Wireless Sensor Networks (WSNs) are becoming big business. U.S.-based analyst Harbor Research predicts the number of wireless devices and sensors shipped in 2010 will be almost 200 million, while ABI Research puts the number closer to 300 million.
If these forecasts are accurate, there needs to be a truly pervasive networking technology that can form networks consisting of hundreds of nodes (even though at present the overwhelming majority of practical wireless networks comprise no more than ten).
These nodes need to be capable of communicating with each other at any time without being compromised by interference from other RF sources. Further, the nodes must be reliable, ultra-low power (because nodes will be typically battery-powered and often hard to access) and low cost to purchase, install and maintain.
To meet these requirements, engineers are faced with developing their own WSNs (a difficult and time consuming business) or selecting a turnkey solution. ZigBee is perhaps the best known. However, in addition to ZigBee, there are some proven proprietary alternatives.
ANT's Practical Wireless Network protocol, for example, is optimized for the star and peer-to-peer networks typical of most WSNs installed to date. More importantly, ANT is a proven technology used in over a million nodes worldwide and employed by several major consumer electronics manufacturers. Let's compare these two WSN technologies in more detail.
Solving the wrong problem
In an attempt to mimic Bluetooth's success as a short-range wireless standard for Personal Area Networks (PANs) in the consumer electronics sector, the ZigBee Alliance has introduced a de facto standard for WSNs. Marketed as "Wireless Control That Simply Works", ZigBee is backed by many powerful silicon vendors, is based on the IEEE 802.15.4 standard Physical (PHY) and Media Access Control (MAC) layers and supports the Alliance's own Network (NWK) and Application (APL) layers.
Like many standards-based protocols, ZigBee has extra features that creep into consortia specs in order to keep all contributing parties happy and ensure interoperability. This tends to increase the protocol's size, lower its efficiency and consequently increase power consumption.
Moreover, ZigBee's backers envisaged that it should be a universal solution for complex mesh networks. However, it turns out that complex mesh networking, whereby a node can communicate with several other nodes directly (see Figure 1), introduces a level of intricacy that's difficult to justify for many practical applications.
Click here for Figure 1.
Figure 1: Mesh networks are overly complex for most practical WSN applications.
While of interest to academics, engineers appreciate that mesh networks are difficult to set up and demand lots of computing resource and electrical power. ZigBee's latest stack, ZigBee Pro, addresses some of these issues.
In reality, practical networks rarely require nodes to communicate with every neighbour. Indeed, virtually all practical networking problems to date have been solved using a pre-determined networking topology such as point-to-point, star and tree (as shown in the examples in Figures 2(a), (b) and (c)).
Click here for Figure 2a.
Figure 2a: Example of a point-to-point network.
Click here for Figure 2b.
Figure 2b Example of a start network.
Click here for Figure 2c.
Figure 2c: Example of a tree network.
For example, Figure 3 illustrates how the network in Figure 1 could be reconfigured to work just as well as a tree network. The key difference is that data is routed via an intermediary before reaching its intended destination. Such practical pre-determined networks will always be more efficient than a complex mesh and hence less costly to set up and maintain, and less power hungry.
Click here for Figure 3.
Figure 3: Tree WSN forming an equivalent of the mesh network in Figure 1
.
ZigBee WSNs are constructed using nodes of varying functionality. These different types of nodes were introduced to reduce costs by allowing the use of the less capable but cheaper types when possible. ZigBee's node family starts with a ZigBee Coordinator (ZC).
This is the most functional device and is used to start the network and provide the bridge to other networks. Then there's a ZigBee Router (ZR). This device is able to run an application function and act as an intermediate router, passing data from other devices.
Finally, the ZigBee hierarchy includes a ZigBee End Device (ZED). This has just enough functionality to talk to its parent node (either the ZC or ZR) but cannot relay data from other devices.
ZEDs can be in a low power sleep mode for much of the time and are the device often used to support ZigBee's claims of long battery life. However, ZR's and ZC's are also needed for a functioning network and these are more complicated and power hungry than ZEDs. So their battery life figures are a little less impressive.
When setting up a ZigBee network it is first necessary to form a subset cluster around a ZC. This also handles requests from neighbouring coordinator nodes wishing to attach their clusters to the mesh (see
Figure 4). Such networks aren't easy to extend on an ad hoc basis, because it is difficult for nodes to casually join and leave the network unless they are the right type.
Click here for Figure 4.
Figure 4: ZigBee requires nodes of varying functionality, complicating network construction.
There are technologies where all nodes are identical and therefore equally capable of acting as "slaves" or "masters" within a practical network and swapping roles at any time. In such a network, nodes can act as transmitters, receivers or transceivers to route traffic to other nodes and can leave or join the network in an ad hoc fashion. In addition, every node is capable of determining the best time to transmit based on the activity of its neighbours, so no special "coordinator" (such as ZigBee's ZC) or "router" (ZR) node is required.
ANT is an example of this type of network solution. It has an inherent ability to support ad hoc interconnection of tens or hundreds of nodes. This means nodes can easily join and leave the network and required system resource overhead is kept low.
