1

EENG 6014, Advanced

Telecommunications Engineering

5. Energy Efficiency in Wireless

Networks

Yihenew Wondie (Dr. Eng.)

May, 2016
AAiT
 Outline
2

 Energy Efficiency in Mobile Networks

 Measuring Greenness
 Energy Saving in Base stations
 Network Planning
 Energy Efficiency in Future Generation Wireless Systems
 Broader Perspective
 Tradeoffs in Cellular Network Design
 Cellular Interference Control and Frequency Reuse

 Wireless Sensor and Green Networks

 Energy Constrained Nodes
 Cross-layer Design in Sensor Networks
 MAC in Sensor Networks
 Energy Efficiency in MAC
3

Energy-Efficiency in Mobile Networks

3
 Measuring Greenness
4

Energy Efficiency Metrics:

 Facility-level: High-level systems where equipment is deployed

 Equipment-level: Evaluate performance of an individual

equipment

 Network-level: Assess the performance of equipment with the

consideration of features and properties
Energy Saving in Base Stations
5

 Minimizing BS Energy Consumption

 Improvements in Power Amplifier: Switch mode PA
 Power Saving Protocols: Sleep Mode

 Energy Aware Cooperative BS Power

Management
 CellZooming: BS can adjust its size according to
network or traffic situation
 Renewable Energy Resources
 Solar and Wind Energy instead of off-grid sites
 Network Planning
6

Heterogeneous Network Deployment:

Based on smaller cells such as micro, pico and femtocells
Advantages:

 Energy consumption of the network can be reduced up to

60% compared to a network with macro-cells only
 Smaller cells use higher frequency bands to provide high
data rates
Draw Backs:
 The efficiency of macro-cell Base Station might be reduced
 Energy Efficiency in Future Generation
Wireless Systems
7

Enabling Technologies:
Green Communication via Cognitive Radio:

 Purpose:Collect information on the spectrum usage

and access the unused frequency bands
 Manage the spectrum optimally and dynamically

Green Communication delivered by Cooperative

Relays:
 Purpose: Improvement in throughput and coverage
 Provide service to more users using less power
 Energy Efficiency in Future Generation
Wireless Systems
8

 Low Energy Spectrum Sensing

 Cognitive Radio: design low complex cyclostationary
detectors instead of energy detectors
 Energy Efficient Resource Management with
Applications in Heterogeneous Networks
 Relay placement, optimal relay selection criterion
 Addressing Uncertainty Issues
 Cognitive Radio: did not take into account the effect of
imperfect sensing
 Cooperative Relays: optimal relays with perfect CSI
 Broader Perspective
9

Ideas to achieve energy efficient cellular network:

 Statistical Power Profiles:
Dynamic algorithms can be designed to switch BS to
different power profile appropriate for that time of the day
 Smart Grid Technology:
Adding measurement sensors which can update the status of
BSs then transmit them to other BSs and smart grid control
system
 Embodied Energy:
Reducing the number of BS since embodied energy accounts
for a significant proportion of energy consumed by the BS
 Key Design Constraints
10

Tombaz, A. Västberg and J. Zander, Energy and Cost Efficient Ultra High Capacity Wireless Access", IEEE Wireless
Communication Magazine, vol. 18, no. 5, pp. 18- 24, October 2011.
 Tradeoffs in Cellular Network Design
11
 DE – EE
12

 DE: a measure of system throughput per unit of deployment

cost
 An important network performance indicator for mobile
operators.
 DE consists of
 Capital expenditure (CapEx)

 Infrastructure costs (base station equipment, backhaul

transmission equipment, site installation, and radio
network controller equipment.)
 Operational expenditure (OpEx)

 electricity bill, site and backhaul lease, and operation

and maintenance cost
 Conflicting Design Rules
13

 DE:
 Large cell radius to save expenditure on site rental, base station
equipment, and maintenance, etc.
 EE:
 Smaller cell radius to save transmit power
 Example :
 By shrinking the cell radius from 1, 000 m to
250 m, the maximum EE of the HSDPA Network
will be increased from 0.11 Mbits/Joule to 1.92
Mbits/Joule, respectively
 Some Practical Considerations
14

 Previous DE-EE tradeoff assumes

 deployment cost scales continuously and proportionally with the cell
radius.
 only transmit power

 In reality
 the equipment cost does not scale proportionally with the target cell
size;
 the total network energy includes both transmit-dependent energy
(e.g. power consumed by radio amplifier) and transmit-independent
one (e.g. site cooling power consumption).
 The relation of DE and EE may deviate from the simple tradeoff curve
and become more complex when considering practical aspect
 BW-PW
15

• Given a data transmission rate R

• Expansion of signal P  WNo (2  1)
W

bandwidth reduces transmit

power and achieves better
energy efficiency.

Converge to
 BW-PW: In Practice
16

 Given a data transmission rate

Expansion of signal bandwidth
reduces transmit power and
achieves better energy
efficiency.

 In practice:
 the circuit power consumption,
such as filter loss, actually
increases with the system BW
 DL-PW
17

 According to Shannon:
Time needed for sending one
1 bit, i.e. delay

Pb  WNotb (2 tbW
 1)
Power needed for
reliable delivery of
one bit
 DL-PW: In Practice
18

 Other device power needed to enable operation:

1

Pb  WNotb (2 tbW
 1)  F (t b )
Other device power

18
 DL-PW: One Step Further
19

 Traffic dynamics
 Delay include both the waiting time in the traffic queue
and the time for transmission

Avg. Power
[Berry and Gallager 2002]

m(P(t), H(t))
A(t)

Min. Avg. Energy Required for Stability Avg. Delay

Open problem with practical considerations
 SE-EE
20

 In single-user scenario
• SE SE  log 2 (1 
P
)
WN0 EE
1
P
• EE W log 2 (1 
N 0 ln 2
)
WN 0
 EE 
P
• SE-EE relationship
 SE
 EE  
(2  1) N 0
SE
0
SE
 Cellular Interference Control & Frequency Reuse
21
 Since advent of CDMA, power BS 1
control + reuse 1 is the basis for
cellular interference management :
 All links transmit on entire bandwidth
(reuse 1).
 Each link sees average of interference
from many small sources.
 Rate adaptation, power control used to
adapt to interference.

 Many effective algorithms:

 Problems are generally convex
 Solutions implementable in current
standards BS 2
 Interference: Friend or Foe?
22

 If treated as noise: Foe

P Increases BER
SNR 
NI Reduces capacity

 If decodable : Neither friend nor foe

 If exploited via cooperation and cognition: Friend

(especially in a network setting)
 Lessons from the Gaussian Interference Channel
23

• Reuse 1 and power control has been the dominant model for
cellular systems, esp. since CDMA
• But, stronger interference conditions in Het Nets requires more
sophisticated methods.
 Two Commonly Used Strategies for Interference
Coordination
24

Power control with Reuse 1:

• Links transmit at same time & freq
• Tradeoff link conditions via power
control
• Basis of 3G and 4G cellular systems
• Works well when interference is low.

Orthogonalization:
• Links transmit at different times or freq
• Tradeoff link conditions via bandwidth
allocation
• Basis of 802.11 standards and earlier
2G cellular systems.
• Works well when interference is high or
unpredictable.
 The New Interference Environment Femtocells &
Heterogeneous Networks
25

4 2

3
1 6 HNB
NB B

UE A2
UE A2
5
UE
Macro UE A1 HNB
NB A
A UE UE
Macro
UE B1 Macro
NB
Macro

Macrocell A NBartment
Apartment A NBartment
Apartment B
Macrocell B

 Next-gen networks scaling capacity via heterogenous

networks:
 Mixtures of femtocells & relays

 A new interference environment:

 Deployment is unplanned.
 Cell selection may be restricted (backhaul, access restrictions)
 Limited ability to coordinate interference control 25
 Dynamic Interference Coordination
26

 Dynamic orthogonalization provides a distributed method for

coordinating interference
 Greatly improved capacity
 Distributed implementation

Source: S.Rangan, R. Madan, “Belief Propagation Methods for Intercell Interference Coordination , July 2010
 27
Wireless Sensor and “Green” Networks

• Smart homes/buildings
• Smart structures
• Search and rescue
• Homeland security
• Event detection
• Battlefield surveillance

 Energy (transmit and processing) is driving constraint

 Data flows to centralized location (joint compression)
 Low per-node rates but tens to thousands of nodes
 Intelligence is in the network rather than in the devices
 Similar ideas can be used to re-architect systems and networks to be green
 Energy-Constrained Nodes
28
 Each node can only send a finite number of bits.
 Transmit energy minimized by maximizing bit time
 Circuit energy consumption increases with bit time

 Introduces a delay versus energy tradeoff for each bit

 Short-range networks must consider transmit, circuit,

and processing energy.
 Sophisticated techniques not necessarily energy-efficient.
 Sleep modes save energy but complicate networking.

 Changes everything about the network design:

 Bit allocation must be optimized across all protocols.
 Delay vs. throughput vs. node/network lifetime tradeoffs.

 Optimization of node cooperation.

 Cooperative Compression in Sensor
29
Networks

 Source data correlated in space and time

 Nodes should cooperate in compression as well as
communication and routing
 Joint source/channel/network coding
 What is optimal for cooperative communication:
 Virtual MIMO or relaying?
 Green” Cellular Networks
30

Pico/Femto How should cellular

Coop
systems be redesigned
MIMO
Relay
for minimum energy?

Research indicates that

DAS
significant saving is possible

 Minimize energy at both the mobile and base station via

 New Infrastructures: cell size, BS placement, DAS, Picos, relays
 New Protocols: Cell Zooming, Coop MIMO, RRM, Scheduling,
Sleeping, Relaying
 Low-Power (Green) Radios: Radio Architectures, Modulation,
coding, MIMO
 Cross layer Design in Sensor Networks
31

 Application
 Network
 Access
 Link
 Hardware

Energy consumption at each layer of the protocol stack must be

considered in the design
Cross-Layer Tradeoffs under Energy Constraints
32

 Hardware
 Models for circuit energy consumption highly variable
 All nodes have transmit, sleep, and transient modes
 Short distance transmissions require TD optimization
 Link
 High-level modulation costs transmit energy but saves
circuit energy (shorter transmission time)
 Coding costs circuit energy but saves transmit energy

 Access
 Transmission time (TD) for all nodes jointly optimized
 Adaptive modulation adds another degree of freedom
 Routing:
 Circuit energy costs can preclude multihop routing
 Key Assumptions
33

 Narrow band, i.e. B<<fc

 Power consumption of synthesizer and mixer
independent of bandwidth B.
 Peak power constraint
 L bits to transmit with deadline T and bit error
probability Pb.
 Square-law path loss for AWGN channel

(4d ) 2
Et  Er Gd , Gd 
G 2
 Multi-Mode Operation
Transmit, Sleep, and Transient
34

 Deadline T: T  Ton  Tsp  Ttr

 Total Energy:

E  Eon  Esp  Etr ( Esp  0, Etr  2 PsynTtr )

Transmit Circuit Transient Energy

 (1   ) PtTon  PcTon  2 PsynTtr

where  is the amplifier efficiency and

Pc  2 Pmix  2 Psyn  PLNA  PIFA  Pfil  PDSP ,
 Energy Consumption: Uncoded
35

 Two Components
 Transmission Energy: Decreases with Ton & B.
 Circuit Energy: Increases with Ton

 Minimizing Energy Consumption

 Finding the optimal pair ( B, Ton )
 For MQAM, find optimal constellation size (b=log2M)
 Total Energy (MQAM)
36
 Total Energy (MFSK)
37

MQAM:
-45dBmJ at 1m
-33dBmJ at 30m
 Energy Consumption: Coded
38

 Coding reduces required Eb/N0

 Reduced data rate increases Ton for

block/convolutional codes

 Coding requires additional processing

 MQAM Optimization
39

 Find BER expression for coded MQAM

 Assume trellis coding with 4.7 dB coding gain
 Yields required Eb/N0
 Depends on constellation size (bk)

 Find transmit energy for sending L bits in Ton sec.

 Find circuit energy consumption based on uncoded

system and codec model

 Optimize Ton and bk to minimize energy

 Coded MQAM
40

Reference system has bk=3 (coded) or 2 (uncoded)

90% savings
at 1 meter.
 MFSK Optimization
41

 Find BER expression for uncoded MFSK

 Yields required Eb/N0 (uncoded)
 Depends on b, Ton a function of b.

 Assume 2/3 CC with 32 states

 Coding gain of 4.2 dB
 Bandwidth expansion of 3/2 (increase Ton)

 Find circuit energy consumption based on uncoded

system and codec model

 Optimize b to minimize total energy

 Comparison: MQAM and MFSK
42
 Medium Access Control in Sensor Nets
43

 Important attributes of MAC protocols

1. Collision avoidance
2. Energy efficiency
3. Scalability in node density
4. Latency
5. Fairness
6. Throughput
7. Bandwidth utilization
MAC Impact on Sensor Networks
44

• Major sources of energy waste

• Idle listening when no sensing events, Collisions,
Control overhead, Overhearing

0.14 0.018
(Joules/Node/Received Event)

(Joules/Node/Received Event)
Diffusion

Average Dissipated Energy

Average Dissipated Energy

0.12 0.016 Flooding

Flooding Omniscient Multicast 0.014
0.1
0.012
0.08 0.01
0.06 0.008 Omniscient Multicast
0.006
0.04
0.004 Diffusion
0.02 0.002
00 00 50 100 150 200 250 300
50 100 150 200 250 300
Network Size Network Size
Over 802.11-like MAC Over energy-aware MAC
 Identifying the Energy Consumers
45

Power consumption of node subsystems

20

15
Power (mW)

10

0
CPU TX RX IDLE SLEEP
SENSORS
RADIO

ETX  ERX  EIDLE  ESLEEP

 Need to shutdown the radio
 Energy Efficiency in MAC
46

• Major sources of energy waste

• Idle listening
• Long idle time when no sensing event happens
• Collisions
• Control overhead
• Overhearing Common to all wireless networks

• Try to reduce energy consumption from all above

sources
• TDMA requires slot allocation and time
synchronization
• Combine benefits of TDMA + contention protocols
 47

AN ENERGY-EFFICIENT MAC PROTOCOL FOR

WIRELESS SENSOR NETWORKS
 Introduction to WSN
48

 Applications:
 Nodes cooperate for a common task
 In-network data processing
 Differences between WSN and ad-hoc network
 Battery powered nodes  Energy efficiency
 Large quantity of densely deployed nodes
 This dense deployment brings high degree of
interactions
 Resources constraint
 Auto configuration and auto organization
 Design Considerations
49

 Level 1 issues
 Collision avoidance-a basic task of MAC protocols
 Good scalability

 Energy efficiency

-Often difficult recharge batteries or replace them

-Prolonging the life-time is important
 Level 2 issues
 Latency, fairness, throughput, bandwidth
 Energy Inefficiency Sources
50

 Collision
 Corrupted packets must be retransmitted and it
increases energy consumption
 Overhearing
 Receive packets destined to others
 Energy Inefficiency sources
51

 Control packet overhead

 Idle listening
 Listening to receive possible traffic that is not sent
Dominant energy inefficiency factor in WSN
 Consumes 50-100% of the energy required for
receiving
 Existing MAC protocol design
52

 Contention based protocols

 IEEE802.11 distributed coordination function (DCF)
 Each node contends for the medium as necessary and
wastes a lot of energy in idle listening
 PAMAS (Power Aware Multi-Access Protocol)
 Asking separate radio channel for RTS/CTS
 Does not address the issue of reduce idle listening
53

 TDMA based protocols

 Advantages
 lower energy conservation when compared to contention
based as the duty cycle of the radio is reduced and no
contention overhead
 Problems
 Requires nodes to form real communication clusters and
managing inter-cluster communication is difficult
 It is not easy to change the slot assignment dynamically,
hence scalability is not as good as contention based
 S-MAC (Sensor Medium Access Protocol)
54

 Design
 Goal
 Reduce energy consumption
 Support good scalability and collision avoidance

 Solutions to energy inefficiency issues

 Collision avoidance - using RTS and CTS
 Overhearing avoidance - switching the radio off when the transmission is not meant
for that node
 Control overhead - Message Passing
 Idle listening - Periodic listen and sleep
 Network assumptions (1)
55

 Composed of many small nodes deployed in an ad

hoc fashion
 Most communication will be between nodes as
peers, rather than to a single base station
 Nodes must self-configure
 Network assumptions (2)
56

 Dedicated to a single application or a few collaborative

applications
 Involves in-network processing to reduce traffic and thereby
increase the life-time
 This implies that data will be processed as whole messages at
a time in store-and-forward fashion
 Hence packet or fragment-level interleaving from multiple
sources only delays overall latency
 Applications will have long idle periods and can tolerate
some latency
 Periodic listen and Sleep (1)
57

 Problem: Idle listing wastes a lot of energy

 Solution: Periodic listen and sleep

 Turn off radio when sleeping

 Reduce duty cycle to ~ 10% (200ms on/2s off)
 Periodic listen and Sleep (2)
58

 Not all neighboring nodes can synchronize together

 Two neighboring nodes (A and B) can have different
schedules if they are required to synchronize with
different node
 Periodic listen and Sleep (3)
59

 If a node A wants to talk to node B, it just waits

until B is listening
 If multiple neighbors want to talk to a node, they
need to contend for the medium
 Contention mechanism is the same as that in
IEEE802.11 (using RTS and CTS)
 After they start data transmission, they do not go
to periodic sleep until they finish transmission
 Collision Avoidance
60

 Problem: Multiple senders want to talk

 Solution: Following IEEE 802.11 ad hoc procedures
 Physical and virtual carrier sense
 Randomized backoff time
 RTS/CTS for hidden terminal problem
 RTS/CTS/DATA/ACK sequence
 Overhearing Avoidance
61

 Problem: Receive packets destined to others

 In 802.11, each node keeps listening to all transmissions from its neighbors
for virtual carrier sensing
 Each node should overhear a lot of packets not destined to itself

 Solution: Letting interfering nodes go sleep after they hear an RTS or CTS
packet
 Which nodes should sleep?
 All immediate neighbors of sender and receiver

 S-MAC lets interfering nodes go to sleep after they hear an RTS or CTS

 DATA packets are normally much longer than control packets

 How long?
 The duration field in each packet informs other nodes the sleep interval

 After hearing the RTS/CTS packet destined to a node, all the other
immediate neighbors of both the sender and receiver should sleep until the
NAV (Ntk Allocation Vector) becomes zero
 Message Passing
62

 Problem: Sensor net in-network processing requires entire

message
 Solution: Don’t interleave different messages
 Long message is fragmented & sent in burst
 RTS/CTS reserve medium for entire message
 Fragment-level error recovery — ACK
— Extend Tx time and re-transmit immediately if no ACK is received
 Advantages
 Reduces latency of the message
 Reduces control overhead
 Disadvantage
 Node-to-node fairness is reduced, as node with small packets to
send has to wait till the message burst is transmitted
63