Atheros and Broadcom Both Guilty of Wireless Problems

Atheros and Broadcom Both Guilty of Wireless Problems

Jim Louderback

Those who live in glass houses should not throw stones.

—Geoffrey Chaucer, “Troilus and Criseyde”

A year ago, I thought 802.11g was a dumb idea, but I’ve changed my tune. With a ratified standard, and 802.11b interoperability, I’ve become a big fan of the 54Mbps wireless networking scheme. Even Cisco and Microsoft have released 802.11g products, so you know it’s here for good.

But this week brought allegations by 802.11 chip maker Broadcom that chips made by rival Atheros can be so noisy—in a proprietary 108Mbit mode—that they block other 2.4GHz wireless devices, including phones, Bluetooth mice and keyboards, and other Wi-Fi devices.

But Broadcom’s not innocent here. The two feuding chip makers—and the third big wireless chipmaker, Intersil—are all guilty of putting marketecture ahead of customers. And that’s bad news for everyone.

Let’s start with Broadcom’s allegations, which they laid out in a private meeting room at this year’s Comdex. The Atheros 802.11g chipset—used primarily by D-Link and NetGear—include the ability to ratchet up the 54Mbps speeds to (they claim) 108Mbps. This “Turbo” mode (called Super G by D-Link) only works when both the access point and the mobile device are running Atheros-based wireless network cards, and when both are set to run at “Super G” mode using configuration software.

Super G gets a significant portion of its performance gain by channel bonding—multiplexing wireless traffic over two channels, instead of one. That doesn’t sound like a big deal, because there are 11 selectable channels for 802.11g Wi-Fi, but that number is deceiving. In fact, there are only three non-overlapping channels, 1, 6 and 11.

There wouldn’t be much of a problem if the channel bonding simply took over channels 1 and 6, and left 11 alone. But the Atheros implementation centers the two channels in the middle of channel 6, which means part of channel one, and part of channel 11 are used up by the channel bonding scheme.

Next page: An in-depth look at Broadcom’s test setup and results.

According to Broadcom, when an access point and client are communicating in SuperG mode, the wireless traffic drowns out or interferes with any other 2.4GHz wireless traffic, including other Wi-Fi networks, Bluetooth mice and keyboards, wireless phones, baby monitors and so on.

To prove the point, Broadcom streamed a DivX version of “The Matrix,” encoded at 2Mbps, between a Broadcom-based, 802.11g enabled PC and Gateway’s Connected DVD Player. The two devices were set to communicate on Channel 1.

Using an oscilloscope, I could plainly see the traffic occurring in the lower third of the 2.4GHz band.

Broadcom then started blasting as much traffic as possible between an Atheros-based hub and notebook, configured for Super G, sitting right next to the Broadcom-based DVD Player and access point.

After a few seconds, the movie pixelated and stopped playing. On the oscilloscope, it was easy to see that the SuperG devices were taking up quite a bit of the 2.4GHz band—enough to bleed into the areas reserved for channels 1 and 11.

“An 802.11g jammer”, said David Cohen, senior product marketing manager at Broadcom, referring to the Atheros products. Although all the wireless products were sitting within a foot of each other, Broadcom’s tests show a similar effect at up to 30 feet away—which could spell trouble for anyone living next door to unsuspecting D-Link and Netgear customers opting for the fastest speeds.

And that’s a problem, according to Cohen, who sees potential for big problems. He wants a warning label, at least, slapped on D-Link and Netgear’s boxes. “Consumers should understand that when you buy a technology, it interferes with other wireless products.”

“I think that’s something we should seriously consider,” responded Bradley Morse, senior vice president for Marketing for D-Link, “but we have to do more thorough testing.”

Next page: What D-Link found in its tests.

Morse and his team, along with engineers from Atheros, have been frantically testing many different configurations since the allegations came out late last week. And he claims that Broadcom’s test is fundamentally flawed.

The company replicated Broadcom’s test, but by putting two totally separate Broadcom-based wireless networks side by side, running in standard G mode. The first network was communicating on Channel 1, the second on Channel 11—as far apart as you can get.

The result: Both networks saw significant performance degradation—as low as 1.2Mbps—simply because they were right next to each other. TurboG or not, claims Morse, if you put two wireless networks next to each other, “the electromagnetic fields in the fabric of space are just saying, ‘Wow, too much.’ ”

Or to translate his LA-speak, stick any two wireless nets next to each other, and the close proximity will slam the performance of both. Remember that Broadcom showed DivX movies encoded at 2Mbps failing—which would have been the result regardless of what 802.11g chips or modes had been enabled, according to D-Link.

“That is a ridiculous statement,” counters Broadcom’s Cohen, rejecting D-Link’s test results as impossible to achieve under the 802.11g spec.

Morse also pointed out that Atheros based products achieve their performance boosts through three additional schemes apart from channel bonding:

Packet Bursting: Sending more packets during a particular time than the 802.11g spec allows, by reducing the wait time between packets.

Compression: Dynamically compressing packets before transmission—which won’t add much benefit for already compressed music and video files.

Fast Frames: Squeezing more data into each frame

According to D-Link, channel bonding can be separately turned on or off by the user. And even when channel bonding has been selected, it can be implemented as a solution of last resort.

In “Dynamic Turbo” mode, channel bonding stays off until the network needs more performance than delivered by the other three schemes — although actual implementations differ in how they implement this dynamic mode.

Next page: Why Jim thinks all wireless vendors are guilty.

But when it comes to proprietary and non-standard 802.11g features, Atheros is not alone. Broadcom’s products include a feature called Xpress that uses packet bursting to increase performance beyond the standard 54 Mbps. Intersil promotes a similar technology called PRISM Nitro.

None of these features are part of the 802.11g standard—although packet bursting is part of a proposed extension called wme.

So is this a tempest in a teapot? Certainly, if you put two 802.11g networks within a few feet of each other, you’re asking for trouble. When it comes to the Atheros channel bonding, I’d recommend that most users disable it—unless they have more than 100 feet separating their house from any others. D-Link and Netgear should clearly warn their customers of channel bonding’s potentially destructive impact.

At work, I’d stick with the standard 802.11g implementations, to guarantee interoperability with all 802.11b and 802.11g devices.

But there’s a broader issue here. Wi-Fi networking has been so successful because it’s so consistent. For users, the guarantee that any two 802.11b (or 802.11g) devices will work together makes it much easier to invest in the products.

Compare the universal interoperability of Wi-Fi with Bluetooth—which has suffered through two years of incompatible products. Wi-Fi’s a raging success, and Bluetooth is still struggling for relevance.

There is absolutely no place in the market for these proprietary add-on features—whether they’re called Nitro, Xpress, Turbo or Super. Granted, the others aren’t nearly as over the top as channel bonding, but by insinuating to the consumer that there are differences between the products, an implied lack of interoperability could scare many away.

I also expect confusion over the claims made by Broadcom, which could also cause potential customers to doubt the reliability of Wi-Fi networks.

So let’s drop all these performance-enhancing technologies until they become real standards. Because those standards are Wi-Fi networking has been the brightest spot in today’s technology market—lifting the fortunes of Atheros; Intersil; and Broadcom, too.

Discuss This in the eWEEK Forum

eWEEK.com Wireless Center Columnist Jim Louderback is editor in chief of Ziff Davis Internet.

Copyright © 2003 Ziff Davis Media Inc. All Rights Reserved. Originally appearing in eWEEK.