Large-scale energy harvesting isn't new. The Swiss have run their whole country on melted snow-water running downhill for more than a century. Nearly half a million Hollanders live in a province called Flevoland, which was completely under water until 1930. Much of that work was done by the wind. A few weeks ago, I visited a saltworks near my home on San Francisco Bay. For over 100 years, sun and wind enabled commercial salt harvesting in the region. For millennia before that, the Ohlone people obtained their salt in much the same way, but on a smaller scale.
In fact, one could argue that the first farmers and pastoralists were really harvesters of solar energy. If you're looking for energy-harvesting technology, there are Web sites describing vertical-shaft windmills in Persia in the ninth century CE. They also describe water-wheels in India (fourth century BCE), Greece (first century BCE), and China (first century CE).
In a way, burning coal or petroleum is a form of energy harvesting—harvesting cretaceous sunlight, if you will, but there's the rub. It's getting continually more expensive to capture and exploit that ancient sunshine. This has led to fresh interest in energy sources other than fossil fuels, from reliable old wind and sunshine, to waves and tides, to all the little bits of energy that typically goes to waste—from the rumbling of a railroad car to the security lighting that discourages nighttime crime.The buzz these days is all about collecting millijoules (mJ) of energy wherever they can be found (which is more like gleaning or scavenging than harvesting), but even that isn't so new. My grandparents had a mantelpiece clock that wound itself by capturing changes in atmospheric pressure.
The latest wind turbines have rotor diameters spanning 160 feet, with the hub 250 feet above ground level (Fig. 1). An intriguing development in the engineering of these behemoths is their use of ultracapacitors. One critical function that must be dealt with is feathering the blades when winds get too strong (Fig. 2).
Blade-pitch motors could be battery-driven, but battery maintenance involves sending workers up the towers in fields of hundreds of towers— an expensive proposition in terms of steeplejack (one who climbs tall structures to do repairs) pay scales and liability insurance. Ultracapacitors last much longer than batteries and can tolerate greater temperature extremes, making them a more cost-effective energy storage device for blade-pitch control.
You may know all about photovoltaic solar. But let's look at what Sandia National Laboratories is doing with parabolic troughs, which use curved mirrors to focus sunlight on a tube that runs the length of the trough. Oil runs through the tube to heat it up, and it's then sent to a heat exchanger to generate steam for a conventional power plant.
There are some big troughs out there. The largest are in the Mojave Desert, near Barstow, Calif. Nine plants ranging in size from 14 to 80 MW generate 354 MW at peak output (Fig. 3). A middle-sized plant may comprise 10,000 modules, where each module has 20 mirrors.
What's new with these plants is mirror-alignment improvement. Sandia's latest work involves a pole with five cameras positioned along it--one for each of the four rows of mirrors and another to vertically center the pole with the trough module (Fig. 4). The actual images are then matched to a mathematically generated ideal image, and the mirrors are adjusted in real time to coax them into the ideal alignment.Microscale Harvesting
The real excitement in energy harvesting is on the microscale. All sorts of schemes can recover wasted energy, from photovoltaic cells on your laptop that recharge your battery overnight (using energy from continuously burning overhead lights) to monitors for bearing-wear in industrial machinery, trucks, and rail cars—monitors that power themselves from the very vibrations they monitor.
In the latter case, the microgenerators are made either from what I call "spring-mounted-slugs-inside-a-coil" or piezoelectrics. What's new are advances in the fabrication of materials. Another approach is represented by Advanced Cerametrics' (ACI) viscous suspension spinning process (VSSP), which adapts rayon-fiber technology for making ceramic fibers. As ACI explains it, the precursor of rayon is cellulose dissolved in caustic soda and water. This is mixed with a slurry of piezoelectric ceramic spun through a spinneret and baked.
A typical single piezoelectric fiber composite bimorph (PFCB) can generate voltages in the range of 400 V p-p with some forms reaching outputs of 4000 V p-p. In practice, with a 30-Hz vibration, ACI's piezo fibers have taken just 13 seconds to produce 1 J. If you're rusty on MKS units, one joule is the kinetic energy in a mass of 1 kg moving at 1 m/s.Practical Designs
Perpetuum's sophisticated microgenerators are based on proprietary magnetic circuits coupled to mechanical resonators. The PMG17 microgenerator targets stationary machines driven by ac motors in industrial plants (Fig. 5). They harvest the "twice-the-line-frequency" vibration that characterizes these motors (see www.perpetuum.co.uk for the application note and datasheet).
If the 700-g mass experiences even 25 mg of vibration energy within a bandwidth from 98 to 100 Hz (or 118 to 120 Hz for the U.S. version), the PMG17 will produce its minimum output power, which is 0.5 mW. Maximum output is 40 mW when things are really shaking, as in one whole g. The output voltage varies from 0 to 22 V ac, so Perpetuum assumes a load on the order of 12 kΩ. A plot in the datasheet shows power output for various vibration and frequency levels (Fig. 6).
To avoid confusion, be aware that "g" in the Perpetuum literature means the acceleration relative to standard earth gravity, not grams. It's helpful to express it that way because vibration in the machines being monitored would typically be measured with an accelerometer.
Slightly different is Perpetuum's PMG27. Designed for helicopters, it's tuned to a 17.2-Hz resonant frequency, which Perpetuum derived empirically. Its output voltage is Zener-limited to 7.5 V. The app note describes a PMG17powered wireless sensor node. It's built around a ZigBee (IEEE 802.15.4) radio based on a now superseded Atmel low-power ZigBee controller with an on-board 10-bit analog-to-digital converter and a Texas Instruments CC2420 ZigBee RF chip. Signal conditioning includes a biasing and gain stage for the accelerometer and an eighth-order active filter to remove aliasing artifacts and reduce noise.
Skipping over the signal-conditioning details in the app note, the new design element introduced by energy harvesting involves sizing the capacitance for the aluminum electrolytic energy-storage capacitor. The capacitor value is critical because the rate at which the system draws down the energy stored in the capacitor limits the rate at which readings can be transmitted (Fig. 7).
The table shows the power budget from the app note, which was developed by calculating the worst-case datasheet values for each stage as described below. It also indicates that the total requirement for Perpetuum's design is roughly 20 mJ. But designers must account for losses in converting the ac output of the microgenerator to the dc stored in the capacitor.
The app note designer chose to work with a switching regulator around a bridge rectifier and a Linear Technologies LTC1877, which has a quiescent drain of 10 µA and a sleep mode in which it draws less than 1 µA. It's capable of efficiencies up to 95% at high loads. But working from the Linear datasheet, Perpetuum's engineers conservatively estimated approximately 75% conversion efficiency at real-life loads. Therefore, the total stored energy requirement for the capacitor was reckoned to be 27 mJ.
The bridge rectifier's design uses ON Semiconductor BAS16 diodes because of their low forward-voltage drop and low reverse leakage. The Zener and the series FET help the storage capacitor charge more quickly when there's little vibration energy to be harvested. The FET keeps that big capacitance offline until the three 22-µF capacitors that form the ripple-filter charge up. This allows VCC to establish itself approximately 30 times sooner than otherwise.
The app note also describes an auxiliary LED diagnostics circuit that provides a useful diagnostic aid at the cost of a marginal percent increase in charging time. The other necessary design inputs are the initial and final capacitor voltages. Then the required capacitance, C, equals twice the required energy in joules divided by ΔV.The PMG17's output is clamped at approximately 11 V p-p by internal Zener diodes. After rectification, the voltage at the output of the bridge would be approximately 10 V peak into an open circuit.
To avoid any issues associated with leakage of the microgenerator's internal Zeners, 8 V was conservatively chosen as the upper voltage. To avoid similar issues with the power supply dropping out of regulation, 3.2 V was assumed to be the lower voltage limit.
Using 27 mJ for the energy requirement and ΔV = 8 – 3.2 V yields 1125 µF. Sizing up to allow for temperature coefficients and manufacturing tolerance, this could be either a standard 1500- or 2200 µF aluminum electrolytic. In the app note, the latter was chosen. For anyone interested in the complete design, the app note contains complete schematics and a bill of materials.Off-The-Shelf Modules
Advanced Linear Devices offers complete energy-harvesting modules that use the company's EPAD floating-gate MOSFETs to meter the energy going into and out of the module (Fig. 8). Those floating gates pre-program the MOSFETs' threshold voltages at the factory. The input device lets a module start charging at essentially 0 V, and the output is limited to a range of either 3.6 to 1.8 V or 5.2 to 3.1 V, depending on the module. So for a nominal 3.3- or 5-V rail, the output of the module is always within the standard operating-voltage range of the microcontroller and associated circuitry selected for the application.