Electronic Design

The Field Of Energy Harvesting Begins To Ripen

Case histories show how the pieces of the energy-harvesting puzzle fit together.

One cannot talk about energy harvesters without discussing wireless mesh networks, sensor batteries (particularly thinfilm batteries), and supercapacitors—along with concepts of power management—nearly in the same breath. Harvesting is a complex and evolving discipline that promises rewards and challenges for engineers who want to take existing skills in new directions.

Most of the technical background information in this report was derived from interviews with companies that presented papers at the NanoPower Forum put on by the Darnell Group, a market analysis organization, in June in Costa Mesa, Calif. For a top-down look at applications, see “Energy-Harvesting Critical Success Factors.”

QUESTIONS OF SCALE
Before anything else, the terms “nanopower” and “harvesting” need to be sorted out. “Harvesting” gets applied indiscriminately to things as diverse as grid-tied solar systems and patient-powered heart monitors.

In one way, “harvesting” sounds like big combines and threshers working through vast fields, collecting tons of produce. Photovoltaic (PV) and geothermal energy harvesting fit that description. In another way, it’s more like gleaning—following after the threshers and collecting what’s been passed over.

Either way, collecting the energy is only a small part of the picture. You then have to store it, which involves power density versus energy density considerations in the storage medium, along with equivalent series resistance (ESR) and charge/discharge characteristics. That, in turn, leads to considerations of power management—not just in terms of how you run the application, but in terms of how you husband those electrons you’ve harvested or gleaned.

On the large scale, harvesting that power management would be something like maximum power-point tracking, But for this article, we’re focusing on the small-scale gleaning companies spotlighted at the NanoPower Forum.

A CASE HISTORY
I don’t know of any explicit design examples of smallscale energy-harvesting systems that are as thorough as what Charles Lakeman of TPL’s Micropower Division presented at the forum, so I’ve adapted that here for its instructional value. Lakeman described a product called EnerPak that combines smart, ultra-low-power charge management circuitry and electrochemical energy storage (Fig. 1).

As Lakeman described the design problem that’s facing the engineer, any wireless sensing application, a class that embraces most of the things people are trying to do with small-scale energy-harvesting today, has three basic modes: data collection, data communication, and idle (sleep) modes. The power demands for each of these modes are significantly different. The default sleep mode draws perhaps a few microwatts. Sense and compute functions draw a few tens of milliwatts or less, while wireless data transmission can require several hundreds of milliwatts, but only in bursts.

To accommodate these disparate power needs, designers usually simply design for a battery that is capable of handling the system’s highest power demands—those for data transmission. That’s not such a good idea, Lakeman said, as it leads to selecting a battery that’s oversized for most of the operational lifetime and capabilities of the system. It’s better to combine a smaller battery with a supercapacitor.

In that synergistic pairing, the supercapacitor delivers energy efficiently. It exhibits high specific power, which allows it to supply the radio (or wireless mesh network node) when it needs to transmit, while the battery stores energy efficiently and provides backup when the harvester isn’t providing enough power. Using a low-impedance supercapacitor as the primary energy-delivery device is much more efficient than oversizing the battery.

Then there’s power management. In the EnerPak, an ultralow- power TI MSP43 microcontroller (MCU) monitors the state of charge of both the battery and supercapacitor. Simultaneously, it dynamically adjusts the operation of the charging module to accommodate any fluctuations in the level of energy delivered by the harvester. Should the incoming energy not be sufficient to recharge the supercapacitors (e.g., in a solar-powered system at night), the MCU switches in the battery. There’s also some IP in the MCU.

“Because energy harvesters only produce very small amounts of power, this circuitry has been designed to operate extremely efficiently to transfer as much of the available power as possible to the energy storage devices without wasting it in the charger,” Lakeman said.

Continue on Page 2

CHARACTERIZATION DATA
Part of the engineering involved in a field this new is determining the operating characteristics of the product. This is more of what Lakeman presented at the forum. In a test of the system’s operation over a period of roughly 68 hours, it charged the battery when there was sufficient incoming energy, e.g., the positive voltage spikes and an overall increase in battery voltage (Fig. 2).

When the PV energy harvester couldn’t collect enough ambient energy, the MCU detected that there was insufficient incoming energy to refresh the supercapacitors and instead used the battery to maintain the output voltage. With the power-management software, even under conditions of diurnal ambient light flux on the PV array, it was still possible to maintain the output voltage, supply current to an external load, and charge the backup battery.

In other tests, TPL characterized EnerPak performance to assess how rapidly it could recover from simulated pulse loads of different levels under various conditions of solar flux. Measurements were carried out using a 1- by 2-in. silicon PV array illuminated with an array of incandescent bulbs. Illumination levels of 1 to 30 klx (corresponding to dull overcast to bright sunny conditions) were used. At 3.3 klx, the PV array delivers approximately 2.5 mA at 1.8 V, or 4.5 mW.

The table shows the time required for the EnerPak to recover its programmed voltage after being subjected to a series of short loads that simulated the transmission demands of a wireless mesh network mote sending bursts of data. Recovery times could be as short as eight seconds on a bright day with a light (50 mW-s) pulse load to as long as 46 minutes on dull overcast days with the heaviest (500 mW-s) pulse loads. Depending on the pulse level, the system’s efficiency in replacing the delivered energy varied from 56% to better than 95%.

To gather further empirical data, TPL bench-tested the Ener- Pak with an actual ZigBee mote, a Crossbow Technology Mica2 wireless sensor radio platform. At 3.2 V, the Mica2 mote consumes 13 mA in transmit mode, 3.8 mA in sense mode, and 267 µA in sleep mode.

The mote was programmed to query the sensor and measure the battery voltage every two seconds and transmit the accumulated data every five sense cycles (10 seconds). At the end of transmission, the processor switched off the radio, and the mote re-entered deep sleep mode.

In the tests with the Mica2, the system used two 1- by 2-in. PV arrays to accommodate the mote’s relatively high average power draw made necessary by the high sampling and data-transmission rates. Figure 3 shows the output voltage from EnerPak under this regime and under different input light levels.

The periodic load profile is evident—four small voltage drops are followed by a larger one corresponding to the transmit pulse. The profile of the system maintaining the output voltage is superimposed on top of these features (bearing in mind the constant current drain in sleep mode).

At low illumination levels, it takes between 50 and 70 seconds to refresh the voltage from a low of roughly 3.2 V to 3.23 V. At higher illumination, this refresh time is on the order of eight seconds.

While the EnerPak examples harvest solar energy, other applications look to recovering vibrational energy. In October, a presentation offered by EoPlex Technologies at the Electronic Design Group’s One Powerful Day, a virtual power-technology series of seminars that was held online, discussed a manufacturing method for piezo-beam vibration harvesters for next-generation, automotive tire-pressure monitoring systems.

By way of design examples, I’ve previously reported on vibrating magnet/spring transducers already being used in watertreatment plants (see “Energy Harvesting Gets Big—And Small” at, ED Online 15844). They monitor and report on the condition of pump bearings and on proof-of-concept testing related to highway structures and railway rolling stock. The critical factor about vibration harvesting in these apps should be fairly obvious—you don’t get much energy transfer unless the system is at resonance.

To illustrate, consider the resonance characteristic from the datasheet for Perpetuum’s PMG17-120 (Fig. 4), which is tuned for applications associated with pumps driven by electric motors in North America that run on 60-Hz power. Typical vibration is about twice the line frequency, and that tunes the magnetic slug and spring system inside the harvester. At the forum, Perpetuum and supercapacitor maker CAP-XX updated the audience about ongoing projects in the U.K. at watertreatment plants (see “Energy Harvester Perpetually Powers Wireless Sensors,” ED Online 20033, and “Ultracapacitors Branch Out Into Wider Markets,” ED Online 20034Ultracapacitors-Branch-Out-Into-Wider-Markets20034).

Continue on Page 3

Perpetuum also makes a PMG27 for helicopters. Based on analyses of the complex vibration patterns of vibrations during typical helicopter missions, it’s tuned to a 17.2-Hz resonant frequency. At the forum, a group from the University of Bristol presented a dedicated self-powered helicopter system.

Some of the most interesting talks at the NanoPower conference dealt with thermal energy harvesting. As with vibrational harvesting, I was struck by another reminder from those basic mechanical engineering classes—in this case a bit of simple thermodynamics.

A Peltier or other thermoelectric device is a heat engine. The heat difference across the junction depends on the head flux through it, which implies the necessity of getting rid of heat on the hot side. Moreover, maximum efficiency for any ?T is never going to be better than the efficiency of a Carnot cycle. Within those parameters, there still appears to be a lot of promise for patient-powered biomedical devices.

GETTING COMPLICATED
So far, all of this may seem a little too basic. But much greater sophistication is certainly out there as well. According to Ferro Solutions’ chief scientist, MIT’s Bob O’Handley (who is one of the go-to guys for magnetostriction), when you sandwich piezos between magnetostrictive layers and pre-stress them with a field, interesting things start to happen.

Other examples of blue-sky research were in evidence at the Darnell conference. IMEC Nederland reviewed research on body-powered and PV-powered patientmonitoring medical applications. A team from the National University of Singapore presented a paper on powerline harvesting, while another team from the University of Colorado at Boulder discussed rectennas and far-field harvesting.

Reporting from New Mexico, TPL evaluated a number of configurations of piezoelectric cantilevers for use with the EnerPak. Figure 5 illustrates the I-V performance TPL observed from a conventional bimorph (bends in two directions) with a metal shim and two multilayer unimorphs with different layer thicknesses.

While the multilayer devices were susceptible to fracture under high loads (low frequency, high acceleration), their performance was noticeably superior to that of the bimorphs at low acceleration values. On the other hand, at high amplitudes, bimorphs delivered voltages that significantly exceed the input voltage level of the control electronics, although at low current values.

Hide comments

Comments

  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
Publish