1

UNIT II ARCHITECTURES

Single-Node Architecture - Hardware Components, Energy Consumption of Sensor

Nodes , Operating Systems and Execution Environments, Network Architecture -
Sensor Network Scenarios, Optimization Goals and Figures of Merit, Gateway
Concepts.

===================================================================
Single-Node Architecture

Principal tasks of a sensor node

1) Sensing/actuation
2) Storage
3) Computation
4) Communication

Nodes must be small, cheap and energy efficient, equipped with right sensors,
necessary computation and memory resources and adequate communication
facilities, all tailored to meet the requirements of a given application.

Hardware components of a sensor node

1) Sensor node hardware overview

5 main components of a basic sensor node:

1) Controller – to process all relevant data, to execute arbitrary code.
2) Memory-to store programs and intermediate data.
3) Communication device-to send and receive information over a wireless channel.
4) Sensors and actuators-devices that can observe changes in the physical environment.
5) Power supply –use batteries and can also recharge them by obtaining energy from
environment.
MEMORY

COMMUNICATIO CONTROLLER SENSORS/

N DEVICE ACTUATORS

POWER SUPPLY

Figure: Components of sensor node hardware

 2

a) Controller

 It is the core or CPU of a wireless sensor node.

 It collects data from sensors ,processes them, decides when and where to send
the data, receives data from other sensor nodes and decides on the actuators
behavior.
 It has to execute various programs ranging from simple applications to time
critical signal processing programs.
 For such processing functionality in WSN,
 if General purpose processors are used-energy consumption is excessive.
 if microcontrollers are used - flexibility in connecting with other devices
like sensors,
instruction set amenable to time critical
signal processing,
low power consumption by going into sleep
state,
inbuilt memory,
freely programmable and hence very flexible,
microcontroller based systems have no
memory
management unit so virtual memory
impossible to achieve.
 if DSPs are used – to process large amounts of vector data
But WSNs need only simple easy to process modulations
that can be handled by communication device itself, also
signal processing tasks related to data sensing are not
complicated.

 if FPGAs –can be reprogrammed to changing requirements; takes more time

and energy
 if ASICs- ASIC is a specialized processor designed for specific applications like
high speed routers and switches. When employed in WSNs, loss of
flexibility occurs but better performance and energy efficient; ASIC
needs expensive hardware for functionality but microcontrollers
use software.

So microcontrollers are the preferred solution for use in WSNs.

b) Memory

o RAM – to store intermediate sensor readings, data packets from other sensor nodes.
Advantage: Fast RAM but loses its content if power supply is interrupted.
 3

o ROM or EEPROM or flash memory– to store program code.

o EEPROM- erases and rewrites contents
o Flash memory – erase and rewrite data in blocks instead of only a byte at a time.
Long read and write access delays, high energy required.
Manufacturing costs and power consumption of sensor node is determined also by
the size of RAM.

c) Communication device
Communication device is used to exchange data between individual nodes.
Wired communication can be used in some cases but mostly wireless communication is
used.

Choice of transmission medium

 May include radio frequencies, optical communication, ultrasound, magnetic

inductance.
 RF –based communication is the most relevant as it best fits requirements of
WSN applications.
 It provides long range and high data rates, acceptable error rates at reasonable
energy expenditure, does not require line of sight between sender and receiver.
 For a practical RF based system, WSNs use communication frequencies between
about 433 MHz and 2.4 GHz.

Transceivers

 A transmitter and a receiver are needed in a sensor node so transceivers are used.
 It converts a bit stream coming from microcontroller to radio waves and vice versa.
 low cost transceivers are commercially available that has all the circuitry needed for
transmitting and receiving (i.e) modulation,demodulation,amplifiers,filters,mixers
etc.

Transceiver tasks and characteristics

To select appropriate transceivers, number of characteristics must be considered.

1) Service to upper layer

 Receiver must offer certain services to upper layer, MAC layer. The transceiver
provides an interface that allows MAC layer to initiate frame transmissions and to
hand over the packet from main memory of sensor node into transceiver. Reversely,
incoming packets must be streamed into buffers accessed by MAC protocol.
 4

 The service is packet oriented ,or provides a byte interface to bit interface to
microcontroller.
2) Power consumption and energy efficiency

 Energy efficiency is the energy required to transmit and receive a single bit.
 Transceivers must be able to switch between different states ,active and sleep.
Power consumption in these states and switching is important.

3) Carrier frequency and multiple channels

 Transceivers are available for different carrier frequencies and must match
application requirements and regulatory restrictions.
 Transceiver may produce many carrier frequencies(channels ) to choose from and
can reduce congestion problems. Such channels are relevant for MAC protocols like
FDMA,multichannel CSMA/ALOHA.

4) State change times and energy

 Transceiver can operate in different modes(i.e)send or receive, use different

channels or in different power safe states. Time and energy required to change
between states are important.
 Turnaround time between sending and receiving are important for MAC protocols.

5) Data rates

 Carrier frequency and used bandwidth along with modulation and coding determine
data rate. It is in few Kbps.
 Different data rates can be achieved by using different modulations.

6) Modulations

 Transceivers support one or many of ASK,FSK or similar modulations.

 If several modulations are available, it is convenient to select at runtime but in real
deployment, dynamic switching between modulations is hard to achieve.

7) Coding

 Some transceivers allow various coding schemes to be selected.

8) Transmission power control

 5

 Transceivers provide control over transmission power to be used. It may need

external circuitry.
 Only discrete number of power levels are available from which actual transmission
power can be chosen.
 Maximum output power is usually determined by regulations.

9) Noise figure

It is defined as ratio of SNR at input of element to SNR at output of element.

It describes as the degradation of SNR due to element’s operation and is given in dB.

10)Gain

 It is the ratio of output signal power to input signal power and is expressed in dB.
 Amplifiers with high gain may achieve good energy efficiency.

11)Power efficiency

 Power efficiency of radio front end is the ratio of radiated power to overall power
consumed by front end.
 Power efficiency of power amplifier is the ratio of output signal’s power to the
power consumed by overall power amplifier.

12)Receiver sensitivity

 It is the minimum signal power at receiver needed to achieve a prescribed bit

/packet error rate(Eb/N0).
 Better sensitivity levels extend possible range of a system.

13)Range

 Range depends on maximum transmission power, antenna characteristics,

attenuation caused by environment. This in turn depends on carrier frequency,
modulation and coding scheme used and the bit error rate acceptable at the receiver.
 Range may be between a few meters and several hundreds of meters.
 6

14)Blocking performance

 Blocking performance of a receiver is its achieved bit error rate in the presence of an
interferer.
 An interferer at higher frequency offsets can be tolerated at large power levels.This
can be improved by a filter between antenna and transceiver.
 Adjacent channel interferer transmits on neighboring frequencies.
 Adjacent channel suppression is the transceiver’s capability to filter out signals from
adjacent frequency bands. This has an impact on SINR.

15)Out of band emission

 It is the inverse of adjacent channel suppression.

 Transmitter must produce little transmission power outside of its prescribed
bandwidth to limit disturbance from other systems.

16)Carrier sense and RSSI

 In many MAC protocols, sensing is done(if channel is busy).Receiver has to provide

that information. The meaning of a carrier sense signal depends on the
implementation.
 RSSI –Received Signal Strength Indicator provided by receiver, gives the signal
strength at which an incoming data packet has been received. It gives rough estimate
about the distance from the transmitter.

17)Frequency stability

 It is the degree of variation from normal center frequencies when environmental

conditions of oscillators change.
 In extreme cases poor frequency stability can break down communication links
(i.e)one node placed in sunlight while its neighbor in shade.

18)Voltage range

 Transceiver must operate over a range of supply voltages.

 Else inefficient voltage stabilization circuitry is required.

Many commercial transceivers are available customized to different regulatory

restrictions on carrier frequency. The major difference between these transceivers and
other communication device is that they lack a unique identifier. the Cost of providing
 7

an identifier is high compared to device’s cost but it is helpful in communication

protocols.

Research is done to improve the commercial designs with better performance and low
energy consumption at reduced cost.

Transceiver structure
RF FRONT END

 Transceiver structure has two main parts:

o Radio Frequency(RF) front end and
o Baseband part.
 RF front end performs analog signal processing in actual frequency band. Baseband
processor performs all signal processing in digital domain and communicates with
sensor node’s processor or other digital circuitry.
 Frequency conversion takes place between these two parts and is done by ADCs or
DACs.
 RF front end is the first stage of interface between EM waves and Digital signal
processing of further transceiver stages.
Some important elements of RF front end are:
1) Power Amplifier(PA) accepts converted signals from IF or baseband part and
amplifies them for transmission over antenna.
2) Low Noise Amplifier(LNA) amplifies incoming signals up to levels suitable for
further processing without reducing SNR. Incoming signal power range may
vary from very weak signal to strong signals. It can be upto 100dB. LNA is
active all time and may consume more transceiver energy and hence must be
managed.
 8

3) Elements like local oscillators or voltage-controlled oscillators and

mixers are used for frequency conversion from RF spectrum to IF or to
baseband. Incoming signal at RF frequencies fRF is multiplied in a mixer with a
fixed frequency signal from local oscillator (fLO).The resulting intermediate
frequency signal has fLO-fRF.

Transceiver operational states

Transceivers can distinguish four operational states.

1)Transmit:

 In transmit state, the transmit part of transceiver is active and antenna radiates
energy.

2)Receive:

 In Receive state, the receive part is active.

3)Idle:

 Any transceiver that is ready to receive but currently not receiving is said to be in
idle state. In this state, many parts of receive circuitry are active and others
switched off.
 Ex. in synchronization circuitry, elements concerned with acquisition are active
while tracking elements can be switched off.
 The major source of power dissipation is leakage.

4)Sleep:

 In this state, significant parts of transceiver are switched off. Transceivers offer
different sleep states like differing in amount of circuitry switched off and in
associated recovery time and start up energy .

The operating software and protocol stack must remember in which state the
transceiver is switched to. Such state changes also dissipate power.
EX. A transceiver waking from sleep state to transmit state needs start up time and
energy. Meanwhile no data transmission or reception is possible.
Problem of scheduling node states so as to minimize average power consumption is
complex.

Advanced radio concepts

1) Wake-up radio
 9

 Waiting for a data and receiving it is a power consuming operation. Receiver circuit
must be powered up so that it can observe the channel. So more power is invested
when a node is waiting for a packet to arrive.
 So to avoid power consumption, a receiver must be a wake up receiver (i.e) it must
raise an event to notify other components of a packet arrival and then the main
receiver can be turned on and receive the packet.
 More sophisticated version would be to decide whether the incoming packet is
actually destined for this node(using proper address information at the start of the
packet) and only then wake up the main receiver.
 Such wakeup receivers are tremendously attractive as they can permanently receive
in a network with low average traffic.

2) Spread-spectrum transceivers

 If there is lot of interference then simple transceivers based on ASK, FSK suffer from
limited performance. So spread-spectrum transceivers can be used.
 Drawbacks:
Complex hardware
High price

3) Ultra wideband communication

 Normally a digital signal is modulated onto a carrier frequency. But in UWB, a very
large bandwidth is used to directly transmit digital sequence as short impulses. They
occupy a large spectrum ranging from few Hz to GHz.
 Sender and receiver must be synchronized to detect the impulses. Hence UWB
communication is fairly resistant to multipath fading.
 This bandwidth will overlap with the spectrum of conventional radio system.
 Because of large spreading of signal, a very small transmission power is enough.
 A very high data rate can be realized over short distances.
 UWB communication can easily penetrate obstacles such as doors which are
impermeable to narrowband radio waves.
 The nature of UWB allows to precisely measure distances.
 Drawbacks in UWB communication
Difficulty in building such transceivers at low-cost and low-power consumption.
An UWB transmitter is relatively simple since it does not need oscillators or related
circuitry found in transmitters for a carrier-frequency-based transmitter.But the
receivers require complex timing synchronization.

Non radio frequency wireless communication

Optical
 10

 Use optical links between sensor nodes.

Advantage: very small energy per bit required for both generating and detecting
optical light .
 Simple LEDs are high efficiency senders.
 Required circuitry for an optical transceiver is simple and small than a radio
transceiver.
 Communication can be concurrent with negligible interference.
 Communicating devices must have line of sight and optical communication is
influenced by weather conditions.
Ex. corner cube reflector
3 mirrors placed at right angles to each other. A beam of light directed at it is reflected
back to its source. This reflection property holds good only if mirrors at right angles.
If a mirror is moved, a signal can be modulated into an incoming ray of light, effectively
transmitting information back to sender.
Advantage:
Mechanical movement of one mirror takes little energy. So passive reading of sensor
node can be done efficiently over long distances.

CORNER CUBE REFLECTOR

Ultrasound

 In some applications, sensor nodes are used in environments where radio or optical
communication is not applicable because these waves do not penetrate surrounding
medium.
 Ex. water medium-Sensors are deployed in marine ground floor and they must
communicate among them. Ultrasound is an option in such cases as it travels long
distances at low power.
 11

Some examples of radio transceivers that are commonly used

Single chip transceivers require only a small number of external parts and fairly low
power consumption.

RFM TR1000 family

 The TR1000 family of radio transceivers from RF Monolithics is available for 916 MHz
and 868 MHz frequency range.
 It works in a 400 KHz wide band.
 Intended for short range radio communication with upto 115.2Kbps.
 On-off-keying or ASK modulation.
 Provides dynamically tunable output power.
 maximum radiated power is given as 1.5 dBm ~1.4 Mw
 It offers received signal strength information.
 Attractive because of its low power consumption in both send and receive modes and
in sleep state.
Hardware Accelerators(Mica Motes)
 Mica motes use RFM TR1000 transceiver and contains a set of hardware
accelerators.
 Microcontroller has tight control over frame formats ,MAC protocols and so on.
Framing and MAC can be computation intensive.Example: for computing
checksums, for making bytes from bits and detecting Start Frame Delimiter(SFD).
Hardware accelerators offer these primitive computations in hardware.

Chipcon CC1000 and CC2420 family

 Chipcon offers different transceivers.

 CC1000 operates in wide frequency range between 300 and 1000 MHz
programmable in steps of 250Hz.
 Uses FSK as modulation, provides RSSI, has programmable output power.
 Interesting feature- possible to compensate for crystal temp drift.
Use in frequency hopping protocols.
 CC2420 is a complex device. It implements physical layer based on IEEE 802.15.4
standard with required support for MAC protocol. Transceiver operates in 2.4GHz
band and has DSSS modem resulting in a data rate of 250 Kbps at a low power
consumption.

Infineon TDA 525x family

 Provides flexible single –chip energy efficient transceivers.

 Provides both ASK,FSK modulation ,efficient power amplifier ,RSSI information,
tunable crystal oscillator ,onboard data filter and intelligent power down feature.
 12

 Interesting feature- self-polling mechanism –can quickly determine data rate.

 Has excellent blocking performance and so resistant to interference.

IEEE 802.15.4 /Ember EM2420 RF transceiver

 Low rate WPAN works in three different frequency bands,868 MHz,915 MHz and 2.4
GHz and employs DSSS scheme.
 For a radiated power of 0.9 mW, supply voltage of 3.3 V, in transmit mode it draws a
current of 22.7 mA, in receive mode it draws a current of 25.2 mA and in sleep
mode,only 12 micro amperes is drawn. In all bands DSSS is used.
 In 868 MHz band, only a single channel with a data rate of 20 Kbps, in 915 MHz,10
channels of 40 Kbps and in 2.4 GHz,16 channels of 250 Kbps are available.
 First two bands use BPSK modulation whereas the last band uses offset-QPSK.

National semiconductor LMX3162

 In AMPS node, radio hardware is a digital baseband processor implemented on FPGA

and for RF front end, this transceiver is used.
 It operates in 2.4GHz band and offers six different radiated power levels from 0dBm
to 20 dBm.
 To transmit data, baseband processor can control an externally controllable Volt
controlled oscillator. The main components of RF front end can be switched off.
(phase lock loop, transmit and receive circuitry).Baseband processor provides timing
information to a TDMA –based MAC protocol.
 For data transmission,FSK with a data rate of 1Mbps is used.

Conexant RDSSS9M

 WINS sensor node has transceiver in band between 902 and 928 MHz and a
microcontroller responsible for processing DSSS signals.
 Data rate is 100 Kbps.
 RF front end offers radiated power levels of 1 ,10 and 100 milli Watt.
 Around 40 sub bands are available that can be selected.
 Microcontroller implements portion of MAC protocol.

d) Sensors and actuators

Sensors

Sensors are roughly categorized into three categories:

1) Passive ,Omni directional sensors
 13

 Can measure a physical quantity without manipulating the environment by active

probing.
 Some of these are self-powered in that they obtain the energy needed from
environment.
 Energy is only needed to amplify their analog signal.
 No direction involved in these measurements.
 Ex. thermometer, light sensors, vibration, microphones, humidity, mechanical
stress or tension in materials, chemical sensors sensitive for given substances,
smoke detectors, air pressure etc.

2) Passive, narrow-beam sensors

 These sensors are also passive but include direction in their measurements.
 Ex. a camera which can take measurements in a given direction but must be
rotated if needed.

3) Active sensors

 Actively probes the environment.

 Ex. radar, sensor or seismic sensors generate shock waves by small explosions.

Theoretical work involves only passive omni-directional sensors. Some practical test
beds involve narrow beam sensors. Active sensors are not treated in literature to any
extent.

Actuators

 Similar to sensor but can control a motor, bulb or some other physical object.
 It converts energy from one form(mostly electrical energy to mechanical energy)
to other and uses this as a force to control a physical object.
 Ex. water sprinklers in fire rescue operation which gets activated when
temperature sensor detects the presence of fire and reports to the fire rescue
station.

e) Power supply of sensor nodes

2 essential aspects in supplying power to sensor nodes.

 1) Storing energy and providing power in required form
 2) Replenish consumed energy by scavenging it from some alternate power source.

Storing energy:Traditional batteries

 14

Batteries are electrochemical storage of energy. A normal AA battery stores about 2.2–
2.5 Ah at 1.5 V.

Types of batteries
Primary batteries – nonrechargeable battery
Secondary batteries –rechargeable battery using an energy scavenging device.

Requirements on batteries:
1) Capacity

 Battery should have high capacity at small weight, small volume and low price.
 Metric to measure capacity is J/Cm3 (i.e)energy/volume.
 Research is done on microscale batteries for example energy is deposited directly
onto a chip.

2) Capacity under load

 Must withstand various usage patterns (i.e) nodes may draw different levels of
power and sometimes draw high current in certain operation modes.
 Mostly larger the battery, more power can be delivered instantaneously.
 But rated battery capacity specified by a manufacturer is valid as long as
maximum discharge currents are not exceeded.

3) Self-discharge

 Self-discharge should be low.

 Battery must last for long time but it is operational only for few months even if
not used.
 Zinc-air batteries have high energy density but has short lifetime.

4) Efficient recharging

 Recharging must be efficient even at low and intermittently available recharge

power.
 Battery must not exhibit any memory effect.
 Some energy scavenging technology produce current only in microAmps region
but current battery technology would basically not recharge at such values.

5) Relaxation

 Relaxation effect is self-recharging of an empty or almost empty battery when no

current is drawn from it.
 This is based on chemical diffusion processes within cell.
 15

 This might increase battery lifetime and usable capacity.

 This may use multiple batteries in parallel and schedule discharge from one
battery to another depending on relaxation properties and power requirement
of operations to be supported.

Unconventional energy stores

1)Fuel cells

 An electrochemical storage of energy .

 It directly produces electrical energy by oxidizing hydrogen or hydrocarbon fuels.
 They have excellent energy densities but current systems need mini size pumps,
valves etc.

2)Micro heat engines

 The more traditional approach for using energy stored in hydrocarbons is to use
miniature versions of heat engines, example, a turbine.
 Shrinking heat engines to desired size needs research on MEMS technology.

3)Radioactive substances

 Radioactive substances are also an energy store.

4)Gold Caps

 Gold caps are high quality and high capacity capacitors store more energy, can be
easily and quickly recharged, do not wear out over time.

DC-DC Conversion

 As its capacity drops, battery’s voltage drops. So less power is delivered to node’s
circuitry. This reduces oscillator frequency and transmission power. Hence such
nodes have smaller transmission range.
 DC-DC converter – it regulates voltage delivered to node’s circuitry.
But in ensuring constant voltage, it draws high current from battery when the
battery is already getting weak. So battery’s death is speeded up. It also consumes
more energy for its own operation. So overall efficiency is reduced.
Advantages: predictable operation during entire life cycle.

Energy scavenging
 16

 Unconventional energy stores transform energy from stored secondary form to

electrical energy.
 It is less direct and not easy as a battery.
 Entire energy is stored in node itself. Once fuel is exhausted, node fails. These are
limited energy stores.
 To ensure long-lasting nodes, energy scavenging is done.
 Several approaches are:

1) Photovoltaics

 Solar cells are used to store energy.

 Available power depends on whether it is done indoors or outdoors,
depends on time of day.
 Resulting power is between 10Microwatts/cm2 indoors and
15megawatts/cm2 outdoors.
 Single cells achieve a output voltage of about 0.6V as long as drawn
current does not exceed a critical threshold which depends on light
intensity. Hence solar cells are used to recharge secondary batteries.
 Trade-offs: complexity of recharging circuitry, solar cell efficiency, battery
lifetime.

2) Temperature gradients

 Differences in temperature, even small, can be directly converted to

electrical energy.
 But practical devices fall very short of theoretical upper limits.
 See beck effect-based thermoelectric generators are commonly considered,
ex. a generator that achieves about 80 microwatts/cm 2 at about 1 V from a
5 Kelvin temperature.

3) Vibrations

 From vibrations of walls or windows due to vehicles running in street,

machinery, ventilations, electrical energy can be obtained.
 Available energy depends on amplitude and frequency of vibrations and
ranges from about 0.1 microwatt/cm3 to 10000 microwatt /cm3.
 Conversion is done based on electromagnetic, electrostatic and
piezoelectric principles.
 a generator based on a variable capacitor
 17

4) Pressure variations

 Variation of pressure can be used as a power source.

 Ex. to fit a piezoelectric generator in the heel of a shoe to generate power.
 It can produce on average 330 microwatt/cm2.

5) Flow of air/liquid

 Flow of air or liquid in wind mills or turbines are another power source.
 Problem is miniaturization.

ENERGY CONSUMPTION OF SENSOR NODES

1) Operation states with different power consumption

 Batteries have small capacity, and recharging by energy scavenging is

complicated and volatile. Hence, the energy consumption of a sensor node must be
tightly controlled.
 For example, consider the energy consumed by a microcontroller per instruction
(i.e) about 1 nJ per instruction. Assume a battery volume of one cubic millimeter and
could store about 1 J. To use such a battery to power a node even only a single day,
the node must not consume continuously more than 11.5 μW. No current controller is
able to work at such low-power levels.
 Designing low-power chips is one of the options for an energy-efficient sensor
node.
But can easily be squandered when the components are improperly operated.
 18

 Most of the time a wireless sensor node has nothing to do. Hence, it is best to
turn it
off. It should be able to wake up again, on the basis of external stimuli or on the basis
of time. So completely turning off a node is not possible, but its operational state
can be adapted to the tasks at hand. Introducing and using multiple states of
operation with reduced energy consumption in return for reduced functionality is
the core technique for energy-efficient wireless sensor node. This model is called as
graded sleep state model.
For example, Advanced Configuration and Power Interface (ACPI) ,PC hardware
introduces one state representing the fully operational machine and four sleep
states of graded functionality . This is Dynamic Power Management (DPM) .
 Transitions between states take both time and energy. In a “deeper” sleep
state,
less power is consumed but deeper the sleep state, more time and energy it takes to
wake up again to fully operational state. Hence, it is better to remain in an idle
state instead of going to deeper sleep states.

 Consider a commonly used model. At time t1, the decision whether or not a
component is to be put into sleep mode should be taken to reduce power
consumption from Pactive to Psleep.
 If it remains active and the next event occurs at time tevent, then a total
energy of Eactive = Pactive(tevent − t1) has be spent uselessly idling.
 Putting the component into sleep mode on the other hand requires a time
τdown until sleep mode has been reached. Assume that the average power
consumption during this phase is (Pactive + Psleep)/2. Then, Psleep is consumed
until tevent.

 Energy required in sleep mode= τdown(Pactive + Psleep)/2 + (tevent − t1 − τdown)Psleep

Energy required in active mode= (tevent − t1)Pactive.
 19

 Energy saved is: energy required in active mode – energy required in sleep mode
(i.e)

 Once processing has to be done, an overhead energy is incurred to return to

operational state.

 Clearly, switching to a sleep mode is only beneficial if Eoverhead < Esaved or if the
time to the next event is sufficiently large:

2) Microcontroller Energy Consumption

i) Basic power consumption in discrete operation states

Embedded controllers have multiple operational states and it is also fairly easy to
control. Some examples are:

ii )Intel StrongARM

The Intel StrongARM provides three sleep modes:

• In normal mode, all parts of the processor are fully powered. Power consumption is up
to 400 mW.
• In idle mode, clocks to the CPU are stopped; clocks that pertain to peripherals are
active. Any interrupt will cause return to normal mode. Power consumption is up to 100
mW.
• In sleep mode, only the real-time clock remains active. Wakeup occurs after a timer
interrupt and takes up to 160 ms. Power consumption is up to 50 μW.

iii )Texas Instruments MSP 430

The MSP430 family features a wider range of operation modes: One fully operational
mode,which consumes about 1.2 mW .
 20

There are four sleep modes in total. The deepest sleep mode, LPM4, only consumes 0.3
μW, but the controller is only woken up by external interrupts in this mode. In the next
higher mode, LPM3 consumes only about 6 μW.

iv )Atmel ATmega

The Atmel ATmega has six different modes of power consumption. Its power
consumption varies between 6 mW and 15 mW in idle and active modes and is about 75
μW in power-down modes.

v)Dynamic voltage scaling

 Another option than discrete operational states is to use a functionality or power

adaptation by adapting the speed with which a controller operates. The idea is to
choose the best possible speed with which to compute a task that has to be
completed by a given deadline.
 One obvious solution is to switch the controller in full operation mode, compute
the task at highest speed, and go back to a sleep mode as quickly as possible.
 The alternative approach is to compute the task only at the speed that is
required to finish it before the deadline because of the fact that a controller
running at lower speed, that is, lower clock rates, consumes less power than at
full speed as the supply voltage can be reduced at lower clock rates. This
technique is called Dynamic Voltage Scaling (DVS) .(scaling down supply voltage
to reduce power consumption)
 This technique is actually beneficial for CMOS chips.
 As the actual power consumption P depends quadratically on the supply voltage
VDD, reducing the voltage is a very efficient way to reduce power consumption. Power
consumption also depends on the frequency f , hence

 Dynamic voltage scaling also reduces energy consumption. When applying

dynamic voltage scaling, care has to be taken to operate the controller within its
specifications. There are minimum and maximum clock rates for each device, and for
each clock rate, there is a minimum and maximum threshold that must be obeyed.
Hence, when there is nothing to process, going into sleep modes is still the only
option and using arbitrary voltages requires a quite efficient DC-DC converter to be
used.

3) Memory

 The most relevant kinds of memory are on-chip memory of a microcontroller and
 21

FLASH memory ; off-chip RAM is rarely used.

 The construction and usage of FLASH memory can heavily influence node lifetime.
Writing is more complicated, as it depends on the granularity with which data can be
accessed (individual bytes or only complete pages of various sizes).
 Consider the energy consumption necessary for reading and writing to the Flash
memory used on the Mica nodes . Reading data takes 1.111 nAh, writing requires
83.333 nAh. Hence, writing to FLASH memory can be a time- and energy-consuming
task.

4) Radio Transceivers

 A radio transceiver has essentially two tasks: transmitting and receiving data
between a pair of nodes. Radio transceivers can operate in different modes, turned
on or turned off.
 For low energy consumption, the transceivers should be turned off most of the time
and only be activated when necessary .This incurs additional complexity, time and
power overhead .
 To understand the energy consumption behavior of radio transceivers, models for
energy consumption per bit for both sending and receiving are required.

i) Modeling energy consumption during transmission

 The energy consumed by a transmitter is due to

i) RF signal generation, which mostly depends on chosen modulation and
target distance and on the transmission power Ptx (power radiated by the
antenna).
ii) Electronic components necessary for frequency synthesis, frequency
conversion, filters etc.
 The transmitted power is generated by the amplifier of a transmitter. Its own power
consumption Pamp depends on its architecture mostly on the power they are to
generate.
 A realistic model assumes that a certain constant power level is always required
irrespective of radiated power, plus a proportional offset:

Pamp = αamp + βampPtx

where αamp and βamp are constants depending on process technology and amplifier
architecture .
 This model implies that the amplifier’s efficiency Ptx/Pamp is best at maximum output
power.
 In addition to the amplifier, other circuitry has to be powered up during
transmission.. This power is referred to as PtxElec.
 22

 The energy to transmit a packet n-bits long then depends on how long it
takes to send the packet, determined by the nominal bit rate R and the coding rate
Rcode, and on the total consumed power during transmission. If, in addition, the
transceiver has to be turned on before transmission, startup costs also are incurred.

 This equation does not depend on the modulation chosen for transmission and
antenna efficiency .
 This model can be enhanced by the effects of Forward Error Correction (FEC) coding
Since FEC just increases the number of bits approximately by a factor of one divided
by the code rate since the coding energy is negligible .

ii) Modeling energy consumption during reception

 The receiver can also be either turned off or turned on. While on,it can actively
receive a packet or can be idle, observing the channel and ready to receive.
 The energy Ercvd required to receive a packet has a startup component TstartPstart.
During actual reception, receiver circuitry has to be powered up.

The last component is the decoding overhead, which is incurred for every bit .
 It needs an appropriate transceiver architecture with fast startup time.

iii) Dynamic scaling of radio power consumption

 Scaling down supply voltage or frequency to obtain lower power consumption

in exchange for higher latency is only applicable to some of the electronic parts of a
transceiver, but the remainder of the circuitry – the amplifier, for instance,
cannot be scaled down still has to be run at high power over an extended period of
time .
 For radio communication, other possible parameters include the choice of
modulation or code leading to Dynamic Modulation Scaling (DMS), Dynamic Code
Scaling (DCS) and Dynamic Modulation-Code Scaling (DMCS) optimization
techniques . The idea is to dynamically adapt modulation, coding or other
parameters to maximize system metrics like throughput or energy efficiency.
 It depends on the hardware’s ability to actually perform such modulation
adaptations and also delay constraints and time-varying radio channel properties
 23

5) Relationship between computation and communication

 This relationship heavily depends on the particular hardware used.

 The ratio of the energy consumption to send one bit compared to computing a single
instruction is between 1500 to 2700 for Rockwell WINS nodes, between 220 to
2900 for MEDUSA II nodes, and about 1400 for WINS NG 2.0 nodes . for the RFM
TR1000 radio transceiver, 1 μJ to transmit a single bit and 0.5 μJ to receive one;
their processor takes about 8 nJ per instruction. This results in a ratio of about 190
for communication to computation costs. In a slightly different perspective,
communicating 1 kB of data over 100 m consumes roughly the same amount of
energy as computing three million instructions .
communication is considerably more expensive than computation.

5) Power consumption of sensors and actuators

 There is a wide diversity of sensors and actuators so it is not possible to provide

guidelines about their power consumption. For passive sensors say light and
temperature sensors, power consumption can be ignored but for active sensors say
radars, power consumption must be considered.
 The sampling rate evidently is important. More sampling requires more energy for
the sensors and also the data has to processed and communicated.

OPERATING SYSTEMS AND EXECUTION ENVIRONMENTS

1) Embedded OS

Traditional tasks of OS:

1) Controlling and protecting access to resources.
2) Managing resource allocation to different users
3) Support for concurrent execution of several processes
4) Communication between the processes.

OS in an embedded system need not perform all of these tasks but support tasks like
i) Energy management which is a much specific need of a WSN.OS can perform energy
management through controlled shutdown of individual components or Dynamic
Voltage Scaling (DVS) techniques.
ii) Handling external components like sensors, radio modems, timers etc easily and
efficiently
iii)handling of asynchronous information (i.e) information available at any point of time.
Also to support entire OS, microcontrollers may not have sufficient resources.
 24

This needs appropriate programming model, clear way to structure a protocol stack and
explicit support for energy management.

2) Programming paradigms and application programming interfaces

i) Concurrent programming
(supporting execution of multiple processes in one CPU at a time)

 Support for concurrent execution is must for WSN nodes as data comes from
arbitrary sources at arbitrary points of time (i.e) multiple sensors and radio
transceivers.
 Ex. a system could poll sensor for any available data and process it then it can poll
transceiver for any available packet and process it.
 This is a simple sequential programming model in which data may be lost when
sensor data or transceiver packets are processed for a long time. So a simple
sequential programming model is not sufficient.

ii) Process-based concurrency

 Most OS support concurrent execution of multiple processes in a CPU so process-

based model is appropriate for supporting concurrency in a WSN node.
 But mapping such model of concurrent processes to a sensor node in a WSN has
some mismatches. Also high overhead is incurred in switching between
processes. Each process requires its own stack space in memory which is not
suited for a WSN.
 25

iii) Event-based programming

 This combines the reactive nature of a WSN node and integrates it into the design
of the operating system.
 The system essentially waits for any event to happen, where an event typically
can be the availability of data from a sensor, the arrival of a packet, or the
expiration of a timer. It only stores the fact that this event has occurred and
stores the necessary information. The actual processing of the information is not
done in the event handler routines, but separately decoupled from the actual
appearance of events.
 An event handler can interrupt the processing of any normal code. It is for the
processing of normal code, which is only triggered by the event handlers.
 But as it is very simple and short, it can be required to run to completion
without noticeably disturbing other code.
 Event handlers cannot interrupt each other but are simply executed one after
each other. It is for the time-critical event handlers where execution cannot be
interrupted.
 Performance of a process-based and an event-based programming model on the
same hardware proved that performance improved by a factor of 8,
instruction/data memory requirements were reduced by factors of 2 and 30,
respectively, and power consumption was reduced by a factor of 12.
 26

iv) Interfaces to OS

In addition to programming model, it needs some interfaces that specify how internal
state of a system can be handled and set. An interface must be accessible from protocol
implementations and closely tied with structure of protocol stack.

An application programming interface (API) consists of a functional interface (functions

include state inquiry and manipulation, sending and transmitting data, access to
hardware (sensors, actuators, transceivers)), object abstractions (wireless links, nodes
etc) and behavioral semantics.

3) Structure of OS and protocol stack

 The traditional approach to communication protocol structuring is to use

layering: individual Protocols are stacked on top of each other, each layer only
using functions of the layer directly below. This layered approach has great
benefits in keeping the entire protocol stack manageable, In containing
complexity,and in promoting modularity and reuse.
 A physical layer information can be used to assist in networking protocols to
decide about routing changes or to assist link layer protocols.Hence,one single
source of information can be used to the advantage of many other protocols not
directly associated with this source . Such cross-layer information exchange
violates layering approach.
 component model: Relatively large, monolithic layers are broken up in to small,
self-contained “ components”. These components only fulfill one well-defined
function each .
 Interactions are not confined to immediate neighbors in an up/down
relationship, but can be with any other component.This solves some of the
structuring problems for protocol stacks and it also fits naturally with an event-
based approach.
 27

 Wrapping of hardware, communication primitives, in-network processing

functionalities all can be conveniently designed and implemented as components.
It uses the notion of explicit wiring of components to allow event exchange to
take place between them.

4) Dynamic energy and power management

To improve energy efficiency, switch individual components into sleep states,

scale down frequency and supply voltage,
select particular modulation and coding.
Decisions must be made by OS, protocol stack or application.
Dynamic Power Management(DPM) :Complicating factor is the energy and time
required for the transition of a component between any two states. It is optimal to
always go into a mode with lowest power consumption which is not possible. So more
advanced algorithms are needed.

i) Probabilistic state transition policies

In policies that regulate transition between sleep states, they consider sensors randomly
distributed over a fixed area and events arrive with temporal and spatial distributions.

This allows to compute probabilities for the time to next event, once an event has
been processed. Some events are missed when the node is in sleep mode. Such
possibilities of missing events are also considered. Probabilistic rules are needed to
decide if to go into such sleep mode.

ii) Controlling dynamic voltage scaling(DVS)

It is rare that only one task is going to be run in an OS and so a scheduler must decide so
as to meet deadlines of each task by receiving feedback from applications.DVS can also
be incorporated into kernel of OS and achieve energy efficiency in mixed loads without
modifying user programs.

iii) Trading off fidelity against energy consumption

Tasks can give accurate results only when completed within its deadline but this is not
possible in WSNs. So always there is a tradeoff between accuracy and time and mostly
accuracy and energy.

5) Case study: TinyOS and nesC

 28

To support concurrency in a sensor node software ,an event-based programming model

is needed but it must not consume more resources and modularity must be supported to
easily exchange events between individual state machines.
TinyOS and nesC address these challenges where TinyOS is the operating system and
nesC is the programming language.

Component

 A Component has semantically related functionality like handling radio interface,

computing routes etc.
 It has state information in a frame, program code for normal tasks and handlers
for events and commands.
 Events and commands are exchanged between different components and
components are arranged hierarchically.
 Events originate in Hardware and pass upward from low level to high level
components
 Commands are passed from high level to low level components.

Example: timer component

Timer Component:

 Component understands 3 commands: init, start and stop.

 It can handle one event (fire ) from another component .
 It issues a setRate command
 It emits a fired event itself.
 Tasks:
Actual computational work is done in tasks.
These tasks can be interrupted by handlers.
Advantages: no need for stack management and tasks are atomic with respect to
each other.
Tasks are concurrent and this can be done by triggering several events and by a
 29

power-aware FIFO scheduler (shuts down a node when no tasks are executing
or waiting).
 Commands and event handlers:
These are for interaction between components.
They must perform simple triggering duties.
Commands must not block for long time but must only request .Event handler
leaves information in its frame and the task will be executed later.
 How components get feedback from another component about a command ?
Split-phase programming approach
Works in two phases:
1) sending a command.
2) explicit information about outcome of an operation delivered by a separate
event.
This needs interface between components.
Interface types - This defines commands and events that belong together.
It can easily express split-phase programming style by putting commands and
their corresponding events into the same interface.

Example: Reorganizing timer component

Timer component is reorganized into a clock interface and stdctrl and timer
interfaces.
 30

Defining modules and interfaces

Example: Building a larger configuration out of two components

Components that have correct interface types can be wired together (i.e) building
a larger configuration from components.
building a timer and hardware clock component wired together into a new
component,completetimer.
TinyOS and nesC offer easy to implement core OS functionalities, communication
protocols and application functions. It helps in combining small highly specialized
components. Code size and memory requirements are small.
TinyOS has become the standard implementation platform for WSNs.
It is beneficial in reprogramming and retasking an existing network.
 31

6) Other examples

contiki10
ecos
mantis

Some examples of sensor nodes

1) Mica Mote family

 built out of research experiments along with Intel collaboration.

 equipped with Atmel microcontroller, radio modem.
 can add more sensor boards to the node.
 specialized enclosures are built for using in rough environments.
 sensors are controlled to microcontroller through 12C bus or SPI.

2) EYES nodes(EnergY Efficient Sensor network )

 Developed by Infineon.
 Equipped with Texas instrument MSP microcontroller, Infineon radio modem,
SAW filter, transmission power control.
 Has a USB interface to PC and possible to add more sensors/actuators.

3)BTnodes

 Developed out of research experiments.

 Equipped with Atmel microcontroller.
 Use Bluetooth as radio technology in combination with Chipcon CC1000
operating between 433 and 915 MHz.

4)ScatterWeb

 Equipped with MSP microcontroller and can range up to embedded web servers .
 Has Bluetooth as well as radio modem .

NETWORK ARCHITECTURE

Sensor network scenarios

1) Types of sources and sinks

 32

 A source is any entity in the network that can provide information, that is, typically a
sensor node or an actuator node .
 A sink is the entity where information is required.
 There are essentially three options for a sink:
it could belong to the sensor network - another sensor/actuator node
it could be an entity outside this network-an actual device like a handheld or PDA
a gateway to another larger network such
as the Internet.

2) Single-hop versus multihop networks

SINGLE HOP COMMUNICATION

MULTI HOP COMMUNICATION

 As direct communication is impossible because of distance or obstacles , multihop

communication can circumvent the problem.
 Direct communication between source and sink is not always possible, specifically in
WSNs, which are intended to cover a lot of ground (e.g. in environmental or
agriculture applications) or that operate in difficult radio environments with strong
attenuation (e.g. in buildings).
 33

 To overcome such limited distances, an obvious way out is to use relay stations, with
the data packets taking multi hops from the source to the sink. The sensor nodes
themselves can act as relay nodes, foregoing the need for additional equipment.
 It improves the energy efficiency of communication.
 As attenuation of radio signals is at least quadratic in most environments (and
usually larger), it consumes less energy to use relays instead of direct
communication.
 When targeting for a constant SNR at all receivers, the radiated energy required for
direct communication over a distance d is cdα (c - constant, α ≥ 2 ,the path loss
coefficient).
 Using a relay at distance d/2 reduces this energy to 2c(d/2)α.
 But energy is actually wasted if intermediate relays are used for short distances d.
Only for large d does the radiated energy dominate the fixed energy costs consumed
in transmitter and receiver .
 Multihop networks may operate in a store and forward fashion. A node has to
correctly receive a packet before it can forward it .
 cooperative relaying - This uses even erroneous reception of packets, for example,
when multiple nodes send the same packet and each individual transmission could
not be received, but collectively, a node can reconstruct the full packet.

3) Multiple sinks and sources

 There are multiple sources and multiple sinks present. Multiple sources should send
information to multiple sinks, where the information has to reach all or some of the
sinks.
 34

4) Three types of mobility

 Wireless communication has ability to support mobile participants.

 In wireless sensor networks, mobility can appear in three main forms:

Node mobility:

 Wireless sensor nodes can be mobile.

 In examples like environmental control, node mobility should not happen; in
livestock surveillance (sensor nodes attached to cattle, for example), node
mobility must happen.
 The network has to reorganize itself frequently to function correctly.
 Frequency and speed of node movement influence the energy required to
maintain a desired level of functionality in the network.

Sink mobility:

 The information sinks can be mobile.

 Mobility of an information sink may not be a part of the sensor network, for
example, a human user requested information via a PDA while walking in an
intelligent building.
 The network with the assistance of the mobile requester must make provisions
that the requested data actually follows and reaches the requester despite its
movements.

Event mobility:

 In applications like event detection and in particular in tracking applications, the

 35

cause of the events or the objects to be tracked can be mobile.

 The observed event is covered by a sufficient number of sensors at all time.
Sensors will wake up around the object, engaged in higher activity to observe the
present object, and then go back to sleep. As the event source moves through the
network, it is accompanied by an area of activity within the network . This has
been called the Frisbee model.
 Nodes that do not actively detect anything are intended to switch to lower sleep
states .

OPTIMIZATION GOALS AND FIGURES OF MERIT

 For all the scenarios and application types, different forms of networking solutions
can be found to optimize a network, to decide which approach better supports a
given application and to turn relatively imprecise optimization goals into
measurable figures of merit.

1) Quality of service

 WSNs differ from other conventional communication networks mainly in the type
of service they offer.
 Additional requirements about the offered Quality of Service (QoS) are made.
 Such QOS can be
i) a low-level, networking-device-observable attribute – bandwidth, delay, jitter,
packet loss rate
ii) a high-level, user-observable attribute like the perceived quality of a voice
communication or a video transmission.
 High-level QOS attributes are more important for WSNs and highly depend on the
application.
 Some generic possibilities are:

1)Event detection/reporting probability - probability that an event that

occurred is not detected or not reported to an information sink .
For example: not reporting a fire alarm to a surveillance station.
This probability depends on the overhead spent in setting up structures in the
network that support the reporting of an event (e.g. routing tables).

2) Event classification error- If events are not only to be detected but also to be
classified, the error in classification must be small.

3) Event detection delay - the delay between detecting an event and reporting it
to any or all interested sinks.
 36

4) Missing reports- In applications that require periodic reporting, the probability

of undelivered reports should be small.

5)Approximation accuracy -For function approximation applications (e.g.

approximating the temperature as a function of location for a given area), the
average/maximum absolute or relative error with respect to the actual function.
6) Tracking accuracy -Tracking applications must not miss an object to be
tracked, the reported position should be as close to the real position as possible,
and the error should be small.

2) Energy efficiency

 Energy is a scarce resource and it must be efficiently used.

 The most commonly considered aspects are:

1)Energy per correctly received bit – the energy spent on average to

transport one bit of information (payload) from the source to the destination,
counting all sources of energy consumption at all possible intermediate hops.
This is useful for periodic monitoring applications.

2)Energy per reported (unique) event - the average energy spent to report one
event. Since the same event is reported from various sources, normalize this
metric to only the unique events .

3)Delay/energy trade-offs- Some applications have “urgent” events, which

needs an increased energy investment for a speedy reporting of such events.
There is the trade-off between delay and energy overhead.

4)Network lifetime -The time for which the network is operational or the time
during which it is able to fulfill its tasks .
Possibilities are:
Time to first node death When does the first node in the network run out
of energy or fail and stop operating?
Network half-life When have 50% of the nodes run out of energy and
stopped operating?
Time to partition When does the first partition of the network in two
disconnected parts occur?
Time to loss of coverage - each point in the deployment region is
observed by multiple sensor nodes. This is the time when for the first time
any spot in the deployment region is no longer covered by any node’s
observations.
 37

Time to failure of first event notification -A network partition can be

seen as irrelevant if the unreachable part of the network does not want to
report any events in the first place. This can be due to an event not being
noticed because the responsible sensor is dead or because a partition
between source and sink has occurred.

3) Scalability

 The ability to maintain performance characteristics irrespective of the size of the

network is referred to as scalability.
 Architectures and protocols should implement appropriate scalability

4) Robustness

 Wireless sensor networks should also exhibit an appropriate robustness.

 They should not fail just because a limited number of nodes run out of energy, or
because their environment changes.
 These failures have to be compensated for.

GATEWAY CONCEPTS
1) The need for gateways

 The network has to interact with other information devices, for example, a user
equipped with a PDA trying to interact with the sensor network via the Internet (the
standard example is to read the temperature sensors in one’s home while traveling
and accessing the Internet via a wireless connection).
 WSN has to be able to exchange data with a mobile device or a gateway which
provides the physical connection to the Internet.
 Either the mobile device/the gateway is equipped with a radio transceiver.
 Possible trade-offs include the percentage of multitechnology sensor nodes required
to serve mobile users in comparison with the overhead and inconvenience to fit WSN
transceivers to mobile devices like PDAs.
 Design of gateways :
1) To regard a gateway as a simple router between Internet and sensor network. This
would entail the use of Internet protocols within the sensor network but WSNs will
require specific, heavily optimized protocols. Thus, a simple router will not suffice as
a gateway.
2) To design the gateway as an actual application-level gateway: on the basis of the
application-level information, the gateway will have to decide its action.

 2) WSN to Internet communication

 38

 Assume that the initiator of a WSN–Internet communication resides in the WSN .

For example, a sensor node wants to deliver an alarm message to some Internet host.
 How to find the gateway from within the network?
 If several such gateways are available, how to choose between them?
 At least if some gateway should be preferred for a given destination host, how it is
done?
 Does a sensor node know to which Internet host to address such a message?
 One option is to build an IP overlay network on top of the sensor network .
Even if the sensor node does not need to be able to process the IP protocol, it has to
include sufficient information (IP address and port number, for example) in its own
packets; the gateway then has to extract this information and translate it into IP packets.
 Gateway must act similar to a Network Address Translation (NAT) device .

3) Internet to WSN communication

 This is simple if the requesting terminal is able to directly communicate with the
WSN. For example, a mobile requester equipped with a WSN transceiver.
 39

 On the other hand, a far terminal requesting the service may not be able to
communicate with any sensor node and so needs a gateway node.
 How to find out that there actually is a sensor network in the desired location and
how to find out about the existence of a gateway node?
 The requesting terminal can send a properly formatted request to this gateway which
acts as an application-level gateway or a proxy for the individual sensor node that can
answer this request.
 The gateway translates this request into the proper intra sensor network protocol
interactions assuming that there is an application-level protocol that a remote requester
and gateway can use and that is more suitable for communication over the Internet than
the actual sensor network protocols.
 There are Web Service Protocols which can explicitly describe services and the way
they can be accessed.
 The Web Service Description Language (WSDL) can be used for extension with the
required attributes for WSN service access (i.e) required accuracy, energy trade-offs, or
data-centric service descriptions.

4) WSN tunneling

 The gateways can also act as simple extensions of one WSN to another WSN. A
larger virtual WSN can be built out of separate parts, transparently “tunneling” all
protocol messages between these two networks and using the Internet as a transport
network .
 Such tunnels can be fixed or mobile network connections.
40