At the January IEEE 802.11 interim meeting in Vancouver, British Columbia, Jan. 12-15, Task Group "n" is expected to issue a call for proposals for the high-throughput amendment to the 802.11 standard. The call would enable proposals to be presented, and the selection process to begin, in May.

The amendment calls for rates of at least 100 Mbits/second, as measured at the interface between the 802.11 media-access control (MAC) and higher layers. Measuring the performance of a wireless LAN this way, rather than in terms of the rate at the physical layer (PHY), is a departure from previous practice within IEEE 802.11. The motivation is that the net data rate experienced by the user in wireless LANs is significantly affected by the many sources of overhead within the 802.11 protocol. The overhead comes primarily from packet preambles, acknowledgements, contention windows and various interframe-spacing parameters. The over-head problem becomes more acute as the raw PHY data rate increases:

The data-carrying portion of packets shrinks, while overhead remains fixed.

For example, the 802.11b standard, with a peak physical-layer rate of 11 Mbits/s, typically delivers a net peak 5 to 6 Mbits/s, while the 802.11a and 802.11g standards, with a peak PHY data rate of 54 Mbits/s, deliver a net peak of around 20 to 24 Mbits/s. The high-throughput extension thus demands, at the very least, a four- to five-times improvement over the maximum achievable with 802.11a/g devices.

Scope of amendment
To achieve the throughput goals, the scope of the allowed changes includes-and indeed, seems to require-both MAC and PHY changes. This defines a minimum scope. But what is the maximum? Could there be an all-new MAC or PHY? And if not, how precisely should the relevant requirement be expressed?

These and related questions dominated discussion during the study group phase. The overwhelming sense of the task group was that the high-throughput protocol should be an extension of the existing 802.11 framework, building on the painstaking progress achieved in multiple amendments over the past several years.

This is expressed in the task group's charter by requiring that 802.11n be written relative to a baseline specification incorporating 802.11 and all of its amendments, both completed and in progress. Thus, the allowed changes will be enhancements: No existing functionality will be subtracted. In addition, only those existing mechanisms that pertain in some way to higher throughput will be altered. Other mechanisms, such as the security framework built up in 802.11i, are out of scope.

Backward compatibility
Backward compatibility with existing 802.11a/g deployments was identified as a must-have requirement in the study group phase. All .11n devices will therefore be backward-compatible, at least with 802.11a/g.

This straightforward statement does not fully explain the requirement, however. There is now an extensive body of experience within the IEEE working group on what backward compatibility means in the 802.11 context. Most recently, the example of IEEE 802.11g demonstrated that mixed-mode operation with legacy .11b devices will result in at least some performance degradation when compared with a green-field deployment consisting of all-.11g devices. Although all devices compliant with the 802.11g specification must be capable of mixed-mode operation, it is perfectly acceptable for the user to set up a network in which legacy devices are not permitted. This may be accomplished, for example, by configuring the access point to deny association to any device that does not contain the new rates that 802.11g adds.

This may seem counterintuitive: The standard mandates that devices be capable of communicating with legacy devices but not that they actually use this capability. But in fact, this corresponds to established practice in 802.11 deployments. The 802.11b networks, for example, often prohibit use of the 1- and 2- Mbit/s modes from the original 802.11 standard, thus excluding legacy devices.

The alternative is to allow a single device at great distance from the access point to dominate the medium and degrade performance for every other device on the network.

The task group's mandate is therefore that some of the modes of operation will be backward-compatible and interoperable with 802.11a/g. This allows for other modes that are optimized for green-field operation and for eventual industry migration to all-.11n networks, without carrying a permanent performance penalty on behalf of legacy devices long after they have disappeared from general use.

Marketing requirements
The end-user focus of 802.11n is also manifested in the development by the Wi-Fi Alliance of marketing requirements for high-throughput devices. The task group has incorporated the resulting requirements into functional requirements and comparison criteria for use in proposal evaluation. This is the first time the Wi-Fi Alliance has taken such a proactive role in the standards-development process. The requirements document is also intended to provide a basis for the eventual creation of Wi-Fi Alliance interoperability test plans.

In addition to the main goal of achieving an increase in actual throughput as measured by the end user, the requirements adopted by the Wi-Fi Alliance include several related requirements. For example, .11n devices should have improved range at existing through-puts, increased resistance to interference and more-uniform coverage within a network.

The Wi-Fi Alliance classified the three major markets as home, enterprise and hotspot. An essential requirement is that the same solution be capable of functioning in all three environments. This enhances the usefulness of the devices and also facilitates user uptake.

The task group's charter includes an ambitious set of high-performance requirements. Within the IEEE 802 process, evidence of technical feasibility is required.

A diverse set of approaches was suggested in the study group phase, including multiple antennas, enhanced modulation and coding schemes, expanded bandwidths and closed-loop methods, with PHY data rates of up to 250 Mbits/s suggested as possible by combining various approaches. Together with improved MAC efficiencies, this allows for net throughputs well in excess of the 100-Mbit/s minimum.

Beyond those suggestions, no preferred directions have been established. The focus has been on enabling solutions that are optimized at the system level.

A significant technical requirement is aimed at providing higher spectral efficiency than existing 802.11a/g systems. This precludes approaches that achieve higher throughputs by merely using more spectrum, such as channel-bonding methods. These approaches do not increase overall network capacity and are an inefficient use of scarce spectral resources.

With the .11g standard, the 802.11 Working Group has achieved a convergence between 2.4-GHz and 5-GHz technology, and it is likely that .11n will be applicable in both bands, as well as new spectrum that opens up. The only explicit requirement in this regard is that .11n devices should be capable of 5-GHz operation.

A final requirement, added by the task group itself rather than mandated in the charter, is that the .11n protocol will enable 100-Mbit/s throughput modes within the existing 20-MHz channels. This does not preclude even higher-throughput modes via the use of wider channels.

Short time line
The estimated time line for 802.11n calls for a finished standard by the end of 2005. This is substantially less time than 802.11 amendments have historically taken. In the present case, however, the genesis of a high-throughput extension came in the January 2002 meeting, so the issues have already been under study for two years.

In addition, the time line established by the Wi-Fi Alliance, based on a poll of member companies, is that high-throughput technology will enter the market within two years. That strong market pull may well overcome the inertia of the IEEE process.

Sean Coffey is manager of the Advanced Technology Group, WLAN, at Texas Instruments Inc. (Santa Rosa, Calif.).

See related chart