High power GaAs H-bridge motor drives

[ehsmeng]

Hi,

Some 10-15 y ago, an incubator company from a Swedish uni was created to produce
http://en.wikipedia.org/wiki/Gallium_arsenide
transistors. These can deal with way higher temperatures and powers than Si-based etc and still have reasonable switching characteristics.

2-3y ago National bought them.

Has anyone used/tested/heard of GaAs for pwm-controling electrical engines in a situation where more power was needed than in an RC car or smth. I'm thinking electric car, bike, wheelchair, etc. Something of that magnitude.

[jpb]

I spent many years working with GaAs FETs (device physics and monolithic microwave integrated circuits) and I am very surprised that you are suggesting that they can work at much higher powers and temperatures.

GaAs can operate at higher frequencies and has the big advantage of having a semi-insulating form so you can put low-loss transmission lines on it for integrated circuits. It is wider band gap so is more radiation hard.
(Edit: the wider band gap also would allow higher temperatures but again SiC is better in such an environment I think)

But if you're looking for very high powers and voltages something like Silicon Carbide is probably more appropriate.

At low frequencies I would be very surprised if GaAs power FETs were a better option than high power silicon mosfets - though my experience was based more around small signal circuits so I may be wrong.

[ehsmeng]

I'm so sorry, I wrote that before the morning coffee. Of course it's supposed to be Silicon Carbide.
And they were bought by Fairchild.
http://investing.businessweek.com/research/stocks/private/snapshot.asp?privcapId=30149548
With this level of abuse of fact, I'd make a great journalist...

The same Q again then With coffee: Are these SiC transistors ready for the limelight? I don't know anyone actually using them.

If they are as magnificent for the humanity as biased texts suggest, we'd be seeing SiC versions of LMD18200 etc popping up? I mean, the timing for that is now, isn't it?

[jpb]

I've been nearly ten years out of the business, but my understanding is that SiC are really only supplied by Cree.

eg
http://www.cree.com/Power/Landing-pages/module-products

though a bit of googling show that ST do them as well:
http://www.st.com/web/en/catalog/sense_power/FM100/CL2062/SC1704?icmp=sc1704_pron_pr2_mar2014

One issue we found with SiC, this was ten years or so ago and we were looking at experimental devices, is that they had major charge trapping effects so that the I(V) curves at dc looked good but they collapsed when measured under pulsed conditions which indicated that the RF power wouldn't be good either.

This may be why Cree now seems to concentrate on GaN for RF applications and restrict SiC to power modules operating at low frequencies (I presume). I only had a brief look at their site so I may have missed SiC RF devices. Though ST also seem to be selling their ones as switches as well.

[T3sl4co1l]

SiC was looking good a few years ago -- if still rather expensive -- but then some jerks went and invented the superjunction process.   Right now, SiC is comparable to SJ MOSFETs for 600V+ applications.  Neither is relevant at lower voltages (sub 400V or thereabouts), nor are IGBTs.  Operating temperature is limited by encapsulation, unless you want to go well out of your way to find the mil spec gold bonded ceramic kind ($$$, lead time ).

SiC schottky diodes are good of course -- if lossy (significant junction resistance) and still on the expensive side.

GaN is pretty cool stuff, but still much too niche-ey for anything but high-falootin' RF.  If they ever release HV devices in GaN, it'll be pretty awesome for high [power / frequency / efficiency] inverter/converter applications.  Right now, about all that's available are the EPC solder-bump dies, which only go up to 200V.  Excellent specs, pricey, probably not easy to solder (fine pitch, but not impossible to manufacture).

Tim

[ehsmeng]

Allow me to share my naive reason for pursuing SiCs.

Say I'm playing with say pwm engine control for a car, bike, wheelchair, ... . If a Si transistor wants to be <100degC and I find a SiC which is happy up to say 300degC, there should in the SiC case be significantly less problems using the same product in winter moscow (-30degC) as in heat wave India (80degC). Worst case I have 220 degrees difference to cool it, and in the Si case 20 degC.

But you're right, if pricing is Nasa level I should drop it. Incidentally, some SiC team is actually approaching Nasa with some SiC version that could do 600degC or smth like that. But I digress. I was just expecting readily available products around now since Fairchild took over.

[jpb]

Having spent 20 years in GaAs research one lesson I learned was that it "is the material of the future ... and always will be!". The problem is that Si keeps getting better and the niche spaces left for the more exotic semiconductors get smaller. In the case of GaAs it was the frequency of operation, when I first started out Si ran out of steam at less than 1GHz but now it goes much higher. GaAs FETs or HEMTs were used for low-noise front ends on consumer satellite receivers and mobile 'phones but they soon went all Si.

There were rumours that with mobile 'phones the manufacturers would submit their 'phones for test with a low noise GaAs/AlGaAs HEMT in the front end and then replace it with Si for the manufacture process! The early "what mobile" type magazines used to list noise figure in their charts I think - not that anyone reading them had any idea what it was!

[SeanB]

The only real application for GaAs is in applications that need the bandgap, like LED's. Even then you find them as a film on a silicon wafer, as that is a lot easier to handle.

[T3sl4co1l]

Yeah no, much better off designing for the temp range to begin with, than to have a crappy thermal design that works because of gimmicks like aerospace grade transistors.

The tempco isn't even all that much better, it just goes further before the transistor stops, er, transisting.  See fig.4:
http://www.cree.com/~/media/Files/Cree/Power/Data%20Sheets/C2M0160120D.pdf
The Rds(ON) rises by about 1.8 times at rated temperature (150C), whereas for a comparable silicon device it would be more like 2.5 times.  Probably it continues rising, so that if you had the same device construction, in a 300C rating, it would be more like 5x at that temperature.  Incidentally, the bare die version of the above is still only qualified for 150C (go figure?).

No free lunch

Tim

[jpb]

The only real application for GaAs is in applications that need the bandgap, like LED's. Even then you find them as a film on a silicon wafer, as that is a lot easier to handle.

GaAs MMICs are still useful at microwave frequencies - things like radar systems but I don't think they have many consumer applications.

[SeanB]

Yes, but there it is the bandgap that gives the low noise and high RF gain again. Specialist niche again where the material is needed and nothing else works as well. Even there SiGe is moving in as well, being both cheaper and easier to work with process wise in large wafers.

[jpb]

Yes, but there it is the bandgap that gives the low noise and high RF gain again. Specialist niche again where the material is needed and nothing else works as well. Even there SiGe is moving in as well, being both cheaper and easier to work with process wise in large wafers.

I don't think you can get a semi-insulating form in SiGe (I'm a bit rusty here). The semi-insulating form of GaAs allows for low loss transmission lines so you can build things like distributed amplifiers with massive bandwidths.

But yes, I agree they will always be niche. When I first went into the area (it was my first graduate job) it was seen as an exciting way to go but the thing is that Si is dirt cheap, you can pull massive crystals (I remember being at a conference where someone was waving around a 12 inch Si wafer where-as GaAs was still at 2 or 3 inch) and it is much easier to work with.

There will always be the push to do everything in Si and it makes economic sense to do so unless it is some thing that just can't be done (i.e. that requires direct band gap or large band gap).

[T3sl4co1l]

There's also InP, which I've seen in papers a lot, but I don't even know what commercial applications it's in.

I know the high-falootin' scopes are all supposed to be monolithic samplers (including schottky based nonlinear transmission lines to generate picosecond pulses), though I don't know what materials they are made on (maybe both?).

Tim