03 July 2006

ZigBee: Delivering Ubiquitous Computing at Last?

Ubiquitous computing has been talked about for many years. In some senses, with the sheer number of microcontrollers (MCUs) that exist in appliances around us, embedded computing is already pervasive. However, today the vast majority of MCUs are self-contained islands of automation that communicate only within the system they serve. For truly ubiquitous computing, MCUs must be able to talk to each other.

Viable Application Areas
But is there value in enabling this so-called ‘ambient microcontroller-class networking’? According to Metcalfe’s Law, the value of the network increases exponentially according to the number of connections. So much for the theory, what are the practical applications?  Well, your digital alarm clock radio has an MCU, and possibly your coffee maker as well. If you change your wake-up time (or your alarm wakes you up early because it has received a severe weather warning), wouldn’t it be nice if your radio told your coffee maker about it?

And there is a strong economic argument for MCU networking besides the convenience it affords. For example, it makes no sense for your dishwasher to start a cycle if there’s no hot water. Your alarm clock radio could also predict when you need hot water, based on your wake up time. With more precise automated control of heating and lighting there is potential to enable significant energy savings. Power generation capacity is geared around satisfying peak demand – if small individual power savings could be made at peak times, hugely expensive capacity additions to electricity generation could be avoided. Security applications are also set to benefit from wireless MCU networking, both in terms of easier installation and their effectiveness.

In industrial situations where large, complex plant and manufacturing capabilities depend on a finely tuned process to improve productivity or save costs, it is seen as being more important that information is shared between microprocessors. Wireless reduces the cost of installation and allows much finer granularity of sensing. The wireless industrial networking alliance (WINA) is a coalition of companies interested in the advancement of wireless solutions for industry. One of the solutions being investigated, both for consumer and industrial control and sensor applications, is the low-cost and low-power standard, ZigBee.

Technology Options
The ZigBee wireless networking standard specifies a software stack for applications and security layers on top of the 2003 IEEE 802.15.4 wireless specification that defines the physical and MAC layers. ZigBee was published in late 2004, by the ZigBee Alliance. Primarily intended for control and monitoring tasks where low data throughput (up to 250 kbps) is required, ZigBee’s short-range reach is around 70 metres. However, ZigBee is not the only potential enabling technology for MCU networking.

Other standards vying for ZigBee’s application space, some of them unconstrained by governing standards bodies, are attempting to outpace ZigBee’s development. Zensys™ Z-Wave™ technology is a proprietary example which has garnered widespread industry support, which is being channeled through the Z-Wave Alliance.

Yet another initiative sees a group within the Internet Engineering Task Force investigating the possibility of implementing a ‘compressed’ version of the potentially heavyweight IPv6 (next-generation Internet Protocol), to reside on top of the 802.15.4 radio standard.

All these protocols provide the basic ‘plumbing’– much like TCP/IP on your laptop. Much of the value-add information services that run on top of these protocols have yet to be developed. The situation is much like the internet – just before the web was invented.

Operating systems are in development for this class of applications. TinyOS is an research-oriented open-source operating system designed for wireless embedded sensor networks. It has its origins in Smart Dust military project, which aims to demonstrate that a complete sensor/communication system can be integrated into a cubic millimeter package.

A Nascent Market
Growing interest shown by major chip manufacturers indicates that this nascent MCU networking market is set for rapid growth. Semiconductor vendors including TI, STMicroelectronics and Freescale have all acknowledged the volume potential for ZigBee products and have announced acquisitions, partnerships and OEM agreements to pursue this market. Oki has also announced its first ARM-based ZigBee solution in collaboration with Integration Associates.

Consequently, although current market predictions for ZigBee are divergent, most analysts are predicting significant growth for ZigBee silicon. A mid-range estimate is that this market will be worth a billion dollars between 2008 and 2010.

From an initial requirement that ZigBee should be a lightweight application that is easy (and cheap) to deploy, the specification for the ZigBee stack has grown to the point where it does begin to stress an 8-bit processor. In fact, the cheapest implementations will be those that integrate ZigBee within the applications processor, rather than separating the wireless functionality in a discrete device. Inevitably, actually doing something useful with the networking capability will demand further CPU cycles. Data has no value until converted into information that can be acted upon. This will drive lots of new and interesting processing. With telemetry-type bandwidth, much of the intelligence will be pushed to the end-node, driving performance requirements higher. Finally, connectivity always throws up security issues, which, depending on the application may require significant steps to be taken. All in all, a low-cost, low-power 32-bit processor is certainly appropriate as a platform for ZigBee-enabled applications. In fact, the primary drivers of cost in MCU cores are the memory arrays and peripherals, not the CPU word width.

Because ZigBee lends itself to integration within the application processor, the availability of good development support for the processor is also an important consideration. Development issues around ZigBee systems should not be underestimated. Complex reliability and security issues remain to be solved. As applications are predominantly deeply-embedded and distributed, debugging and system-level modeling capability is very important.

With the standards, technology and industry support now in place to enable microcontroller networking, a burgeoning market beckons. With most analysts believing it’s a case of ‘when’ rather than ‘if’ this market takes off, the exact timing will depend on delivery of killer consumer and industrial automation applications.