Subscribe to EDN

IR’s GaN platform: Is the long run of the silicon MOSFET ending?

March 3, 2010

APEC MOSFETs Part 2: Is Silicon’s long run with MOSFETs over?

(Here’s Part 1: New MOSFETS target server power supply efficiency.)

Every year at APEC, the discrete semiconductor companies roll out their next generation of faster power MOSFETs with lower RDS(ON). Clearly these improvements can’t go on forever, but where will they end? Betting against improvements in silicon performance would seem to be a losing game. But International Rectifier,  the company that introduced the first hexagonal MOSFET in 1979, has now declared that silicon’s long reign is over, to be supplanted by gallium nitride on silicon. The company first announced back in 2008 that it was the first to master the art of building gallium nitride devices on silicon – this announcement marked the introduction of the first actual product based on the GaN platform, which the company calls “GaNpowIR”.

GaN device structures are not new – they’ve been around for 20 years in the RF realm, using a sapphire substrate. What IR has done is figure out how to make the ultra-fast switching times of GaN commercially viable for power electronics, and they have done that by figuring out how to build GaN devices on a silicon substrate. IR claims that GaN on silicon heralds the end of silicon’s long reign in MOSFET switching devices.

Speaking at APEC, IR’s technology consultant Dr. MikeBriere looked back at the impact that the first commercially viable MOSFET –the hexagonal MOSFET — had on the power space: It made possible the high-volume commercial switching power supply. From the HEXFET to the super FET to the superjunction FET, the figure of merit for the devices (RDS(on) x Qg) has improved a couple of orders of magnitude over the past 30 years. But, Briere argued,  this rate of improvement can’t go on forever, and the physical limitation of the technology will be reached. That’s where Briere sees silicon MOSFET technology now: It’s taking more and more investment to get less and less returns, and as a result, it’s becoming commercially less viable for semiconductor companies to invest in silicon power devices.

Enter GaN on silicon technology. It operates in native depletion mode, although it can be made in enhanced mode and switches at up to 5MHz. The voltage roadmap calls for it to reach from 20V up to 1200V.  The iP2010 and iP2011, the first products, are integrated power stage devices that integrate a PowIRtune driver IC matched to a multi-switch monolithic GaN-based power device, all mounted in a flip chip package, that the company claims more than double the switching frequency of state-of-the-art silicon-based integrated power stage devices. 

From the product announcement: “The iP2010 features an input voltage range of 7V to 13.2V and output voltage range of 0.6V to 5.5V with an output current up to30A. The device operates up to 3MHz. Operating up to 5MHz, the pin-compatible iP2011 features the same input and output voltage range but is optimized for an output current up to20A. By offering multiple current rating devices in a common footprint, IR provides flexibility for meeting different customer requirements in terms of current level, performance and cost.”

 

More product specs:

iP2010 Specifications

 

 

Part Number

 

 

Package

Vin Range

(V)

Vout Range (V)

 

Iout Max

(A)

Sw Freq

Range

(kHz)

iP2010TRPBF

7.7 x 6.5mm LGA

7 – 13.2

0.6 – 5.5

30

250 – 3,000

iP2011 Specifications

 

 

Part Number

 

 

Package

Vin Range

(V)

Vout Range (V)

 

Iout Max

(A)

Sw Freq

Range

(kHz)

iP2011TRPBF

7.7 x 6.5mm LGA

7 – 13.2

0.6 – 5.5

20

250 – 5,000

 

 

These parts target the server power supply market where some customers are willing to pay a hefty premium for power efficiency. Pricing begins at US $9.00 each and US $6.00 each respectively in 2,500-unit quantities for the iP2010TRPbF and iP2011TRPbF.

Posted by Margery Conner on March 3, 2010 | Comments (2)

March 9, 2010
In response to: IR’s GaN platform: Is the long run of the silicon MOSFET ending?
Andy T commented:

Nothing but economics keeps them from entering the market today at 1kV


March 4, 2010
In response to: IR’s GaN platform: Is the long run of the silicon MOSFET ending?
Andy T commented:

The most important Figure of Merit in all this, to me, is COST relative to paralleling up Silicon MOSFETS. Unless you are miniaturizing your power stage and have zero choice in available area, the GaN just pushes power density, a scenario for possibly being a thermal nightmare. According to IR, the Qg*RDSon is 33% better on their GaN device (otice I did not say "on GaN"), while Rds is 3X better (with promises it'll go to 10X in a DECADE). So, for the same Rds, if that's what you're after in high current apps (to reduce the thermal problem and improve efficeincy), you can parallel 3 Si MOSFETs today, spreading the power out to ease thermal design. These paralleled FETs will require a driver that's 33% more powerful in pushing the gate charge around in a monolithic MOSFET than in IR's GaN device. If you read the PR, there seem to be TWO chips inside the IR GaN devices...one of them appears to be a DRIVER chip. That's cheating, kids. We're not seeing the native beefits of GaN, IMO...in fact, the mere presence of an internal driver might indicate a WORSE FOM.... And, getting back to my own figure of merit, COST, the paralleled Si solution is $4, while the GaN is $9. You can buy an awesome driver chip for 1/3 that cost difference, and save a bunch of money on the thermal headaches associated with a smaller-packaged solutions, as well. IR mispriced this, IMO, by a factor of TWO. I'm not convinced this technology is quite ready yet, but it looks to me like somebody at IR had to make lemonade fro a possible lemon to serve to its investors until the technology is ready for prime time in a few years. From my perspective as a seasoned product definer and strategist, maybe now we'll see the Si MOSFET guys motivated to integrate driver chips onto their MOSFETs (0.7V threshold voltage please guys...I, and I'm sure a lot of designers, will want to drive the "gate" from a CMOS ASIC) - it's about time...

POST A COMMENT
Display Name
captcha

Before submitting this form, please type the characters displayed above. Note the letters are case sensitive:

Advertisement
Advertisement
Advertisement
About EDN   |   Site Map   |   Contact Us   |   Subscription   |   RSS
© 2012 UBM Electronics. All rights reserved.
Use of this Web site is subject to its Terms of Use | Privacy Policy

Please visit these other UBM Canon sites

UBM Canon | Design News | Test & Measurement World | Packaging Digest | EDN | Qmed | Pharmalive | Appliance Magazine | Plastics Today | Powder Bulk Solids | Canon Trade Shows