The Wireless LAN Rush

The call is echoing across the world: "No more copper!" Everything seems to be going wireless, including the local area network (LAN). A number of approaches to the wireless LAN (WLAN) have emerged, however, creating confusion and uncertainty as to which standard to embrace. The situation is further aggravated by a potential for interference between WLANs using the same radio spectrum. Understanding the options and system implications of each approach is the first step toward making the wireless vision a reality.

Three major approaches have reached the level of industry standards: IEEE 802.11, HiperLAN, and Bluetooth. In addition, the unique needs of the home market sparked industry development of the HomeRF approach. Other approaches often mentioned in conjunction with networking, including wireless ATM and wireless local loop, are not so much networking schemes as they are wireless Internet access mechanisms. Still, they may be the right approach for some applications. Then there are the evolving approaches, such as ultra wideband (UWB), that may represent future generations of WLAN. All merit consideration.

Understanding the structure and application of the various alternatives is the place to begin. You can get more technical detail on each of the standards from TechOnLine's courses (Broadband Wireless Access), lectures (Wireless Local Area Networks), and feature articles (Broadband Wireless Access). You can also get a quick overview from the technical summary table. But the lay of the LAN can only be seen from a high-level viewpoint.

An 802.11 network's main application is data transmission. The baseline media access method uses a carrier sense multiple access/collision avoidance (CSMA/CA) scheme to manage use of the RF channel. This scheme calls for a node to listen on the channel before attempting transmission (RF channels are half-duplex). If the channel is busy, the node waits until it is free, then uses a random delay before checking again. This approach lowers the likelihood of a channel conflict, but makes no allowances for timeliness of data transmission. An optional extension to the baseline method, the point coordination function, supports time-division multiplexing for time-bounded, connection-oriented applications such as telephony.

Figure 1a:  The ad-hoc network structure in the 802.11 protocol

The 802.11 protocol supports two types of network structures. The ad-hoc network (Figure 1a) is one formed from a collection of peer nodes all using RF links. This network has no formal structure; all nodes can communicate with all other nodes. Several algorithms are available to prevent this from being total chaos, however, including a spokesman election algorithm that selects a master from the collective and makes all others slaves. Another possibility is to use broadcast and flooding to all other nodes to establish an addressing scheme. A good example of an ad-hoc network is one that is formed when a group gets together at a meeting and everyone has WLAN-enabled PCs. They can form an ad-hoc network at the meeting to share data.

Figure 1b:  The infrastructure network structure in the 802.11 protocol

The infrastructure network has a formal structure (Figure 1b). It uses fixed access points (AP), which are RF-enabled nodes on a hard-wired LAN. The structure allows mobile nodes to communicate with the access points to join the network. Mobile units can move freely within the area covered by the access point radios, typically a range of 100 meters for the 2.4 GHz band. The RF link is intended to operate with units moving at pedestrian or vehicular speeds.

A group of RF-connected nodes, regardless of structure, is called a basic service set (BSS). The ad-hoc grouping forms an independent BSS. With the structured network, however, you can configure the sets to form an extended service set (ESS). This is accomplished by spacing the APs, which are connected through a distribution system such as an Ethernet LAN, so that their areas of coverage overlap. Wireless groupings within each AP's range form a service set, and each node of the set, upon initialization, receives an address that includes the access point's identity. Mobile units are free to roam both within and among the service sets. The protocol handles addressing issues by forwarding data from the addressed AP to the AP currently supporting the mobile node.

Because the two RF versions of the 802.11 represent changes to the PHY layer only, it's easy to assume that 802.11a represents a growth path for 802.11b networks. That's not the case, however. The 802.11a PHY's signal characteristics trade signal range for bandwidth, with the result that the range for the 5-GHz links is less than half that of the 2.4-GHz links. To cover the same area, then, the 802.11a network would need more than four times as many access points. Instead of replacing 802.11b, then, 802.11a would handle high-density, data-intensive computing clusters while 802.11b provides wider geographic coverage for less demanding transactions.

The HiperLAN/1 standard was intended from the outset for both data and multimedia traffic. It allows the setting of channel access priorities for data streams so that they can be ensured of timely delivery. The priority is based on a residual lifetime for the packet that represents the maximum time interval allowed for delivery. This lifetime must accommodate the multiple hops of an extended service set, however, so packet priority must increase as a node roams.

The HiperLAN/2 standard is just becoming a commercial reality. It is a higher-speed network, allowing 54 Mbps at ranges up to 30 meters. Like the 802.11 family, however, the HiperLAN family does not represent a growth path. HiperLAN/2 uses the 5 GHz band, but has an entirely different network structure from HiperLAN/1. HiperLAN/2 is connection-oriented, using time-division multiplexing to allow duplex traffic and orthogonal frequency division multiplexing to break the high-speed data stream into several parallel, interleaved data streams on different channels. Prior to data transmission, the mobile node and the access point establish the channel for the connection using signaling protocols. If the mobile node moves to another AP's region, a new channel gets established.

The connection-oriented nature of HiperLAN/2 provides the opportunity for both point-to-point and point-to-multipoint data channels. The point-to-point channels are bi-directional; the multipoint channels are one-way transmissions from the AP to the mobile terminals. The connection approach also allows HiperLAN/2 to offer quality-of-service (QoS) features. The QoS can be as simple as a relative priority over other data channels or handled in terms of bandwidth, jitter, delay, bit error rate, and the like. With its combination of high bandwidth and QoS features, the HiperLAN/2 can easily handle mixed voice, video, and data transmissions simultaneously.

The other two versions of HiperLAN are not yet fully defined. A point-to-structure forms the basis of HiperLAN/3, also called HiperAccess. This is a longer-range variant of HiperLAN that is intended to provide users with high-speed (25 Mbps, typical) access to a wide variety of fixed networks. The HiperLAN/3 supports many network types, including IP and ATM networks. HiperLAN/4, called HiperLink, is intended to be a short-range, high-speed (155 Mbps at 150 m) interconnection between HiperLANs and HiperAccess.

HomeRF's shared wireless access protocol (SWAP) uses a combination of data access methods. The basic data structure is a time-division multiple access (TDMA) with eight time slots available for voice traffic and the remainder of the frame for data traffic. If voice channels are not in use, their bandwidth is made available for data. Within the data frame, a connection-oriented data service for streaming media has first priority for time slots. Access to the remaining bandwidth for asynchronous data follows the CSMA/CA approach.

The HomeRF access protocol allows simultaneous establishment of host/client and peer-to-peer communications in either an ad-hoc or a structured topology. Nodes can be one of four types: a connection point, a voice terminal, a data node, or a combination voice and data node. The connection point serves as the home gateway to the public switched telephone network and the host for host/client communications. The voice terminal uses the TDMA to communicate with the connection point. The data node uses the CSMA/CA service to communicate with the connection point or with other data nodes. The combination node, well, is a combination of the voice and data nodes. The network will support a total of 127 nodes of combined types.

Figure 2:  HomeRF handles multimedia, data traffic, and voice using a frequency-hopping spread-spectrum RF link.

Although HomeRF supports three data types, voice is a particularly important component of its intended application environment. HomeRF provides for as many as eight simultaneous voice channels and makes a special effort to keep voice quality high in the presence of severe interference. It uses the unique approach of having the radio's frequency hop occur before the final time slot in a data frame (Figure 2). This time slot is available for re-transmission of lost voice packets so that the replacement packet uses a different frequency than the lost one, minimizing the chances of the data being lost a second time.

The data structure of SWAP is designed to simplify reuse of existing design elements to lower system cost. The radio section uses the same components as Bluetooth. The voice packets follow the digital enhanced cordless telecommunications (DECT) system used in standard cordless phones, so silicon and software for cordless phones can be easily converted to a HomeRF design.

The Bluetooth RF link uses a rapid frequency-hopping broadband modulation scheme that allows it to sustain a number of simultaneous physical channels with a range of about 10 meters. Time division duplexing makes the channels bidirectional. The numerous channels let a Bluetooth master establish as many as three synchronous connection oriented links (SCO) to each slave, one asynchronous connectionless link (ACL) for point-to-multipoint traffic, or a combination of the two types. Bluetooth slaves can support three SCO links to a single master, or one to each of two masters. The ACL link can be either symmetric 432 kbps channels or an asymmetric 721 kbps forward channel with a 57 kbps return channel.

A Bluetooth master can communicate with as many as seven slaves simultaneously out of a field of 256 devices. Devices not actively communicating with the master stay in a park mode until the master establishes a channel to them. Slave nodes in a Bluetooth setup cannot communicate directly with one another; they can only communicate to the master.

Figure 3:  The Bluetooth piconet and scatternet

It's not quite that simple, though. A Bluetooth device can simultaneously serve as a master in one piconet and a slave in another (Figure 3). This duality allows for an extended topology, called a scatternet, in which several Bluetooth piconets link together for the exchange of data. The extended topology can become very complex and take on some of the characteristics of a true network.

The various nets in Bluetooth are an ad-hoc structure. When a Bluetooth master is activated it begins a device discovery process that determines what nodes are within its range and the services they offer. It then establishes its connection topology with the slaves. The net exists only as long as the master is present. When it moves out of range, the net collapses.

One such PHY improvement under consideration is the UWB radio, which transmits a series of very short pulses to represent the data. These short pulses translate into a broad frequency spectrum with very little power in any particular band. They also allow for high data rates, with as much as 10x improvement over existing WLAN PHYs. UWB, under development by companies such as XtremeSpectrum and Pulse-Link, is still a long way from commercial implementation, however, as it has not yet received regulatory approval.

Changes are continual in the network protocols, as well. Along with emerging issues such as security and authentication, there is a keen interest in the industry in making sure the network types at least coexist where possible. If the networks target the same application, as do 802.11 and HiperLAN, coexistence may not be an issue; only one or the other WLAN will be present at any time. When the application spaces are different, however, there may be advantages in having them coexist.

A prime example of the need for coexistence is the conflict between 802.11 and Bluetooth. They have complementary applications, with Bluetooth handling links among peripheral devices and 802.11 linking the computing devices. It's easy to envision an office environment in which a computer uses Bluetooth to talk to a PDA or local printer and 802.11 to access the office network. Unfortunately, Bluetooth's frequency hopping scheme prevents a mixture of the two. Both use the same frequencies in a spread-spectrum approach. Bluetooth changes frequencies some 600 times faster than 802.11, however, and does not use collision avoidance, making the chance of interference with the 802.11 network quite high. The IEEE 802.15 working group is looking into approaches for solving the problem.

They are not the only organizations working on making improvements. The HomeRF group is polishing its specification, the HiperLAN2 Global Forum is working with that standard, and the Bluetooth Special Interest Group (SIG) seeks to continually improve its approach. The Wireless LAN rush is in full swing, and the players are still jockeying for position.

About the Author Richard A. Quinnell is a technology journalist and writer who has covered the electronics industry for 15 years. Prior to writing about electronic design, he was an EE working on projects as varied as satellites, surgical lasers, and submarine communications.