The announcement comes as the IEEE-802.11g Task Force has failed yet again to reach a consensus on a separate standard to create a middle ground between the approaches.

The task force wants to define a 22-Mbit/second and faster wireless LAN to run in the 2.45-GHz band. That would roughly slot in between today's .11b, which hits rates from 1 to 11 Mbits/s at 2.45 GHz, and the emerging .11a, at 6 to 54 Mbits/s at 5 GHz.

With the advent of chip sets like Envara's, time may have run out for the IEEE effort, which has been ensnared in both political and technical debates, some analysts said.

The delays fuel speculation that the .11g work might prove irrelevant in the long run.

"I don't understand .11g and the logic behind it anyway," said Craig Mathias, principal at consulting firm Farpoint Group (Ashland, Mass.). "You still need a different access point. It's a whole different radio. At best, .11g is an end-of-life kicker at this point. It's just not that important in the cosmic scheme of things."

In addition, the cost of 5-GHz radios is coming down fast, said Mathias, thanks to fierce competition spurred by Atheros' $35 all-CMOS solution. "If you can get 802.11a components at the same price as 802.11g components, who'd go for 802.11g?" he asked.

Intersil Corp. announced its $35 Prism Indigo silicon germanium/CMOS chip set in June, to analyst comments that the company was taking a hit on cost to remain in the same price range as Atheros — speculation that Intersil vehemently denies.

The bottom line, as Mathias sees it, is whether or not the falling cost of 802.11a kills 802.11b. "Many companies that have made a heavy investment in .11b are very worried about this, as the window to get a return on their investment is going to be much shorter than they thought," he said. Mathias believes .11a won't make .11b obsolete, he said, and that the best bet is to "put some users on each to make the best possible use of the airwaves."

Receptive market

Such mixed-network solutions, and the large installed .11b base in homes and public spaces, will likely supply a receptive market for Envara's solution. "Replacing the business infrastructure will take time and money; we need backward compatibility, as .11b will dominate for at least the next three years," said Izik Kirshenbaum, Envara's co-founder and chief executive.

"Where Envara distinguishes itself from other dual-mode solutions is its low-cost focus," said Navin Sabharwal, an analyst with Allied Business Intelligence. "Others are providing a DSP-based architecture with a high degree of programmability," he said, pointing to nBand and Systemonic, "while Envara's baseband solution is hardware-based and hence low-cost. In addition, it's providing its own RF chips, thereby allowing for higher integration." Many companies are offering MAC/baseband-only solutions.

With its genesis in Kfar Saba, Israel, Envara started out as hLAN, a HiperLAN2 advocate. But finding the standards process for HiperLAN2 excruciatingly slow, hLAN changed its name to Envara and took up the 802.11a flag.

Kirshenbaum called foul to wide claims that 5-GHz operation implies poor propagation characteristics relative to 2.45-GHz. "We have studies and white papers which show that on a building-by-building basis — not in free space — the propagation characteristics of signals in the frequency range of 1.7 up to 6 GHz have more to do with the building than the frequency." In effect, upgrading to Envara's 5-GHz-only solution means no penalty in range and performance, the company maintains.

Envara has been demonstrating its $35 two-chip set for 5-GHz wireless LANs in Japan. Next comes a dual-mode, 802.11b/802.11a solution. It's due out in late 2002 and will sell for $30 — "$5 less than our 5-GHz-only solution," said Kirshenbaum.

Envara's two-chip dual-band solution, called the Wind, puts the MAC and baseband on one chip and the dual-band RF on the other. The MAC and baseband are based on ARM9 and ARM7 RISC processors, respectively, for the control planes, with the data plane done completely in hardware. A single analog-to-digital converter completes the design. Total power consumption is 500 milliwatts.

"The RF chip design is tricky, as we can't show a lot here due to our patents," said Kirshenbaum. Key innovations, however, include a zero-loss front end and an enhanced-zero-IF downconversion technique. The external power amplifier will be sourced from the general market.

Done in a silicon germanium process, which Kirshenbaum claims is cost-competitive with RF CMOS, the RF chip (including the external power amp) consumes about 1.2 W. "For the power amplifier, we're looking at 0.5 W to deliver 200-mW output radiated power," said Kirshenbaum. Achieving that will give the Wind a total power consumption of under 1.5 W.

"Much of the power and size savings come from the zero-loss front-end design," said Kirshenbaum, "as well as the addition of the 2.45-GHz front end in such a way that we can use most of the 5-GHz blocks. That way we don't have to double the size of the RF chip."

The typical front end to date comprises a transmit-receive switch and an antenna-diversity switch, which together have an insertion loss of up to 3 dB, said Kirshenbaum. "Other people have tried to lower this loss by changing the parameters of the switches, while leaving the switch function a separate component. We've managed to implement the switching functions on-chip to effectively eliminate the 3-dB loss," he said.

Kirshenbaum described the enhanced-zero-IF feature as a technique with all the benefits of direct conversion but none of the downside. "Everyone's going to direct conversion zero IF to eliminate all those off-chip components and filters to simplify the design and lower cost," he said. "But direct conversion has problems with I/Q inverse/quadrature balancing to allow 54- Mbit/s, 64-QAM quadrature amplitude modulation transmission, as well as problems eliminating dc offsets in the signal path."

It also gives poor performance in adjacent-channel rejection, he added, as all the heavy filtering is eliminated.

Envara says that its version avoids all the external components, eliminates all the balancing and offset issues, and meets the .11a spec on IP3 points and rejection.

"It seems like they have a good team with good specialists — and they've gone out to external sources to get the expertise they don't have internally," said Allen Nogee, an analyst with Cahners In-Stat. "Their solution will be very interesting if they can keep it at the prices they say."

Envara will introduce its 5-GHz-only chip set, along with a reference design that supports its dual-mode chip set, in the second quarter. The dual-mode chip set is scheduled to follow in the fourth quarter.

Other companies that have publicly announced their dual-band intentions include Intermec and TDK, both using Atheros' 5-GHz AR5000 chip set. For its part, Intersil said it plans to announce a combo solution before the end of next year.

"Companies have to be very careful in their predictions," said Farpoint Group's Mathias. "It's an extremely competitive market at 5 GHz right now. If there's one slip in the design or in making a batch of chips, you're toast."

'The leader'

For 5-GHz devices, Mathias vouches for Atheros and Intersil, since "Atheros was first, and Intersil is the leader today in WLANs in general," he said.

While Atheros has no immediate plans for a .11b/.11a combination chip, president Rich Redelfs said, "we do have customers who have publicly discussed the building of .11b/a products based on our AR5000 chip set — specifically, Intermec and TDK."