Reliability is the central issue for DARPA's triple play on GaN

"A key objective of this program will be a rapid transition of the technology developed into military systems," said DARPA in its proposal document, a sentiment echoed by DARPA program manager Mark Rosker, who described the technology as a "coming revolution" at the CS Mantech event in New Orleans. DARPA also aims for a "great" improvement in understanding the physical reasons behind device failures and the development of physical models to predict performance, as well as a reproducible device and MMIC fabrication process, and improved thermal management and packaging.

Track 1 is focused on the development of an X-band transmit/receive module containing both power-amplifier and low-noise amplifier MMICs. If all of the program options are exercised, this contract could, ultimately, be worth a whopping $59.4 million.

For Raytheon, teaming up with Cree is all about accelerating the development of GaN devices. By combining the capabilities of both companies, the technology should become available to the military and commercial worlds much sooner, they claim. Raytheon s part of the project will take place at Raytheon RF Components in Andover, Massachusetts, while Cree s efforts will be conducted at its Durham headquarters in North Carolina, as well as at its Santa Barbara Technology Center in California.

For NGST, the primary contractor on Track 2, the focus is on moving its GaN research and development work into volume production - the demand from DARPA is for at least 384 three-inch wafers to be delivered over the course of the program. Historically, NGST has been a key member of DARPA s programs to develop III-V devices, having worked on the MIMIC GaAs-development program in the early 1990s and also on the successful development of InP-based MMICs that are used today in advanced satellite applications.

According to Dwight Streit, NGST s vice-president of foundation technologies, the expectation is that GaN will sit beside GaAs and InP MMICs as a complementary technology. NGST s ultimate goal is to produce a Q-band (> 40 GHz) module with a continuous-wave power output of 20 W (see tables).

Under the Track 3 effort, which kicked off in mid-February, device development will take place at both TriQuint and BAE Systems, with the relevant teams from both companies working together on the process. TriQuint s Tony Balistreri, the firm s research and development program manager in Hillsboro, Oregon, says that the manufacturing process will be finalized at TriQuint.

Balistreri and Rosker highlight reliability as the key challenge for all three teams. Track 3 is to produce a wideband high-power amplifier module operating at frequencies of 2-20 GHz for applications such as jamming and electronic attack. Balistreri says that this window takes advantage of TriQuint s strength in X-band power output and efficiency, while BAE Systems brings expertise in devices operating at higher frequencies and higher gain to the table.

"Developing this band gives us a path to higher frequencies," said Balistreri. If successful, the development should lead to modules being manufactured for a range of commercial applications such as local multipoint distribution services, two-way satellite links and cellular back-haul.

While it is clear that in all three tracks SiC will be the substrate material of choice, a crucial part of the initial two years (denoted Phase II) of this project will involve gaining a more thorough understanding of the effects of different substrate types on device performance. For example, BAE Systems has vast experience in making GaN-based low-noise amplifiers on native GaN substrates.

Nitronex, also a Track 3 partner, is focused on GaN-on-silicon transistor development, and Balistreri says that the Raleigh, NC, company is on board to help the reliability drive. The team wants to understand why GaN-on-silicon devices have superior lifetimes, and how this knowledge can then be applied to GaN-on-SiC devices.

The TriQuint team also has two academic partners that will take on crucial roles. Michael Shur from the Rensselaer Polytechnic Institute in Troy, NY, is one of the world s leading exponents in the physical modeling of devices, while Jesus del Alamo at the Massachusetts Institute of Technology will lend his experience to the reliability drive through extensive physics-of-failure analysis.

Wideband operation will enable multiple functions in a single system that transmits and receives a range of communications at the various frequencies it employs. For TriQuint, the 18 month technological goals (table 2) call for a wideband device operating at 40 V with 39 dBm continuous-wave output power, 12 dB gain and a power-added efficiency (PAE) of 60%. TriQuint s state-of-the-art performance is exhibited by a 40 V, 1 mm device with an output of 5 W/mm at 10 GHz, 11-12 dB gain and a PAE of a little under 50%.

While the performance achieved so far is fairly close to the X-band goals, significant development is required to push up to the higher frequencies needed for the wideband MMIC. "Device performance is a challenge, but it isn t the main one," said Balistreri. "The number-one challenge is reliability." To improve reliability there are two key steps: optimization of epitaxial structures and the development of a manufacturing process that maintains reliability.

"Getting to 104 and 105 [hours reliable performance] maybe tough," said Balistreri "but reaching 106 is the real tricky part." All devices degrade under stressful conditions such as high-temperature operation or RF under bias, with effects such as a degradation of drain current with time manifesting.

Like NGST, TriQuint sees the development of a volume-capable manufacturing process as a priority, and as a recognized volume supplier of MMICs already, it is putting its own money and resources into this aspect of the program: "We are a large volume producer of III-V devices, and we expect to become a large supplier of GaN devices as well," said Balistreri. "At the end of the day, TriQuint expects to have both a GaN foundry process as well as standard GaN products."

While DARPA s development of GaN has led to inevitable comparisons with the GaAs MIMIC program of the early 1990s, there is some disappointment within the industry regarding the amount of funding allocated to the effort, which is smaller in scale than the GaAs program. In comparison, the GaN teams are working with what might be termed a "tiger" budget.

However, it seems clear that DARPA is trying to fast-track GaN development in another way, with the agency wanting to see GaN technology in live systems as soon as possible. While there is usually an intermediate step between the technology-development programs and system insertion, that is not the case with the GaN projects. In the second phase of the current efforts (denoted "Phase III"), project partners are under instructions to come up with business plans that will identify system-insertion points at the earliest opportunity.

By taking this relatively aggressive approach, it is reckoned by some that three to five years may have been taken out of the entire commercialization process. Module development and engineering may commence in around five to six years, while volume production should become a reality within a decade.