Multichannel time-synchronised mesh networks based on 802.15.4 radios address many of the challenges involved in designing wireless sensor networks, write Lance Doherty, Jonathan Simon and Thomas Watteyne from Linear Technology
While both low-power sensors and wireless networking are now ubiquitous, there are still relatively few examples on the market of low-power wireless sensor networks.
This is not for lack of interest - wireless sensors promise to cut installation costs dramatically, and provide access to data previously impossible or prohibitively expensive to obtain.
With such strong cost motivation, why have wireless sensor networks (WSNs) not come to dominate the market?
Two factors - radio power and network reliability have kept WSNs largely "niche" products.
Data collection applications typically involve large numbers of sense points forwarding data to a central host that may respond with a process set-point or other configuration changes. There are several technologies competing to fill this role, including satellite, cellular, Wi-Fi, and IEEE 802.15.4-based solutions.
Satellite and cellular have the highest energy cost per packet, and are generally unsuitable for battery powered applications.
These solutions make sense only for applications sending at a very low data rate (e.g., one data packet per day) with known good connectivity.
Wi-Fi (IEEE 802.11b, g) sensors are now widely available. The energy cost for a Wi-Fi packet is much lower than cellular, but still high for battery powered operation.
Connectivity and coverage remain important concerns, as the density of access points necessary for reliable communication with a fixed sensor is typically high.
The IEEE 802.15.4 standard defines a physical layer (PHY) and Medium Access Control (MAC) layer for short-range, low-power operation, ideal for WSNs.
IEEE 802.15.4 forms the basis of several proprietary and standards-based protocols including ZigBee unsynchronised single-channel networks, and WirelessHART1 time-synchronised multichannel networks.
With the 802.15.4 standard, sending a few bytes of sensor data, with routing, cryptography, and other headers uses under 30 μC of charge, including receiving a secure link-layer acknowledgement.
Reliability is a concern with wireless systems. The wireless channel is unreliable in nature, and a number of phenomena can prevent a transmitted packet from reaching a receiver, these can be exacerbated as radio power decreases.
Interference occurs when multiple transmitters send at the same time and frequency. This is particularly problematic if they cannot hear each other (the "hidden terminal problem"). Backoff, retransmission, and acknowledgement mechanisms are required to resolve collisions. Interference can come from within the network, another similar network operating in the same radio space, or from a different radio technology operating in-band - a common occurrence in the band shared by Wi-Fi, Bluetooth and 802.15.4.
A second highly varying phenomenon called multipath fading can prevent successful transmission even when the line-of-sight link margin is expected to be sufficient.
This occurs when multiple copies of the transmission bounce off objects in the environment (ceilings, doors, people, etc.) all traveling different distances. When interfering destructively, fades of 20-30dB are common.
Multipath fading depends on the transmission frequency, device position, and on every nearby object - predicting it is practically impossible.
Because objects in the environment move, the multipath effect changes over time, and it is unsafe to rely on any particular connection between devices, or path, to be useable forever.
Figure 1 shows the packet delivery ratio on a single wireless path between two industrial sensors over the course of 26 days, and for each of the sixteen channels used by the system.
At any given time some channels are good (high delivery), others bad, and still others highly varying. Importantly, there was no period observed where a channel was good on all paths everywhere in the network.
Because of interference and multipath fading, the key to building a reliable wireless system is to exploit channel and path diversity.
In a mesh network, sensors can forward radio packets from peers, extending the range of the network far beyond the range of a single radio, and providing the network with immunity to any single radio link failure.
Using a time division multiple access (TDMA) scheme mitigates self-collision and allows for predictable network scaling.
Several mesh protocols are available for sensor networking: 802.11s for Wi-Fi, various ad-hoc routing protocols such as those used in ZigBee, and Linear Technology's Time Synchronised Mesh Protocol, which forms the basis of the WirelessHART and 802.15.4e standards.
Key selection criteria for a WSN technology are cost of ownership and flexibility. Wireless technologies reduce installation costs dramatically compared to wired solutions, but battery changes needed for lifetime goals increase total ownership costs.
Protocols must be designed to scale both in density and total number of sensors, to tolerate interference, and survive the loss of individual devices.
WSNs must be designed to work reliably in environments with wide temperature variations, and with link-layer packet delivery ratios (PDR) down to about 50%.
For industrial applications, the reliability target is typically to receive at least 99.9% of the generated data, as missing data can trigger expensive alarm conditions.
The system must support some minimum number of sensor data packets per second with bounded latency - "stale" packets are typically discarded and count against reliability.
Multichannel time-synchronised mesh networks based on 802.15.4 radios address many of the challenges involved in building reliable, low-power wireless sensor networks.
Authors are Lance Doherty, systems engineer, Jonathan Simon, systems engineering director and Thomas Watteyne, systems engineer in the Dust Networks product group at Linear Technology.
