The soon-to-be-approved Draft 2.0 for 802.11n will protect 2.4 GHz legacy networks well, but is tuned for 5 GHz: A large part of the horsetrading involved in the IEEE Task Group N's work between Jan. 2006 and Jan. 2007 centered on making sure that Draft N didn't beat up its older relatives, 802.11a, b, and g. The constraints placed on 2.4 GHz will naturally steer new network deployments into 5 GHz, a better band for video streaming around the home as well.
I spoke with Atheros, Broadcom, and Metalink recently, along with the Wi-Fi Alliance, to look at how the current Draft N will protect legacy networks, and where N will take the consumer markets. (Last month, I wrote lengthily about some of the technical issues that make 5 GHz appealing, but may also constrain it.)
The industry as a whole hasn't been happy with me for much of 2006. That's ok. Writing this site isn't a popularity contest, and to the credit of the many product managers and PR folks, they've always been willing to engage me politely on the points.
But it's a relief to say, at last, that I'm confident that the Draft N gear sold to date using chips from major manufacturers can be fully upgraded to Draft 2.0 via software and firmware patches. I'm finally hearing the kind of full endorsement that I wanted consumers to have before they purchased early 802.11n gear that had a chance of being incompatible at the highest possible speeds with a final standard.
"The changes that were made are all changes that can be accommodated with software driver changes," said Bill McFarland, chief technical officer at Atheros. "All of our products that we've ever shipped since Day One will be upgradable to the final draft 2.0," said Bill Bunch, Broadcom's director of product management for wireless LAN.
Wha's not being said at the moment, of course, is whether the current generation of equipment and that shortly to appear will also be upgradable to the final version of 802.11n. However, I believe bookmakers would take a bet against final upgradability--there will be too much at stake, even more than in this last period, to make that error. The industry prediction is that the path from Draft 2.0 to a final, ratified draft will be long--Oct. 2008 is the projected ratification date--but that few changes, none significant, are likely.
Those who own current generation Draft N gear, or are about to buy it, should know that there's no known timetable for the release of firmware updates that improve interoperability among devices from different makers, fix known glitches, and make changes to support elements of Draft 2.0 that aren't in current devices.
Broadcom and Atheros both noted that while they can have firmware updates completed in the near future, their customers--the device makers--have to integrate and release those upgrades in their own due time. Atheros's McFarland said, "I see no reason why, at least physically speaking, it couldn't be done within two months" on Atheros's end. But, he said, some makers may choose to push out a simple patch, while others integrate it with driver and host software releases. Tim Higgins of SmallNetBuilder spoke to major equipment manufacturers and compiled their responses.
The current draft of 802.11n supports three separate mechanisms to keep Draft N devices from stamping all over legacy 802.11a, b, and g networks through its use of wide channels. The older standards support 20 MHz channels, and can carry a maximum of 54 Mbps of raw data per channel. 802.11n ups that to 65 Mbps, but, with at least two discrete data streams and double wide channels--40 MHz wide--has a raw rate of 300 Mbps. (I don't know how 65 times 4 equals 300, either, folks.)
Now 2.4 GHz already has a lot going on in it: 802.11b and g, of course; Bluetooth; microwave ovens; industrial, medical, and scientific (ISM) stuff; licensed ham radio in the lower part; other licensed purposes in other parts. Well, it's rather crowded in there. (2.4 GHz has about 83.5 MHz in the US spread over 11 staggered channels, where we have 555 MHz available in the 5 GHz band spread over 23 distinct channels.)
It could be argued that 802.11n looks just like two b or g channels. But that's slightly specious because it's not likely that in your immediate area, you would suddenly see every neighbor add two access points at the same time, which would be the effect of 802.11n rolling out wide channels by default in 2.4 GHz. "We're not talking about whether 2.4 GHz will increase density, but what's the rate and what's the consequence when you achieve that limit," said Broadcom's Bunch.
As a result, some equipment makers are choosing to not allow wide channels in 2.4 GHz, Apple foremost among them. Apple explained this in a briefing with me a few weeks ago by noting that their wide support of Bluetooth was incompatible with using wide 802.11n channels, as Bluetooth--a frequency hopping standard that operates across the entire 2.4 GHz range--could see its throughput and even consistency affected.
Atheros's McFarland noted that that company's customers "obviously don't want to disrupt Bluetooth in some severe way" but they're willing to take some advantage of the wide channels, too. "40 MHz accelerates the problem of congestion and interference, but we already had the problem pretty good already," McFarland said. Broadcom's Bunch noted that the IEEE "took a fairly conservative approach wtih the algorithm while still allowing 40 MHz channels to be used," which lets equipment makers and customers choose what works in different circumstances.
For manufacturers who choose to deploy wide channels in 2.4 GHz, there will be three mechanisms that try to mitigate problems with legacy networks; one of these mechanisms won't be needed in 5 GHz, which is another factor in its favor.
Intolerance. We usually look on intolerance as a fault, but it's actually an advantage. The "intolerant bit" in Draft N allows any device on the same network to signal that it wants only 20 MHz channels to be used. That's how Apple will control its environemnt. "This device sending this packet would really like not to have 40 MHz going on around it in the 2.4 GHz band," said McFarland.
Clear-channel assessment (CCA). Every device that wants to transmit over a wide channel has to check that the additional 20 MHz of spectrum it might use isn't in use by a legacy network. McFarland noted that this mechanism is already in place for regular channels--"listen before talk"--and this extends it to the additional channel, allowing 20 MHz transmission if that part is clear.
Politeness. With politeness, there's a very low threshold that disables the use of 40 MHz channels when any legacy network is overlapping in the extended 20 MHz section. "If it's sending any amount of real data traffic at all" only regular channels can be used, McFarland said. Politeness extends for 30 minutes from the last check, and must be continuously monitored.
There's no reason to be very polite, but neither to be intolerant in the 5 GHz band, chipmakers pointed out. The intolerance mechanism isn't required by the standard because there are enough channels to go around that there's no good reason for any given device to tell others to use regular channels.
And politeness goes out the window, too: the mechanism is still active, but it's less likely for those who use the default auto-channel selection option that will be found on perhaps all dual-band Draft N gateways that they will have traffic in the extra 20 MHz used in a wide channel. With 23 channels in play in the US, it's likely that a relatively empty segment should be available even in the heart of a city. "The probability of you being able to aquire a channel that is not being used is obviously significantly higher," said Ron Cates, vice president of marketing at Metalink, a firm specializes in media streaming using Draft N.
Those interviewed expect that dual-band gateways are likely to be much more prevalent than was the case with 802.11a by some orders of magnitude. Because 802.11a didn't hit the market until just before 802.11g shipped, there was no speed advantage to using 5 GHz, and dual-band a/b/g adapters weren't available immediately. Intel delayed its dual-band support for quite some time as the initial Centrino launch included just an 802.11b radio due to timing issues with the standard.
Intel wasn't caught flatfooted this time, and both Apple and Intel announced their general support for 802.11n within a week: Apple already shipping adapters with the N part disabled until they were secure in the standard being stable; Intel committing to dual-band a/b/g/n in their Centrino platform, with the first models due later this month.
Broadcom's Bunch noted about 5 GHz, "It does point the way to dual band products being more available than they have been. The basic issue you had since the beginning is that there's almost always a price premium, particularly on the infrastructure. On the notebook side, the price premium between a single band and a dual band has been quite low for some time." Bunch said that all PC makers that ship Broadcom Draft N adapters include dual-band N, which makes adoption of 5 GHz much simpler.
Ultimately, the desire to stream video will make 5 GHz dominant for home media adapters, just as Apple has signaled by including both 2.4 GHz and 5 GHz support in its Apple TV's Draft N adapter. Cates of Metalink noted that Internet Protocol Television (IPTV), the way in which providers are bringing TV to the home through fiber optic and other pipes, requires either 20 Mbps for older MPEG2 compression or 10 Mbps for newer MPEG4 compression to deliver full high-definition quality. That just won't fly in 2.4 GHz. The limitations of 2.4 GHz "compel us time and time again to tell our customers to configure" using 5 GHz if it's video they want to use, said Cates.
One small problem in 5 GHz appeared with Apple's initial release of the revised AirPort Extreme Base Station. While the gateway and the adapters that are built into Macs support 5 GHz, they only allow the use of eight of the 23 channels: the lowest four and the highest four. The reason? Apple won't state this, but it's because the middle 15 require the use of two protocols for broadcasting at the lowest necessary power level coupled with detecting and avoiding broadcasts on frequencies used by active military radar installations. That avoidance was part of a grand compromise in the US that opened 255 MHz of additional 5 GHz bandwidth.
McFarland noted that Atheros has been producing silicon that conforms to these standards for some time; it's the earliest firm that committed to 802.11a back when there were doubts that 802.11a could even be built using the least expensive chipmaking process, CMOS. He said, without specifically commenting on Apple's choice or plans, that adding radar avoidance and dynamic power control is as simple as a firmware upgrade for any modern Wi-Fi chipset.
Cates of Metalink said that, "If you're detecting the presence of signals there, you want to avoid those channels. That's the beauty of the 5 GHz band -- there's so many to choose from." Cates said that the radar in question in the US tends to be fixed and broadcasting continuously, making it simple to avoid. In Europe and Asia, Cates said, there's more use of bands also employed by civilians, making it perhaps more of an issue there. Cates said future mechanisms would allow gateways to signal clients to change channels, which would obviate increased density of both military and unlicensed 5 GHz use.
Frank Hanzlik, executive director of the Wi-Fi Alliance, said that the group is well on track to deliver its first announcements of certified Draft N products by June, which tracks their goal for this program when announced last year.
In terms of the particulars of what will be tested, that's still up in the air. "We don't have an approved test plan yet; we're still in that process," Hanzlik said, although the group will clearly be looking at legacy and Draft N interaction, and devices' conformance with all the politeness and avoidance behaviors that are in the draft.
Beyond certification comes the emergence of faster Draft N devices that deploy more data streams. The first generation of gear can generate two simultaneous data streams that are directed by two or more transmitting antennas to occupy unique physical paths--so-called spatial multiplexing. The spec requires at least two streams, but can support up to four. The predictions are that those additional streams come slowly, with no near-term availability as the cost structure and demand isn't in place.
And, finally, comes ratification, now targeted for Oct. 2008. At which point, another, faster standard will likely already be under discussion and the subject of much debate.