The greed for speed

Evolving 802.11: When faster isn't necessarily better

By Maury Wright, Editor in Chief -- EDN, 2/19/2004

Go down to your local electronics superstore, and you can witness what could be labeled the Starburst Wars. Companies that sell 802.11 WLAN (wireless-LAN) products believe that speed sells, and they all want to feature a big maximum-bandwidth number on their packaging—typically in a prominent, brightly colored starburst.

In reality, chip makers and system vendors could better serve consumers by worrying more about range. But speed is sexier. As a result, both designers and consumers these days need to navigate carefully. The next revision of the 802.11 standard is probably two years off, but the players at the chip level are pushing proprietary ways to boost data rates in current products. Moreover, all will be trying to push their ideas into the next standard: 802.11n.

The presidential campaign has nothing on this race, but let's try and sort through the rhetoric, decide exactly what the vendors have to offer, and figure out how it matters.

The speed race is taking place despite the fact that it's hard to find applications that justify even the 54-Mbps data rates provided by both the 2.4-GHz 802.11g and 5-GHz 802.11a standards (see sidebar "Speed or range?"). On the other hand, much faster rates could be used to carry video around a home.

Discerning the chip story isn't a necessity only for system designers. Smart consumers must sort through the underlying technologies and figure out which products contain which chips if they want to take advantage of the proprietary features; most end-product vendors are shipping devices based on chips from multiple vendors.

However, sorting the 802.11 participants at the chip level is a trying task. The companies all approach the market differently in terms of target applications, the proprietary technologies that they're using to extend baseline performance, or both. In addition, at least a half-dozen vendors claim to be among the top two or three players as measured by chip shipments. Meanwhile, the market is evolving in both the frequency band and the 802.11 flavor that's accounting for the volume shipments.

The companies that appear to be shipping the greatest volume of chips of various flavors include Texas Instruments, Atheros, Broadcom, and GlobespanVirata (formerly, Intersil and soon to be Conexant).

Looking back, GlobespanVirata has been the clear leader in 802.11b volume. It's amazing that Intersil was in a hurry to sell off the market-leading WLAN unit to GlobespanVirata, which now, in turn, will sell it to Conexant. Perhaps the unit will find a symbiotic home among Conexant's other technologies. Unless obstacles appear, the sale should close by the end of this month.

In 2003, Texas Instruments also held a dominant position in 802.11b chips, although many sold under the Intel banner. Intel's Centrino wireless technology relies on ICs from TI and others because Intel's WLAN efforts have yet to yield an 802.11 product in production quantities.

Broadcom, meanwhile, got to market first with 802.11g chips and likely shipped the greatest number of chips for that flavor in 2003. Indeed, the company angered some by selling a forward-looking 802.11g implementation before the IEEE had ratified the standard.

Atheros pioneered 802.11a technology and was the early leader there. The company also got to market first with an integrated 802.11a/g offering and is likely the leader in installed base of dual-band chips.

Other WLAN players include Agere, Philips, and Marvell. Start-ups such as IceFyre and Airgo Networks are trying to garner market share. And Intel presumably will ship its own chips as well.

In retrospect, Broadcom's gambit of launching a prestandard 54-Mbps product in the 2.4-GHz band probably begat the speed-rating battle going on today—although Atheros had long promoted a turbo mode in its 802.11a products. Vendors that sold Broadcom's early 802.11g products found success by prominently featuring the number "54" on the packaging. So today, several players are engaged in pushing the number on the box even higher. The battle will be waged at claimed speeds above 100 Mbps.

Today, Atheros is touting its Super G technology as capable of 108-Mbps rates, and products with the technology have been available for some time. For instance, witness the package for D-Link's DI-624 router, with its prominent claim of 15-times faster operation and 108-Mbps maximum data rates (Figure 1). Competitors such as Linksys and Netgear use similar approaches on their packaging.

As for Super G technology, Atheros recently found itself at the center of controversy. Among other enhancements, Super G employs channel bonding—using two channels simultaneously to send data—for some of its performance gains. Detractors, including Broadcom, claim that Super G is a bad neighbor and severely degrades the performance of nearby WLANs. Atheros claims that its implementation dynamically adjusts to not interfere. I haven't had the chance to test these claims, but you can find online reviews that support both arguments in different test cases at www.smallnetbuilder.com.

Taking a less controversial path, GlobespanVirata has chosen to rely on a combination of increased efficiency in the MAC (media-access controller), the ability to burst data packets, and data compression in its Nitro XM enhancement to the Prism family.

The effectiveness of compression always depends on the data being transferred, of course. Compression doesn't yield much improvement if a user transfers zipped data files or even music compressed with a lossy algorithm, such as MP3. However, MAC optimizations, such as packet bursts, can reduce latencies caused by media contention and deliver improvements, especially in big file transfers. GlobespanVirata claims its 802.11g extensions will enable best-case 140-Mbps rates.

The company has added another twist to its newest product with a feature called Direct Link. The technology allows two enhanced clients to directly transfer data—without Client A's first transmitting to the router and then the router's transmitting to Client B. The system must set up the transfer via normal media arbitration with the router, but the actual transfer then happens at essentially double the rate because the data must traverse only one span.

For its part, Broadcom will shortly offer its own 100-Mbps or faster offering, reportedly to be called Afterburner.

Naturally, all of the vendors extending the 802.11 standard claim that their offerings are backward-compatible with 802.11b/g. And, of course, users need to have matching enhanced products to realize the performance gains.

For now, the other 802.11-IC vendors are concentrating on other features or application targets to win market share. TI has already had success in PDA and cell-phone designs, and the company's dominant position in cell phones should position it for more success in that arena with 802.11. The company also has strong market positions in portable music players and digital cameras—both of which are ripe for 802.11. TI has the flexibility to produce dedicated 802.11 chips or integrate 802.11 into its OMAP (Open Multimedia Applications Platform), which is featured in many DSP-enabled products.

Agere also has a significant cell-phone business that it may leverage to sell 802.11 into the phone market. However, Broadcom is ahead of TI and Agere with the AirForce One true single-chip-802.11b implementation (Reference 1). The market for handheld systems is likely to remain an 802.11b market for a while, although the market for desktop systems is headed quickly to 802.11g. Following that trend, Atheros announced a single-chip 802.11g implementation just as this article went to press.

Agere, Texas Instruments, and Marvell are all pursuing integrated 802.11a/g implementations. TI has a unique approach to dual-band designs that may minimize cost and push such implementations into the mainstream. The company has developed the wOne software package for routers, which allows a single MAC and a single multiband radio to simultaneously support both 802.11a and 802.11g.

The technology essentially time-slices between the bands, handling traffic for one band in one instant and for the other band in the next. A router equipped with wOne might not perform to the level of one equipped with dedicated chips for each band, but the technology allows a router design to simultaneously support both types of traffic and costs the consumer just a $20 premium above a single-band router. Today, 802.11a/g routers sell for more than double the price of single-band routers, and most go for more than $200.

Meanwhile, Agere is touting a range extension for its newest multimode WaveLAN chip. Like most of the other players, the company doesn't specify exactly what techniques will deliver the extension, other than optimized receiver sensitivity and transmitter power. Atheros is also talking about range extension but offering little explanation of how its technology will work.

I believe that word of mouth will ensure rapid success for any product that actually delivers a significant range advantage. Home users in particular are far more likely to return a product due to range problems than to data-rate problems.

Lately, some newcomers have arrived on the scene with new approaches to the range dilemma. IceFyre, for instance, claims that its Class F switch-mode power amplifier delivers greater range and reduces power consumption. The company also offers matching MAC and PHY (physical-layer) chips, but it may find more success with its amplifier given the crowded market.

Start-up Motia plans to be the first to bring smart-antenna technology to 802.11b/g. The company claims that its adaptive-array antenna technology can double or even quadruple effective range. If only the router contains such an antenna, clients will get a doubling or a tripling of range, according to Motia. If both the router and the client possess the technology, range will quadruple.

Motia plans to offer the technology as an appliqué—a function that requires no change to the underlying MAC or transceiver technology. In fact, products that have an accessible antenna port could conceivably accept a Motia retrofit. Nevertheless, the company will also work on winning spots in the all-important reference designs from the major chip makers.

Motia claims that, in volume, its Javelin chip will add less than $10 to the cost of a product. The chip is unique in that it does not use DSP technology but rather works at 2.4 GHz in the analog domain to combine the input from four antenna elements and produce an optimized signal. The company hopes to see Javelin-based products on shelves after midyear.

While all of this activity in 802.11 products takes place, the IEEE working group for the 802.11n standard is defining criteria for the next standard, which will seek to formally extend data rates beyond 100 Mbps and stay backward-compatible with all 802.11 gear. The group will start considering technology proposals in the second quarter, and most IC vendors expect to see new products around mid-2006.

It's now difficult to say what improvements might make it into 802.11n. The standard might incorporate some of the MAC-enhancement and -compression techniques now appearing in extended 802.11g products. Some parties predict a change in modulation. Others argue for channel bonding, although the prevailing opinion is that bonding should be an option only after an upgrade within the channel limits. Almost everyone agrees on one thing, however: The standard will incorporate some form of MIMO (multiple-input multiple-output) technology.

MIMO is simple to explain at a high level but almost impossible to fathom for those of us familiar with Shannon's theories on how much data can pass through a channel. MIMO relies on multiple antennas on the transmitting and receiving ends of the wireless link to essentially simultaneously pass multiple channels of data in the same frequency band, using the same modulation techniques. The technique depends on the spatial diversity of the antennas to somehow discern and reassemble multiple streams from multiple antennas on the receiving end.

Ideally, MIMO theory states, you can achieve a direct multiplier of channel capacity simply by adding antennas—without using any additional spectrum. In reality, the math involved limits near-term implementations to two or three antennas that can double or triple the available bandwidth.

Today, Airgo Networks is driving MIMO technology. The company has prototype MAC chips and transceivers that can move real channel rates—not throughput achieved by compression and other tricks—to 108 Mbps (Figure 2). The company claims that network vendors will this summer ship products equipped with its MIMO technology and that prices may be just 50% higher than those for 802.11 gear.

Furthermore, Airgo argues that MIMO solves range issues as well, because it breaks a single data stream into multiple slower streams for parallel transfer. The slower rate on each link means that successful transfers work in the face of higher noise and thus can extend over longer distances.

I began research on this article ready to declare that WLANs were going to fail in the quest to move HDTV streams around a home. And I'm still not ready to accept that possibility as a done deal. But antenna improvements and schemes such as MIMO have convinced me not to count Wi-Fi out just yet.

Author Information

You can reach Editor in Chief Maury Wright at 1-858-748-6785, fax 1-858-679-1861, e-mail mgwright@edn.com.