Download Lecture3-wireless

CS 410/510 Sensor Networks
Portland State University
Lecture 3
Wireless Communication
Source Acknowledgements
• Alberto Cerpa and Deborah Estrin
• Alec Woo and David Culler
• Jerry Zhao and Ramesh Govindan
5/25/2017
Nirupama Bulusu
2
Outline
• IEEE 802.15.4 Wireless Communication
Standard
• Single Hop packet loss characteristics
– Axes
• Environment, distance, transmit power, temporal
correlation, data rate, packet size
5/25/2017
Nirupama Bulusu
3
IEEE 802.15.4: Why the need?
• Sensor and Personal Area Networks
require
– Low Power Consumption
– Minimal Installation Cost
– Low Overall Cost
• Existing Technologies
– Wired
– 802.11 (WiFi) and Bluetooth
History
• Combination of Two Standards Groups
– ZigBee Alliance: “an association of companies
working together to enable reliable, costeffective, low-power, wirelessly networked,
monitoring and control products based on an
open global standard.”
– IEEE 802 Working Group 15
• Task Group 4 formed in December 2000
– Low-rate Wireless Personal Area Network
System Layering
High-Level Characteristics
Network Layer Guidelines
• 802.15.4 Specification does not address
Network Layer
• Expected to be self-organizing and selfmaintaining to minimize cost to user
• Two Network Topologies Supported:
– Star Topologies
– Peer-to-Peer Topologies
Topology Formations
Data Link Layer
• Two Parts
– Logical Link Control (LLC)
• Standard among many 802.x standards
• Communicates with MAC through SSCS
• Proprietary LLC’s can communicate directly
– MAC Sublayer
• Data Service - Common Part Sublayer
• Management Service – Management Entity
MAC Frame Format
Superframe Beacons
• Time between beacons divided in 16 time slots
• Can be used to provide bandwidth guarantees
• Contention-free period and duration of
superframe announced in beacon
Additional MAC Features
• Channel Access Mediums
– Slotted CSMA-CA
– Unslotted CSMA-CA
• Acknowledgements
• Security
– No security
– Access Control Lists
– Symmetric Key Security
Physical Layer
• Two Potential Physical Layers
– 868/915Mhz
– 2.4Ghz
– Direct Sequence Spread Spectrum
– Same Packet Structure
• 27 Frequency Channels Total
• Dynamic Channel Selection left to network
layer
Physical Layer Packet Structure
Other Physical Layer Features
• Modulation
– 868/915 – Binary Phase Shift Keying
– 2.4 – Offset Quadrature Phase Shift Keying
• Sensitivity and Range
– 868/915  -92 dBm
– 2.4  -85 dBm
– 10-20m typical range
MicaZ and Sun SPOT Platforms
Outline
• IEEE 802.15.4 Wireless Communication
Standard
• Single Hop packet loss characteristics
– Axes
• Environment, distance, transmit power, temporal
correlation, data rate, packet size
5/25/2017
Nirupama Bulusu
18
Zhao’s Study of Packet Loss
• Hardware
– Mica, RFM 433MHz
• MAC
– TinyOS Mac (CSMA)
• Encoding
– Manchester (1:2)
– 4b/6b (1:1.5)
– SECDED (1:3)
• Environment
– Indoor, Open Structure, Habitat Environment
5/25/2017
Nirupama Bulusu
19
Indoor is the Harshest
5/25/2017
Nirupama Bulusu
20
Indoor is the Harshest
• Linear topology over a hallway (0.5/0.25m spacing)
• 40% of the links have quality < 70%
• Lower transmit power
– yields smaller tail distribution
• SECDEC
– significantly helps to lower the heavy tail
5/25/2017
Nirupama Bulusu
21
Packet Loss and Distance
• Gray/Transitional Area
–
–
–
–
ranges from 20% to 50% of the communication range
Habitat has smaller communication range?
Other evidence (Cerpa et al., Woo et al.)
RFM: BAD RADIO??
5/25/2017
Nirupama Bulusu
22
ChipCon Radio (Cerpa et al.)
Mica On Ceiling
• Higher transmit power doesn’t eliminate transitional
region
– Range in (a) and (b) are the same?
• Indoor RFM result is worst than that in Zhao’s work
– cannot even see the effective region
5/25/2017
Nirupama Bulusu
23
Can better coding help?
• SECDED is effective if start symbol is detected but
does not increase “communication range”
– Bit error rate (BER) is higher in transitional region
• Missing start symbol is fatal
Nirupama
Bulusu
– Better coding for start
symbol?
5/25/2017
24
Loss Variation (Cerpa et al.)
• Variation over distance and over time
– binomial approximation for variation over time?
• Zhao shows that SECDED helps decrease the
5/25/2017
Nirupama Bulusu
variation over distance
(but very large SD here)25
Packet Loss vs. Workload
• Packet loss increases as network load increases
– But what is the network load?
– How many nodes are in range?
• Not sure!
• Is 0.5 packets/s already in saturation?
• Difficult to observe is it hidden node terminal
5/25/2017
Nirupama Bulusu
26
Packet Loss vs. RSSI
• Low packet loss => good RSSI
– But not vice versa
– Too high a threshold limits number of links
• Network partition??
5/25/2017
Nirupama Bulusu
27
Other Findings
• Correlation of Packet Loss
– correlation at the gray (transitional) region for indoor
– Habitat: much less
• Independent losses are reasonable
• 50%-80% of the retransmissions are wasted
– Neighbor = hear a node once
• Asymmetric links are common
– > 10% of link pairs have link quality difference > 50%
– Cerpa et al.
• Moving a little bit doesn’t help
• Swap the two nodes, asymmetrical link swaps too
– i.e. not due to the environment
5/25/2017
Nirupama Bulusu
28
Packet Size (Cerpa et al.)
• Loss over distance is relatively the same
for different packet size (25 bytes and 150
bytes) at different transmit power
5/25/2017
Nirupama Bulusu
29
Lessons to Take Away
• Who to blame?
– Radio?
• Similar results found over RFM and ChipCon radio
• Hardware calibration! Yeah! 
– Base-band radio
• Multi-path will remain unless spread-spectrum radio is used
– But 802.11 is also not ideal (Decouto et al. Mobicom 03)
• What is the effective communication range?
– What does it mean when you deploy a network
• What defines a neighbor?
• Why study high density sensor network?
5/25/2017
Nirupama Bulusu
30