Smart antenna boosts IQ of WLANs, startup says
By Patrick Mannion -- EE Times -- August 18, 2003 (11:36 a.m. EST)
MANHASSET, N.Y. — A Palo Alto, Calif., startup says it can deliver wireless' holy grail: a cost-effective multiple-input, multiple-output smart antenna based on orthogonal frequency-division multiplexing. So armed, Airgo Networks Inc.'s first-generation, three-antenna, four-chip design can double the throughput and quadruple the range of 802.11 wireless LANs, the company said.
"This is one of the most significant announcements of the year and will probably be the future when it comes to improving wireless performance," said analyst Craig Mathias, principal at Farpoint Group (Ashland, Mass.). "It's about 18 months ahead of my forecast for seeing MIMO [multiple input, multiple output] in WLAN designs."
Research in optimal receive algorithms and a dream team of advanced RF IC and software designers are the story behind Airgo's AGN100 chip set. Its single-chip CMOS baseband/media-access controller, the AGB100BB, was built in a 0.13-micron process. Fabricated in a 0.5-micron silicon germanium process, the AGN100RF front end-one for each antenna-is dual-band and 802.11a/b/g compliant, with support for the current drafts of 802.11e, .11i and .11h, Airgo said.
The Airgo design has already drawn attention from the IEEE 802.11n working group, formed in July to study proposals for next-generation WLANs with data rates in excess of 108 Mbits/s. The group is scheduled to deliver a standard in 2006. Between now and then it will grill Airgo, primarily over issues of compliance with legacy WLAN technologies, power consumption and of the cost-in both dollars and real estate-of implementing multiple analog transmit-and-receive paths.
The chip set is sampling now, with production expected by year's end. The company has not disclosed pricing but founder and chief executive officer Greg Raleigh says Airgo's solution would entail a $20 to $50 street price premium over existing WLAN designs.
A multiple-input, multiple-output antenna using orthogonal frequency-division multiplexing (OFDM) has long been considered the ultimate in wireless-channel optimization. The MIMO concept came to the fore in the early 1990s in research at Bell Labs that eventually led to the development of the Blast MIMO system. But it was too large, said Raleigh, and has yet to be deployed commercially.
The basic concept of the MIMO space-time-coding concept starts with the generation of multiple signals containing the same data and transmitting them with slight variations, such as delays (time), through multiple antennas that are separated (space). Multiple antennas on the receive side detect the signals and use advanced algorithms to regenerate the original signal. While doing his doctoral thesis, Raleigh hit upon what he called "very sophisticated mathematical models" that he said proved the multiplicative effect of MIMO was limited to at least three things: the number of input antennas, output antennas and multipath components present. "Using those precise models, we were able to mathematically derive the expressions for information capacity-without the need for simulation, as we already knew the results," he said.
According to Raleigh, the big design issue in MIMO-OFDM-and the area where most of Airgo's pending patents lie-is how to build an "optimal" receiver. This he defines as one that achieves the lowest possible error rate and highest possible capacity. Raleigh's mathematical models are the foundation for such a receiver. "Of hundreds of potential algorithms, we cracked the code on the best four or five," he said. Airgo filed patents on all of them, along with techniques to realize those algorithms cost-effectively.
While guarded on the scheme Airgo employed, Raleigh did say that it takes into account the instantaneous coherent-state information, as well as the interference signature across the receive array. Then, he said, the scheme coherently digitally downconverts the signals and puts the channel state and interference information into an optimal receiver that basically does the soft decoding and determines the output.
"There are about four classes of algorithm that would be considered optimal for this, any of which would result in $50 to $100 of raw silicon," said Raleigh. "That's totally impractical." Airgo's secret sauce lies in selecting the algorithm that enabled the design team to reduce that to 60 cents. This was important, Raleigh said, "because the MIMO receiver has to be a tiny percent of the cost of the chip, as there's lots of other stuff on there."
At the core of the radio-frequency (RF) design lies the principle of function sharing to avoid having completely separate (and costly) radios. "In general, you have to share RF components; you can't have a multiband radio that has separate IFs [intermediate frequencies]," said Raleigh. "And when you have this many antennas you have to find ways to share your frequency-synthesizer resources."
For reasons of sensitivity, signal integrity and noise-the reduction of phase noise is critical in OFDM systems-Airgo opted for SiGe, which generally is more expensive than CMOS. But Raleigh maintains that the cost premium is falling as more SiGe foundries come online. The use of SiGe was critical in getting the voltage-controlled oscillators and resonators in the synthesizer on-chip, he added.
Despite the advanced algorithms and RF design, Airgo still faces an uphill battle, since the combination of multiple analog front ends and complex algorithms that require large baseband processors is not conducive to the low-cost designs that are imperative for today's WLANs.
"There's a fundamental trade-off in terms of how much to do digitally with more-advanced signal processing, vs. how much you do on the analog side by trying to combine two or more transmit-and-receive chains," said Sheung Li, product line manager at Atheros Communications, a leading WLAN chip vendor. "It's a philosophical debate."
In Sheung's view, multiple analog chains should be avoided. He argues instead for digital processing, saying that more resources can be applied there to regenerate the signal-at lower cost.
"People are expecting WLANs to be built into everything-and cost almost nothing," he said. "That's why it's very difficult to just throw on additional analog." In the Airgo offering, MIMO entails "3x the analog and 3x the antennas-I don't get the math [for the claim of low cost]."
Sheung also pointed to the trend toward light access points, most of which now use Atheros' solutions. Airgo is "also going against that trend," he said. In addition, Sheung dismissed Airgo's competitive analysis, which he said compares the AGN chip set's performance with first-generation devices from Atheros and others. Atheros is currently prepping its fourth generation. "Any modern .11g/a chip set, with enhancements such as our SuperG or SuperA/G, would outperform what MIMO can currently deliver," he said.
Raleigh described the AGN100 as the company's first generation. "We expect orders-of-magnitude improvements in our next generation," he said. "The future of RF systems is not just in single antennas, but how one can integrate multiple antennas and multiple receivers to get a cost-effective architecture."
With respect to power, Ramesh Harjani, chief technical officer of Bermai Inc., was skeptical. "With multiple outputs, you have to halve [with two antennas] the ouput power from each or your power goes up." Carl Temme, Airgo's director of marketing, agreed, saying that the total power was equivalent to competing cards. "Our mini-PCI implementation [for example] consumes less than 2 watts peak."
Despite the skepticism, Raleigh believes the quality of the team means Airgo can deliver the goods. Along with Raleigh, the group includes two other co-founders: V.K. Jones, vice president of DSP and ASIC design engineering, and Richard van Nee, director of WLAN product engineering. Raleigh said van Nee, a Bell Labs veteran, is widely considered the founder of wireless LANs. From Philips, the founders pulled Steffen Hahn, who developed the RF ICs used in Apple's Airport and Intel's Centrino chip. Marketing's Temme and Frank Howley hail from Atheros. "The crew there is first-rate," said analyst Mathias, "they know what they're doing."
commsdesign.com |
