ZigBee: Wireless Technology for Low-Power Sensor Networks

Technologists have never had trouble coming up with potential applications for wireless sensors. In a home security system, for example, wireless sensors would be much easier to install than sensors that need wiring. The same is true in industrial environments, where wiring typically accounts for 80% of the cost of sensor installations. And then there are applications for sensors where wiring isn't practical or even possible.

The problem, though, is that most wireless sensors use too much power, which means that their batteries either have to be very large or get changed far too often. Add to that some skepticism about the reliability of sensor data that's sent through the air, and wireless sensors simply haven't looked very appealing.

A low-power wireless technology called ZigBee is rewriting the wireless sensor equation, however. A secure network technology that rides on top of the recently ratified IEEE 802.15.4 radio standard (Figure 1), ZigBee promises to put wireless sensors in everything from factory automation systems to home security systems to consumer electronics. In conjunction with 802.15.4, ZigBee offers battery life of up to several years for common small batteries. ZigBee devices are also expected to be cheap, eventually selling for less than $3 per node by some estimates. With prices that low, they should be a natural fit even in household products like wireless light switches, wireless thermostats, and smoke detectors.

Figure 1:  ZigBee adds network, security, and application-services layers to the PHY and MAC layers of the IEEE 811.15.4 radio

Although no formal specification for ZigBee yet exists (approval by the ZigBee Alliance, a trade group, should come late this year), the outlook for ZigBee appears bright. Technology research firm In-Stat/MDR, in what it calls a "cautious aggressive" forecast, predicts that sales of 802.15.4 nodes and chipsets will increase from essentially zero today to 165 million units by 2010. Not all of these units will be coupled with ZigBee, but most probably will be. Research firm ON World predicts shipments of 465 million wireless sensor RF modules by 2010, with 77% of them being ZigBee-related.

In a sense, ZigBee's bright future is largely due to its low data rates—20 kbps to 250 kbps, depending on the frequency band used (Figure 2)—compared to a nominal 1 Mbps for Bluetooth and 54 Mbps for Wi-Fi's 802.11g technology. But ZigBee won't be sending email and large documents, as Wi-Fi does, or documents and audio, as Bluetooth does. For sending sensor readings, which are typically a few tens of bytes, high bandwidth isn't necessary, and ZigBee's low bandwidth helps it fulfill its goals of low power, low cost, and robustness.

Figure 2:  ZigBee's data rates range from 20 kbps to 250 kbps, depending on the frequency used

Because of ZigBee applications' low bandwidth requirements, a ZigBee node can sleep most of the time, thus saving battery power, and then wake up, send data quickly, and go back to sleep. And, because ZigBee can transition from sleep mode to active mode in 15 msec or less, even a sleeping node can achieve suitably low latency. Someone flipping a ZigBee-enabled wireless light switch, for example, would not be aware of a wake-up delay before the light turns on. In contrast, wake-up delays for Bluetooth are typically around three seconds.

A big part of ZigBee's power savings come from the radio technology of 802.15.4, which itself was designed for low power. 802.15.4 uses DSSS (direct-sequence spread spectrum) technology, for example, because the alternative FHSS (frequency-hopping spread spectrum) would have used too much power just in keeping its frequency hops synchronized.

ZigBee nodes, using 802.15.4, can communicate in any of several different ways, however, and some ways use more power than others. Consequently, ZigBee users can't necessarily implement a sensor network any way they choose and still expect the multiple-year battery life that is ZigBee's hallmark. In fact, some technologists who are planning very large networks of very small wireless sensors say that even ZigBee is too power hungry for their uses.

A ZigBee network node can consume extra power, for example, if it tries to keep its transmissions from overlapping with other nodes' transmissions or with transmissions from other radio sources. The 802.15.4 radio used by ZigBee implements CSMA/CA (carrier sense multiple access collision avoidance) technology, and a ZigBee node that uses CSMA/CA is essentially taking a listen-before-talk approach to see if any radio traffic is already underway. But, as noted by Venkat Bahl, marketing vice president for sensor company Ember Corp. and vice chairman of the ZigBee Alliance, that's not a preferred approach. "Having to listen burns power," says Bahl, "and we don't like to do that."

Another ZigBee and 802.15.4 communications option is the beacon mode, in which normally sleeping network slave nodes wake up periodically to receive a synchronizing "beacon" from the network's control node. But listening for a beacon wastes power, too, particularly because timing uncertainties force nodes to turn on early to avoid missing a beacon.

Fortunately, this in-your-face strategy leads to very little RF interference. That's largely because ZigBee nodes have very low duty cycles, transmitting only occasionally and sending only small amounts of data. Other ZigBee nodes, as well as Wi-Fi and Bluetooth modules, can easily deal with such small, infrequent bursts.

ZigBee's talk-when-ready scheme doesn't suit all purposes, however. For example, in a network of thousands of tiny sensors dropped into a war zone to monitor enemy troop movements, the power savings provided still might not be enough. With each network node sending data periodically—and with transmissions repeated numerous times through other nearby nodes of a mesh network configuration in order to reach a network controller—large numbers of packet collisions and retransmissions could waste power and significantly shorten sensor node battery life. If the sensor batteries are very small and power-limited, that's especially problematic.

Although contention for airwave access isn't generally a problem for ZigBee, it can be. Sensor-network company Dust Networks, in fact, says contention issues are keeping the company from turning to ZigBee—for now, at least—even though Dust remains a member of the ZigBee Alliance. "Each ZigBee device needs to contend for airspace with its neighbors," says Dust director of product management Robert Shear, "so there's inevitably some contention and some inefficiency." To avoid ZigBee's access contention, Dust uses contention-free TDMA (time division multiple access) technology. ZigBee, through the 802.15.4 MAC layer, provides guaranteed time slots in a scheme that somewhat resembles TDMA, but only as part of an optional "superframe" that's more complex and less power-efficient than TDMA.

ZigBee has still more power-saving tricks up its sleeve, however. For example, it reduces power consumption in ZigBee components by providing for power-saving reduced-function devices (RFDs) in addition to more capable full-function devices (FFDs). Each ZigBee network needs at least one FFD as a controller, but most network nodes can be RFDs (Figure 3). RFDs can talk only with FFDs, not to other RFDs, but they contain less circuitry than FFDs, and little or no power-consuming memory.

Figure 3:  ZigBee networks can contain as many as 65,536 nodes in a variety of configurations

ZigBee conserves still more power by reducing the need for associated processing. Simple 8-bit processors like an 8051 can handle ZigBee chores easily, and ZigBee protocol stacks occupy very little memory. An FFD stack, for example, needs about 32 kbytes, and an RFD stack needs only about 4 kbytes. Those numbers compare with about 250 kbytes for the far more complex Bluetooth technology.

From ZigBee's relatively simple implementations, cost savings naturally accrue. RFDs, of course, reduce ZigBee component costs by omitting memory and other circuitry, and simple 8-bit processors and small protocol stacks help keep system costs down. Often, an application's main processor can easily bear the small additional load of ZigBee processing, making a separate processor for ZigBee functions unnecessary.

But the main strategy for keeping ZigBee prices low is to have big markets and high volumes. The ZigBee Alliance, by making ZigBee an open standard and by vigorously promoting interoperability among ZigBee devices, expects that ZigBee will be very big in applications such as home and building automation. The alliance is currently working on interoperability procedures for those particular applications, which it expects to complete later this year along with ZigBee Specification 1.0.

One reason for optimism about ZigBee adoption for home automation and security is its ease of use. ZigBee networks are self-forming, making it easy even for consumers to set them up. "In the residential space, there's no configuration involved," says the ZigBee Alliance's Heile. "You take something out of the box, put the batteries in, and maybe do something as simple as button-press security—bring two devices close together, push the buttons until the green lights come on, and you're done."

ZigBee networks can also self-form in commercial and industrial settings, but professional installers will have tools that provide additional control, particularly for security. ZigBee security is flexible, says Heile, to give both consumer and professional users what they need. "You don't have to have 128-bit public-key encryption for a smoke detector," he says, "but if I'm in a high-rise office complex, that's exactly the level of security I'm going to have for my fluorescent light fixtures. If you're in a high-rise building on Fifth Avenue, you don't want someone going down the street and turning your lights off."

By offering interoperability, however, ZigBee adds capabilities that proprietary products can't. For example, says Ember's Bahl, interoperability allows the ZigBee nodes of a lighting system to work with the ZigBee network of an HVAC system, or vice versa. "Philips Lighting is really excited about this," Bahl, says, "because it turns them from a ballast manufacturer into the infrastructure backbone of a building-automation system."

Needless to say, many of the major semiconductor companies, and especially those that are big in embedded systems, are eagerly anticipating ZigBee's entry into mass markets. Freescale Semiconductor (until recently known as Motorola's Semiconductor Products Sector) is already providing ZigBee-ready technology to select customers. Other semiconductor companies, including AMI, Atmel, Microchip, Philips, and Renesas, are members of the ZigBee Alliance.

ZigBee will likely be slow to penetrate the industrial market for wireless sensors, however. According to market research firm ON World, it will take five to seven years to convince industrial customers of the reliability, robustness, and security of wireless-sensor systems. ON World does predict significant long-term growth of ZigBee in industry, though. By 2010, the company projects, RF modules used in industrial monitoring and control will reach 165 million units, up from 1.9 million in 2004. About 75% of those, ON World predicts, will be based on ZigBee and 802.15.4.

Eventually, ZigBee could go into a wide variety of applications. In household appliances, it could help monitor and control energy consumption. In automotive applications, it could provide tire-pressure monitoring and remote keyless entry. ZigBee could also be used in medical devices or even in computer peripherals, such as wireless keyboards or mice.

Concern is increasing, though, that ZigBee could turn into a one-size-fits-all technology that doesn't fit any application particularly well. Some skeptics, for example, worry that an attempt to make ZigBee all-encompassing could make the ZigBee protocol stack too large for ZigBee's twin goals of very low power consumption and very low cost. If that happens, then ZigBee's low-power, low-data-rate niche—narrow as it is—will have proven to be too broad. And then, perhaps, we'll need yet another wireless standard to go with the burgeoning number we already have.