An AA-Sized Vibration- Based Microgenerator for Wireless · PDF fileAn AA-Sized...

9
64 PERVASIVE computing Published by the IEEE Computer Society 1536-1268/07/$25.00 © 2007 IEEE An AA-Sized Vibration- Based Microgenerator for Wireless Sensors C hemical batteries power most mod- ern, portable electronic devices. Such batteries require replacement or charging once consumed. So, nonex- haustive power sources would be more convenient. One readily available option is vibration energy. Even though the power from miniature vibration-based generators is very small, improvements in integrated circuit (IC) technol- ogy have made miniaturized vibration-powered systems feasible. Vibration-powered wireless sen- sors obtain power from ma- chine vibrations, human move- ment, or other forms of motion. These sensors could soon play an important role, particularly in monitoring applications in which battery replacement is inconvenient. Eventually, if such systems’ effi- ciency continually improves, vibration might be able to provide power for mainstream pervasive computing devices. Here, we demonstrate the feasibility of incor- porating micro power transducers (MPTs) with a voltage multiplier and rectifier to make a micro power generator (MPG) that is the same size and shape as an AA battery. The AA-sized module includes a voltage multiplier and a large capaci- tor to produce the DC output. To find the most energy-efficient voltage multiplier to use with this MPG, we use PSpice simulations to analyze the performance of a doubler, a tripler, and a quadru- pler circuit in terms of input power, energy effi- ciency, and charge time. A start-up circuit and a regulator, which are external to the MPG module, apply regulated power to an application circuit after the MPG’s voltage output exceeds a preset threshold. We’ve used the MPG with the start-up circuit and the regulator in a vibration-powered wire- less thermometer system, which can transmit temperature measurements every 20 seconds when continuous vibrations with an input acceleration of 4.63 m/s 2 and a frequency of 70.5 Hz are present. We chose the combination of storage capacitor size and comparator threshold voltage so that the regulator applies power to the application circuit when sufficient energy is available for reliable operation. When the vibration is insufficient to drive the appli- cation circuit, the start-up circuit gracefully shuts down this application circuit; it resumes when the storage capacitor has sufficient energy for operation. The application has a total power consumption of 27.6 W, which the MPG com- pletely delivers. To the best of our knowledge, our MPG is the smallest vibration-to-electrical transducer that can power off-the-shelf ICs. It weighs 10.5 g, about half the weight of a typi- cal alkaline cell. (The “Comparison with Other Induction-Based Microgenerators” sidebar compares this MPG with other induction-based implementations.) Micro power transducers The MPG consists of two MPTs and a voltage multiplier circuit. You can connect the MPTs in series or in parallel, depending on whether you This vibration-to-electrical transducer has an AA-size form factor and generates a DC voltage that can power off-the-shelf integrated circuits. POWER GENERATION Steve C.L. Yuen, Johnny M.H. Lee, Wen J. Li, and Philip H.W. Leong Chinese University of Hong Kong

Transcript of An AA-Sized Vibration- Based Microgenerator for Wireless · PDF fileAn AA-Sized...

64 PERVASIVEcomputing Published by the IEEE Computer Society ■ 1536-1268/07/$25.00 © 2007 IEEE

An AA-Sized Vibration-Based Microgeneratorfor Wireless Sensors

Chemical batteries power most mod-ern, portable electronic devices. Suchbatteries require replacement orcharging once consumed. So, nonex-haustive power sources would be

more convenient. One readily available option isvibration energy. Even though the power fromminiature vibration-based generators is very small,improvements in integrated circuit (IC) technol-ogy have made miniaturized vibration-poweredsystems feasible. Vibration-powered wireless sen-

sors obtain power from ma-chine vibrations, human move-ment, or other forms of motion.These sensors could soon playan important role, particularlyin monitoring applications inwhich battery replacement is

inconvenient. Eventually, if such systems’ effi-ciency continually improves, vibration might beable to provide power for mainstream pervasivecomputing devices.

Here, we demonstrate the feasibility of incor-porating micro power transducers (MPTs) with avoltage multiplier and rectifier to make a micropower generator (MPG) that is the same size andshape as an AA battery. The AA-sized moduleincludes a voltage multiplier and a large capaci-tor to produce the DC output. To find the mostenergy-efficient voltage multiplier to use with thisMPG, we use PSpice simulations to analyze theperformance of a doubler, a tripler, and a quadru-pler circuit in terms of input power, energy effi-ciency, and charge time. A start-up circuit and a

regulator, which are external to the MPG module,apply regulated power to an application circuitafter the MPG’s voltage output exceeds a presetthreshold.

We’ve used the MPG with the start-up circuitand the regulator in a vibration-powered wire-less thermometer system, which can transmittemperature measurements every 20 secondswhen continuous vibrations with an inputacceleration of 4.63 m/s2 and a frequency of70.5 Hz are present. We chose the combinationof storage capacitor size and comparatorthreshold voltage so that the regulator appliespower to the application circuit when sufficientenergy is available for reliable operation. Whenthe vibration is insufficient to drive the appli-cation circuit, the start-up circuit gracefullyshuts down this application circuit; it resumeswhen the storage capacitor has sufficient energyfor operation. The application has a total powerconsumption of 27.6 �W, which the MPG com-pletely delivers. To the best of our knowledge,our MPG is the smallest vibration-to-electricaltransducer that can power off-the-shelf ICs. Itweighs 10.5 g, about half the weight of a typi-cal alkaline cell. (The “Comparison with OtherInduction-Based Microgenerators” sidebarcompares this MPG with other induction-basedimplementations.)

Micro power transducersThe MPG consists of two MPTs and a voltage

multiplier circuit. You can connect the MPTs inseries or in parallel, depending on whether you

This vibration-to-electrical transducer has an AA-size form factor andgenerates a DC voltage that can power off-the-shelf integrated circuits.

P O W E R G E N E R A T I O N

Steve C.L. Yuen, Johnny M.H. Lee,Wen J. Li, and Philip H.W. Leong Chinese University of Hong Kong

want a higher output current or voltage.The MPT is the key component to con-vert ambient mechanical energy fromvibration energy to electrical energy.1 Itcontains an inner housing, a microelec-tromechanical systems (MEMS) springwith spring constant k, an N45-gradingrare-earth permanent magnet with massm and magnetic field intensity B, and acopper coil of length l. The inner hous-ing secures the spring with an attachedmagnet. Figure 1a shows the key com-ponents. Figure 1b shows the outer andinner housing, the magnet, and the res-onating spring; the principle is similar tothat of a shake-driven flashlight.

The MPT generates AC power whenthe system vibrates. If the generatorhousing vibrates with amplitude Yt, themagnet vibrates with amplitude Zt. Therelative movement between these twovibrating parts causes magnetic flux lines

to cut through the coil, inducing voltagein the coil loop according to Faraday’slaw of electromagnetic induction. Thesystem’s average power output P is

(1)

where �e = (Bl)2/(2Rm�n) is the electricaldamping factor, R is the load resistance,Y0 is the input vibration amplitude, � isthe input vibration angular frequency,and �n is the spring-mass system’s nat-ural resonance frequency. � = �m + �e isthe system damping factor, �m = d/(2m�n)is the mechanical damping factor, and dis the mechanical damping ratio.2 At res-onance, maximizing the average power

and voltage output gives

(2)

(3)For an MPT using an electroplated cop-

per spring, some approximate typicalparameters include spring constant k = 40N/m, N45 magnet with mass 192 mg,first-mode resonance frequency f1 = 72Hz = 456 radians, damping ratio d = 0.01,system damping ratio � = 0.723, magneticfield intensity B = 0.36 tesla (3,600 gauss),coil length l = 30 m (� 1,700 turns), andinput vibration amplitude Y0 = 200 �m.

The MPT’s key design issue is how tocontrol the resonance frequency via thespring shape and material. After study-ing various materials, we found that cop-per was best because it has a relatively

VBlY

= 0

2

ωξ

n

Pm Y

=ξ ω

ξe n0

2 3

24

Pm Y

=

⎝⎜⎞

⎠⎟

−⎛

⎝⎜⎞

⎠⎟⎡

⎢⎢

⎥⎥ +

ξ ωω

ω

ωω

en

n

02

3

3

22

1 2ξξ ωωn

⎝⎜⎞

⎠⎟

2

JANUARY–MARCH 2007 PERVASIVEcomputing 65

C ompared to the previously reported induction-based imple-

mentations of C.B. Williams and his colleagues1 or Rajeevah

Amirtharajah and Anantha Chandrakasan,2 our system achieves

3.5 to 20 times better power density. Also, it’s possible to manu-

facture our mechanical resonating springs using an electroplating

technique compatible with microelectromechanical systems. This

feature has an advantage over commercial systems such as Perpet-

uum and Ferro Solutions for volume manufacturing and enables

precise tuning of mechanical resonating frequencies. When com-

bined with a voltage multiplier and AC-to-DC conversion circuitry,

the mechanical resonator fits in a standard AA-sized form factor

package for vibration-to-electrical power conversion. A standard

2,500 mA-hr, 1.5 V AA battery could operate circuitry drawing

27.6 �W for approximately 15 years, so this type of technology is

best suited for low-power applications where battery replacement

is difficult.

Shad Roundy developed piezoelectric generators,3 which pro-

vide better power density and higher voltages. Roundy used his

generator to drive the University of California, Berkeley’s MICA mote

node (a mote is a low-power computing node), but turning on the

MICA node’s power switch for first-time activation requires human

intervention. In our micro power generator, the start-up circuit

handles this electronically. Furthermore, a continuous-vibration

source must be available to operate Roundy’s generator, whereas

ours degrades gracefully when the voltage is insufficient. Although

Nathan Schenck and Joseph Paradiso have developed a scheme

that includes a start-up circuit for a piezoelectric-based gener-

ator,4 our start-up circuit uses hysteresis and the designer can pre-

cisely set the turn-on and turn-off points via resistors.

REFERENCES

1. C.B. Williams et al., “Development of an Electromagnetic Microgenera-tor,” IEE Proc. Circuits, Devices and Systems, vol. 148, no. 6, 2001, pp.337–342.

2. R. Amirtharajah and A.P. Chandrakasan, “Self-Powered Signal Process-ing Using Vibration-Based Power Generation,” IEEE J. Solid-State Circuits,vol. 33, no. 5, 1998, pp. 687–695.

3. S.J. Roundy, Energy Scavenging for Wireless Sensor Nodes with a Focus onVibration to Electrical Conversion, doctoral dissertation, Dept. of Me-chanical Eng., Univ. of California, Berkeley, 2003; http://engnet.anu.edu.au/DEpeople/Shad.Roundy/paper/ShadThesis.pdf.

4. N.S. Schenck and J.A. Paradiso, “Energy Scavenging with Shoe-Mounted Piezoelectrics,” IEEE Micro, vol. 21, no. 3, 2001, pp. 30–42.

Comparison with Other Induction-BasedMicrogenerators

low Young’s modulus and a high yieldstress compared to silicon. Brass, tita-nium, and 55-Ni-45-Ti are alternativesfor different applications; for example,55-Ni-45-Ti would be suitable for situ-ations requiring a lower resonance fre-quency. We selected a resonance fre-quency of 70 Hz on the basis of theavailable materials, their mechanicalstrength, the packaging-size constraints,and our desire to optimize the MPT’spower density through equation 3. Pos-sible vibration sources with a frequencyof approximately 70 Hz include the baseof a 5-horsepower, three-axis machinetool with a 36-inch bed (10 m/s2 maxi-mum acceleration), high-volume air con-

ditioning (HVAC) vents in an officebuilding (1.5 m/s2), and a notebook com-puter reading a CD (0.6 m/s2).3 An exam-ple of an application consistent with thesedesign parameters is monitoring a ma-chine tool’s operating temperature.

We used a high-aspect-ratio MEMSelectroplating technique to manufac-ture a spiral spring. With this process,you can make springs as thick as 150�m. First, we put a gold layer on thesubstrate; this layer acts as a conductingseed layer for the copper electroplatingstation. We then used a lithographictechnique to secure an SU-8 negativephotoresist mold on the gold layer.After electroplating the copper, we re-

leased the spring from the substrate.This lithographic electroplating tech-nique can produce springs that are thin-ner and have smoother edges than thoseproduced through laser cutting, result-ing in a lower, more precise resonancefrequency. A lithographic process is alsomore suitable for mass production be-cause you can make many springs si-multaneously in a single batch. Figure 2shows scanning electron microscopeimages of an electroplated spring.

Voltage multiplierEach MPT’s AC output voltage, Vrms =

450 mV, can’t directly drive a conven-tional digital circuit. Therefore, weintroduced a voltage multiplier to stepup and rectify the two MPTs’ AC out-puts (connected in series) to producea DC output.4 Figure 3 shows circuitdiagrams for the three multipliers weanalyzed.

We built a prototype circuit using 10 �F, 10 V capacitors (KEMET typeT491) and Toshiba 1SS374 silicon epi-taxial Schottky barrier diodes. Wechose the 1SS374 diode because it hasa low forward voltage (0.23 V), and wechose the capacitor for its size and lowleakage. A 1,000 �F Panasonic FCseries aluminum electrolytic capacitor(diameter = 10 mm, height = 12 mm)stores the MPG-generated electricalenergy and smooths out the voltagemultiplier’s output.

Start-up circuitThe start-up circuit applies power to

an application circuit only after the

66 PERVASIVEcomputing www.computer.org/pervasive

P O W E R G E N E R A T I O N

16 mm

Powermanagement

circuit

Micropower

transducer

11 mm 11 mm

48 mm

14 mm

(a)

(b)

Innerhousing

Magnet

Spring

Figure 1. Micro power generator: (a) the power management circuit(which consists of a voltage multiplierand a storage capacitor) and the micropower transducer (with the magnet inthe center), and an assembled AA-sizedMPG containing the power managementcircuit and two MPTs; and (b) the micropower transducer’s inner structure, showing the coil wound around the innerhousing.

MPG’s voltage output exceeds a presethigh threshold voltage, Vth(H). Withoutthe start-up circuit, the application cir-cuit would begin to operate and con-sume a large current well before theMPG’s output reached the minimumvoltage required for correct operation.The voltage multiplier’s output voltagethen wouldn’t be able to continue rising.

The start-up circuit in figure 4 is acomparator with hysteresis. The com-parator’s negative reference voltage,along with the three resistors connectedto the comparator’s positive input, setsits hysteresis—that is, the values ofhigh threshold voltage Vth(H) and lowthreshold voltage Vth(L).5 The start-upcircuit’s output switches the NMOStransistor on when the supply voltageexceeds Vth(H) and switches that tran-sistor off when the supply voltagedrops below Vth(L). When that tran-sistor turns on, the start-up circuit pro-vides a return path between the appli-cation circuit’s system ground and thepower ground.

Figure 5 shows the start-up transientresponse of the start-up circuit and thesystem supply. The start-up circuit’s in-put capacitor must be large enough toprovide sufficient energy for at least one

activation of the application circuit. Forexample, a wireless RF thermometer(described later) consumed 212.62 �Jper activation. So, a 1,000 �F capacitoris sufficient, because the amount of en-ergy available in a charged 1,000 �Fcapacitor with Vth(L) = 1.4 V and Vth(H) =1.9 V is

(4)

MPG modelWe tested the MPG and the start-up

circuit using a vibration drum and asignal generator. The vibration fre-quency was 70.5 Hz, and we measuredthe input acceleration as approximately4.63 m/s2 using an accelerometer at-tached to the vibration drum. The am-

plitude, according to a laser vibrome-ter, was 250 �m. Using a stroboscopeand an oscilloscope, we observed twotransducers. The system generated themost energy when the magnets in theMPTs experienced both translationaland rotational vibration. This vibra-tion mode provides the highest rate ofchange in magnetic flux.

Using varying resistive loads, we mea-sured the voltage-current characteristicsof two coils in series. The output powerwas 20 to 120 �W for 1 k� to 30 k�loads, and the power density of the twoMPTs was 53.1 �W/cm3.

Table 1 compares different microgen-erator designs and technologies. Our sys-tem generally has better power densitythan other electromagnetic approaches.The MEMS capacitor and piezoelectricapproaches have better power density.We modeled two coils as an AC source,with internal resistance Rint = 1,030�.Using this model, we performed severalPSpice simulations with different volt-age multipliers (a doubler, a tripler, anda quadrupler) to find the one that would

E iv dt

Cdvdt

v dt

C v dvV

V

=

=⎛⎝⎜

⎞⎠⎟

=

0

0

th

th

( )

(

L

H))

th th( ) ( )H L

J

= ⎡⎣ ⎤⎦ − ⎡⎣ ⎤⎦⎧⎨⎩⎪

⎫⎬⎭⎪

=

CV V

2

825

2 2

μ

JANUARY–MARCH 2007 PERVASIVEcomputing 67

Figure 2. Images of an electroplatedspring, from a scanning electronmicroscope: (a) low and (b) high magnification.

(b)(a)

(c)(a) (b)

Doubler 1

ACinput

ACinput

ACinput

1 C1 2

0

D2

D2

C1

C3

C3

D3D3

D1

D2

D1C2

C1

D1

C2C43

Vout

Vout

C2 D4 Vout

Doubler 2

Figure 3. Schematics of a voltage (a) doubler, (b) tripler, and (c) quadrupler.

best meet our output voltage and powerrequirements.

The following performance metrics canhelp quantify the converters’ performance:

• Available stored energy. The MPGdoesn’t usually generate enough powerfor continuous use. It stores inputenergy on a storage capacitor and re-leases it as necessary. This method is theduty-cycled approach. Leakage issuesaside, you can apply the MPG to anyoff-the-shelf circuit with the start-up cir-cuit, provided that circuit’s averagepower consumption is lower than theMPG’s and the storage capacitor is large

enough. The start-up circuit controls theamount of energy stored in storagecapacitor Cstorage through Vth(H) andVth(L). Available stored energy is thestored energy available to the applica-tion circuit; we calculate ASE usingequation 4.

• Start-up time. ST is the elapsed timerequired to charge the storage capac-itor’s potential from 0 to Vth(H) usingthe MPG.

• Recharge time. RT is the elapsed timerequired to charge the storage capac-itor from Vth(L) to Vth(H) using theMPG.

• Average input power. Input energy(IE) is necessary to charge Cstorage

from Vth(L) to Vth(H). Average inputpower is the input energy per unitcharge time. Mathematically, AIP =IE/RT. Lower AIPs are desirable.

• Energy efficiency. EE is the availablestored energy per unit of input energy.Quantitatively, EE = (ASE/IE) � 100%.Higher EE means better performance.

• Output load (capacitive and resis-

tive). The voltage multiplier’s outputconnects to a storage capacitor and aresistive load. A duty-cycled ap-proach requires a large storage capac-itor, typically 1,000 �F, so the load ispredominantly capacitive. The resis-tive load represents the load from thestart-up circuit and the applicationcircuit in shutdown (sleep) mode.Typical loads are small, in the rangeof 50 to 500 k� (equivalent to thestart-up circuit’s load). We didn’tinclude the energy dissipated in theresistive load when calculating theoutput power.

We performed several PSpice simula-tions with a test voltage multiplier,source resistance Rint, and an AC source(Vp–p = 2.4 V, f = 70.6 Hz), choosing theparameters that best represent the actualMPG. The voltage multiplier’s storagecapacitor was 1,000 �F. We used threepredefined groups of threshold voltagesVth(L) and Vth(H): (1.0 V, 1.4 V), (1.4V, 1.8 V), and (1.8 V, 2.2 V), with ASE of0.48 mJ, 0.64 mJ, and 0.8 mJ, respec-tively. Using PSpice, we compared volt-

68 PERVASIVEcomputing www.computer.org/pervasive

P O W E R G E N E R A T I O N

Tripler output

1.2MΩ

1.2 MΩ

1.5MΩ

1,000μF

1.8MΩ

349kΩ Application

circuit

Vref

Figure 4. The start-up circuit. (Referencevoltage Vref sets the circuit’s thresholdvoltage. The voltage multiplier here is atripler.)

TABLE 1A comparison of different vibration-to-electrical transducers.

Input amplitudeor acceleration, Power output Power and vibration (power, output density

Design frequency voltage Vout) Volume (cm3) (�W/cm3) Technology

Williams et al.6 — — — 10–15 Electromagnetic

Amirtharajah and 2 cm, 2 Hz 400 �W, 160 2.5 ElectromagneticChandrakasan7 180 mV rms

(root-mean-square)

Meninger et al.8 2.52 kHz 8.6 �W 0.075 114.6 Microelectro-mechanical system capacitor

Roundy’s first 2.25 m/s2, 85 Hz 207 �W, 1 207 Piezoelectricdesign3 12 V DC

Roundy’s third 2.25 m/s2, 85 Hz 1,700 �W, 5.1 335 Piezoelectricdesign3 12 V DC

Two micro power 4.63 m/s2, 80 Hz 120 �W, 2.262 53 Electromagnetictransducers 900 �V rmsin series

age multipliers in terms of average inputpower, energy efficiency, start-up time,and recharge time for different resistiveloads. In general, low-order voltage mul-tipliers have a lower input power thanhigher-order voltage multipliers. Forexample, for the same input energystored on the storage capacitor, a dou-bler requires less input energy than a trip-

ler or a quadrupler. However, the dou-bler can’t achieve Vth(H) � 1.8 V underload. As Figure 6 shows, the tripler ismore efficient than the quadrupler forconverting power but takes slightly moretime for start-up and recharge.

We chose the tripler for the wirelessthermometer application because it hasthe highest efficiency for (Vth(H) = 1.8 V,

Vth(L) = 1.4 V). Its AIP is roughly 100�Js–1, which is approximately equal tothe two MPGs’ maximum power output.The tripler’s start-up time was approxi-mately 32 seconds, with a recharge timeof approximately 18 seconds for loadslarger than 200 k�. Energy efficiency foraverage input power below 100 �W was53 percent.

JANUARY–MARCH 2007 PERVASIVEcomputing 69

Voltage

V th (H)

V th (L)

VoltageStartup point

First charging cycle Usage cycle Usage cycleCharging cycle(application circuit in sleep mode)

Voltage acrosscapacitor

Time

Time

MPG

MPT

VMStartup

Power regulator MCU

13

1. MCU I/O pin2. Comparator output pin3. Power regulator output pin

MCU Microcontroller unitMPG Micro power generatorMPT Micro power transducerVM Voltage multiplier

Start-up circuit output

V th (on)

Comparator

(b) (c)

(a)

2

Figure 5. Transient response of the start-up circuit and system supply during start-up and usage: (a) input voltages to the start-up circuit (top trace) and the corresponding start-up circuit output (bottom trace), (b) an oscilloscope trace showing voltages at different points, and (c) a block diagram of the system.

Wireless-thermometer systemWe developed an MPG-powered wire-

less thermometer to achieve low powerconsumption, size, and component count.Figure 7 shows the system diagram forthis thermometer. The system includes aTexas Instruments MSP430 microcon-troller unit (MCU), a TI TMP100 tem-perature sensor, a custom-made surfaceacoustic wave (SAW) RF transmitter, anda TI TPS60311 charge-pump-based reg-ulator. The MSP430 is a 16-bit MCU de-signed for ultralow-power applications.It has a long-period timer (3.67 s) clockedwith a low-frequency crystal (76.8 kHz),and we connected its general-purpose I/O(GPIO) to the TMP100 for temperaturereadings. The custom-made SAW trans-mitter9 gives a high degree of flexibilityfor controlling transmission power andantenna impedance matching. The trans-mitter operates at a carrier frequency of433 MHz with a maximum transfer datarate of 4,800 bps. A compact helical an-tenna radiated the RF signal. The TMP-100 thermometer chip (� 3 V), theMSP430 microcontroller (� 1.8 V), andthe RF transmitter (� 3 V) require a 3 Vpower supply, so we used the chargepump regulator to further step up the 1.8

V input to 3.0 V. The regulator has aninput range of 0.9 to 1.8 V and up to 90percent efficiency, with a quiescent cur-rent consumption of 2 �A. It has a typi-cal output ripple of 30 mV. We built thewireless thermometer with a start-up cir-cuit so that the MPG could connectdirectly to it.

The operating sequence is as follows:

1. When we vibrate the system, theMPG charges output storage capac-itor Cout. We assume that the initialvoltage of Cout is 0.

2. The start-up circuit switches off theapplication circuit until the voltageacross Cout rises higher than Vth(H),at which point the application cir-cuit can operate.

3. The application circuit performsMCU initialization, temperatureconversion, and data transmission.

4. After the application finishes, thecircuit sleeps for a preprogrammedperiod and waits for the next acti-vation if there’s sufficient energy.Otherwise, it waits for the nextcharge-up of Cout above Vth(H).

We measured and analyzed the wire-

less thermometer’s power consumption.We placed a 2.0 V power supply in serieswith a 22 k� resistor to emulate theMPG by limiting the current entering the1,000 �F storage capacitor to 27 �A (at1.4 to 2.0 V). We measured the instan-taneous current consumption from thevoltage across a 3� resistor, and we cal-culated the instantaneous power con-sumption by multiplying the currentacross the resistor by the voltage acrossthe storage capacitor.

Table 2 shows the detailed measure-ments. The total energy consumed for a single activation was 212.62 �J. Ini-tial system start-up and MCU initializa-tion consumed 30 percent of the total en-ergy. This consumption is avoidable forconsecutive temperature measurements if the MPG generates more power thanthe application circuit consumes in sleepmode (22.38 �W). Temperature con-version consumed 17 percent of the totalenergy. Wireless transmission consumedthe most energy: 53 percent; it took 1.42 s for a single measurement. Whilethe MPG’s storage capacitor charged, thesystem had a static power consumptionof 4.5 �W. This is the minimum systemcurrent consumption, so the MPG mustsupply more than this amount to contin-uously operate the circuit.

The MPG directly powered the wire-less thermometer. The start-up circuitprovided correct operation for 50 trans-mission cycles when we randomly ap-plied vibration to the system. The start-up circuit didn’t apply power to thethermometer until the MPG raised thevoltage on the 1,000 �F storage capaci-tor above 1.8 V. When the voltageexceeded this value, the start-up circuitapplied power to the wireless thermo-meter. The MCU then acquired the tem-perature, transmitted it, and returned tosleep mode.

When the vibration on the MPG iscontinuous, the storage capacitor is large

70 PERVASIVEcomputing www.computer.org/pervasive

P O W E R G E N E R A T I O N

0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.450.40 0.50× 105

15

20

25

30

35

40

45

50

55

60

Resistive load (Ω)

Ener

gy e

ffici

ency

(%)

Doubler

Tripler

Quadrupler

Figure 6. Energy efficiency versus resistiveload for Vth(H) = 1.8 V and Vth(L) = 1.4 V(from PSpice simulations).

enough so that a single operation doesn’treduce the voltage across the storagecapacitor to a value lower than 1.4 V.When the vibration stops, the voltageacross the storage capacitor eventuallyfalls below 1.4 V, at which time the start-up circuit cuts power to the wireless-thermometer circuit until it again reachesa value greater than 1.8 V.

The voltage regulator’s start-up timewas 600 �s, and the MCU’s start-up timewas 545.4 �s. For first-time activation,the MPG took 32 seconds to produceenough energy to drive the wireless RFthermometer with an input accelerationof 4.63 m/s2 at 70.5 Hz. It took the MPG18 seconds to generate enough energyfor subsequent measurements. The RFtransmission range between the host andthe transmitter was 15 meters.

We are working to improveour generator’s poweroutput and reduce its size.As developments in mi-

croelectronics continue to reduce powerconsumption and motion-based gener-ators improve in efficiency, we believethat this technology will becomeincreasingly more pervasive. We expectto see vibration-powered devices in bio-medical, sensing, data-logging, signal-processing, and consumer electronicsapplications in the future, because they

have the compelling feature of not re-quiring batteries.

ACKNOWLEDGMENTSWe thank the Hong Kong Innovation and Technol-ogy Commission, Brilliant System, DAKA Develop-ment, and Varitronix International for funding thisproject (ITF/185/01). We also greatly appreciate theMagtech Industrial Company (Hong Kong) for do-nating the permanent magnets needed for this proj-ect. Special thanks go to Thomas K.F. Lei and Gor-don M.H. Chan for their contributions to this work.

REFERENCES

1. N.N.H. Ching et al., “A Laser-Microma-chined Multi-modal Resonating PowerTransducer for Wireless Sensing Systems,”Sensors and Actuators A: Physical, Apr.2002, pp. 685–690.

2. J.M.H. Lee et al., “Development of an AASize Energy Transducer with Micro Res-onators,” Proc. IEEE Int’l Symp. Circuitsand Systems (ISCAS 03), IEEE Press, 2003,vol. 4, pp. 876–879.

3. S.J. Roundy, Energy Scavenging for Wire-less Sensor Nodes with a Focus on Vibra-tion to Electrical Conversion, doctoral dis-sertation, Dept. of Mechanical Eng., Univ.of California, Berkeley, 2003; http://engnet.anu.edu.au/DEpeople/Shad.Roundy/paper/ShadThesis.pdf.

4. P. Horowitz and W. Hill, The Art of Elec-tronics, 2nd ed., Cambridge Univ. Press,1999.

5. “SOT23, 1.8V, Nanopower, Beyond-the-Rails Comparators with/without Refer-ence,” data sheet, Maxim Integrated Prod-ucts, 1999; http://pdfserv.maxim-ic.com/en/ds/MAX917-MAX920.pdf.

6. C.B. Williams et al., “Development of anElectromagnetic Microgenerator,” IEEProc. Circuits, Devices & Systems, vol. 148,no. 6, 2001, pp. 337–342.

JANUARY–MARCH 2007 PERVASIVEcomputing 71

System ground

System power = 3 V

Start-up circuit and application circuit

2.3 V DC(unregulated)

AA-sized micropower generator

Start-upcircuit

MSP430 16-bitmicrocontroller

unit

TMP100temperature

sensor

RFtransmitter

Micropower

transducer

Data

Data Power

Voltagetripler

TPS60311power

regulator

Figure 7. A system block diagram of awireless-thermometer application.

TABLE 2Energy consumption measurements for system start-up and the application circuit.

Energy consumption (�J) Power consumption

Parameter Time taken Estimated Measured calculated

Regular start-up 600 �s N/A 49.0 81.7 mW

Micro power start-up 545 �s N/A 15.0 27.6 mW

Temperature conversion 1.41 s 33.2 36.1 25.6 �W

Wireless transmission 4.17 ms 109 113 27.0 mW

Sleep (t = 30 s) t 20t 22.4t 22.4 �W

Total energy consumption measured (213.1 + 22.4t) = 884 �J

Average power consumption calculated 884/32 = 27.6 �W

7. R. Amirtharajah and A.P. Chandrakasan,“Self-Powered Signal Processing UsingVibration-Based Power Generation,” IEEEJ. Solid-State Circuits, vol. 33, no. 5, 1998,pp. 687–695.

8. S. Meninger et al., “Vibration-to-ElectricEnergy Conversion,” IEEE Trans. VeryLarge Scale Integration Systems, vol. 9, no.1, 2001, pp. 64–76.

9. J. Nie, “SAW Based Transmitter DesignNotes,” 20 Aug. 2003; www.rfm.com/support/apnotes/sawbasedtransmitter.pdf.

For more information on this or any other comput-ing topic, please visit our Digital Library at www.computer.org/publications/dlib.

72 PERVASIVEcomputing www.computer.org/pervasive

P O W E R G E N E R A T I O N

the AUTHORSSteve C.L. Yuen is a principal engineer at Kontel Data System. He completed thework on the project described in this article while he was a graduate student at theChinese University of Hong Kong. His research interests include mixed-signal IC sys-tems and low-power sensor electronics. He received his MPhil in computer scienceand engineering from the Chinese University of Hong Kong. He’s a member of theIEEE. Contact him at [email protected].

Johnny M.H. Lee is an engineer at SAE Magnetics. He completed the work on theproject described in this article while he was a graduate student at the Chinese Uni-versity of Hong Kong. His research interests include power microelectromechanicalsystems and optoelectrical design. He received his MPhil in mechanical and auto-mation engineering from the Chinese University of Hong Kong. Contact him [email protected].

Wen J. Li is a professor in the Chinese University of Hong Kong’s Department ofMechanical and Automation Engineering. His research interests include micro-electromechanical systems and nanoscale sensors and actuators. He received hisPhD in aerospace engineering from the University of California, Los Angeles. He’s amember of the IEEE. Contact him at [email protected].

Philip H.W. Leong is a professor in the Chinese University of Hong Kong’s Depart-ment of Computer Science and Engineering. His research interests include reconfig-urable computing and signal processing. He received his PhD in electrical engineer-ing from Sydney University. He’s a senior member of the IEEE. Contact him at [email protected].

www.computer.org/internet/

• Autonomic Computing• Roaming• Distance Learning

• Dynamic Information Dissemination• Knowledge Management• Media Search

Engineering and Applying the InternetIEEE Internet Computing reports emerging tools,

technologies, and applications implemented through the Internetto support a worldwide computing environment.

In 2007, we’ll look at: