Archive for the ‘Power Devices’ Category

Inside Cree’s SiC Power MOSFET

Monday, March 5th, 2012 by Carav

Contributed by: St.J. Dixon-Warren

Cree is the market leader in silicon carbide (SiC) device technology. They sell both material and a variety of discrete devices including LEDs and Schottky diodes. Recently, they launched a silicon carbide power MOSFET device, the CMF20120D, targeted at solar inverters, high voltage DC/DC converters, and motor drives. According to the datasheet, the CMF20120D is rated for VDS up to 1200 V with RDS(ON) of 80 mOhms, and with ID(MAX) at 33 A. The device is packaged in a standard TO-247-3 package.

The material properties of SiCs bring certain benefits. Foremost, an SiC has a much larger bandgap than silicon, hence SiC devices can operate at higher temperatures without suffering from thermally induced intrinsic conduction effects. An SiC device also has a higher breakdown voltage, meaning that devices can support higher voltages for a given device geometry. In addition, an SiC chip has higher thermal conductivity and higher electron drift velocities, enabling higher power and higher speeds.

Since the CMF20120D is the first SiC MOSFET available easily on the open market, we felt that an analysis was warranted. In this brief article, we share some of the highlights of our investigation.

Figure 1 shows a high resolution photograph of the CMF20120D die following decapsulation in Chipworks’ lab. The active region of the die is 3.46 mm x 3.46 mm. There are four wire bonds connecting to the source metallization area, while a fifth wire bond is connected to the gate contact pad on the left side of the die. Some simple measurements show that the source metallization area corresponds to 61% of the die area, while the gate pad and gate interconnect tracks correspond to 11% of the die area. The balance of the area is comprised of the inactive region around the periphery of the die.

Figure 1. CMF20120D Die Photograph

A detailed view of the corner of the source metallization area is shown in Figure 2. Each dark rectangle corresponds to a source contact to the SiC substrate. The contacts are surrounded by the polysilicon gate that lies beneath the metal. Each gate cell is 10 µm x 17.4 µm wide, hence a simple calculation estimates there to be 58,000 gate cells on the die. Consequently, when the device is conducting the highest rated drain current of 33 A, then each cell would carry ~0.6 mA.

Figure 2. CMF20120D Source Metallization Corner and Gate Cell Area

Figure 3 shows a cross section through the edge of the CMF20120D SiC die. The cross section location is shown in Figure 1. The device is a vertical drift MOSFET, hence the source metal and gate array are located on the top side of the 370 µm thick die, while the drain is located at the bottom side.  The operation of a discrete power MOSFET is described on Wikipedia. When the device is off, then the full 1200 V rated voltage will be blocked between the source and drain, however, when the device is on, the source to drain resistance will be 80 mOhms. The switching time, according to the datasheet, is on the order of 30 ns to 80 ns.

Figure 3. CMF20120D Die Cross Section

The top side of the die features the source metal and a poly silicon gate array. Figure 4 shows a cross section through a single cell in the gate array. The transistor is comprised of source contacts to N+ diffusions that lie within the P-body. The N- drift region is formed using a layer of SiC epi plus the full thickness of the die.

Figure 4. CMF20120D MOS Gate Cell Cross Section

A more detailed view of the edge of a MOS gate is shown in Figure 5. The active channel lies between the edge of the N+ diffusion and the N- drift region.

Figure 5. CMF20120D MOS Gate Cell Cross Section Detail

The CMF20120D is the first easily available SiC vertical MOSFET device. This analysis shows that Cree has adopted a very conventional design for this discrete device. The next step will be the development of a SiC power IC that incorporates both control circuitry and power transistors into a single chip.

Related Analysis:

CREE CMF20120D Silicon Carbide Power MOSFET 1200V 80 mΩ Z-FETTM MOSFET N-Channel Enhancement Mode Exploratory Analysis

Efficient Power Conversion EPC1015 Enhancement Mode GaN-on-Silicon Power Transistor Exploratory Analysis

International Rectifier IP2010PBF GaN Integrated Power Stage Process Review Report

ON Semiconductor demo a great metaphor for industry

Tuesday, February 14th, 2012 by rwilliamson

Contributed by Rob Williamson.

In a previous post we discussed the massive push of really innovative products into the LED lighting market that, in North America anyway, doesn’t really exist yet. But consumers aren’t early adopters in a low-thought purchase like light bulbs.  Good thing technical people are always looking for the bleeding edge you say. This is not always so in the power industry where things designers know, trust, and are on the periphery are often best left alone.

The very technical people that are building power supplies for the latest washing machines, handheld electronics, and server farms have a hard time letting go of what they know. Moreover, they often have a hard time justifying potentially higher costs and risks on a new feature that their end customer doesn’t value. It is hard to balance a 3% improvement in vampiric power draw against an increase of the BoM cost on that particular feature.

We saw this theme a few times at APEC 2012. There was the LED lighting discussed above, the reluctance of some designers to adopt digital controllers even when the application might be better suited to use them. But none more saliently than a fantastic demo in the ON Semiconductor booth.

Space Invaders running on a Commodore Vic 20

ON Semi was showing a part that they still sell in high volume called the TL431: Programmable Precision Reference / Shunt Regulator, This is a part that dates back to the Commodore Vic 20!  The sign above the display said, “Time for an Upgrade?”  What stronger challenge is that?  Editorial note – we at Chipworks are very partial to this kind of demonstration because the entrance hallway to our corporate head office includes a computing museum with a slightly more modern Commodore 64.

This particular dinosaur was actually a fully working demo. No emulator hidden under the counter here. ON Semi was selling the point that, although there is demand for the product, they have something new, the NCP431 that has modern features that should be considered when designing a power supply. With more and more power components 100% focused on lowering their parasitic draw while also improving peak efficiency to well over 90%, they are pointing out that some of the peripheral devices are now an important design consideration for the overall supply efficiency. So upgrade!

ON Semi demonstrated a few other products at the show, including the NCP1246 fixed-frequency current mode controller that, according to their press release, “achieves extremely low no load input power consumption in AC-DC adapters”.  The latest technology eliminates the need for resistors to bleed power for safety but detects the signal and switches off, reducing power demands. This impressive part works with the NCP4353 or NCP4254 secondary side switch mode power supply (SMPS) controllers to deliver input power of <10 milliwatt in a 65 watt notebook adapter. It also includes constant voltage and constant current regulation and optional built in LED driver for an adapter indicator light. Booth staff advised that it is sampling with customers and will be in full production at the beginning of Q2.

On the LED lighting side they were showing off the NCL30000 Power Factor Corrected TRIAC Dimmable LED Driver.  It isn’t brand new, but still the device performed well and dimming performance was impressive. ON Semi advises us that the device has over 90% efficiency and, like others, is doing primary side regulation to reduce the overall module part count.

HV AC/DC LED Drivers, a great potential!

Thursday, February 9th, 2012 by rwilliamson

Contributed by Jefferson Chua and Rob Williamson

It is no surprise that High Voltage LED Drivers is one of the hottest topics at the APEC. In fact, a lot of people (in the average range of 300) flock to HV LED sessions in order to understand more about this technology while exhibitors, such as Fairchild, TI, Monolithic Power, NXP and many others try to showcase their HVLED portfolio. This industry is no cosy duopoly but a fiercely competitive one with dozens of billion dollar suppliers touting their latest products.

So with all these driver companies vying for sockets, why don’t I have a single LED light bulb in my home? Several speakers at the 2012 APEC conference either had a slide on the subject, or spoke to issues during their talk, so we will summarize here.

The power industry has a perception problem in the mainstream. Nobody understands it and LED is a clear example as to why:

1)      “They” use terms like 50,000 hour bulb life – which is confusing a consumer isn’t going to do that math at point of sale. Moreover, it is misleading since the buyer will think that their average bedroom light, that is used about 2 hours per day, will now last their entire lifetime (when the capacitors won’t)

2)      They tout fundamental changes in the national power consumption (lighting was shown during a plenary session at APEC to use 3% of the total grid so there are better places to put our environmental mindshare)

3)      They cite the low efficiency of traditional lighting but if a consumer is smart enough to know where the efficiency loss goes (heat), and if they live in a cool climate, then they are probably smart enough to know that waste heat is (a relatively expensive) benefit 75% of the year.

4)      They use terms like “color” to explain the light quality when compared to incandescent, they are all white. And a white that most of us aren’t really happy with anyway.

5)      And finally, since we can’t do the math, the bulbs have high cost compared to a $.25 incandescent one.

Despite these challenges, it still seems more than obvious that LED lighting is the future. Proof of the strange dichotomy in the power industry that innovates like crazy and making a real difference in small increments.

The history of LED lighting drivers started innocently and simply enough – according to Matt Reynolds of TI, people used to take a buck-boost converter and convert it to an LED driver. While that may work on Gen 1 LEDs, it is not going to work on Gen 2 and 3 LEDs.

He then talked about the retro-fit dilemma of having 250 billion sockets needed to fill, and if only (I/they) could get 1% of market share. In the rest of the technology industry there is no such thing as 1% market share (to quote the movie Antitrust, technology success is binary, you are either a one or a zero), but in power technology the market is crowded yet everyone seems to be successful, and gains are measured in small increments after all.

So what innovations are going on to make a good SSL LED IC Driver? Lower cost, high (improved)-efficiency, reliability, good LED current regulation, good dimmer decode, and  good system protection are among those mentioned.

One vendor solution for LED (Fairchild) - click to enlarge

LED drivers have been the weakest link in LED systems because of component failure, PCB board defect, lightning/surge failure, solder junction failure and so much more. Reynolds then talked about the possible solutions to these LED driver challenges such as the use of GaN, high-voltage LED stacks, and AC LEDs.

Having said all of this, it is no wonder why companies are showing off their chipsets and reference boards for LED lighting with new controller products from such as ON Semiconductor, NXP, Fairchild, TI, and Monolithic Power.

To finish up, here is a demo video on the dimming quality of today’s chips in the Monolithic Power booth.

After all, the mythical 1% market share in the case of LEDs, is a whole lot of sockets.

Transphorm’s GaN power device takes on Infineon’s CoolMOS

Thursday, February 9th, 2012 by rwilliamson

Contributed by Jefferson Chua.

Transphorm, which offers the industry’s first qualified 600V GaN device, states that GaN is the choice of power device at higher frequencies. That is, GaN’s frequency-loss ratio shows that it has twice the advantage over SiC and 14x advantage over Si. Yifeng Wu (V.P. Product Development) took the stage at APEC 2012 to explain.

Yifeng Wu, VP of Product Development at Transphorm

Comparing performance

According to Wu, VP of product development, GaN replaced SiC LEDs within 3 years of the first product, going from research to a fully qualified product in 5 years. GaN then entered the RF market and is rapidly growing from the first power amplifier in 1993 to production in 2003. The next market is power, where GaN has gone from the first devices in 2000 to now being in production.

GaN Commercialization

Wu then stated that the GaN high band-gap transistor not only sustains high electric field with a short drift region, its GaN buffer layers are suited for low leakage and high breakdown voltages, and is normally off, which is ideal for use in the power market.

Block Diagrams of Device

Transphorm proceeded with an implementation of a typical boost converter using their proprietary GaN power transistor. The presentation compared the performance of GaN with Infineon’s CoolMOS to show the benefits of their technology. This comparison spanned almost half of the talk but here are some highlights.

A graph showing the di/dt between the GaN transistor versus the CoolMos shows that when both devices were tested at 450A/microsecond, Transphorm’s GaN HEMT came out with little ringing while CoolMos was unstable.

Comparing coolMOS with GaN

During the question and answer portion, Wu stated that Transphorm’s 600V GaN power device has a better breakdown voltage compared to the Infineon CoolMOS. Its thermal characteristic is similar to the CoolMOS while its input capacitance is 1/5 of a CoolMOS.

But can the technology compete with the CoolMOS for all applications?  Not necessarily. Firstly, the expectation is that it will only be price competitive in the next “5 to 10 years” so the focus for GaN FETs by Transphorm is in the higher value markets. Additionally, Wu indicated that they have only qualified their solution to 150 degrees and that it should not be used for applications with a high spike design. However, to quote the answer to an audience question about yields, “yields are good and if you place an order for as much as you want, we’ll ship it.”

While we haven’t looked at a Transphorm device yet, we have looked at a CoolMOS, and if you place an order on this link, we’ll ship it!

Die Photo of Infineon IPB50R299CP CoolMOS Chip

Texas Instruments Product Ecosystem has Power Covered

Wednesday, February 8th, 2012 by rwilliamson

We have been documenting the success of Texas Instruments by the simple method of watching their design wins in mobile devices. TI certainly leads the globe in analog foundry capacity, and with the acquisition of National Semiconductor gets the benefit of some (already paid for) space at the 2012 APEC conference.

We took the chance to sit down with Michael Gilbert, Senior Product Marketing Manager for Isolated Power.  Long title, but he presented some of TI’s roadmap and coverage in the power sector and they seem to have the breadth and technology to continue their growth (dominance?).

They were right on top of the news makers in GaN by announcing several FET drivers that were specifically architected for GaN devices. This is a new market and Mr. Gilbert was, “not aware of anyone other than Transphorm and EPC that were producing in any kind of volume”. That said, TI has invested in what is a very promising technology.

Comparing the size of Si and GaN Devices

The designs specifically handle GaN challenges like, “never going above 6V at the gate and getting to zero very fast”.

Specifically they have the LM5113 100V integrated half bridge, high-side driver, the new LM5114 low-side driver, and the new UCC27511 high-speed 4-A/*-A low-side driver. You can see the influence of some National Semiconductor part numbers here. They are also launching a “coming soon” 5.4 MHz, 4.5-V to 60-V synchronous buck controller for GaN FETs or MOSFETs expected to be in the hands of clients towards the end of H1.  The point is, TI is investing heavily in the latest upstart technology.

TI Digital Power Ecosystem

In terms of the overall coverage, TI is acknowledged as having a huge analog portfolio, but from their ecosystem slide they certainly have coverage in the digital power ecosystem, whether this is with traditional analog chips, analog solutions with digital interfaces or mixed signal MCUs. And they continue to invest.

At the show they announced their next generation digital controller for isolated power, the UCD3138 and it is touted as the first highly integrated controller optimized for AC/DC and isolated DC/DC power supplies. Rather than re-purposing existing technology, the device was specifically designed for this application to limit die size and keep the cost very competitive. It includes 32K of flash, an ARM7 32-bit processor, high speed precision data converters, multiple programmable hardware control loops, analog peripherals and various communications engines.

UCD3138 Architecture

As a digital device, they have the expected GUI-configurable programmability and also includes advanced monitoring capabilities. Several development kits are available from TI.

  • Development kits:
    • UCD3138PFCEVM-026: Universal input, 400-Vout AC/DC PFC development kit, configurable into single- or two-phase interleaved and bridgeless topologies
    • UCD3138PSFBEVM-027: 400-Vin/12-Vout DC/DC phase-shifted full bridge
    • UCD3138LLCEVM-028: 400-Vin/12-Vout DC/DC half-bridge resonant LLC
    • UCD3138HSFBEVM-029: 48-Vin/12-Vout DC/DC hard-switching full-bridge
  • Reference designs:
    • Universal input, 12-Vout 600-W AC/DC reference design (PFC plus LLC, and PFC plus phase-shifted full-bridge)
    • 48-Vin/12-Vout 1/8 brick DC/DC reference design (hard-switching full bridge)

More to come from APEC – stay tuned!