Raytheon Primes GaN For Electronic Warfare
In an industry first, the Office of the Secretary of Defense has rated Raytheon's GaN process as ready for defence production. Compound Semiconductor finds out more.
Painstaking process development from Raytheon is helping to drive GaN devices into defence applications.
In the race to build better electronic warfare, the US defence industry is ploughing more and more cash into developing GaN RF devices. These chips can be used in myriad applications including long-range radar, jammers and satellite communications systems.
To drive development forward, the Department of Defence (DoD) awarded so-called Title III contracts to key GaN players Raytheon, TriQuint, Cree and more recently Northrop Grumman, to increase yield, lower costs and improve time-to-market cycles for defence and commercial GaN integrated circuits. Earlier this month, Raytheon emerged as the first to complete the GaN production improvement programme.
“The award follows a decade of investment for Raytheon and the government," says Joe Smolko, director of Microelectronics, Raytheon IDS Advanced Technology Programs. “The Title III program is a capstone on development activity; there's been lots of materials science, process development and device work and in the last three years we have really wanted to establish economically viable production for GaN RF circuit technology."
And the company has. Attaining the highest level of manufacturing readiness - a production capability of '8' - of any organisation in the defence industry, Raytheon raised its GaN process yields by a hefty 300% and slashed costs by more than 75%. Process development was demonstrated on X-band power amplifier MMICs, FETs and MMICs and yields are in line with mature GaAs processes.
“If you look at the bottom line now, the GaN process is very well suited to power amplifiers and if you look at power amplifier cost in terms of dollars per Watt, GaN easily wins," says Smolko.
The Raytheon team started work on the Title III program in late 2009 focusing on L band to Ku band MMIC designs. As Smolko explains, in the first nine months the team processed 24 wafers with a base line production process to evaluate initial yields and reliability.
“The wafers were processed with a given mathset we had developed with government, to benchmark the technology at the beginning of the programme and frankly, the yield wasn't bad," he says. “Reliability was pretty good too but there were some pretty obvious and more subtle opportunities for improvement."
Having set 13 key performance parameters, including RF yield, line yield, cycle time and cost, the team went onto process some 150 wafers in batches. Most experiments were executed in series, so results could be incorporated into subsequent experiments.
According to Smolko, easy wins included passive component processing and the team also incorporated an improved gate process ways to raise yields.
“With every group of wafers we would carry out Pareto analysis, root cause analysis and basically find out where the dust bunnies were, and then come up with the methods and means to mitigate these," he says. “We did this through a sequential process and over the years improved yields by more than 300%."
At the same time reliability was verified at each step of the improvement process and the team also looked at ways to streamline manufacturing processes. “We looked at how we cue wafers, how we moved them through the process line and with a combination of actions reduced cost by 75% from the baseline to the end of the programme," explains Smolko.
So three years and more than two hundred 100 mm GaN wafers later, Raytheon has successfully completed the Title III GaN production improvement program. Competitors Cree and TriQuint are in the process of completing similar programs - Raytheon started earlier – and Northrop Grumman has just come on board.
Yields of Raytheon's GaN process are now on par with mature GaAs process levels, and while Smolko remains tight-lipped on the company's next GaN moves, he asserts the technology will 'figure prominently' in future roadmaps across a broad array of systems.
When asked if the technology will supplant GaAs, he answers: “Largely yes to wherever it makes sense. There's still probably some applications where it doesn't make sense to migrate from GaAs to GaN but more and more systems are going to tend towards using GaN."