Very small networked sensors, known as motes, have many applications, including healthcare and structural monitoring of both military and civil aircraft and buildings. However, providing them with power is problematic. Energy harvesting is a technique for converting otherwise unused energy into an electrical form. Large examples of this are grid-connected wind turbines and photovoltaics. A well-known miniature example is the Seiko Kinetic watch. Improved wireless power supplies are increasingly sought after as electronic systems proliferate because they allow portability. Even for non-portable applications they can avoid the need for costly wired installations, particularly where sources of wired power are not locally available.
Batteries in their various forms have so far been the primary solution. But they frequently dictate a greater than desirable size, and sometimes cost, of the devices in question, and introduce an unwanted maintenance burden of replacement or recharging. Moreover, if ubiquitous computing with motes is to become a reality, a perpetual source of power is required. Three possible solutions are to use local energy supplies with higher capacity, to deliver power wirelessly from an active source introduced for this purpose, or to extract power from ambient sources in some way. This last option is the more attractive, in that it does not require the installation of any additional infrastructure.1
The energy-harvesting principles behind wind turbines and solar arrays can also be used in portable devices. The most successful applications to date include the self-winding watch and the solar-powered calculator. In addition to kinetic energy and light energy, thermal gradients and electromagnetic radiation could also be harvested to power small devices. However, the application in which energy-harvesting is to be used often dictates the types of energy that can be harvested. As an example, solar power could not be harvested in a body-implanted device. In this case, harvesting energy in the form of motion or vibration is the most universally applicable approach.2
The mechanism of the self-winding Seiko Kinetic watch is shown in Figure 1. An unstable mass swings on a shaft as the wrist of the wearer moves. This rotation is passed through a high gear ratio to turn a relatively conventional, although miniature, electromagnetic generator. The average power required by a watch is around 1μW. One of the major challenges facing researchers in this area is the further miniaturisation of the power supplies below the size of the wristwatch generator. Many pervasive computing applications, such as implantable medical sensors, would only become feasible if devices could reach sizes of no more than a few cubic millimetres. This miniaturisation requires a different approach from that taken by Seiko due to differences between the scaling laws of varying physical phenonena.
The Seiko Kinetic watch. (Courtesy Seiko Watch Corporation.)
An obvious example of differences in scaling is that volume scales as the cube of length (L) and area with the square. In the same way, electrostatic forces (between charges) tend to scale somewhere between L1 and L2, and electromagnetic forces (between currents) between L3 and L4. Consequently, as energy-harvesters get smaller, it is more advantageous to make them using electrostatic rather than electromagnetic transducers. This brings the additional advantage that the transducers are also compatible with MEMS (microelectromechanical systems) processing. The principle of the electrostatic transducer is shown in Figure 2. The force between the charges on the two electrodes allows the volume of the electric field to increase when the plates are separated at constant charge, thus causing an increase in stored electric potential energy.
The electrostatic transducer, showing the charges on the electrodes +qand −q, the electric field E, a constant voltage V and time varying voltage Vt.
Our work on energy-harvesting at Imperial has concentrated on this type of variable capacitance electrostatic device. Many typical inertial energy-harvesters rely on using a mass-spring resonant system to extract the maximum energy from a particular motion. However, these devices can have very limited bandwidth, meaning that they derive energy from vibrations of only a narrow range of frequencies. For sensors worn or implanted in the body, the motion which drives them tends to be erratic and rich in spectral content. Several work-arounds for this are possible. One is to produce a large bandwidth generator to harvest vibrational energy of a wide range of frequencies. The second is to produce a generator which can tune its resonant frequency—in other words, adapt itself to harvest better whatever frequency of vibrational energy is available—always to maximise the energy generated.
We have created an electrostatic constant charge generator harvester with an inherently wide bandwidth. A photograph of our finished device is shown in Figure 3. The device is non-resonant and makes use of a non-linear snapping action to generate power only when the input acceleration is a maximum. It is easy to tune because the damping force can easily be set by altering the precharge voltage. It is capable of generating 2.4μW when operated at a frequency of 20Hz with an excitation amplitude of 1.1mm.3 This amount of power generation is typical for generators reported by the research community. However, the amount of power generated is highly application-dependent,4 with machine vibration (e.g. motors) exhibiting vibrations which could produce over 1mW/cc. Commercial energy-harvesting generators have now begun to appear5 and have found applications in process plant monitoring.
MEMS electrostatic energy harvester fabricated at Imperial College London.
In summary, we are the first to make a broadband generator capable of being easily adapted during operation to successfully generate power from human body motion. The key was to realise that it would be possible to generate as much power from a constant charge generator as constant force. Future work on these devices will concentrate on miniaturisation and on-line optimisation so that the generators are able to adapt to different vibration environments.