(Updated Wed. morning) A game is being played with announcements from Task Group N: Yes, there is unity. Of a sort. In an interview last week, Atheros's chief technology officer said the Draft 1.0 accepted by Task Group (which will eventually produce the 802.11n next-generation standard to move Wi-Fi forward) was essentially complete with small details to work out. He said there was a very small risk that major changes would be required. Bill Bunch, director of product marketing at Broadcom, confirmed that view in an interview Wednesday morning.
Atheros, Broadcom, and Marvell all had silicon to show in some form at the Consumer Electronics Show (CES) in Las Vegas in early January; Intel, I didn't hear about, although they've announced both 802.11n and WiMax support for laptop designs later this year. Commitment to silicon, even in small quantities, is expensive, and changes in certain aspects of the radio's functionality would almost certainly require "retaping" circuit designs--making changes to produce new chips that have to go through significant testing. (Paul Callahan has details on where Atheros, Broadcom, and Marvell are at with chip designs, reporting that Marvell's chipset is nowhere near ready.)
Airgo's chief Greg Raleigh said in an interview on Tuesday, however, that "There's been an unprecedented effort to manipulate and monopolize the standards process." Raleigh maintains there are a few key issues for performance and backwards compatibility that are simple and won't change the cost for chip or device production. "It literally takes a couple of weeks" to make these changes in the spec, he said. Several chip competitors that are trying "to prevent any improvements" because they're "struggling to ship chips based on immature versions of the standards."
Atheros and Broadcom said, in contrast, that they preplanned for some changes in a critical backwards compatibility area. "It strikes me that no chipmaker is going to build a chip without thinking about these issues," said Bunch. "We have built a fundamental toolkit into our silicon so we can go in a number of different directions. We think we have all the flexibility we need to solve the generic problem."
Raleigh said Airgo hasn't committed to silicon yet because they believe a Draft 2.0 or 3.0 will be required before they are comfortable putting their fabless--that's chip-fabrication-plant-less--efforts into what they're calling their Gen N chip, which will succeed three generations of chips they've sold into the marketplace through Belkin, NetGear, and others. Cisco and Motorola are chairing an ad hoc group in Task Group N to put together a proposal for some changes to Draft 1.0, Raleigh said. Thousands of comments from the ballot that will go out are also expected, one source said.
The most significant changes have to do with the optional double-wide channel mode that 802.11n will use as one component in improving Wi-Fi's net speed. 802.11n will use MIMO (multiple-in, multiple-out) antenna technology for range and multiple simultaneous data streams in the air, but the 40 megahertz (MHz) wide channel has a neat doubling effect without fancy footwork. Existing 802.11a, b, and g channels are 22 MHz wide; double that, and you roughly double raw bandwidth at one swoop. Some companies won't implement 40 MHz channels in some or all of their chips for reasons of cost or power use--think cell phones that won't process more than several Mbps, anyway--and 40 MHz-wide channels aren't yet permitted in certain countries.
The problem with 40 MHz comes from 802.11b/g having just three nonoverlapping 22 MHz channels in the 2.4 gigahertz (GHz) band. These channels, numbered 1, 6, and 11, do, in fact, overlap, but somewhat faintly at signal strengths found in consumer devices. In most urban areas, there's already so much Wi-Fi in use, that every channel has someone "talking" on it, whether one of these three or 2, 3, 4, 5, 7, 8, 9, or 10--and 12, 13, or 14 in some countries. (802.11a in the 5 GHz band has 23 channels now, though some are designated for particular purposes. These channels are effectively non-overlapping: with so many and with channel spacing, 802.11n operating in 5 GHz won't cause the same hardships that N in 2.4 GHz might.)
For home users, therefore, it's important that the existing scarce 2.4 GHz spectrum not have even more demands placed upon it, which is why where a 40 MHz channel sits (by default or by choice) becomes a critial part of 802.11n. Task Group N still needs to decide whether it will decide where to place 40 MHz in the 2.4 GHz band and how to adapt to co-channel and adjacent channel networks, new and old.
Placement means that, unlike Atheros's widely disdained Turbo mode in their Super G extensions to Wi-Fi, you don't stick two channels in the middle of the band, which is one possibility on the table. Turbo grabbed channels 5 and 6--using much less than 40 MHz--but the technique caused performance degradation across the whole band. (That's why Turbo in shipping devices is either turned off or set to "dynamic"--disabled when it detects other networks--since about 18 months ago.) Task Group N might decide to not dictate how to specify how 40 MHz channel interact with legacy networks, too, leaving it up to each manufacturer. "That's up to the committee to decide whether they want to specify the gory details of how one manages the spectrum," said Broadcom's Bunch.
Airgo knows quite a bit about the 40 MHz interference problem because they've had a product on the market that uses 40 MHz--NetGear uses Airgo's third-generation chips in its RangeMax 240, which has a raw data rate of 240 Mbps in its most expansive mode (two streams, 40 MHz channel). The device uses a Wi-Fi channel and then extends out to one side (on either end of the band) or both sides (on a middle channel). Signals look like Wi-Fi on the same channel, making co-channel interference less of an issue. But that's not the case for networks on adjacent channels.
Tim Higgins of Tom's Networking back in January showed that the firmware available at that time caused significant problems in backing down to 22 MHz when nearby, separate, existing networks are detected. Airgo never publicly critiqued Higgins's review, but sent around some of the issues with both broad statements and methodology. The review has been updated three times to address methodology, particularly in measuring consistency in downlink speed. None of those changes affect the adjacent network interference.
In email to me at the time, Airgo justified a number of decisions on how they dealt with existing nearby networks, but also said that many of those decisions would be changed in firmware upgrades. For instance, in January the device could only automatically choose a new channel at startup; new firmware would allow it to switch channels while operating if other networks started up, although it would monitor these new networks for five or more minutes before switching to avoid causing unnecessary network disruption. (No new firmware is listed as of today at NetGear's support page for the model.)
I turned for a sanity check to a colleague who attends Task Group N but doesn't have a horse in the race: this person works neither for a chipmaker nor a consumer equipment manufacturer, but must remain anonymous because of their day job. This colleague stated that there certainly was an unresolved issue about playing nice, but my colleague views this as one of a number of minor issues. Since he or she isn't involved in making silicon his assessment and Raleigh's aren't incompatible: if there's no cost considered with an existing silicon commitment to earlier drafts, the changes could be very minor in wording and overall effect. But there's no particular way to know at this point whether Airgo's competitors have included the flexibility necessary in their chip designs to address whatever might be required for 40 MHz adaptation to legacy networks.
Now, of course, Airgo has the most to lose from commoditization of their key advantage: high-range, high-throughput backwards-compatible Wi-Fi. Their third-generation chips, already on the market in some products, can achieve 100 Mbps of real throughput out of a raw rate of 240 Mbps. (Atheros's CTO said that 802.11n products should hit as high as 75 percent throughput of their raw data rate, or as much as 450 Mbps out of 600 Mbps, but more likely as high as 150 Mbps out of the base 200 Mbps version.)
There's nothing to touch Airgo's throughput at this point, and they need to have chips ready for market at essentially the same time as competing chipmakers. Likewise, competing chipmakers have every incentive to be faster off the mark than Airgo, even if there's risk involved, because any contract they can sign with a consumer manufacturer to get products to the market a month or two or three faster is gold for that product maker--if the product works as advertised and can be fully firmware upgradable to final standards. I don't think there's a risk that we'll see devices on the market that use draft 802.11n and can't be upgraded; that would be far too embarassing to manufacturers, and destroy consumer confidence. Broadcom's Bunch said, "What we are saying is our products are fundamentally designed from day one to track the standard. The reason we're not guaranteeing is there is some small possibility that there is something significant found that we can't track."
Now, if you remember 802.11g, Broadcom managed to score an enormous coup in late 2002 by signing Apple, Buffalo, and Linksys as customers. They captured almost the entire 802.11g chip market between January and June 2003 because there simply was almost no other silicon available. There were huge problems with the early devices (but not the underlying chips) because the standard was in flux: it didn't get finalized until summer, and I saw enormous interoperability problems between Apple and Linksys equipment even though they were practically the same reference design. Despite the several firmware upgrades required to reach stable interoperable status in summer, consumers apparently liked the higher speeds enough.
However, an important point is that the first silicon Broadcom shipped in December 2002 was firmware upgradable to the final 802.11g standard, a standard that Broadcom helped shape but was in no position to dictate. Bunch noted, "Every single 11g chipset that we produced based on earlier drafts were all upgradable to the exact final standard after ratification." This helped anoint Broadcom in the Wi-Fi space, and was another push for Atheros to leave its 802.11a-only roots and become a fierce competitor through feature differentiation in the 802.11g market.
I suspected that Airgo might have another reason to want a different 40 MHz solution to win out: their third-generation chips might be compatible enough with the 802.11n offering to allow them interoperability, even if that cost some performance. Right now, all MIMO devices in the market promise no better throughput than 802.11g-level performance with whatever 802.11n products ship. Higher performance levels require compatible, proprietary equipment.
However, Raleigh said something I found surprising: He said the his 3rd gen chips have a solution "that's frankly not quite as good as what's on the table with 802.11n." He said, "The solutions that are on the table for 11n are superior to that," referring to what the ad hoc group is looking at to suggest for the next Task Group N draft. Raleigh said he wants silicon-based solutions that would far exceed what could be done in firmware to preserve the quality of legacy 802.11 networks. (Some elements of Airgo's previous generations will probably have forward compatibility with 802.11n, I should note, but the extent will certainly depend on the final draft and what Airgo can offer through firmware for older devices.)
If this legacy interference issue doesn't delay the spec, then Raleigh and company may wind up slower off the mark than their larger competitors. Even if does, Airgo as a smaller firm may not be able to produce designs and chips as rapidly as their larger competitors retool for any changes. Pushing through a particular approach to spectral management would give them a position of technology leadership, of course, and one can't underestimate their fleetness despite their size: Raleigh says their 2 1/2 years of building MIMO chips will let them have the best "product-to-price performance."
Bluntly, Airgo is saying that its competitors would risk destroying the performance or utility of existing 802.11b/g in order to capture part of Airgo's current proprietary, higher-margin business. I asked Raleigh if that characterization were correct, and he agreed with it. But he didn't think the plan to bake Draft 1.0 with minor changes would work. "I don't these guys are going to have the political power to clog the process at this point because the changes are highly visible and quite small," Raleigh said.
Broadcom's Bunch said in response, "The implication that we would damage our 11g business just to take his small share is very strange." He noted, "We cannot have the bad neighbor policy...It's a very dense world out there." Bunch strongly rejected any suggestion that Broadcom's spectral management approach to interference wouldn't fully consider existing networks' integrity.
Where the standard may wind up in the short-term is only slightly hazy. Draft 2.0 will certainly appear after comments are incorporated into Draft 1.0; that draft might be voted on in May. Even Raleigh thinks that a Draft 2.0 or 3.0 might be complete enough to base silicon on. Similar standards have taken much longer and gone through many more drafts; 802.11g ended up at Draft 8.2 heading for ratification.
What's certain is that Airgo has put its cards on the table: Their third-generation chips lack characteristics that they want built into 802.11n, and they believe their four prime competitors have lacked the savvy to build those characteristics into their proposed chips. It may be that Airgo gets its way, and specific spectral management behavior is added to 802.11n. But that might not benefit them: that behavior might be something that Atheros, Broadcom, Intel, and Marvell have already counted on, and which they might bring to market months ahead of Airgo.