Just a few years ago, the term "network" referred to a mainframe in a back room connected to a series of dumb terminals. Today, the network is a worldwide array of computers connected to increasingly more intelligent clients ranging from personal computers to personal digital assistants (PDAs) to cell phones. Traditionally, these clients connect to the worldwide array over wire or some other physical medium, such as a cable or copper twisted pair. Wired connections provide reliable, high-speed information pipelines, but they tie clients to a location. Wireless technology frees clients from their limiting umbilical cord to the network. A wireless front end connects the client to the wired network over a wireless link. This approach frees clients from cables and wires, enabling the flexibility

of mobility and letting users decide how and where they want to work instead of having the work space define how and where users work. Such flexibility ranges from the elimination of cables to connect PCs to peripherals, such as printers and speakers, to PDAs enabling a nurse to immediately access patient records from anywhere in a hospital. Wireless connections also enable automatic network connections without human interaction; for example, at the end of the day, a repair truck could automatically report mileage and inventory status from a parking lot. Several proposed wireless standards promote the growth of wireless connectivity. The IEEE 802.11 standard targets professional and wireless-LAN applications. Two other standards, Bluetooth (www.bluetooth.com) and Home RF's (www.homerf.org) Shared Wireless Access Protocol (SWAP), are more relaxed specs targeting cost-conscious consumer markets. Their originators based these standards on 802.11 technology but relaxed power requirements and transceiver complexity and functions to reduce cost. Each of these three standards, 802.11, Bluetooth, and SWAP, operate in the 2.4-GHz spectrum, and 802.11 will this year offer a 5-GHz option. All three employ RF-spread-spectrum technology (see Table 1 ). Other organizations involved in wireless technology are the Wireless LAN Alliance (www.wlana.com) and the Wireless LAN Interoperability Forum (www.wlif.com). Although the standards committees target each standard to specific applications, significant overlap among them still exists. In some markets, such as the home network, the competition could be fierce, and selecting the right standard (or proprietary protocol) requires a careful look at the technical issues of implementing each. 2.4-GHz standards Among the 2.4-GHz standards, 802.11, the business wireless-LAN standard, offers a secure and robust platform for connectivity. In July 1997, the IEEE adopted the 802.11 standard supporting 1- and 2-Mbps data rates in the 2.4-GHz band with frequency-hopping spread-spectrum (FHSS), direct-sequence spread-spectrum (DSSS), or infrared physical layers. The 802.11 standard supports both data (synchronous) and voice or time-bounded data (asynchronous). The point-coordination function, a protocol scheme similar to time-division multiple access, supports asynchronous data, guaranteeing that a client owns a certain amount of bandwidth. The 802.11 standard also supports wired equivalent privacy, an optional security scheme that you implement by embedding RSA's (www.rsa.com) RC4 security algorithm in the media-access controller (MAC) either as software or as a core of 200 to 300 gates. Two new 802.11 extensions, slated for release this December, are "draft-available," meaning that the extensions are probably stable enough to base implementations upon and that you may see parts supporting these extensions this summer. The extensions will continue to use the same MAC layer as before but different radio (physical) layers from before. Task Group A, (802.11a) jumps from 2.4 to 5 GHz, uses an orthogonal frequency-division-multiplexing scheme, and requires mandatory support of 6-, 12-, and 24-Mbps data rates. It also provides optional "turbo" modes as fast as 54 Mbps to allow for future extension as technology progresses to enable the multimedia network and its diverse types of traffic. Implementation becomes more difficult at higher speeds as the modulation scheme shifts from binary phase-shift keying (PSK) and quadrature PSK to 16 quadrature-amplitude modulation (QAM) for 24 Mbps and 64 QAM for speeds greater than 24 Mbps. Only a DSSS physical layer will be available. 802.11a will also support convolution, coding, and as many as eight channels at 20-MHz spacing. Additionally, Task Group A has set a goal of making the physical layer the same as the European Standards Telecommunications Institute broadband radio-access network to allow both standards to use the same radio and drive down costs. Task Group B (802.11b), designed to deliver Ethernet speeds, uses a complementary-code-keying waveform, the result of a compromise between schemes that Harris and Lucent presented. The standard specifies a 2.4-GHz data rate, but that rate is negotiable from 1 or 2 Mbps at 400 ft to 11 Mbps at 150 ft and 20 dBm. Major challenges for designs operating at 11 Mbps include modularizing and shrinking the radio, as well as antenna issues, such as whether to use an onboard or an integrated antenna, how to achieve an actual coverage of 360-, and whether to use multiple antennas (antenna diversity) to improve the chance of getting a signal with low interference or to avoid nulls. Bluetooth The original conception of Bluetooth targeted point-to-point short-range links for voice applications, such as cell phones to PDAs or a hands-free automobile-adapter kit that eliminates the need for cables and adapter sockets. Its functions have expanded to connecting both the personal network/work space and the distinct personal networks to each other. Additionally, Bluetooth devices could form an ad hoc network with multiple devices exchanging information or relaying information to other devices to extend the 10m range. For example, a company's private-branch exchange could route calls to users' desk phones or cell phones through a Bluetooth network of access points or gateways throughout the office. The Task Force had expected to have by last month completed Version 1.0 of the spec. To preserve interoperability of products, Bluetooth partners have agreed to release no products before finalization of the standard. Shortly after the release of Version 1.0, developer kits will become available. For example, VLSI offers such a kit and slates broad distribution for June. VLSI based the kit on the VLSI Velocity Rapid Silicon Prototyping Platform; plug-in modules offer baseband, radio, and software. The kit also supports development of both ASICs and software above the protocol stack. Using configurable cores and FPGAs, designers can develop products using hardware rather than simulation. Ericsson and VLSI developed the baseband processor that the plug-in modules use, basing it on Version 0.8 of the standard. The baseband processor will not be generally available except as part of the development kit. The two companies hope to offer baseband processors based on Version 1.0 of the standard as samples in the third quarter with volume availability, pending the standard's approval, in the fourth quarter. Target price for implementing Bluetooth in volume is $5 to $7. The challenge in meeting this goal lies in the fact that the combined price for the baseband, radio, and flash technologies for implementing Bluetooth may still run $10 to $15, even by 2001. Vendors hope to reduce total system cost by integrating non-Bluetooth functions onto Bluetooth chips, thus compensating for the $5 to $10 anticipated overrun with other cost savings. Further increasing the challenge is the difficulty of integrating mixed RF and digital technologies onto a single chip. Although several companies hope to offer Bluetooth-enabled products by year's end, volume ramp up is not expected until 2000 to 2001. Market-analyst company Dataquest (www.dataquest.com) anticipates that most of the market growth—250 million units by 2002—will occur in the cell-phone market. Discussion is under way regarding extending the range to 100m by amplifying the transmitter and increasing receiver sensitivity. HomeRF The HomeRF group has proposed the SWAP standard, an attempt to take the best parts of 802.11 and marry them with the European digital-enhanced-cordless voice standard to create a hybrid data/voice standard. HomeRF considered 802.11 too expensive for the home market and relaxed the radio requirements by cutting complexity and features to hit consumer price targets. The major difference between HomeRF and the other wireless standards is that a PC acts as the communications center among clients in the home. RF design The 802.11, Bluetooth, and HomeRF standards share the same two fundamental subsystems that comprise a wireless design: the RF, or radio, side and the baseband side, which comprises a modem/baseband converter, a MAC, and memory. Several companies, such as Harris and Philips, offer products for the RF portion. Both companies have four-IC chip sets. Harris offers the Prism II chip set (Picture1), and Philips supports both FHSS and DSSS with the SA2410 power amplifier, the SA2420 RF driver, and either the UMA1022 synthesizer and SA639 mixer for FHSS or the UMA1021 and SA1630 IF transceiver for DSSS (Picture2). Because Bluetooth and Home RF are based on the 802.11 radio, the same chip set can serve in most 2.4-GHz applications. Designing a radio requires solid system analysis before you begin design. If you want to achieve the necessary receiver sensitivity for 802.11, you have to keep bit-error rates low and have good matching between ICs and filters to avoid too many losses. Routing issues include ensuring that the synthesizer or local oscillator does not read frequencies into the signal. The true challenge and the one that will hold back widespread adoption of wireless technology is making RF cost-effective. The Philips chip set currently costs $12 (100,000), and the Harris chip set costs $36.03 (10,000). However, with $5 cost projections for a Bluetooth implementation, both companies have their work cut out for them to reduce costs. One method of cutting costs is to optimize chips for Bluetooth or HomeRF. The 802.11 radio is more demanding to ensure better reliability and lower bit-error rates than consumer applications may require. By using a less demanding radio, Bluetooth and HomeRF applications can also cut costs for the external matching components and filters. Additional integration options include integrating memory into the modem or MAC or integrating the modem and radio (omitting the MAC for systems that already have a CPU). On the MAC side of design, companies such as Oki Semiconductor offer ICs such as the MSM7730 802.11-compliant MAC with an integrated modem and a physical-layer-to-radio interface. The MAC can also interface with an off-chip, 16-bit processor to handle higher levels of MAC processing, such as security. Partitioning is a key issue for MAC design, especially determining where in the stack you run functions and how much latency the system can tolerate as data passes both ways over the bus to a host CPU. You need software at several levels: the MAC software, mezzanine software, application software, and the operating-system shell. Oki supplies MAC software as firmware and has partners such as Symbionics that sell off-the-shelf higher level software. For applications in which time to market is the driving factor, you can leverage the expertise of companies that have struggled to implement the latest technologies by purchasing modules. Modules can provide you with a working, FCC (www.fcc.gov)-certified radio and MAC and may be available in volumes as large as 100,000 to 1 million. You will pay a premium for modules, but their immediate availability can significantly speed your time to market and allow you to add value at higher levels, such as roaming, network management, plug-and-play operability, power management, and security. Modules, such as Lucent's Wave LAN (www.wavelan.com) family and Digital Wireless' WIT2410, offer standard bus interfaces, such as PCMCIA, and supply everything you need but the antenna. You can use the same module for both access points and clients. Lucent offers $295 to $495, 802.11-compliant, 2-Mbps and turbo-version cards for both PC and embedded applications. Two antenna connectors make it easier to embed a module and secure comprehensive signal coverage. Digital Wireless' 9-mm-thick WIT2410 offers seamless roaming and dimensions smaller than a business card. Price is $275 (1000), and a complete developer's kit is available. Key design issues Putting together a wireless front end is not trivial, and you must balance several factors, including power consumption, transmission range, effective multi path rejection, and interoperability among clients using the same spectrum. Power is a key issue for mobile wireless applications, and there is still much room for improvement. Each of the standards support various power-down or sleep modes. Checking with the access point to see whether data is waiting, say every 400 msec, can significantly reduce power consumption; however, maximum response time consequently increases to 400 msec. Semiconductor manufacturers will also continue to reduce power consumption through greater integration. The challenge involves integrating mixed-signal technologies, such as RF, filters, and VCOs, and moving from dual-conversion radios using IF and RF to single-conversion radios that require fewer filters. Road maps for these chips project that you'll see single-chip, 2.4-GHz offerings as soon as next year. Range is complex, depending highly upon the particular standard, output power, and physical operating environment. (Metal in the area of operation raises multi path and echo issues.) Greater range has the primary benefit of decreasing overall system cost by reducing the number of access points your design requires to comprehensively cover an environment. One way to increase range is to increase transmission power. However, increasing transmission power in mobile clients significantly reduces battery life. An alternative approach is to increase the receiver sensitivity of the access points, which increases the range at which they can hear clients. In this way, you maintain the power efficiency of the mobile unit while increasing its range and directing cost and power increases to the access points. Another key consideration is how well the radio or signal processor handles echoes, especially if your design can work in such diverse environments as small rooms and huge warehouses. Multi path interference is a function of frequency and the physical environment, depending on whether the device is in a traditional office with walls that absorb signal energy or an open airplane hangar surrounded by metal that reflects signals. The multiple reflections of a signal can add to or subtract from a signal, creating nulls or shadows that may cause reflections to cancel out a signal below the sensitivity of the receiver. An FHSS device in a multi path null can avoid the null at a particular frequency by hopping to another frequency. You may have to move a DSSS device from the null to keep a connection, but users may not always have this option. Effective multi path-rejection schemes in the radio firmware or the signal processor are often the key to a design. Interoperability Interoperability among clients using the same spectrum is essential for smoothly and efficiently using bandwidth in a wireless network. Interoperability is important on the physical, MAC, and network levels. On the physical level, you have to know how well the radio acquires a signal and under what kind of interference. For the MAC layer, the efficiency of data processing from the physical layer is critical. And, for the network layer, you have to know how well the device hooks in and authenticates. You will likely concentrate most of your effort on the network layer. For example, multiple clients that share the same part of the spectrum or access point have to share the limited available bandwidth. Additionally, the greater the range of an access point, the potentially greater number of clients contending for that bandwidth. Devices that share bandwidth must also cooperate with each other, which can create interoperability problems among clients from different vendors. The ability to roam can also be a critical part of expanding the coverage of a wireless network, considering that many networks are larger than the maximum range of one access point. In the 802.11 standard, each client currently manages roaming, or the ability to move from one cell to another. Each available access point sends a beacon message. Clients compare incoming beacons and choose the strongest one. When the client moves out of range of the access point, the signal drops below a threshold, and the device begins looking for a new cell. The 802.11 committee is currently addressing interoperability issues between Vendor A and Vendor B access points to support smooth handoffs without packet loss. To achieve this goal, access points must execute the handoff over the network backbone to avoid possible hiccups if the client remains responsible. In either case, the ideal result is a transition transparent to the user. Testing for interoperability is more difficult than surveying a wired site, and you need to update traditional tools, such as "pings" or "sniffers," which monitor real-time traffic, to work wirelessly. For assistance with interoperability testing, neutral operability labs, such as the one at the University of New Hampshire (Durham, NH, www.iol.unh.edu), can be helpful. Too many standards? The variety of available standards enables engineers to select the technology best suited for their applications, avoiding the waste of paying for and supporting unnecessary functions. Such a plethora of standards is both exciting and daunting. But what are the real differences between the standards? Engineers could debate protocol security, robustness, kindness to other systems in providing fair access to the medium, and other issues. Yet each of these standards is still in flux and repositioning potential applications to grab more markets. The challenge for the standards committees will be knowing when to stop: It's not that Bluetooth cannot run a wireless-LAN network, for example, but that to encumber it with the functions and cost necessary to do this task would make it less appealing to the cell-phone market. Also note that going with the official standards may not always link you with the market leader. For example, with its proprietary 2.4-GHz, 2-Mbps, FHSS Open Air Standard, Proxim claims to own more than 40% of the current 2-Mbps market for wireless LANs. You may also require higher data rates than the current standards support. In this case, a proprietary offering not bound by a slowly moving committee can keep your products on the leading edge. For example, Radio LAN, a manufacturer of wireless-end-user and OEM cards, promises to provide cost-effective but proprietary 100-Mbps rates at 5 GHz within a year. Depending on your application, going with a recognized standard may be unimportant. Perhaps your data-rate needs are so low—for example, a bar-code reader taking inventory in a warehouse—that higher range facilitates the most cost-sensitive design. Unfortunately, such a variety of options also spreads confusion. The difference between Home RF and Bluetooth is subtle and perhaps one that consumers will be unable to understand. And, as each standard expands its functions over time, increasing range, bandwidth, and interoperability, the application space each covers begins to overlap, adding further confusion. To further exacerbate the problem, industry participants still perceive risk in the market concerning the benefits of wireless technology. Consistent coverage can be a problem, and the network is less predictable than a wired network. Additionally, wireless technology is slower than wired; mainstream 100-Mbps wireless is still in the future, and wired is already here. Wired is the standard way of doing things, and many small-office/home-office users are unaware that wireless is an option. In this light, the path to wireless connectivity is difficult. Many obstacles remain. Some companies view these obstacles as problems. For others, they represent opportunity. Infrared: a viable alternative to RF In many ways, RF is superior to infrared (IR) technology. IR targets connecting the personal work space and "walkup" users—those who "walk into" a range of a wireless network. It connects immobile devices at distances of of 1m or less and requires an unobscured line-of-sight connection. RF, on the other hand, can communicate over or through walls. IR also has a shorter range than RF and offers less bandwidth. IR is also more difficult to "embed." An IR hands-free phone adapter in a car would require an IR access point bracketed within line of sight and thus visible to the driver. IR's key advantage over RF, however, is that it is here today and already widely deployed. The Bluetooth (www.bluetooth.com) standards group hopes that Bluetooth will replace IR. End-user products, however, may be unavailable for another year and will cost significantly more than those employing IR. Hewlett-Packard (www.hp.com) anticipates that implementing RF in a printer will cost $20 compared with $6, including hardware and firmware, for IR. To use RF devices, PC users must bear the added (versus invisible) cost of an RF transceiver, giving IR, which is standard in many laptops, a significant cost and availability advantage. Additionally, no guarantee exists that any of the RF standards in progress will meet the aggressive schedules set for their release. The initial 802.11 standard, for example, took five years to settle, much longer than anticipated. The Infrared Data Association (www.irda.org) is addressing many of the drawbacks of IR. Very Fast IR (VFI), ratified this January, increases data rates to 16 Mbps and will appear in products next year with no projected cost increase over 4-Mbps Fast IR technology. Instead, VFI relies mainly on changes in firmware and an upgrade to the red diode that generates the IR signal. Advanced IR, scheduled for ratification in July, is also in the works to support reflections (bouncing off walls) to ease line-of-sight restrictions and to extend range with negotiable data rates as high as 115 kbps at 10m. Additionally, multiple devices will be able to access each other. FHSS versus DSSS The FCC does not require users to have a license to operate in the 2.4-GHz operating band, making it a popular band. However, this lack of restriction means that other devices that might interfere with yours may be operating in the same vicinity. For this reason, 802.11 employs spread-spectrum technology. Spread spectrum does not solve interference problems but does improve connection survivability. Two popular methods for using the spread spectrum are frequency-hopping spread spectrum (FHSS) and direct-sequence spread spectrum (DSSS). FHSS splits the available spectrum into separate bands. Both the access point and client "hop" between frequencies based on the same pseudorandom pattern, transferring a piece of data during each hop. Whenever interference corrupts the signal, the devices can resume data transfer after the next hop to a new frequency that is clear. Bandwidth drops each time the device encounters a blocked frequency. However, interference does not break a connection, which leads to the FHSS mantra that slow is better than failure. Intelligent hopping (avoiding frequencies that are known to be blocked) can increase throughput. DSSS, in contrast, spreads the signal across a wider bandwidth than FHSS, creating a lower power density across the spectrum. DSSS encodes each bit of data into 11-chip sequences, which provide both security and robustness against noise. The FCC allows the use of as many as 83 FHSS frequencies in the 2.4-GHz range and as many as three DSSS channels. Two users on the same channel must share the bandwidth, thus reducing overall potential throughput. Because FHSS allows more channels in the same frequency band, FHSS systems can support a greater aggregate bandwidth for coverage. The 802.11 standard supports as many as 26 collocated FHSS networks for higher aggregate bandwidth through the use of organized and orthogonal hopping patterns that reduce collisions. DSSS, on the other hand, can scale to only three channels. In FHSS and DSSS channels of equal data rates, DSSS devices achieve a slightly higher throughput because of higher processing efficiency: FHSS has the added overhead of hopping frequencies and re-establishing the connection, as well as lost bandwidth when encountering blocked frequencies. DSSS also offers greater range than FHSS and is the only physical layer that the 5-GHz 802.11a extension supports. For mobile applications sensitive to battery life, FHSS offers more power efficiency. DSSS applications traditionally use linear power amplifiers, whereas FHSS applications can use nonlinear power amplifiers, thus achieving higher power efficiency. At 2 Mbps, FHSS can be cheaper than DSSS, but at higher bit rates, FHSS hits a cost wall, and you must move to DSSS. In general, FHSS will probably drive the low-cost, lower-bit-rate markets, and DSSS will drive the high-frequency, higher-bit-rate markets. Interoperability and cooperation Several wireless standards could possibly survive the test of time, creating interoperability challenges. For example, how will a Bluetooth (www.bluetooth.com) personal digital assistant (PDA) talk to a HomeRF (www.homerf.ogr) printer? Perhaps a market for black boxes will emerge to translate between standards, such as adapters or dongles that plug into a Universal Serial Bus port. Conceptually, these black boxes already exist: The network itself facilitates interoperability. For example, the Bluetooth PDA could talk to a PC, which then talks to the printer over Home RF. To address interoperability issues such as these, the Open Services Gateway Initiative (www.osgi.org/osgi_html/osgi.html) formed on March 1. The group aims to create "an open standard for connecting the coming generation of smart consumer and small business appliances with commercial Internet services." The standard defines gateways that will provide a central point from which companies can deploy and manage. Regardless of the translator, however, all devices will have to "play well together" for the wireless market to achieve success. Learn from the interrupt-request wars of PC cards: It is not essential that all devices interoperate, that is, talk directly with one another, but that each device cooperate; that is, be considerate of other devices using the same spectrum and therefore avoid contaminating the spectrum. HTML clipboard