Start-up stakes its future on InP for cell-phone handsets

Following Brewer s arrival, Xindium radically changed direction to develop InP for power amplifier (PA) applications. Then, two months ago, the company lost its entire research team along with co-founder Feng in acrimonious fashion after an apparent difference of opinion with Brewer over the commercialization of products. Xindium is now focused squarely on PA solutions, and Brewer says that InP can make an impact in handsets sooner than might be expected. "There are products that [handset OEMs] can t ship today because the PA doesn t work properly," he said. "Every major compound semiconductor company has got InP-based PAs on the near-term road-map, and so does every handset manufacturer."

By short-term, Brewer means the next two to three years at most, but he believes that manufacturers will make their move in the next year. "This is one of those things that s been creeping along below the radar for quite a while, but our epitaxy vendors have got people demanding high-power, high-breakdown voltage InP constantly. Our processing partner has the same issues."

"Last December the very strong demand was underforecast," explained Brewer, adding that the two components that proved particularly difficult to obtain at that time were color displays and PAs. The smaller handset ODMs saw their supply squeezed at the expense of larger customers, and good lead times and delivery were hard to come by, he added. This would allow a smaller company like Xindium to find its niche in the market.

Xindium already has one GaAs-based quad-band PA product on the market, and Brewer says that this portfolio will be extended to three or four by early next year, with essentially the same product but in a different package. "We re selling die into module manufacturers in Asia. There s a whole group of companies that build all sorts of modules for things like Bluetooth, but they can t buy GSM die - so they re left out of the GSM market."

Brewer says that although it was not part of the company s original plan to commercialize GaAs-based RF devices, it has been an important factor in forcing the start-up into commercial mode, selling products, and getting its name in front of handset makers.

This may be crucial now, but however well Xindium performs in the GaAs PA market, its future products based on InP will ultimately determine whether or not the company is a success. "Our market is InP for EDGE and W-CDMA. We will live or die on that," Brewer told Compound Semiconductor.

Although EDGE phones are on the market and networks are being built to accommodate their higher data throughput, Brewer says that these handsets do not have "true" EDGE capability. "Nobody can build an EDGE PA right now that meets the output power and the multi-time-slot transmit requirements," he claimed. "EDGE phones are shipping, and they all do up to 100 kbit/s in the downlink, but in the uplink they re all limited to half the necessary transmit power and one time slot - essentially no faster than GPRS is today."

And this is where InP-based PAs come in. One of the key physical advantages of InP compared to GaAs technology is its high thermal conductivity. "What s limiting transmit power and speed in EDGE is the PA heating up," explained Brewer. "InP is a 50-60% better heat conductor than GaAs, and it also has significantly better speed and bandwidth. So you can make smaller transistors that generate less heat but the right amount of power."

Brewer appears to be under no illusions regarding the technical challenge of developing InP-based PAs for commercial EDGE applications. "An EDGE PA is probably the hardest thing ever done in a MMIC at these power ranges, in terms of linearity and efficiency," he said.

Cost is another issue with InP, but Brewer believes that InP PAs can make an impact in high-end handsets: "There will always be room [in the US market] for a $500 handset that has the latest, whizziest bits in it. And in that segment there s enough room on the bill of materials for PAs to go in at the high $3 range. In 2005 this will not be an issue for InP."

That will mean extending the brittle, awkward-to-handle InP substrates from 4 inches to 6 inches in diameter, a technical feat that seems unlikely when the substrate has to be thinned to just 100 μm. Brewer thinks that an answer may lie in epitaxial techniques, where a thin film of InP is deposited onto a GaAs substrate. "The epitaxial challenge in 2005-6 is to work on 4 inch diameter wafers, but to have a 6 inch solution. I think that GaAs substrates are probably the solution." While Xindium concentrates on its material solution to the heat problem, other PA makers are trying to solve it through advances in packaging technology. "We re now seeing flip-chip PAs, when a few years ago this would have been viewed as heresy," said Brewer. "But we know of two vendors that have produced flip-chip prototypes to improve heat conduction in GaAs PAs."

Although Xindium is without its Urbana-Champaign-based research team and pure research is on hold for a few months, Brewer says that the company already has its epitaxy structure for InP devices worked out, and that the main challenge is to co-ordinate manufacturing. Taking charge of the technical side of the company for the moment is Keith Manssen, Xindium s vice-president of engineering and a 20-year veteran of the RF and cellular industry.

Of the departed technical team under Feng s control, Brewer said that while it was a "very, very capable" group of researchers, it lacked expertise with regard to commercial products. "They did really good work for us. The difficulty is the small-group collegial environment, which can be quite insular," he said. "Nobody in Champaign knew anything about GSM. The guys that we now have [in Crystal Lake] have all been working on GSM PAs for 10 years."

"Research mode is always more comfortable, the time lines are less rugged. But when you re in commercial mode it s a case of get the thing working and get it tested. Get the technology to market as fast as you can."