The attraction of wireless connection for everyday objects is obvious, but it contains the paradox that, the more "everyday" the application you contemplate, the more removed you are from the expertise that we used to regard as necessary to getting and keeping wireless connectivity up and running. How much simpler could installation and maintenance become if you could wirelessly connect process-control and factory-data sensors and avoid all of the convoluted and damage-prone field wiring that such systems require; interconnect networks of environmental sensors in the field, perhaps temporarily, for research projects; or wirelessly connect heat, light, and security sensors and the controls for building services?

Building control leads to the frequently quoted ultimate mundane application: the wireless light switch. The idea is that you have lighting power only in a building's roof void, with networked wirelessly connected nodes performing switching. You can position the battery-powered, stand-alone switch, which has a life of years, anywhere within range of the wireless net. You cut the fixed installation cost of the building, and you can add, remove, or reposition switches at will. The network can carry data or control information; the same benefits accrue to a variety of applications from ambulatory-patient monitoring in medical systems to metering in utility supply.

Such is the promise of ZigBee and its underlying RF standard, IEEE 802.15.4. However, engineers must present applications such as the light switch to end users with no knowledge of an underlying RF technology, and these applications' manufacturers most likely have previously been electromechanical-product providers. Can you reconcile these needs with 2.4-GHz operation?

The well-documented 802.15.4 PAN (personal-area-network) standard explicitly fills a gap in the progress toward ever-higher-data-rate approaches in other standards. The IEEE intends it as a low-data-rate, short-range standard that exploits the techniques for data transmission in other packet-based network standards but that largely hides their complexities from system designers and completely hides them from end users. As its IEEE numbering implies, it is part of the Ethernet family tree of standards. an ever-growing labyrinth of services that a forthcoming article in EDN Europe will explore. Users will inevitably compare 802.15.4, a low-data-rate, wireless standard, with Bluetooth, but the two are distinct. Bluetooth is a higher-data-rate service for device-to-device communications, and it can carry traffic, such as audio, on a high-duty-cycle basis. It offers limited networking capability, being structured in clusters of a maximum of eight devices. ZigBee, on the other hand, targets low data rates, low duty cycles, and extended networks, and it can overlay a high level of security.

The 802.15.4 standard describes three licence-free radio bands. Two "low" bands include one for use in Europe that carries a single 600-kHz channel centering on 868.3 MHz and one for the United States with 10 channels within a 902- to 928-MHz band. Additionally, the standard employs 16 channels in the worldwide 2.4-GHz ISM (industrial, scientific, and medical) band. It uses DSSS (direct-sequence-spread-spectrum) signals and BPSK (binary-phase-shift-keying) modulation in the low bands, and O-QPSK (orthogonal-quadrature PSK), with a 4×4-point constellation, at 2.4 GHz for high power efficiency and anticipated good coexistence with other 2.4-GHz services. The standard provides speeds as high as 250, 40, and 20 kbps at the 2.4-GHz, 915-MHz, and 868.3-MHz bands, respectively. Allowed power levels vary by geography, typically permitting a maximum level of –3 dBm, but you can set this level depending on the range, bit rate, and tolerable packet-error rate you require for an application.

In network terms, devices can be controllers: hub nodes or FFDs (full-function devices). Alternatively, they can be peripheral: RFDs (reduced-function devices). You can set up star topologies, in which RFDs cluster around an FFD, or mesh topologies, in which multiple FFDs "talk" to each other, and each of which can have associated RFDs "hanging" from their nodes (Figure 2). The mesh topology gives the standard its robust networking characteristics; the key words are "self-configuring" and "self-healing." The protocol integrates autonomous discovery of network resources, attachment to the network, and network configuration. You can think of the 802.15.4 standard as wireless miniature Internet. The analogy is inexact, but useful similarities exist. The 802.15.4 standard uses the same addressing structure as the IP (Internet Protocol), and the objectives for the network are like those that originally defined the Internet. The network environment must be robust, and, as a packet-based protocol, must reliably route a data packet to its destination. Multiple routes exist in a mesh by which a message can reach its destination, and the network employs this mesh to yield a reliable "over-the-horizon" communication fabric.

However, in certain aspects, this standard represents "packet-routing Lite." A data-communications packet-routing scheme must maintain large routing tables so that the network can optimise message routing. Because ZigBee 802.15.4 operates with minimal resources, on the other hand, it reduces the number of tables it acquires in the process of network discovery. In a similar way, the routing algorithms are optimised for power demand; when you need to send a packet, you can "flood" all the available connections in the mesh and dispatch a packet to each one to find its way through the FFDs in the net. (RFDs don't participate in network-routing operations.) However, such flooding would be inefficient in power usage, and it would also tie up scarce processing resources in redundant operations. ZigBee 802.15.4 aims for an optimum strategy that yields guaranteed transmission but at minimum power.

The packet structure allows for CSMA (carrier-sense, multiple-access) operation, which ought to work well with the low-duty-cycle applications that ZigBee's designers had in mind. Alternatively, if you need a deterministic response, you can use a packet structure with reserved slots that you allocate to specific nodes on the network. This structure allows for acknowledgment of all messages. A "beacon" feature gives you even higher power efficiency, allowing short wake-up times for receivers and guaranteeing successful message exchange.

Despite its lite attributes, ZigBee supports a high level of authentication and security at several levels with an AES (Advanced Encryption Standard)-algorithm-based, 128-bit key scheme. It requires the high security levels to secure control building-automation and industrial-control schemes. You might also use such systems to feed data into a corporate IT structure and, if you fail to protect that data, attackers could potentially hack the network as an open gateway. Similarly, if you use a network for medical monitoring, confidential patient data could be part of the traffic.

It would not be a typical IEEE standard if someone were not already working to extend its capabilities, and a suffix-b proposal is already at the task-group stage that seeks to extend the O-QPSK-modulation scheme to low-band operation to increase the throughput and power efficiency of those transceivers.

In the last few weeks, vendors have announced a number of new chips sets for the RF standard, and purchasers have a wide choice of products at the IC, module, and IP (intellectual-property) levels.

You now see the terms "802.15.4-compliant," "ZigBee-ready," and "ZigBee devices." Figure 1, which you see in any ZigBee-related publication, does not fully follow but has strong similarities to the OSI 7-layer model. The fully ratified 802.15.4 IEEE standard deals with the PHY (physical) and MAC (media-access-control) layers or, broadly, the radio itself and its parametric control, plus part of the data-link layer. ZigBee will define the higher layers, eventually extending to the application level. It will have a limited number of application "profiles" in a fashion similar to those Bluetooth defines. Initially, the ZigBee Alliance will this year ratify the standard without those extensions.

Therefore, an 802.15.4-compliant radio should be just that: a standard air interface and modem that you can operate by and configure according to a set of commands the standard outlines. The ZigBee Alliance has not yet finalized ZigBee's higher level functions, so if a vendor declares its product ZigBee-ready, the company is implying that you will be able to upgrade its software to behave as it would in the approved version. For now, the vendor may write it to what it expects ZigBee will contain, or it may be a proprietary implementation. Likewise, using the silicon that has recently become available, you can design to comply with the draft ZigBee proposals, or you can use the radio standard to build a network complying with your own requirements.

In many cases, a proprietary implementation may be good enough for your purposes; if you require only a self-contained, RF-based network that you can configure or extend using only products from your own line, then you can design one today with the 802.15.4 products that are now on the market. If you are building a system that will use the "open" nature of the standard—for example, that much-noted light switch in which, ultimately, the consumer should expect to buy any brand from the hardware shop and have it work—then you need to pay careful attention to the ZigBee-compatible aspect.

For modest-volume products or for early development, you will likely use preassembled modules that implement the 802.15.4 function; for higher volume production, you might prefer a chip-level approach for cost reasons. As with Bluetooth, you have a choice of architectures. A typical offering has a number of discrete blocks that may come as multichip implementations; most will over time migrate to single-chip approaches. The transceiver at the front end has an associated modulator/demodulator; a software-protocol stack then drives a baseband controller. In a similar fashion to Bluetooth, a number of options exist for where that software resides.

You can port the software to a separate microcontroller that is also running the application that it requires to communicate. In that case, you must ensure that the application processor has enough free code space and processing power to add the extra functions. A microcontroller core with just enough power to run the protocol stack can be part of the transceiver-chip set for a stand-alone approach. This approach might be the optimal one if you want to, with minimal intervention, add RF networking to an application. Or, for simple applications, a single chip with its own embedded microcontroller might have enough spare processing capacity to host your application code as well as the protocol stack.

In all cases, however, remember that the design is for an RF device operating at gigahertz frequencies. Although the use of the ISM bands have become commonplace, signal propagation at these frequencies is difficult. To get an effective link or network requires pc-board design, signal routing, an effective antenna, and an enclosure that accommodates the gigahertz signals. Laboratories lacking a full range of RF-test equipment are likely to undertake ZigBee designs. In that case, they should probably precisely follow reference designs, noting that any component substitutions or detail changes such as pc-board material can drastically affect performance.

Antennas for use in the ISM bands are subjects of their own; you have the choice of providing a pc-board track antenna (easier at 2.4 GHz than at 868 MHz, with dipole length of about 5 cm as opposed to 17 cm) or compact "chip" (ceramic) antennae for propagating directly from the pc board. Alternatively, you can use an external antenna. Motorola has produced a primer on the subject that describes the geometry of a number of well-tried antenna configurations (Reference 1). The range of a ZigBee/802.15.4 link is a complex function that depends on the RF environment; the power level, which you may choose to optimise for battery life; the data rate; the physical location of the transceiver; and a variety of other parameters, such as the chip set's receiver sensitivity. The 802.15.4 standard specifies a minimum sensitivity, but most announced receivers exceed it with figures of around –90 dBm. The standard gives a formula for calculating range, but it is strictly an "all-other-things-being-equal" calculation for line-of-sight operation, and any given equipment design most likely requires practical evaluation. In practice, you may choose to operate with a range of only a few metres, such as in a medical patient-monitoring application, or when you have a closely spaced array of sensors. With the maximum power and a high-gain, directional antenna, you might reach kilometres of range. Tens to hundreds of metres is the likely usable spread.

Similar optimsation choices apply to battery life. In many applications, the mains-powered FFDs that form the backbone of the network are always receiving but remote, typically low-duty-cycle sensors are usually in deep-sleep mode, drawing microamps, and awake only in response to an event, to report a data reading, or to periodically verify a time-out interval. The ZigBee standard also minimises power on those low-duty-cycle events; a node can awake, sign on to the network, pass a small data payload, and close down again in tens of milliseconds. This mode of ZigBee operation leads to long battery life. The ZigBee Association publishes a graph showing that, for a regular check-in cycle, if you power a node from a small alkaline cell, any reporting cycle with a longer interval than 14 sec yields an average power drain that is less than the self-discharge rate of the battery.

This standard, then, promises to be one in which you can concentrate on wirelessly enabling an application and get the benefits of flexible wireless networking with almost no effort. If you choose to go beyond the default settings, you can configure the detail of your network in many ways.

Vendors in this sector divide into a number of start-ups, energetically promoting both silicon and module products, and the established semiconductor suppliers, especially those with a track record in RF or mixed-signal products. Some of this group have with little fanfare added 802.15.4 products to their portfolios. Deal-making measures occur between these two groups, and start-ups provide specialist expertise to design silicon that a semiconductor major manufactures and the silicon that both parties offer.

Look for silicon offerings from vendors such as Atmel, Chipcon, CompXs, Freescale (formerly Motorola), Infineon, Philips, Mitsubishi, Oki, Renesas, Xemics, and ZMD. You can find a complete list of members of the ZigBee Alliance on its Web site, although not all of them have announced products.

Atmel offers the low-band AT86RF210 Z-Link transceiver and is developing a flash microcontroller based around its AVR RISC core. Silicon vendor Chipcon has developed the 2.4-GHz CC2420 transceiver with hardware MAC and AES-128 encryption support; dedicated supplier Ember uses and resells this chip for its EmberNet product line (Figure 3). Start-up Ember offers both module and chip products, including a ready-to-operate module that you can upgrade, if necessary, to the final ZigBee specification. Ember this year acquired a design team and ZigBee/802.15.4 IP from Cambridge Consultants, which had in turn carried out considerable initial development on mesh networking and will continue to operate as design consultancy for project work in the same field. Chipcon will shortly announce the 2430, a flash-memory-equipped version for the transceiver chip.

Roke Manor Research has, similarly, had early access to silicon from Motorola and has used this silicon as the basis of IP that it is offering for module-level products. Acknowledging that high-volume manufacturers will eventually design and produce their own products, Roke's Technical Sales Manager Elaine Burrows notes that the company has been able at an early stage to bring its expertise in RF systems to bear and to refine IP that circumvents the problems in that area. The offering from Freescale includes the 2.4-GHz MC13192 RF data modem with the less integrated alternative of the 191 transceiver chip. It supports the Z-Stack ZigBee software that Figure-8 Wireless developed and that the Roke IP also uses. Freescale offers the 13192DSK-A00 development kit, which it based on the RF silicon and an HCS08 microcontroller. Freescale will also be developing integrated sensor products that are wireless-ready for ZigBee/802.15.4.

Sensor company Crossbow is taking a similar approach with its MICAz wireless measurement system, a fully configured module for wireless monitoring of a number of physical parameters, such as temperature and vibration. CompXs, which has produced silicon in association with Oki, also produces a full development environment in a "low-integration" form for development work, a range of module products, and instrumentation in the form of a protocol tester. A similar breadth of products is available from Helicomm, a company that has alliances with Chipcon and ZMD for silicon. It also offers its product as IP, as does design company Jennic. Jennic has developed a series of hardware and software IP modules, covering an 802.15.4 RF transceiver for the 2.4-GHz band with a claimed receiver sensitivity of –93 dBm, a modem, a baseband controller, and a software stack. The company has built an integrated chip that includes the complete RF modem with a controller core using an open-source RISC design, memory, and interfacing circuitry. First silicon, from IBM's RF CMOS process, is due in the next few weeks; Jennic's Technical Director Colin Faulkner says that the company's mission is to sell the IP in whatever combination a project requires. However, if the demonstrator chip turns out to be a good fit for a given application, the company could quickly build something like it.

Fewer compliant radios exist for the low bands, but one comes from mixed-signal vendor ZMD. The ZM44101transceiver has a "thin" MAC implementation in which the protocol stack runs on the host application's controller.

In general, the parameters you will likely compare among vendors centre on the familiar ones of sensitivity and power drain, but you should be cautious because vendors base many of their published figures on simulation and are for stand-alone chips. You must decide the optimum architecture for your project and estimate the total power demand to get a true comparison. In the same way, you will see many similarities in technique, if not in execution, that the various vendors have learned through integrating RF sections for products such as cell phones and wireless LANs. Such techniques include low- or zero-IF receivers, for example.

To end with the invidious comparison with Bluetooth: The view of the ZigBee Alliance and many of its members is that this history should be much simpler than the first round of Bluetooth developments. ZigBee includes more built-in features, default configurations, and RF lessons. ZigBee's aspiration of "wireless that just works" looks achievable.

You can reach Editor Graham Prophet at +44 118 935 1650, fax +44 118 935 1670, e-mail gprophet@reedbusiness.com.