12 July 2011
Power without responsibility: The challenges involved in energy harvesting
You can find out a lot about an electric motor simply by moving it. The vibrations produced don't just result in an irritating noise; they can also provide a handy guide to the motor's health. If you attach an intelligent sensor to the motor, you can maintain a constant watch on its health at little ongoing cost. The problem is powering them, even though they are close to a large energy source in the form of mains electricity.
You could provide wired power to the devices and use the same cable to receive data. But that presents the user with the problem of installing and managing the extra electrical cable. The combination of battery power and wireless communications makes installation cheaper, but batteries need replacing and, if you are not careful with the design, they need replacing frequently. Where you may have sensors strewn across a factory, simply finding their physical locations to replace lithium coin cells at the right time is more than a little problematic.
The motor's vibration provides another option: an energy source for a sensor node. It makes the device practically autonomous, especially if it has enough power to use wireless communication to relay data to a collecting computer.
As a result, energy harvesting and wireless communications tend to go hand in hand. The bad news is that it is tough to harvest enough energy from the environment, even to drive such supposedly low-power protocols as ZigBee. One of the most efficient vibration harvesters made to date occupies just 0.1cm^3, but generates, on average, less than 50µW. In receive mode, even a well designed ZigBee node can consume 30mW or more.
Physics tends to conspire against you in many of these applications, as the properties you want – such as compactness – are exactly the wrong ones for efficient energy harvesting.
Ideally, a small sensor node would use MEMS devices – micromachined cantilevers sound ideal – but their small size points to a high resonant frequency of vibration. Not only that, efficient harvesting implies the ability to resonate with a wide range of frequencies, but this is hard to achieve: a MEMS cantilever will typically only respond well to a narrow range of frequencies, although high vibrational forces will expand the useful bandwidth of a harvester.
Specialists such as Perpetuum have found the answer revolves around having a decent amount of mass to play with and focusing on the right frequencies, although some tolerance needs to be designed in to stop changes in loading from sending the target frequency out of range. One of the company's microgenerators, roughly the size of a small apple, concentrates on either 100Hz or 120Hz – twice the power supply frequency of large AC motors and a common source of vibrational energy.
Researchers at CEA-LETI claim to have found a way to encourage micromachined resonators to respond to frequencies in the hundreds of Hertz range, rather than the kiloHertz area. However, it still needs a seismic mass of 100g or more to couple enough vibration into the harvester. The team used a pattern of interdigitated fingers to build a structure that can harvest energy in two stages. When the structure vibrates, the capacitance between the fingers changes. If those capacitors are charged by a voltage, it is possible to extract energy from the rapid changes in the capacitance. The output in experiments was around 10µW per gram of seismic mass.
Like a self-winding watch, it would be useful to harness the vibration from everyday activities, such as walking, perhaps even breathing, to power body-worn sensors. But the body's vibrations are usually much lower in frequency and tougher to target. This makes the choice of power source for medical applications tougher.
Electromagnetic radiation is arguably the best source of environmental energy, witnessed by millions of square metres of solar panels. Unfortunately, it is hard to capture significant amounts of energy without a lot of real estate.
Perhaps the most widespread use of electromagnetic harvesting is the humble RFID tag – these absorb just enough energy from the field used to trigger them to power a simple response. The idea is now being extended to wireless transmitters that can power devices in the home or in the body, delivering the energy through megahertz or gigahertz frequency RF transmitters. For short-range applications, such as implantable sensors, gigahertz frequencies seem to offer the best trade offs between efficiency and size. For wireless loudspeakers and other consumer gadgets, the typical frequency is in the sub-10MHz range.
While researchers at Infineon Technologies dabbled in thermogenerators using doped polysilicon because of the material's low toxicity, the material of choice for thermogeneration remains bismuth telluride, which can make the most of what are typically tiny differences in temperature. Even at the hottest part on the body – around the neck – the difference between skin and clothes is usually only about 5K. Conversion efficiency with the best materials is fairly poor – generally less than 5%.
The thermogenerator works by placing p- and n-doped strips of semiconductor next to each other and wiring the load across them. You need multiple strips in parallel to achieve a usable quantity of energy, but some of the heat flowing through the strips is converted through the Seebeck effect, which causes charge carriers to diffuse from the hot to the cool end.
The Seebeck effect might seem to be a useful way to recover heat from electronic systems themselves. Unfortunately, the best efficiencies are achieved by limiting the flow of heat from the hotspot, so a converter makes a very poor heatsink. This means any converter needs to be some way from the electronics in need of cooling.
Other than RFID tags, the most widely used example of energy harvesting relies on pressure. EnOcean, for example, shipped close to 1million of its wireless switches by the end of 2009. These use a piezoelectric material to convert the pressure from someone pushing a switch into just enough energy to transmit a message to a nearby basestation.
Similarly, piezoelectric floor tiles are being trialled as a way to recover energy from the footsteps of passers-by. The East Japan Railway Company said a couple of years ago it would install an energy-converting floor in its Tokyo station. Israel-based Innowattech aims to recover energy from the pressure of the trains themselves to power nearby signs and lights. Late last year, the company installed a trial monitoring system near Haifa's main rail station.
EnOcean's approach to wireless in its switch illustrates the kind of thought process that goes into communication design for wireless energy-harvesting nodes. It sounds counter-intuitive, but the switches can only transmit because this minimises power consumption. Although the instantaneous power draw of a transmitter is usually higher than that of an active receiver, the long-term average consumption is usually much lower. The device only transmits when it has something to send and, in the case of the EnOcean devices, this is at the same time that energy is supplied to the circuit by someone's finger.
Having a system that can only transmit presents some configuration issues – it's hard to verify automatically that a switch is still working, but it may be the only choice. The problem with having a receive channel is that the node does not know when a message will turn up and simply has to keep listening. This makes it hard to power down and, realistically, a sensor node that uses energy harvesting generally has to operate on a very low duty cycle. It may use a large capacitor or supercapacitor to store energy harvested over a long period so that it has enough 'fuel in the tank' to perform computations and send a message when it has to.
Even techniques such as mesh networking that are intended to reduce power consumption can cause problems because, although they can make it possible to reduce transmission power, they demand that nodes listen for and relay data when they might be
running low on stored energy themselves.
A number of manufacturers have dealt with the problem by looking at the duty cycle of the wireless protocol itself, although this has demanded a move away from standards such as ZigBee which do not allow the necessary level of control. They may use IEEE802.15.4 for the basic transmission protocol and ZigBee-style packets, but then diverge from its standardised behaviour. In 2009, the ZigBee group announced that it was working on enhancements to support networks of energy-harvesting devices, but has yet to publish a specification.
Restric listening time
The best way to save on receive power is to restrict the amount of time the node has to be actively listening. ZigBee allows that through the beacon mode: it broadcasts a signal to tell nodes to start transmitting. If you can synchronise to that, it means the receiver can be shut down much of the time. However, there is only one such beacon signal which makes the trade off of responsiveness versus power efficiency difficult to decide. Ideally, you would have a combination of beacons or the ability to move the network into frequent and infrequent update modes.
As energy harvesting designs become more widespread, we will see further tension between accepted standards, created in a time when power was not important, and the needs of systems where every microWatt is critical.