Content uploaded by Kosanovic Mirko
Author content
All content in this area was uploaded by Kosanovic Mirko
Content may be subject to copyright.
978-1-4244-4383-3/09/$25.00 ©2009 IEEE 65
Power Management and Energy Harvesting Techniques
for Wireless Sensor Nodes
Mile K. Stojčev1, Mirko R. Kosanović2, Ljubiša R. Golubović3
Abstract – Wireless Sensor Networks, WSNs, are large
networks composed of small sensor nodes, SNs, with limited
computer resources capable for gathering, data processing and
communicating. Energy consumption represents a barrier
challenge in many sensor network applications that require long
lifetimes, usually an order of several years. Sensor nodes, as
constituents of wireless sensor networks, are battery driven
devices and operate on an extremely frugal energy budget.
Conventional low-power design techniques and hardware
architectures only provide partial solutions which are insufficient
for sensor networks with energy-hungry sensors. This paper
surveys several techniques used in today’s wireless sensor
networks with order to surpass the problem of energy
consumption, power management and energy harvesting. It
provides an insight into how various power reduction techniques
can be used and orchestrated such that satisfactory performance
can be achieved within a given energy budget.
Keywords – Sensor node, Wireless Sensor Networks, power
management, harvesting techniques.
I. INTRODUCTION
Collections of tiny, inexpensive wireless sensor nodes
(modules), organized in clusters and networks deployed over a
geographical area, capable to integrate continuous and
unobtrusive measurement, computing and wireless
communication, have attracted much attention during the last
decade in forming the concept of smart spaces. One of the
many challenges associated with sensing multiple parameters
from the environment, by using wireless sensor networks, is to
how to transmit data and power the sensors. Batteries provide
the most obvious power source of sensor nodes. In spite of the
fact that battery technology is mature, extensively
commercialized, and completely self-contained, even for
relatively large battery capacity and moderate communication
traffic requirements, the mean time to replacement or
recharging is only two or three years. For deployment with
hundreds of sensors, this means that a battery will need a
replacement every few days, what represents an unsuitable
rate for many applications. Several solutions to the power
problem exist, such as reducing power consumption to the
point where batteries can elongate the sensor module’s
lifetime. Another solution is energy harvesting–EH (or energy
1Mile Stojčev is with the Faculty of Electronic Engineering,
University of Nis, Aleksandra Medvedeva 14, 18000 Nis, Serbia, E-
mail: mile.stojcev@elfak.ni.ac.rs
2Mirko R. Kosanović is with the High Technical School,
Aleksandra Medvedeva 20, 18000 Nis, Serbia, E-mail:
mirko.kosanovic@vtsnis.edu.rs
3Ljubiša Golubović is with the Faculty of Technical Science,
Svetog Save 65, 32000 Cacak, Serbia, E-mail:
ognjen.golubovic@gmail.com
scavenging) - that is extracting energy from ambient sources.
Common energy ambient sources for energy harvesting
include mechanical energy resulting from vibration, stress and
strain; thermal energy from furnaces and other heating
sources; solar energy from all forms of light sources, ranging
from lighting to the sun; electromagnetic energy that is
captured via inductors, coils and transformers: wind and fluid
energy resulting from air and liquid flow; human energy
which depend of human movement by foot, human skin and
blood; and chemical energy from naturally recurring or
biological processes. This solution assumes that the wireless
sensor node completely alone, can capture and accumulates
energy as it becomes available. In most cases, these energy
sources provide energy in very small packets that have
previously been difficult to capture and use. Because of that,
capturing, accumulating, and storing of small packets of
electrical energy requires high energy efficiency. The
harvesting circuit must stay in active mode permanently, to be
ready to capture harvestable energy whenever it becomes
available, and to be capable to provide an output as the
application requires. The power consumption of the harvester
has to be very small so that the energy consumed by this
circuit is much smaller than the energy provided by the
ambient sources. The second key component of the harvester
is its high energy retention, i.e. the capability to store the
captured energy for as long as possible with minimal leakage
or loss. Energy harvesting circuits must have extremely high
energy retention, due to the infrequency of the energy capture
activity. Low harvesting activity levels mean that it may be
many hours before enough energy has been stored by the
energy harvesting circuit to trigger some activities of SNs,
such for example data transmission, sensing data, collecting
data, etc.. The energy harvesting circuit must also economize
the stored energy in order to provide correct operation for the
intended application.
This article starts from the fact that WSNs are ideally suited
for long-lived applications deployed at large densities for low
cost. The article discusses some promising techniques and
research directions for alleviating the energy problem in
wireless sensor node, including power management, energy
aware sensing and environmental energy harvesting. Its aim is
to point to some global viewpoint, concerning power
reduction and energy harvesting problems, as useful design
concepts for sensor node designers with order to provide long-
lived sensor networks. The remainder of the article is
structured as follows. Section II concentrates on sensor node
system architecture. The workload profile of the sensor node
is briefly discussed in Section III. Section IV deals with power
management techniques currently implemented in sensor
nodes. The three main techniques used for energy harvesting
66
are shortly presented in Section V. Finally concluding remarks
are given in Section VI.
II. SYSTEM ARCHITECTURE OF A SENSOR NODE
System architecture of a typical wireless SN is pictured in
Fig. 1. The sensor node is comprised of four subsystems: i)
computing subsystems consisting of microprocessor or
microcontroller; ii) a communication subsystem consisting of
a short range radio for wireless communication; iii) a sensing
subsystem that links the node to the physical world (external
environment) and consists of a group of sensors and actuators;
and iv) power supply subsystem which houses the battery,
DC-DC converter, and energy harvester.
ADC
input 1
input n
sensor
1
sensor
n
control signals
Signal
conditioning
. filters
. PGA
. MUX
Procesor
. MCU core
. RAM
. Flash
. Timer
Actuators
Transceiver
unit
Power supply
. battery
. solar power panel
. DC-DC convertor
. energy bar resting electronics
. motion electronics (optional)
output 1
output k
Fig. 1. System architecture of a typical wireless sensor node
A. Computing Subsystem – Microcontroller (MCU)
Most computing subsystems of SNs’ are implemented as
CMOS MCU fully static devices which operate from very low
frequencies from 1 kHz up to 32 kHz, to a maximum speed
from 1 MHz at 1.8 V DC up to 100 MHz at 5 V DC that
depends on the technology. In spite of a hefty current
consumption at 1 mA/ MHz the current draw may still be 100
μA at 32 kHz, when the MCU is running continuously, what
is not sufficient to achieve multi-year battery life [1]. In this
approach, the MCU is put into a power-savings mode, such as
idle, sleep or stop mode.
Early MCUs required an external event to toggle a pin.
Modern MCUs can wake-up from internal-timer events or
external I/O pin events. Using the wake-up timer allows the
MCU to enter in the power-savings mode for 99.9% while
running 0.1% of the time.
static current ( A)μ
Idd (μA)
dynamic
current (μA/MHz)
MCU frequency (MHz)
I
f
Fig. 2. Dynamic versus static current
When the MCU is built entirely of clocked CMOS logic
circuitry, the current consumption is perfectly linear of clock
speed, with zero offset current. This was the case when there
were simple ROM-based MCUs without any analog circuitry.
Today’s modern mixed-mode MCUs are flash-based and
packed with analog circuitry. The active mode current is
composed of two elemental components: static current and
dynamic current. The dynamic current consumption is the
increment current change versus a change in clock frequency
(see Fig. 2).
The static current is a current component that is
independent of operating frequency and is composed of
analogue-block currents, flash-module current and leakage
current.
TABLE I
POWER CONSUMPTION FOR SOME COMMON CPU
CPU Power
suplay
[V]
Power
Active
[mW]
Powe
down
[µW]
Sensor
Node
4-bit CPU
EM6603 1,2-3,6 0,0054 0,3
EM6605 1,8-5,5 0,012 0,9
8-bit CPU
ATtiny 261V/
461V/861V
1,8-5,5 *0,38
mA @
1,8V
1MHz
*0,1
PIC16F877 2-5,5 1,8 3 CIT
MC68HC05PV8A 3,3-5 4,4 485
AT90LS8535 4-6 15 45 WeC
Rene
ATmega163L 2,7-5,5 15 3 Rene2
Dot
ATMega103L 2,7-3,6 15,5 60 Mica
IBadge
C8051F311 2,7-3,6 21 0,3 Parasitic
ATmega128L 2,7-5,5 26,7 83,15 Mica
Mica2Dot
Mica2
BTnode
PIC18F452 2-5,5 40,2 24 EnOcean
TCM
80C51RD+ 2,7-5,5 48 150 RFRAIN
16-bit CPU
MSP430 F149 1,8-6 3 15 Eyes,BSN
MSP430F1611 1,8-3,6 3
1,5
15
6
Telos
SNoW5
MC68EZ326 3,3 60 60 SpotON
32-bit CPU
AtmelAT91
ARM Thumb
2,7-3,6 114 480
Intel PXA271 2,6-3,8 193 1800 iMote2
Intel StrongArm
SA1100
3-3,6 230 25 WINS
µAMPS
Most analogue blocks, independently of the MCU clock
frequency, have a significant current drain while the analogue
block is powered. The flash memory typically draws current
to power the array and read from flash cell. In some cases
67
memory blocks draw more current than the CPU, especially at
low clock speed. The leakage current depends very much on
the process technology. The active-mode static current might
be 10 times the current in a power-savings mode. To minimize
the active static current the SN must operates at a low duty
cycle, what means that faster speed results in substantial
power savings. In addition, as the lithography advances the
dynamic current goes down, but the static leakage current (for
low-voltage submicron technology) tends to increases. Thus,
low duty-cycle operation is more beneficial for the more
advanced technologies.
Table I shows power consumption of the most popular CPU
installed in some standard sensor nodes [2-4].
Performances of analogue peripherals represent also
important consideration in managing power consumption. For
example, a high-performance 300 ksamples/s 10-bit SAR
ADC can complete 12 acquisitions in about 40 μs. The faster
the ADC, the shorter the acquisition time and the sooner the
MCU can go back to sleep. High-end ADCs also reduce the
effective duty-cycle required for data-acquisition. In general,
as technologies move toward deep-submicron (< 90 nm),
performance will continue to increase, while power savings
modes become more important.
By analyzing Table I we can conclude that CPUs with 4-,
8-, 16-, and 32-bit data bus width are implemented in SNs. 4-
bit CPUs were used in first SN`s generation, mainly intended
for acquiring on/off signals (light detection, temperature,
movement). The second generation of SNs is typically
realized with 8-bit CPU. In average the power consumption in
active mode of operation varies from 3 mW up to 30 mW, and
in power down mode is about 10 µW. Modern SNs use 16/32-
bit CPU with larger number of power down modes, and are
intended for multimedia data acquisition (voice, image). The
power consumption of 32-bit CPUs in active mode is >100
mW.
B. Communication Subsystem – Radio
The SN’s radio provides wireless communication with
neighboring nodes and the outside world. Several factors
affect the power consumption characteristics of a
communication subsystem, including the type of modulation
scheme, data transfer rate, transmit power, and the operational
duty cycle. Table II shows power consumption of the most
popular radio modules (transceivers) used by sensor nodes [2],
[5], [6]. From communication aspect of wireless SNs
operation, the physical layers can be considered to be in one
of the five states:
a) Off- the only power consumption is leakage current, but
coming out of the off-state can take a long time (many ms).
b) Sleep/ Standby- the SN may be consuming as little as
(100-300) μW and can wake-up quickly unless the main
crystal oscillator is turned off.
c) Listen- the SN is listening for a packet to arrive, so most
of the radio receiver must be on. State-of-the-art power
numbers for SN communication modules in this mode are
within a range from 9 mW up to 40 mW, respectively.
d) Active Rx- similar to the Listen state, but use of
additional circuitry may push power consumption for
transceiver to 50 mW.
TABLE II
POWER CONSUMPTION FOR SOME COMMON RADIOS MODULES
Type Clock
[MHz]
Rx
power
[mA]
Tx
power
[mA/dBm]
Power
down
[µA]
low-power radio modules
MPR300CB 916 1,8 12 1
SX1211 868-960 3 25/10
TR1000 916 3,8 12/1,5 0,7
CC1000 315-915 9,6 16,5/10 1
medium-power radio modules
nRF401 433-434 12 26/0
CC2500 2400 12,8 21,6
XE1205 433-915 14 33/5 0,2
CC1101 300-928 14,7 15 0,2
CC1010 315-915 16 34/0 0,2
CC2520 2400 18,5 17,4/0 <1
CC2420 2400 19,7 17,4/0 1
CC1020 402-915 19,9 19,9 0,2
CC2430 2400 19,9 19,9
PH2401 2400 20 20
nRF2401 2400 22 10/0 0,4
CC2400 2400 24 19/0 1,5
CC2530F32 2400 24 29/1
RC1180 868 24 37/0
LMX3162 2450 27 50
STD302N-R 869 28 46/0
MC13191/92 2400 37 34/0 1
hugh-power radio modules
ZV4002 2400 65 65/0 140
e) Active Tx- in the transmit state, the SN’s active
components include the RF power amplifier, which often
dominates in high-power transmit systems. State-of-the-art
power consumption for SN transceiver module is in average
40 mW at 0 dBm Tx power.
By analyzing Table II we can conclude that radio modules
are used for all three ISM bands: 433.05 - 434,79 MHz, 902 -
928 MHz i 2400 - 2483,5MHz. Having in mind that the duty
cycle ration between transmit (Tx) and receive (Rx) mode is
usually 1:1000, we decide to classify radio modules according
to the current consumption in receive mode (Rx power). The
first group, called low-power is characterized by current
consumption less then 10 mA. For the second group, called
medium-power, the current consumption is within a range
from 10 mA up to 50 mA. In the last group, referred as high-
power, the current consumption is >50 mA.
C. Sensing Subsystem
Sensor transducers translate quantities from the non-
electrical (physical) domain into the electrical domain
(electrical signals). According to the type of output they
68
produce sensors can be classified as analogue or digital
circuits. There exists a diversity of sensors that measure
environmental parameters such as temperature, light intensity,
humidity, proximity, magnetic fields, etc.
There are several sources of power consumption in a sensor
including [7]: i) signal sampling and conversion of physical
signals to electrical ones; ii) signal conditioning; and iii) A/D
conversion. Table III lists power consumption of some
common of-the-shelf sensors [2], [8].
Several factors need to be considered when selecting
sensors for use in tiny wireless SNs: a1) volume; a2) power
consumption; a3) suitability for power cycling; a4) fabrication
and assembly compatibility with other components of the
system; and a5) packaging needs, as sensors that require
contact with the environment, such as chemicals, add
significant packaging considerations.
TABLE III
POWER CONSUMPTION FOR SOME COMMON SENSORS
Sensor type Sensing Power [mW]
consumption
micro-power
SFH 5711 Light sensor 0,09
DSW98A Smoke alarm 0,108
SFH 7741 Proximity 0,21
SFH 7740 Optical Switch 0,21
ISL29011 Light sensor 0,27
STCN75 Temperature 0,4
low-power
TSL2550 Light sensor 1,155
ADXL202JE Accelerometer 2,4
SHT 11 Humidity/temper. 2,75
MS55ER BarometricPressure 3
QST108KT6 Touch 7
SG-LINK(1000Ω) Strain gauge 9
medium-power
SG-LINK(350Ω) Strain gauge 24
iMEMS Accelerometer 30
OV7649 CCD 44
2200-2600 Series Pressure 50
high-power
TI50 Humidity 90
DDT-651 Motion Detector 150
EM-005 Proximity 180
BES 516-371-S49 Proximity 180
EZ/EV-18M Proximity 195
GPS-9546 GPS 198
LUC-M10 Level sensor 300
CP18,VL18,GM60 Proximity 350
TDA0161 Proximity 420
ultra high-power
FCS-GL1/2A4-
AP8X-H1141
Flow control 1250
FCBEX11D CCD 1900/2800
XC56BB CCD 2200
In principle, due to diversity of sensors there is no typical
power consumption number. In general, passive sensors such
as temperature, touch, seismic, etc., consume negligible power
relative to other SN’s subsystems. However, active sensors,
such as level sensors, proximity, pressure, humidity, flow
control, imagers, etc., have usually acquisition times longer
than transmission times, and accordingly they consume
significantly more energy then the radio.
We divide sensors (see Table III) into five groups. The
on/off sensors belong to the micro-power group with power
consumption <1 mW. The second group, referred as low-
power, characterizes power consumption less than 10 mW and
small amount of linear signal processing. Sensors of medium-
power group have consumption within the range from 10 mW
up to 50 mW and are realized with mixed circuits (analogue
and digital electronics). In high-power group of sensors, some
kind of dedicated signal processors are implemented. This
possibility makes the sensor of this group to be SMART
devices. The power consumption of this group is from 50 mW
up to 1 W. The last group, ultra high-power, characterize
consumption > 1 W. Due to higher power consumption for the
last two groups harvesting electronics is usually obligatory.
D. Low-power versus Ultra-low-power Sensor Node Design
As it was previously mentioned the top consideration in the
design of wireless sensor node is that energy consumption is
paramount. During this, one possible differentiation between
low-power design and ultra-low-power sensor node design is
that the former tries to maintain performance while reducing
power, but the latter has very minimal performance
requirements and sacrifices everything to minimize power
consumption.
E. Energy Sources
Usually, wireless sensor nodes utilize a combination of
energy storage and energy scavenging devices.
Capacitors may also be used in these systems to effectively
lower the impedance of a battery or energy harvester in order
to allow larger peak currents or to integrate charge from
energy harvester to compensate for lulls, such as night-time,
for a solar cell. Current capacitors, such as Ultra-capacitors,
store up to 10 mJ/mm3, which is less than 1 % of the energy
density of lithium cells.
F. Batteries Issues
From the system’s perspective, a good micro-battery should
have the following features [9]: 1) high energy density; 2)
large active volume to packaging volume ratio; 3) small cell
potential (0.5 – 1.0 V) so digital circuits can take advantages
of the quadratic reduction in power consumption with supply
voltage; 4) efficiently configured into series batteries to
provide a variety of cell potentials for various components of
the system without requiring the overhead of voltage
converters; 5) rechargeable in case the system has an energy
harvester.
69
A number of small batteries are being developed until now
for wireless communications. It seems that three cell
chemistries currently dominate the growing wireless sensor
network application market: Nickel-Metal Hydride (NiMH),
Lithium Ion (Li-Ion), and Lithium Polymer (Li-polymer).
Each battery type has unique characteristics that make it
appropriate, or in-appropriate, for a SN. Knowing the specific
characteristics of each cell chemistry in terms of voltage,
cycles, load current, energy density, charge time, and
discharge rates is the first step in selecting a cell for a SN. The
following discussion gives a short overview of the
characteristics, strengths, and weaknesses of each of the three
cell chemistries.
Nickel-Metal Hydride (NiMH): Characteristics of NiMH
batteries include a nominal voltage of 1.25 V, 500 duty cycles
per lifetime, less than 0.5 C optimal load current, an average
energy density of 100 Wh / kg, less than four-hour charge
time, typical discharge rate of approximately 30 percent per
month when in storage, and a rigid form factor. NiMH Battery
systems excel when lower voltage requirements or price
sensitivity are primary considerations in cell selection. NiMH
Systems can be configured with up to ten cells in a series to
increase voltage, resulting in a maximum aggregate voltage of
12.5 V [10].
Lithium Ion (Li-Ion): Li-ion battery characteristics
include a nominal voltage of 3.6 V, 1000 duty cycles per
lifetime, less than 1 C optimal load current, an average energy
density of 160 Wh / kg, a less-than-four-hour charge time,
typical discharge rate of approximately ten percent per month
when in storage, and a rigid form factor. These characteristics
make Li-Ion battery systems a good option when requirements
specify lower weight, higher energy density or aggregate
voltage, a greater number of duty cycles, or when price
sensitivity is not a consideration. Li-Ion battery systems can
be configured up to seven cells in series to increase voltage,
resulting in a maximum aggregate voltage of 25.2 V [10].
TABLE IV
BATTERY PARAMETERS
Voltage Nominal cell voltage
Capacity The amount of electrical charge that can
be stored
Specific
Energy
The volume-related content, measured in
energy/weight
Energy
Density
The volume-related content, measured in
energy/volume
Internal
resistance
Characterizes the ability to handle a
specific load
Self discharge The internal leakage, and aging effects
Re-charge
cycles
The number of charge cycles before
performance degrades
Charging
procedure
Type of charge circuit required
Lithium Polymer (Li-polymer): Li-polymer cells have
similar performance characteristics when compared with Li-
Ion cells, but have the advantage of being packaged in a
slightly flexible form. However, this flexibility is often
misleading, as Li-polymer cells should remain flat when
installed in a device, not even bending for installation in the
battery system. Characteristics of Li-polymer cells include a
nominal voltage of 3.6 V, 500 duty cycles per lifetime, less
than 1 C optimal load current, an average energy density of
160 Wh / kg, less than four-hour charge time, typical
discharge rate of less than ten percent per month when in
storage, and a semi-rigid form factor. Li-Ion cells can be
configured up to seven cells in series to increase voltage,
resulting in a maximum aggregate voltage of 25.2 V [10].
The crucial battery parameters are given in Table IV and V
[11].
TABLE V
BATTERY TYPES
Type Voltage Energy
density
Specific
energy
Self
discharge
Lead-acid 2.0 V 60-75
Wh/dm3 30-40
Wh/kg
3-20%/
month
Nickel
Cadmium
1,2 V 50-150
Wh/dm3
40-60
Wh/kg
10%/
month
Nickel
Metal
Hydrid
1.2V 140-300
Wh/dm3
30-80
Wh/kg
30%/
month
Lithium-
Ion
3.6 V 270
Wh/dm3
160
Wh/kg
5%/
month
Lithium-
polymer
3.7V 300
Wh/dm3
130-200
Wh/kg
1-2%/
month
III. WORKLOAD PROFILE OF SENSOR NODE
As is shown in Fig. 3 a typical workload profile for a SN
consists of two distinct phases [12]:
1. low workload - corresponds to the state of a wireless SN
in the absence of intruders. SNs periodically wake-up, sample
their sensors in order to detect any intruders, and, in their
absence, go back to sleep. To cope with high energy
efficiency in this phase a SN should provide: a) ultra low
power sleep mode; and b) rapid wake-up capability.
2. high workload - represents the state when intruder
activity is detected. During this phase the SN performs
significant amount of computation and communication with
other SNs.
power
consuption
low workload low workload
high workload
T
TT
τ
ττ
wakeup
period
shutdown
period
sampling sensors
sleep period ≅ 1 x 10
-3 duty cycle
time
Fig 3 Two phases of SN’s operation
70
A. Node Level Energy Minimization
The following two approaches are used for reducing energy
consumed by a SN [13]:
1. duty cycling - consists of waking-up the SN only for the
time needed to acquire a new set of samples and then
powering it off immediately afterwards
2. adaptive-sensing strategy - is able to dynamically change
the SN activity to the real dynamics of the process.
In designing the SN software modules (OS drivers)
intended for manipulation with duty cycle control, special care
should be devoted to a choice of the following two operating
parameters:
a) wake-up latency: it is a time required by the sensor to
generate a correct value once activated. For example, if the
sensors’ reading is performed before the wake-up latency is
elapsed, the acquired data is not valid.
b) break-even cycle: is defined as the rate at which the
power consumption of SN with implemented power
management policy is equal to that of one with no power
management.
More details concerning the hot-topic, problematic, how to
extend the lifetime of sensor units with energy-hungry
sensors, the readers can find in [14].
IV. POWER MANAGEMENT
One of the biggest problems in SN design is management
of energy and/or power management.
In general most wireless sensor nodes have at least two
modes of operation: a) an active mode, useful processing
(sensing, digital signal processing and/or communication)
takes place, and b) an idle mode - when the system is inactive.
It is acceptable to have higher power consumption in active
mode as a trade-off to increased performance, but any power
consumed when the system is idle is a complete waste and
ideally should be avoided by turning some parts of the sensor
node off.
Conventional low-power design techniques and hardware
architectures only provide point solutions which are
insufficient for WSN’s operation as a typical representative of
highly energy-constrained systems. Energy optimization, in
the case of sensor networks, is specific and much more
complex, since it involves not only reducing the energy
consumption of a single sensor node but also maximizing the
lifetime of an entire network.
Designers address the energy problem at three levels:
hardware, operating system, and application. At the hardware
level, low-power circuit designs are used in order to reduce
energy consumption. In addition, hardware devices can
include some specific power management policies at the
system’s higher levels. At the operating system (OS) level, we
can observe the applications’ and devices’ combined resource
demands. Existing OS-level energy management tends to
focus on individual devices. At the application level, you can
save energy by making the applications energy aware. An
energy-aware application can decrease its power consumption
by reducing its activity, and it can give hints to the device
manager or change its device-access pattern to create energy-
saving opportunities for hardware.
Until recently most strategies for energy management
assumed that data acquisition part of the sensor node
consumes significantly less than data transmission. But, when
this assumption does not hold, effective energy management
strategies should includes policies for an efficient use of
energy-hungry sensors, too.
A. Techniques for Power Management
Many techniques have emerged for saving active and
leakage power in SNs. It is not uncommon to find multiple
techniques at use in different parts of a SN design. In the
sequel we will give a quick overview of these approaches:
a) Clock gating: is one of the earliest techniques for
reducing dynamic power. It can increase static power because
the clock-gating cells need to be fast and designers implement
them with large, low-threshold transistors. This method
simply shuts off the clock to portions of the SN that are
inactive. Originally, designers used clock gating at the block
level as a way of creating a standby mode. More recently,
designers have employed fine-grain clock-gating, down to the
level of individual latches. Control circuitry can simply decide
not to issue a clock pulse on a cycle when the data in latch
does not change.
b) Voltage islands: if some blocks can be slower than other,
it make senses to run the slower blocks at lower frequency and
turn down the supply voltage until these blocks just meet
timing.
c) Power gating: involves turning off the supply voltage to
a block in order to stop both static- and dynamic-power
consumption. This technique involves a relatively complex
mechanism which relate to determining how to sequence the
shutdown and power-up cycles and whether it is possible to
anticipate activity of the block early enough with aim to
perform the power-up sequence. The designer must isolate the
block from surrounding circuitry during power transitions.
d) Dynamic voltage frequency scaling: is a mixture of
voltage islands and power gating. The designer adjusts the
voltage and clock frequency of each block in the fly so that it
is just meeting its deadlines for the current task. This
technique requires fairly detailed knowledge of the
application’s performance requirements. The whole SN
system must meet timing at every legal combination of block
operating frequencies.
e) Dynamic threshold voltage control: dynamically controls
the threshold of individual sets of transistors, thereby
choosing a leakage-versus-speed point that just matches the
requirements of the block on the selected path. Today, this
approach is primarily used by only a few advanced-processor
vendors.
In general, to achieve efficient energy reduction in SN it is
necessary:
i) Reduce at an absolute minimum the energy needed for
data transmission.
ii) All processes running in the SN should be optimized for
speed and duration.
71
iii) Component not needed to support the process running at
any point should be switched off, while for processes that
have to run continuously, the focus is on reducing energy
consumption.
Basic components of energy management blocks are:
1. Threshold detector- is responsible for monitoring
whether spontaneous sensor information is available.
2. Timer- periodical processes are controlled by efficient
timers.
3. Control logic- implemented as a FSM (Finite State
Machine) intended for controlling the available energy
sources, i.e. generating control signals for clock/power
switching on/off, adjusting voltage/clock frequency, etc, on a
block-by-block basis.
V. SOURCES OF ENERGY HARVESTING
Batteries can only store a limited amount of energy, which
places an upper bound on network lifetime. An emerging
technique that promises to circumvent this limitation is
environmental energy harvesting (scavenging). The process of
extracting energy from the surrounding environment and
converting it into consumable electrical energy is termed as
energy harvesting or power scavenging [16]. In general,
harvesting sources are used to increase the lifetime and
capability of SNs by augmenting the battery usage.
Energy harvesting is most applicable to applications that
demand small amounts of continuous power or that have short
periods of high-power use, which previously harvested and
stored energy can provide for. SNs are typical candidate
devices for such applications. Scavenging energy from the
environment will allow the wireless SNs to operate nearly
indefinitely, without their battery dying.
The added advantage of using energy scavenging devices is
that they are usually small. For example, there are SNs which
do not use any self-contained energy source; they only
scavenge energy from the surrounding. Such nodes can be
very small since they do not have to carry their energy sources
with them. However, the supply of energy may be interrupted
at a period of time since the power obtained from the
surrounding can not be guaranteed all the time.
There are many types of energy harvesters each offering
differing degrees of usefulness depending on the application
[13]. The various sources for energy harvesting are wind
turbines, photovoltaic cells, human body, thermoelectric
generators and mechanical vibration devices such as
piezoelectric devices or electromagnetic devices [15]. Table
VI shows power outputs for typical energy scavenging
devices [2]. The classification of energy harvesting can be
organized on the basic of the form of energy they use to
scavenge the power, and in general, we can distinguish three
types of harvesting sources from surrounding [16]:
1. Photovoltaic Cells - Perhaps the best known harvesters
(transducers) are solar or photovoltaic cells. This is a device
that converts light energy into electrical energy. The form of
energy exploited is typically light energy obtained usually
from sunlight. From locations where the availability of light is
guaranteed and usage of batteries and other means of power
supply are not feasible or expensive, usage of solar cells is a
convenient solution. While designing sources which scavenge
solar energy the first thing which we must consider is the
power supply requirements for SNs. The second, we must
have in mind such factors as availability of day light, period
of dense cloud and snow cover, effects of operation at higher
latitudes, characteristics of the solar cells used and the
intensity of the incident light. Solar radiation is the most
abundant energy source and yields around 1 mW/mm2 (1
J/day/mm3) in full sunlight or 1 μW/mm2 under bright indoor
illumination. Solar cells have conversion efficiencies up to 30
%.
2. Mechanical Vibration – When a device is subjected of
some movement, three type of energy can be generated:
vibration, kinetic or mechanical. All types of this energy can
be harvested, because they may be converted into electrical
energy using the following mechanisms:
a.) piezoelectric – piezoelectric materials convert
mechanical energy from pressure, vibrations or force into
electricity. This property is considered by many researchers in
order to develop various piezoelectric harvesters in order to
power SNs in different WSN applications. Crucial property of
piezoelectric materials is that it varies with age, stress and
temperature. The possible advantages of using this kind of
harvesters are the direct generation of desired voltage since
they do not need a separate voltage source and additional
components. But, they have some disadvantages because
piezoelectric materials are brittle in nature and sometimes
allow the leakage of charge [16].
b.) electrostatic – the principle of harvesting is based on
changing the capacitance of vibration-dependent varactors.
Vibrations separate the planes of an initially charged varactor,
and the mechanical energy is converted into electrical energy.
Electrostatic generators are in essence mechanical devices that
produce electricity by using manual power. The main benefit
of using the electrostatic converters is their ability to integrate
them into microelectronic-devices, what means that they do
not need any smart surrounding components. From the other
hand, disadvantage of using electrostatic converters is that
they need an additional voltage source intended for initial
charging of the capacitor [16].
c.) electromagnetic – electromagnetic induction is the main
principle in electromagnetic energy harvesting.
Electromagnetic induction is defined as the process of
generating voltage in a conductor by changing the magnetic
field around the conductor. One of the most effective ways of
producing electromagnetic induction for energy harvesting is
with the help of permanent magnets, a coil and a resonating
cantilever beam. Electromagnetic induction provides the
advantage of improved reliability and reduced mechanical
damping as there would not be any mechanical contact
between any parts and no separate voltage source is required.
However, the great disadvantage of this type is that
electromagnetic materials are bulky in size and are
complicated to integrate with SNs.
The electrostatic and piezoelectric harvesters are capable of
producing voltage from 2 to 10 V, whereas the
electromagnetic harvesters have limitation of producing a
max. 0.1 V voltage amplitude [16].
72
3. Thermoelectric Generators – thermoelectric generators
use the principle of thermoelectricity in order to produce a
required electrical energy. The phenomena of creating electric
potential following a temperature difference and vice-versa
can be termed as thermoelectricity. It is well known that a
voltage is generated when there is a temperature difference
between two junctions of conducting material. Thermal
energy harvesting uses temperature differences or gradients to
generate electricity, e.g. between the human body and the
surrounding environment. Devices with direct contact to
human body can harvest the energy radiated from the human
body by means of thermoelectric generator.
For applications where duty-cycling is acceptable, solar
cells or other power scavenging sources can be used to trickle-
charge a capacitor or battery after which the stored energy can
be used at much higher-power rates than the charging pace.
However, the supply of energy may be interrupted at a period
of time since the power obtained from the surroundings
cannot be guaranteed all the time.
TABLE VI
POWER OUTPUT FROM VARIOUS ENERGY SCAVENGING TECHNOLOGIES
Harvesting technology Power Density
Solar cells – direct sun 15 mW/cm2
Solar cells – cloudy day 0,15 mW/cm2
Solar cells – indoors 0,006 mW/cm2
Solar cells – desk lamp < 60 W 0,57 mW/cm2
Piezoelectric – shoe inserts 330 µW/cm2
Vibration – microwave oven 0,01-0,1 mW/cm2
Thermoelectric – 10 oC gradient 40 µW/cm2
Acoustic noise – 100 dB 9,6-4 mW/cm2
Passive–human powered system 1,8 mW
Nuclear reaction 80mW/cm3
1E6mWh/cm3
VI. CONCLUSION
The rapid development of low power electronics has made
it possible to create wireless networks of hundreds or even
thousands of devices of low computation, communication and
battery power. The networks can be used for example as
distributed sensors to monitor large geographical areas in
remote surroundings. In these applications, devices have their
own batteries to provide energy. Since every message sent and
received, input quantity sensed, and computation performed
drains the battery, special care is required in the utilization of
power. Achieving sensor lifetime of several years and
providing nontrivial application functionality represents one
of the highest challenges for designers. This article present
research directions for alleviating the energy problems in
development of wireless sensor networks, including wireless
sensor architecture, power management techniques, and
environmental energy harvesting approaches.
ACKNOWLEDGEMENT
This work was supported by Serbian Ministry of Science
and Technological Development, project No. TR-11020-
“Reconfigurable embedded systems”
REFERENCES
[1] K. Berringer, “High-performance Mixed-signal MCUs in Low-
power Applications”, EPN, Vol. 37, No. 7, July 2008, pp.12.
[2] M.V. Marcos Augusto, Diogenes Cecilio, M. Jose, “Survey on
Wireless Sensor Network Device”, http://perso.ens-
lyon.fr/isabelle.guerin-lassous/Enseignement/survey-WSN-
devices.pdf ,15.03.2009.
[3] Atmel Corporation Datasheets, www.atmel.com/products/
15.03.2009.
[4] Tatiana Bokareva, “Mini Hardware Survey”, WEB page,
http://www.cse.unsw.edu.au/~sensar/hardware/hardware_survey
.html, 15.03.2009.
[5] Texas Instruments Datasheets, www.ti.com, 15.03.2009.
[6] J. Polastre, R. Szewczyk, D. Culler, “Telos: Enabling Ultra-Low
Power Wireless Research”, In Proceedings of IPSN/SPOTS, Los
Angeles, CA, April 2005.
[7] V. Raghunathan, C. Schurgers, S. Park, M. Srivastan, “Energy-
aware Wireless Microsensor Networks”, IEEE Signal
Processing Magazine, Vol. 40, No. 3, March 2002, pp.40- 50.
[8] OMEGA Catalog, Complete Measurement, Control and
Automation Handbook and Encyclopedia, Omega Engineering,
2008.
[9] F. Pistoia, “Battery Operated Devices and Systems: From
Portable Electronics to Industrial Products”, Elsevier,
Amsterdam, 2008.
[10] T.R. Crompton, Battery Reference Book, 3-ed, Newnes, Oxford,
2000.
[11] Eliasson Jens, “Low-Power Design Methodologies for
Embedded Internet Systems”, Doctoral Thesis, EISLAB, Lulea
University of Technology, Sweden, 2008.
[12] V. Raghunathan, S. Ganerival, M. Srivastava, “Emerging
Techniques for Long Lived Wireless Sensor Networks”, IEEE
Communication Magazine, Vol. 44, No. 4, April 2006, pp.108 –
114.
[13] C. Alippi, G. Anastasi, M. Di Francesco, M. Roveri, “Energy
Management in Wireless Sensor Networks with Energy-hugry
Sensors”, IEEE Instrumentation & Measurement Magazine,
Vol. 12, No. 2, April 2009, pp.16-23.
[14] R. Want, K. Farkas, Marayanaswami, “Energy Harvesting and
Conservation, Guest Editor’s Introduction”, IEEE Pervasive
Computing; Mobile and Ubiquitous Systems, Vol. 4, No. 1,
January-March 2005, pp.14-17.
[15] G. Park, C.R. Farrar, M.D. Todd, W. Hodgkiss, R. Rossing,
“Energy Harvesting for Structural Health Monitoring Sensor
Networks”, Technical Report, Los Alamost National
Laboratories, LA, February 2007.
[16] S. Chalasani, J. Conrad, “A Survey of Energy Harvesting
Sources for Embedded Systems”,
http://www.citeulike.org/user/macgyveremir/article/3041664,
15.03.2009.