Getting started with GaN FETs

[jmw]

I've read about how GaN FETs are how Finsix and Innergie build their miniaturized laptop power supplies that switch up into the VHF range. Are there any books, papers, articles, etc. on theory and practical use, how to build designs with these? At the very least they seem to require special driver ICs and it sounds like one does not just simply replace silicon with GaN, crank up the frequency and go. What's a good place to get started for the curious, assuming a decent foundation with MOSFETs and power electronics?

[filssavi]

I doubt you will find anything wide band gap at the hobbist level, as how to use such devices is still an open research question

If you want to use wide band gap as they are supposed to be used you have to crank up both frequency, but more importantly edge rate (with 600V GaN devices you can get up to single digit nanoseconds raise and fall times) to reduce the switching losses

Now the layout must be done right (that’s an euphemism) to avoid turning off tv’s and pc’s in a radious of 50Km with EMI

Another problem with GaN devices is that they are very very very fragile, any gate overshoot (that includes ringing) and the device is dead, if soldering is not done right (and they are a pain in the a** to do even with hot air) they die, if a mosquito farts across the room they die etc

Lastly even if wide band gap can work at high temperatures you have to keep them cool as the on state resistance can double or even triple and you kill your efficiency that way

The magnetic components insulating system must also be made to withstand the possible partial discharges from the high raise times or you’ll kill the inductor...

[jbb]

Texas Instruments and EPC have some power reference designs which could be interesting.

I hear that for GaN a) the gates must be very carefully driven woh regulated voltages (because required gate drive voltage and fatal gate voltage are close) and b) they cannot take unclamped over voltage (because they have no PN junction in the right place to avalanche).

For both GaN and SiC devices, layout for minimum loop areas is absolutely critical. This also helps with EMI.

[spec]

Here is a good overview from, Efficient Power Conversions(EPC) who specialize in GaN devices: 

https://epc-co.com/epc/Portals/0/epc/documents/publications/GaN%20Transistors%20for%20Efficient%20Power%20Conversion%20-%20Chapter%201.pdf

GaN transistor's 'natural' mode is depletion, but enhancement mode types are now available:
 
http://epc-co.com/epc/Portals/0/epc/documents/datasheets/EPC2036_datasheet.pdf   (£1.00UK from DigiKey)

http://epc-co.com/epc/Portals/0/epc/documents/datasheets/EPC2016C_datasheet.pdf    (£2.00UK from DigiKey)

Notice the low gate threshold voltage and remarkably low parasitic capacitances (compared to Si MOSFETs).

These two characteristics open the way for some interesting design approaches. For example, the low capacitance greatly improves the open loop frequency response of the traditional constant current circuit using an opamp and a Si MOSFET.  GaN transistors also look interesting for high quality audio amplifier output stages, where the horrifically high and varying parasitic capacitances of Si MOSFETs are a major problem.

[T3sl4co1l]

Incidentally, GaN isn't very useful above about 100W per power stage, due to physical size constraints.

That is: the switching loop inductance is proportional to physical size of the layout.  This inductance combines with switch capacitance to give a characteristic time constant / cutoff frequency, and impedance, of the power stage.  If that impedance is very far from the load impedance (Vsupply / Ipeak), it is at risk of being excited by switching transients, where destructive peak voltages or currents ensue (voltage being the critical one, most times).

High voltage GaN helps with this, since their capacitance is already quite low and the tolerable inductance is that much higher.  Still, it can't go much beyond the low kW, at least right now.

In any case, needless to say, layout is critical here.  Two-layer boards are nearly useless.  Don't play with GaN unless you're playing with 4+ layer boards!

This is not news by any means -- it already applies to silicon devices, this being the limiting factor against large MOSFET modules, say.  They are available in lower ratings, but IGBTs take over in the 100s of amperes range.  Speed also isn't a big priority in high voltage converters, to a point (although right now, the limiting factor is definitely the device -- in 2kV+ devices, minority carrier lifetime is multiple microseconds, and we'd be happy to have that a bit under 1us).

This is fundamentally why, for example, commercial FM transmitters are made from myriad blocks wired together: small amplifiers, running from modest supplies, all working together to generate the final (~50kW?) output.

Likewise, power converters, operating at high frequency, must be divided into a multi-channel design (usually multi-phase, for the inverse reason why the broadcast transmitter is multi-channel and in-phase!).  This, too, already affects silicon, but also inductors and capacitors which are otherwise impractical at scale (hence the N-phase Vcore supplies on PC motherboards, GPUs and such).

Tim