WiMAX and WiBro emerge as major targets for GaN

Like many early-stage technological innovations, wide-bandgap transistors could legitimately be described as a solution in search of a problem. The power output, high-voltage and high-temperature operation of GaN, as well as the robustness and reliability of SiC-based microelectronics, are all very impressive. Such devices have attracted lots of attention and funding for research from the military sector, but genuine commercial applications remain scarce.

While third-generation cellular was generally regarded as the application that would usher GaN transistors into the mainstream, it now seems more likely that the emerging broadband wireless access protocol WiMAX will provide that opportunity. WiMAX (also known as IEEE standard 802.16e) will have to compete with many other broadband alternatives, but could be the preferred last-mile solution in areas that are underserved by wired broadband or high-speed cellular coverage.

Kevin Linthicum, the chief technology officer (CTO) at GaN transistor manufacturer Nitronex, certainly believes that WiMAX offers a big opportunity. The Raleigh, NC, company - one of the pioneers in GaN microelectronics - has just released its first commercial products for the WiMAX market, having sampled them to customers for the past year.

The Nitronex strategy is to target the application areas where silicon LDMOS is weakest. "In chasing LDMOS we ve given up trying to compete on cost," Linthicum told delegates at the Key Conference in Palm Springs, CA, in early November. "What we need to do is focus on areas where LDMOS might struggle."

In practice, this means targeting high-frequency applications. Initially the WiMAX protocol will operate at 2.5-2.7 GHz and 3.4-3.6 GHz, with next-generation systems set to use the 5.725-5.85 GHz band. More specifically, Nitronex is looking to penetrate applications in WiMAX base stations rather than target customer premise equipment.

Crucially, Nitronex has already snared one customer in the high-technology breeding ground of Korea. In early 2006, RFHIC, a company based in Suwon that sells RF components and modules, is expected to deploy WiMAX products from Nitronex that feature its 50 W GaN-on-silicon transistors.

The power GaN HFET devices measure 1 mm × 6 mm, and the chips feature a gate periphery of 36 mm. The amplifiers produce around 70 W in continuous-wave power at 28 V, with efficiency peaking at around 65%. Nitronex fabricates the devices by MOCVD to deposit a stress-mitigating transition layer between the silicon base and a 0.8 μm semi-insulating GaN buffer (figure 1). Linthicum says that at the moment the full process gives Nitronex a known-good-die yield of approximately 75%, while the company has recently used field-plate technology to increase power and gain characteristics.

Although Nitronex manufactures its GaN transistors on a silicon base, the substrate material is not quite the same as the conventional (100) crystal structure used in CMOS and LDMOS processing. Instead, Nitronex uses float-zone silicon with a (111) crystal orientation. The good news for Nitronex is that this is a scalable technology, and a future change to 6-inch-diameter wafers is expected. However, something that will prove more problematic for the company will be finding a toolset supplier that can provide GaN deposition equipment that is compatible with the larger wafers.

This is because LED makers dominate the world of GaN device manufacturing, and in this business much smaller substrates are typical. While pioneers such as Cree have begun the switch to 3 inch production (Cree now makes about 60% of its chips on 3 inch substrates), 2 inch fabrication remains a common choice.

However, switching to 6 inch GaN-on-silicon will not be necessary to serve the requirements of the relatively small WiMAX and W-CDMA base station markets, and more applications will need to be found to justify any drive towards greater manufacturing volumes at the larger diameter.

At the moment it appears that the WiMAX drive is being spearheaded in Korea, which has its own particular "flavor" of WiMAX, known as WiBro (short for wireless broadband). The Korean communications company KT Corporation is pioneering the deployment of this standard, which uses the 2.3-2.4 GHz frequency band.

KT joined the board of the WiMAX Forum, a global consortium pushing the technology, in November, and has since launched the first commercial WiBro service in Korea - a demonstration for industry executives and dignitaries at the Asia-Pacific Economic Co-operation s annual event in Busan, south-east Korea. The WiMAX Forum president, Ron Resnick, said of that development: "KT has launched the world s first mobile broadband network that is based on the soon-to-be-ratified IEEE 802.16e standard, which is at the core of mobile WiMAX."

Products based on the WiBro standard are set to feature in the first mobile WiMAX systems, with a release due in late 2006, and Linthicum says that the Korean market is already exerting a considerable demand on Nitronex technology.

"The WiMAX and military markets will be the early adopters of GaN HFET power transistors," added Linthicum, who expects to see revenue from these sectors within six months. Commercial deployment in W-CDMA base stations will take at least another 6-12 months to bear fruit, with the pace of design-ins determined by the need to introduce new architectures and broader-bandwidth power amplifiers (PAs). "The cellular market is very conservative," Linthicum said, indicating that the likelihood of penetrating this sector with GaN technology may be even further off.

Cost will be one of the key factors that decides whether or not GaN makes an impact. Direct comparison with silicon LDMOS is difficult to quantify, since the cost of silicon die can be such a fast-moving target. For Linthicum, a more relevant comparison can be made with GaAs: "GaN will need to debut in the range of commercial GaAs PHEMT pricing in order to be adopted in commercial markets," he told Key Conference delegates.

Of course, Nitronex is not the only company to be targeting WiMAX applications with GaN amplifiers. Eudyna Devices sees WiMAX as a critical market for GaN, and the Japanese company has three product offerings in each of the 2.5-2.7 GHz and 3.4-3.6 GHz bands, ranging from 30 W to 180 W output.

Oki Electric, a second Japanese company with WiMAX in its sights, is taking a similar approach to that of Nitronex. Oki says that the high-performance amplification of its GaN-on-silicon HEMT, which has a 56 GHz cut-off frequency and an fmax of 115 GHz, results from the use of a so-called recessed gate structure (figure 2).

"We can contribute to the acceleration of WiMAX and next-generation wireless communication systems," said Oki s CTO, Harushige Sugimoto. "Volume shipments of such products are planned to start from 2007."

While Oki and Nitronex are taking the silicon route to cutting transistor cost, Eudyna has chosen a more radical approach to making GaN devices more competitive. It plans to slash costs by using n-type SiC substrates designed for optoelectronic applications instead of the conventional, and much more expensive, semi-insulating SiC material.

LED manufacturers grow GaN on top of n-type SiC, and by using an AlN layer grown by hydride vapor phase epitaxy before MOCVD of the active GaN layers, Eudyna believes it has solved the twin problems of parasitic capacitance and dielectric loss that are met with the unconventional approach.

Eudyna s Yasunori Tateno expects that the cost of devices can be drastically cut using this method (figure 3). "From a cost reduction point of view, we must use n-type SiC substrates," he told Key Conference delegates. Tateno added that Eudyna was planning to release devices manufactured on n-type SiC by the end of 2006. Initial results have been impressive, with amplifiers showing an output power of 101 W and a gain of 15.5 dB; but whether Eudyna s customers will be happy to rely on the use of such unconventional materials remains to be seen.

Tateno was keen to emphasize the high efficiency of GaN HEMTs for W-CDMA cellular applications, saying that if all 100,000 base stations in Japan had their (relatively) inefficient silicon LDMOS devices replaced with GaN-based PAs, the savings could be huge - both in terms of dollars and the environment.

"If total base station efficiency was improved from 15% to 25%, electricity consumption would be reduced by 400 million kW hours, equivalent to 600 million kg of CO2," claimed Tateno. Such a wide-scale deployment is far from a realistic scenario, partly because the key figure of merit is not only the efficiency of the PA chips, but the efficiency of the entire system that is based around it.

At the Compound Semiconductor IC Symposium, which was co-located with the Key Conference, an improvement in GaN-based PA efficiency proved to be one of the most popular talks of the week. Nitronex and researchers from the University of California San Diego (UCSD), collaborating with Qualcomm and Nokia, described a W-CDMA base station amplifier with overall power-added efficiency of 50%, which is thought to be the highest figure yet reported. "Measurement of the high-voltage envelope amplifier used in this work shows efficiency of 77% under W-CDMA signals," claimed the team. It concluded that combining GaN HFETs with advanced amplifier architectures would lead to "dramatic improvements" in base station PAs.

The key to the development is not just the use of a GaN transistor from Nitronex, but also the advanced amplifier architecture that it enables. In the UCSD set-up, an envelope-tracking bias system improves both linearity and efficiency. The design means that the amplifier operates closer to saturation, with the transistor maintained at a lower temperature. The dynamic peak voltage also reaches higher values than can be used in a constant drain bias voltage configuration.

According to Linthicum, envelope tracking is just one of several advanced designs that GaN transistors can enable. Others include Doherty configuration Class F amplifiers, as well as digital predistortion amplifiers.

While the combination of these novel architectures and the material properties of GaN may never quite knock LDMOS off of its pedestal in cellular applications, it appears that WiMAX is now the key application sector for wide-bandgap RF microelectronics, and the one where its commercial viability will first be proved. If WiMAX takes off, it could be just the problem that GaN has been looking for.