Gallium Nitride Power Transistors Priced Cheaper Than Silicon | Hacker News Reader

Gallium Nitride Power Transistors Priced Cheaper Than Silicon

spectrum.ieee.org

74 pointsrbanffy17 comments

This article has more meaty info:

http://spectrum.ieee.org/semiconductors/materials/the-toughe...

Highlights:

> In the GaN FET, on the other hand, the two-dimensional electron gas already exists naturally. So a positive voltage applied to the drain immediately pushes current from source to drain. Thus the amount of current is varied by applying a negative voltage to the gate, which restricts the number of electrons available to flow from source to drain. A large enough negative voltage turns off the flow altogether. Thus in contrast to a silicon FET, which is normally off, a GaN FET is normally on.

(Somewhat misleading; silicon FETs that are "normally off" are enhancement-mode transistors. Of course, there are depletion-mode MOSFETs, and of course JFETs. There are even power MOSFETS that are depletion mode, see here: http://www.ixys.com/documents/appnotes/ixan0063.pdf In any case, this confirms the new transistor to be a depletion mode device. This has implications for biasing which could be inconvenient in some applications. In any case, it means it's not simply drop-in replacement for enhancement-mode MOSFETs in existing designs.)

> One of us (Mishra) has succeeded in making bipolar GaN transistors. But they are not yet as reliable as the FETs because at the moment it is very difficult to make p-type material good enough to use in a bipolar transistor. Applying electrical contacts to the material, as is necessary to connect the device into a circuit, often wiped out the semiconductor's p-type character.

If these challenges are solved, the idea of a new kind of BJT is exciting.

Biasing is indeed non-trivial, and most designs require temperature compensation to keep drain current constant. In fact, it's pretty easy to fry a GaN part without proper bias sequencing at power-on and off. At least it is with the parts I'm using.

Edit: parts like MAX881R [1] make it pretty easy though :)

[1] http://datasheets.maximintegrated.com/en/ds/MAX881R.pdf

I've used the MAX881 for some lower power GaAs, but it's limited to a few mA gate current. TI has an opamp I used that can drive capacitive loads with +/- 30 mA, so it makes a good gate driver, whilst allowing a decent amount of gate bypass capacitance. For temp comp I have found once you get the temp vs IDS curve, it's identical for devices within a lot, and only need to offset for pinchoff, which varies device to device within the same lot.

Both depletion-mode and enhancement-mode GaN devices exist now. The former can be turned into a "normally off" device by combining it with a low voltage Si FET to form a cascode [1]

[1] Slide 36: http://www.irf.com/product-info/ganpowir/gangeneral.pdf

Great for the high side though :-)

I've been working with GaN parts for RF after being out of the hardware game for a few years. GaN was around but nowhere near as prolific as it is now. I thought SiC would have made more inroads than it has. GaN parts just seem incredible in every way (to someone like me at least) - amazing performance, efficiency, power handling in such a small package, more impressive impedance matching on wideband parts - it's just a shame the basic problem of harmonics still hasn't been whisked away by magic :)

It's been eons since I touched this stuff, but the article said that they're on GaAs on Si wafers for a substrate. Is that the usual these days?

ISTR that SiC was always a PITA to grow on decent/cheap substrates, which may have held it back. That and it didn't have the whole blue LED industry driving it.

Still issues getting the heat out, but yes, GaN is pretty amazing. 10 years ago I was load pulling and matching sub 1 Ohm (24 mm) HFET die for 12 watts. Now you can get 120W in the same periphery.

Packaging is indeed critical to get heat out. There have been some cool (no pun intended) developments on this front such as the DirectFET [1], LFPAK [2]

[1] http://www.irf.com/technical-info/whitepaper/directfet.pdf [2] http://www.nxp.com/documents/leaflet/939775016838_LR.pdf [3]

Interesting. I have never seen RF power parts in flip-chip, but I suppose there is no reason why they couldn't. Now Cree did have a 25W, 3 GHz FET in SO-8 package, which is kind of strange given package inductance, but they are have moved to QFN now.

These are pretty cool, this is a datasheet for the 100V part : http://epc-co.com/epc/Portals/0/epc/documents/datasheets/EPC... which in this case is an enhancement mode transistor (so 'N' channel, or positive Vgs to turn it on)

The thing that should smack you in the eye when you read the datasheet that that it has a Theta(j-ambient) of 1100dC/W [1] which is why this tiny part is limited to 1A continuous current (1 amp x 1 amp x .065 ohms == .065 watts and a temperature rise of 71 degrees C. But it can be switching probably close to 80 watts while doing that.

I'm a bit surprised though that it doesn't list the switching time.

[1] You can get that down to 100dC/W if you solder it to a 1" square on 2oz copper but then what is the point of having such a tiny switch :-)

There is that guy growing sheets of diamond, if he isn't off'd by the diamond cartels. I suppose diamond semiconductors will supplant GaN some day. I'm read a few years ago about cold cathod parts, but I guess it never panned out.

Good article until the end

"Lidow says EPC decided to first go after applications at 200 V or less in order to pursue new applications silicon can’t easily reach, a category that includes virtual reality and small medical imaging systems."

Sorry for the expression, but "dafuq did I just read"

What does VR has to do with a 100V power transistor? nothing

Small medical imaging systems may benefit from this (X-Rays? Mini-MRIs?) but they're not essential (also medical devices are kind of non-price sensitive, so)

If I am understanding correctly GaN mainly used in power components since it can handle larger voltages. Is it possible that with GaN getting cheaper that it might start moving towards other component areas that Si dominates? Or is GaN's usefulness primarily in handling power?

GaN provides a couple of benefits compared to Silicon:

1. It offers a lower on-resistance for the same breakdown voltage as a silicon device [1].

2. It offers a lower gate charge which allows you to turn it on and off at a much higher frequency than Silicon devices [2]

(1) means you have lower losses in a power circuit with GaN vs Si (leading to more compact - smaller heat sinks - power supplies for example)

(2) means power supplies can run at much higher frequencies. At higher frequencies, you can use smaller inductors and capacitors which means more compact circuits.

A practical example of GaN in action is FinSix' Dart power supply [3]

[1] Slide 5 of http://www.irf.com/product-info/ganpowir/gangeneral.pdf [2] http://m.eet.com/media/1179793/si-vs-gan-vs-sic-which-proces... [3] http://finsix.com/dart/

That's what I am wondering. There was a mention of RF, but don't go into detail. If they can use this advancement in the CPU / GPU area, I think a new "golden era" might happen. New CPUs and GPUs each year would be nice. =)

Thats part of my other questions. How small is the current GaN technology? How much can it build on silicon technology?

Reply on news.ycombinator.com