Microwave Journal
www.microwavejournal.com/articles/28179-executive-interview-jim-milligan-director-of-rf-for-wolfspeed
Jim Milligan, director of RF at Wolfspeed

Executive Interview: Jim Milligan, Director of RF for Wolfspeed

Wolfspeed will remain with Cree, after the Infineon deal was blocked by the U.S. government

April 13, 2017

What can you tell us about the U.S. government’s concerns that led Cree to terminate the proposed acquisition with Infineon?

Cree’s February 16 press release states that the U.S. government’s Committee on Foreign Investment in the United States (CFIUS) had national security concerns for which alternatives were not able to be identified. Cree has shifted its focus back on growing the Wolfspeed business. Further information can be found on Cree.com.

Describe the Wolfspeed business, specifically the RF segment, and how you see its outlook.

The RF business segment offers for sale both catalog bare die and packaged products, or components, as well as foundry services for customized products. These products or services are based on GaN semiconductor device technology on a SiC substrate (GaN on SiC).

GaN on SiC has become known in recent years as an emerging, high performance semiconductor technology that offers higher output power, bandwidth and efficiency than incumbent device technologies such as silicon LDMOS or GaAs.

As mentioned, we also offer foundry services where external companies can design — or have our MMIC team design for them — custom products using validated in-house models and layout rules, via process design kits, that are manufactured in our dedicated GaN fabrication facility. These custom products can be as simple as bare GaN die for 4G telecom applications or complex MMICs that can be realized as multistage power amplifiers, low noise amplifiers or a variety of control functions (switches, phase shifters, attenuators, etc.) for emerging defense applications.

At present, we offer four released GaN processes covering DC to 18 GHz, from which customers may select the specific process that best meets their cost and performance requirements. We’re also very excited about a new millimeter wave process in pre-release, which we’ve made available to early adopters, that will provide solutions for many of the higher frequency 5G architectures (e.g., massive MIMO arrays) being considered by telecom OEMs and network operators.

Our GaN RF business outlook has never been brighter. The worldwide RF GaN market for 2017 is projected to be in the range of $300 million, with a compound annual growth rate of 14 percent over the next five years, and we are extremely well positioned to be a significant player in that space.

Prior to the announcement of the Infineon deal, your strategy seemed to be to develop products for defense and aerospace applications and serve as a foundry for companies pursing the infrastructure market. Is this still your go-to-market approach?

We remain strategic in our positioning of the technology where it can provide the most value. The mix between which markets are serviced with catalog products or foundry services is based on the best way to service a given market segment, given our strengths along with those of our customers.

You’re designated a trusted foundry by the U.S. government and have a strong reputation in the defense market. What’s the outlook for this market and opportunities for GaN insertion?

As you know, the high performance capabilities of GaN have been something the defense market has been focused on for the last 10+ years. As such, defense contractors were many of the earliest GaN adopters to make use of the technology. Today, GaN is ubiquitous across the defense community as the go to technology for next generation systems.

In the defense radar market, we are seeing strong growth via new and upgrade programs. Emerging radar systems are taking advantage of the performance benefits offered by GaN, via solid-state upgrades to historically tube-based systems. These newer systems will have the ability to detect, track and classify smaller targets at longer ranges. We see a significant number of opportunities across ground-based, airborne and naval systems addressing a combination of radar, EW, comms and air traffic control functions.

We see a lot of discussion about RF energy, such as solid-state cooking and RF heating. Are you considering pursuing any of these applications?

Yes, this is a very exciting new market space and one we think will benefit greatly from the use of GaN. We are actively engaged with customers currently developing systems for solid-state cooking, industrial heating and drying, etc.

The cover feature in our February issue discussed Ericsson using GaN in point-to-point radio systems. Point-to-point radio opens the door to many millimeter wave bands, which are also being discussed for fixed wireless access and 5G. Where do you see the opportunities for your GaN technology at millimeter wave?

As you say, we see GaN as being extremely beneficial to upcoming millimeter wave applications, particularly in the 28 and 38 GHz bands. How and where GaN will be applied will be dependent on system architectures and application requirements.

Architectures such as massive MIMO or distributed antenna systems (DAS) requiring elemental beamforming, where size is critical, will play nicely into GaN, given its higher power density and smaller physical size for a given power level. It will also depend on the optimum antenna array size to satisfy a given application requirement. Given a trade between array size and output power per element, larger arrays at lower powers may favor SiGe or GaAs versus GaN, which may be more applicable to smaller, higher power arrays.

The optimum approach will vary depending on system power requirements, packaging constraints, beam coverage, etc. We’ll likely see a mix of technologies used in the end, but we’re confident GaN will be a significant player in this space.

Remind us of your GaN process portfolio and the new processes you’re developing.

A list of our currently released process technologies is shown in the table:

Process

Gate Length (μm)

Bias (V)

Breakdown Voltage (V)

Power Density (W/mm)

Max Frequency (GHz)

Efficiency (%)

Wafer Size (in)

Substrate

G28V3

0.40

28

>120

4.5

8

75

4

SiC

G28V4

0.25

28

>120

4.5

18

75

4

SiC

G40V4

0.25

40

>120

6

18

75

4

SiC

G50V3

0.40

50

>150

8

6

75

4

SiC

 

We currently run these processes on 4-in (100 mm) GaN on SiC substrates. Being vertically integrated, we also manufacture both the substrates and the epitaxy used in these processes.

As mentioned earlier, we also have a 0.14 µm process in development that will provide good performance through 40 GHz. This process is currently available to early adopters and is scheduled for full release during calendar 2017.

Do you see capacity expansion on the horizon?

We are continuously focused on improving the price and performance points for our products. The transition to 150 mm wafers will be undertaken in that context, based on our volume/capacity model. While there can be hype with regard to 150 mm GaN, as with any technology, the right time to transition is based on a variety of factors, including fab loading. The GaAs industry was scarred with a number of fabs that made the move too early and didn’t have enough loading to realize a cost benefit. We are evaluating when the best timing for that conversion would be for us.

Your main RF competitors (e.g., MACOM, Qorvo, WIN) have multiple technologies in their semiconductor portfolios, while your RF segment relies on GaN — although you have a sister unit pursuing a totally different market area. Do you plan to focus solely on GaN for RF or might you expand your technology base to meet your growth objectives?

While having a diverse portfolio can have advantages, it can also be dilutive to the speed at which any one technology can be improved.

Conversely, we’ve chosen to focus on one technology and strive to be the best in the world at it. For us that is, of course, the SiC and GaN material systems. Being vertically integrated and focusing exclusively on only SiC and GaN has allowed us to mature and produce high performance, very reliable products for our customers. Having access to all elements of the process (crystal growth, epitaxy, device processing) allows us to push this technology forward quickly. It’s also beneficial to a manufacturing line to have that single technology focus.

There is debate on GaN on Si versus GaN on SiC for RF applications. Are the markets large enough for both to thrive?

Yes, there has been a lot of play in the industry regarding the two GaN implementations.

As you may know, for all the debate, GaN on SiC currently services greater than 95 percent of the RF GaN market, even though GaN on silicon has been available for the better part of 15 years.

The notion of GaN on Si admittedly sounds attractive, because silicon is largely synonymous with low cost, and SiC undoubtedly costs more to produce than silicon. That said, the silicon used in GaN on silicon is not high volume, low cost CMOS silicon. It’s high resistivity silicon that is not only more expensive to produce but has more loss than a semi-insulating substrate, such as SiC, as frequency increases and as junction temperature increases.

Silicon also has a much higher thermal resistance. One of the primary advantages of GaN is its higher power density. The beauty of SiC is its very high thermal conductivity. So, in one sense, GaN on SiC is a match made in heaven, very high power density and the ability to cool it. Our semi-insulating SiC substrates have a thermal conductivity of 4.9 W/cm-K, while silicon has a thermal conductivity of only 1.5 W/cm-K. Thus, the die size of GaN on silicon devices has to be larger in order to keep the junction temperatures under control. This means that the devices are typically 20 percent larger than an equivalent device in GaN on SiC. This negates much of the purported cost advantage.

The other advantage of GaN on SiC is that GaN and SiC are much more closely matched in lattice constant than GaN and Si. So exotic buffers and strain relief layers are needed in GaN on Si to keep the epi from cracking or warping the wafers, and that problem gets much worse as the wafer diameter increases. These buffers also add extra cost.

So the net result is GaN on Si starts with a cheaper substrate, but then adds extra cost in buffer layers during the epi process, then must make a significantly larger die to handle the thermal limitations. While I won’t say this washes the cost out completely versus a SiC substrate, it largely makes the difference in the final cost come down to a question of performance. Clearly, the market values the better performance obtained with GaN on SiC over GaN on Si, based on its much higher adoption rate.

That said, I do believe GaN on Si has a place in the GaN ecosystem, but it will likely be in lower power, lower frequency applications due to the thermal and frequency limitations discussed.