reasonably high frequencies (a few hundred kHz) if a 600/650V rating is enough, but if you need more voltage than this the switching loss goes up and speed goes down quickly!

[Steve Ward]

On an "academic" level, GaN is still superior to SiC in terms of junction charge loss vs on-state resistance, so in theory you can make the *most* efficient switchers with GaN, and in fact, that's what I do for work lately. However, you'd better ditch any kind of "through hole" packaging and stick with laminated multi-layer designs with very low total inductance (<10nH for a half-bridge/de-coupling cap loop).  Even considering resonant loads that can charge the output junctions for "free", it still costs VARs which have a cost via the circuit resistance, so GaN wins by reducing junction charging time/current.

However, the downsides of GaN are indeed smaller voltage tolerances and possibly higher thermal resistance, and well, it costs a lot more than Si or even SiC.  I've used GaN to drive Tesla coils, but only because I have the parts on hand for work projects, the cost/performance ratio applicable to TCs is not very compelling.  Perhaps for a small coil aiming for minimal heatsinking and the cost isnt too high, it would be handy.