It's also difficult to get 802.11n devices to work well together with 802.11b and other devices in the 2.4-GHz band when using 40-MHz channels that double the data rate, because of potential interference issues. Techniques for dealing with this problem are still under discussion, and Draft 3.0 will probably resolve them, says McFarland. Some of these techniques are high-level-politeness algorithms based on measuring the amount of traffic in the environment: If there's a lot of traffic, channel width remains at 20 MHz. If traffic is low, channel width can expand to 40 MHz to take advantage of greater bandwidth. Another possible method calls for actively "listening" to detect any nearby access points or networks and moving to 20-MHz mode if any are detected, regardless of the amount of traffic. TGn is discussing modifications of this variant because it's potentially problematic.
Other, more fine-grained methods for preventing interference, such as CCA (clear-channel assessment), operate on a packet-by-packet basis. To prevent collision, before a 40-MHz-wide packet is transmitted, CCA checks to make sure that both channels are clear and that the packet can be transmitted on that whole 40-MHz frequency range. This option will probably remain. The spec now requires the use of both high-level-politeness and more fine-grained mechanisms. How the high-level algorithm will work is under discussion.
"At the MAC [media-access-control] layer, the concern is how devices share the airwaves on a packet-by-packet basis," says McFarland. "The biggest enhancement here in Draft 2.0 versus 1.0 is packet aggregation." Traditionally, in 802.11x, a packet is transmitted and retried until an acknowledgment is returned. In 802.11n, mandatory packet aggregation combines a large number of packets into one superframe, sends it, and gets back a block acknowledgment specifying which packets were and were not received correctly. Only packets that failed must be retransmitted, resulting in a more efficient system and making better use of the high data rate. Additional differences in Draft 3.0 are likely to include changes to some of the spec's optional modes and features, including MIMO methods.
When a product implements one of the spec's optional features, the 802.11n spec provides for a negotiation associated with that feature so both devices can determine which options they include, says McFarland. But complications may arise in network configurations. For example, if the network requires beam forming, currently optional, both access points and clients must support it. Clients without that feature can still interoperate, but the network doesn't have the resulting performance enhancement.
For the first phase of its Certified for 802.11n Draft 2.0 program, the WFA defined a set of features that corresponds closely to most of the mandatory features of 802.11n Draft 2.0. The program also tests some of the spec?s optional features if the device under test implements them.
In its second phase, currently targeting summer 2008, the program will test some optional features that the IEEE is considering for inclusion in the spec. The WFA has not yet decided how to label these capabilities. One possibility is a profile concept, or sets of features that correspond to the requirements of classes of devices, such as one set for PCs and data-centric devices; another for consumer devices that require video and audio streaming; and another that might include handheld devices requiring voice, VOIP, and telephony features. "This [arrangement] would give us better combinations of mandatory and optional features that make sense for that usage class," says the WFA's Hanzlik.
Available 802.11n chip sets differ in whether they target use in access points or client equipment and enterprise or consumer devices. They also have performance differences and varying MIMO configurations. Chip sets for 802.11n consist of CMOS baseband MAC engines, a separate CMOS radio, and an external discrete power amp.
Single-chip 802.11n silicon will likely become available within the next year or so. Legacy Wi-Fi a/b/g single-chip silicon already exists in CMOS SOCs (systems on chip), says Kevin Mukai, Broadcom's senior product-line manager for 802.11n products. Single-chip silicon will be more of a challenge for 802.11n products, because the a/b/g SOCs contain only one RF chain per chip, but 802.11n requires multiple RF chains.