Placing Wireless Sensors Anywhere |

Placing Wireless Sensors Anywhere

Placing Wireless Sensors Anywhere

Placing Wireless Sensors Anywhere - Delivering Low Power & Reliable Wireless Sensor Networks over Long Distances

By Lance Doherty, Jonathan Simon, Thomas Watteyne and Ross Yu, Dust Networks Product Group, Linear Technology

One of the visions of the Internet of Things is to be able to measure and instrument things that have never before been measured. Whether monitoring aging infrastructure, such as bridges, tunnels or power transmission lines, or providing real-time parking and traffic information, these applications call for wireless sensor networks (WSNs) to deliver wire-like performance and yet be practical to deploy. These WSNs need to be able to scale to large numbers of wireless nodes, or motes, and in many cases, cover long distances.

Keys to Wide Spread Usage

In order for WSNs to be deployed on a wide-scale basis, they must be practical to deploy and run reliably for many years, often over a decade. To enable this, WSNs must meet a number of key requirements:

·         Place a sensor anywhere– Measurement points need to be placed where optimal for sensing, but not necessarily optimal for communication. As a result, sensor nodes are often placed in locations that do not necessarily have convenient access to communications or power infrastructure, and often are in challenging RF environments (e.g. close to ground, in tunnels, under cars or deep within machinery).

·         Low maintenance– The network must be largely self-maintaining, and any physical maintenance (e.g. battery replacement) must not incur additional ‘truck rolls’ or technician visits. For example, in smart parking applications, battery-powered sensors embedded in the street surface are permitted only if they can be replaced at the same interval as regular road repair, which occurs no more frequently than every 5-7 years. In other applications, WSNs are deployed for more than a decade.

·         Communications reliability – Must be able to reliably communicate with the sensors despite the fact that they may be located in a rough RF environment.

·         Scalability - A network needs to be suitable to a variety of similar yet unique deployments which cover a fairly wide range of network sizes (both number of motes and area coverage), depth (i.e. number of radio hops a node can be from a data egress point), data traffic level, etc.

Building a Predictable Network on an Unpredictable Medium

Low Power is Difficult without Making Tradeoffs - There are many approaches in wireless sensor networking that target low-power operation. Some wireless networks, such as ZigBee, achieve low power only on the sensing devices at the edge of the network but require line power for any routing nodes. Other networks introduce a basic form of duty cycling, called ‘beaconing’, in which the entire network shuts down to a low-power sleep mode for extended periods of time, but sacrifice network availability and overall network capacity. However, for the types of applications talked about for the Internet of Things, wireless sensor networks must be able to accommodate much larger networks and publish at regular data intervals. The challenge, therefore, is to provide low power without sacrificing reliability or network availability.

RF Is Unpredictable – Radio (RF) is an unpredictable communications medium. Unlike wired communications where the communications signal is shielded from the outside world by cabling, RF propagates in the open air and interacts with the surrounding environment. There is the possibility of other RF transmission sources to cause active interference. Much more common is the effect of multi-path fading, where the RF message may be attenuated by its own signal reflected off of surrounding surfaces and arriving out of phase. Mobile phone users experience multi-path fading every day when their phone has poor signal strength in one spot, but can improve it by moving a few centimeters. Furthermore, the effects of multi-path change over time, as nearby reflective surfaces (e.g. people, cars, doors) typically move. The net result is that any one RF channel will experience significant variation in signal quality over time. However, since multi-path fading affects each RF channel differently, using channel hopping for frequency diversity minimizes the negative affects of multi-path fading. The challenge for WSNs, then is the ability to use frequency hopping over large networks with multiple hops.

Time Synchronized Channel Hopping Mesh Networks

Low-power, reliable wireless sensor networks are a reality with Time Synchronized Channel Hopping (TSCH) mesh networks, pioneered by Linear Technology’s Dust Networks, and have been proven in some of the toughest environments around. TSCH is already a foundational building block of existing industrial wireless standards, such as WirelessHART (IEC62591), and is an enabling piece of emerging Internet Protocol-based wireless sensor networks standards.

In a TSCH network, each node has a common sense of time that is accurate across the network to within a few tens of microseconds. Network communication is organized into time slots to enable low-power packet exchange, pair-wise channel hopping and full path diversity.

Low-Power Packet Exchange - The use of TSCH allows motes to sleep at ultralow power between scheduled communications. Each device is only active if it is sending a packet or listening for a potential packet from a neighbor device. Even more importantly, since each node knows when it is scheduled to wake up, each node is always available to relay information from its neighbors. Therefore, TSCH networks often reach duty cycles of <1% while keeping the network completely available. Furthermore, since each packet transaction is scheduled, there are no in-network packet collisions in a TSCH network. Networks may be dense and scale without creating debilitating RF self-interference.

Pair-Wise Channel Hopping - Time synchronization enables channel hopping on every transmitter-receiver pair for frequency diversity. With a TSCH network, every packet exchange channel hops to avoid inevitable RF interference and fading. In addition, multiple transmissions between different device pairs can occur simultaneously on different channels, increasing network bandwidth.

Full Path and Frequency Diversity - Each device has redundant paths to overcome communications interruptions due to interference, physical obstruction or multipath fading. If a packet transmission fails on one path, a mote will automatically retry on the next available path and a different RF channel. Unlike other mesh technologies,

a TSCH network does not require powered routers and time-consuming path rediscovery.

Figure 1. Path and Frequency Diversity - If communication fails on the 'green' arrow, node D retries on the 'purple' arrow using another channel

TSCH-based networks are successfully deployed today in such applications as smart-parking applications, computer data centers to monitor energy efficiency, and in industrial plants. Many applications, such as pipeline monitoring, structural monitoring of bridges and tunnels, as well as power transmission line monitoring require the WSN to span long distances. And yet, the ability to establish and successfully maintain a low-power, reliable wireless network over such distances represents one of the more challenging network topologies. By definition, a deep-hop network means that the messages from nodes furthest away need to traverse many hops to reach their destination. While this enables a single network to cover a large geographic area with relatively low-power transceivers, it sometimes raises the concern of whether a long network can successfully sustain regular data traffic from all its motes, and do so with acceptable latency and current consumption.

Case Study – A deep hop mesh network

To characterize such a network, a 100-mote, 32-hop deep network was constructed and measured using Dust Network’s SmartMesh IP network. Each of the 100 motes generated and sent a data packet every 30 seconds with the expectation that each packet is received within 30 seconds latency (i.e. before the next packet from the same node is generated).

Figure 2.A deep hop network – the motes in gray are within range of mote 50

The deep network is constructed from real wireless devices in which seven devices (given IDs 1 through 7) communicate directly with the manager. Devices 8 through 10 communicate through the first seven motes, and the remaining devices (devices 11 through 101) are within range of the three IDs lower and higher. For example, device 50 is within range of devices 47, 48, 49, 51, 52 and 53. In this topology, the minimum number of transmissions (hops) required to reach device 101 is 32, although in practice most packets take more hops.

At the time this article was written, this network has been running continuously for 52 days. In total, 17 million data packets have been collected, requiring a sum of over 400 million individual transmissions, due to the hop depth and retries. Of the 17 million packets sent, none were lost, yielding a data reliability of 100%. About 25 thousand of these packets are “health reports” – diagnostic information sent periodically by the nodes.

Number of Nodes


Packet Generation Rate

~1 data packet / 30 seconds from each node

Hop Depth

32 hops

Number of Data Packet Sent

17 Million over 52 days

Number of Raw Individual Transmissions (including retries)

Over 400 Million

Packets Lost

0 (100% data reliability)

Table 1. Data Reliability of a Deep-Hop Network

Analyzing Latency & Current Consumption

Each packet is timestamped when it is generated at the sensor mote and again when received at the manager, so the latency of each packet can be monitored. Over a period of 90 minutes in this network, the data distribution is plotted in Figure 3. As expected, motes with the higher IDs, which are deeper in the network, have longer latency and more variation per packet, as the route options increase exponentially with depth. Despite this, data packets from the furthest mote (ID 101) all arrived at its destination in less than the targeted 30 second latency.

All motes keep an internal count of the battery charge consumed and report this information in periodic reports to the manager. From this information, the average current throughout the network can be plotted as in Figure 4. The motes with low ID numbers show the highest current consumption, because they carry the traffic from motes further away. But as can be seen, even the most heavily loaded routers in this 32-hop deep network had average current consumptions of a few hundred microamps. At such low current consumption, the routing nodes can be powered with a pair of lithium D-cell batteries yet last over 15 years.

Figure 3.Packet Latency- packets in a deep network are delivered reliably within the targeted 30 second latency


Figure 4.Average Mote Current – Even the most heavily loaded routers in this deep network consume only a few hundred microamps


SmartMesh IP networks, based on Time Synchronized Channel Hopping, routinely deliver >99.999% data reliability and very low power consumption in challenging applications. With 10-15 years of operation on reasonably small lithium batteries, wireless sensors can pragmatically placed anywhere, enabling real city-scale IoT applications.