How can wireless sensors enable the autonomous homes of tomorrow? By Christian Mauderer, System Design Engineer, Software and Hardware Development, embedded brains.
Home automation embodies the ‘convenience factor’ in new constructions and the ‘value add’ when modernising existing buildings. A typical application may be an automatic roller shutter control; the roller shutter closes in the summer under direct sun exposure in order to keep the heat out of the building but, if installed on the outside of the building, returns to its original resting position in windy conditions in order to prevent damage.
But a roller shutter is not the only factor contributing to a comfortable room temperature in modern buildings; much more attention must be placed on insulation. Insulation should be as seamless as possible, for example thermal bridges caused by cut-outs for power cables are frowned upon. This is the collision point between insulation and automation: to measure wind and sun exposure an exterior sensor is required, which is usually connected with cables and therefore creates thermal bridges.
Wireless sensor networks
The use of wireless sensor systems is one alternative, which has the advantage of not requiring a cable that cuts through the insulation. The installation is also much more flexible. One light sensor on each side and a wind sensor on the roof are very easy to install and require no additional cables. Depending on the size of the house, the indoor system requires only one or two central receivers.
Problems quickly arise, however, if one considers maintenance requirements. Power supply may be the greatest challenge. In the past, a reliable power supply was easily ensured via the power cable; now, the radio nodes must draw energy from another source. A battery is a common alternative. It is a perfectly acceptable solution for small buildings, where one or two nodes may require a battery change once every two years. This is a much more labour intensive process for larger buildings with many sensors. At a stated battery life span of two years, two nodes out of 50 would, on average, fail every month. Failures would of course increase toward the end of two years, so that all batteries could be replaced at the same time, but every single sensor node will need to be accessed in order to do this.
The maintenance issue can be reduced significantly if radio sensor nodes could gain the required energy from their environment, completely autonomously. The central nodes in the building then collect and evaluate the data.
Collecting energy from the environment - or ‘energy harvesting’ - is simple in principle; energy converters creating even small amounts of electricity are possible energy sources. Some classic examples for this are light or thermal variations, and PV cells or Peltier elements would be appropriate converters. Vibrations could also be used (via electro-mechanical converters). With the above mentioned wind or light sensors, energy may be directly derived from the measurement process. A wind sensor can derive energy from the rotation of an anemometer, if the load is relatively weak and fairly even. In case of a light sensor, the PV cell is an ideal energy source. Windless or dark times can be recognised merely by a temporary lack of data. Failures can be recognised by redundant sensors or a long-term lack of data. A light sensor should have enough energy - at least during daylight hours - to transmit data. And weeks of absolutely no wind are also fairly rare.
Depending on the energy source and respective light and wind conditions, harvesting sources often achieve a power output of only 10μW. Operating a wireless MCUsystem with this amount of energy is still a challenge, even if the collected energy is stored and then withdrawn in larger portions.
The radio standard is a decisive parameter because the majority of energy is required for the data transmission. While more and more universal radio standards are developed for low-energy applications (e.g. ZigBee) and existing standards are modified (e.g. Bluetooth Low Energy), most of them still require too much energy for a harvesting operation. The use of a radio protocol that is specifically tuned toward the respective application can achieve significantly better results.
In comparison to protocols using the 2.4GHz band (e.g. Bluetooth), the range can also be significantly improved with the use of a low frequency band of 868MHz, while maintaining the same transmission power. The downside of using a smaller bandwidth is not decisive for typical sensor applications as they create only very small data volumes. The problems commonly known from the 434MHz band, such as interference with other devices due to excessive load on the band, are highly unlikely for the 868MHz band due to more stringent access regulations. These access regulations usually limit the permitted maximum transmission time depending on the subband, but this limit poses no problem to sensor applications, since nothing is transmitted most of the time anyway. Beside the radio standard the MCU is the biggest consumer. But similar to the transmitter, it can remain in an energy-saving status most of the time and must only be activated for measurements or data preprocessing.
In automation applications, only the most current values are usually of interest. If measurement results must be allocated to specific times, this can easily be solved by an additional time stamp. Normally, this can be done with a timer operated from the controller’s internal clock. Energy can be saved here as well, if the internal clock - which is usually relatively fast and energy-hungry - is shut down in the standby phases and if an energy-saving low-power oscillator is used instead. This can reduce the controller’s standby current during such phases to less than 1µA.
As mentioned above, the protocol’s influence on energy consumption must not be ignored. Many protocols require long periods to send or receive data. Since there is usually insufficient energy for this, they are not ideally suited for radio nodes operated by energy harvesting.
In home automation systems, receivers are usually mains-powered and so are always ready to listen to new sensor messages. One solution could be for one nominated node to be supplied from either a battery, a large PV cell or a combination of several sources, so that is is always activated and able to forward data to the centre. All other nodes can then transmit data at a random time to improve energy management significantly.
The data packages for harvesting nodes must be temporarily stored on the node with a better power supply, in order to transmit controller information into the other direction. Data can then be transmitted once the harvesting nodes are active again. For this, the harvesting node must allocate a small slot in its receiver buffer after every message.
If systems are to be used in various locations, it is important that they are modular and easily adjustable. Specifying only one energy source, for example solar, is not suitable for wind sensors, as wind can occur also at night. Permanently installed sensors often satisfy demands only partially. A universal radio standard usually requires significantly more energy as additional administrative data is often generated even for unused functions. This requires additional fields in the header, which increase the data package and the transmission time required for it; therefore, the energy may not even suffice for one data package.
Compact sensor nodes
Using its experience in the development of radio modules and MCU systems of all sizes, embedded brains has developed a base platform for radio sensor nodes. The ELEPHANT (Extensible Low Energy Sensor Platform for Harvesting Applications and different Network Topologies) radio sensor node is based on an STM32 controller with an ARM Cortex-M core and an energy-saving radio chip. Energy supply can be adjusted to different harvesting sources as needed. With a universal connection for sensor systems, the system can be expanded in a modular manner. The base node can also be easily equipped with PC interfaces to allow for a seamless connection to a home automation system. It is therefore very well suited for classic home automation tasks as well as new, truly innovative applications.
The use of the Contiki operating system, which is specifically designed for sensor networks, allows the utilisation of a great variety of off-the-shelf software components and various protocols.