Transcript San Onofre Nuclear Generating Station

Radio Channel Quality in
Industrial Sensor Networks
Photo: San Onofre Nuclear Generating Station
Daniel Sexton, Jay Werb
SICon 05
February 9th 2005
1
2
“Generic”
Mesh Network Features
•
•
•
•
•
Bi-directional  Acknowledgements and commands
Multi-hop  Extend solutions and resiliency
Self-healing  Reliable and flexible
Standards  IEEE 802.15.4
Secure  Symmetric link to link encryption
3
Bridge nodes interface
the sensor network to
the IP world
Anatomy of an installed
system
Mesh nodes form a
reliable backbone for
routing sensor data
Sensors connect to
end nodes to access
the network
4
Value Proposition
Δ = 25.61°C
Motor Replaced
5
IEEE 802.15.4
Standard
• Desirable features
–
–
–
–
–
250 kbps → Flexibility, low duty cycle
2.4 GHz → International
Lightweight MAC → low complexity
Simple ASIC → many sources → low cost
Network Security
• But will it work in a harsh factory environment?
6
Does the promise match the reality?
• Research focuses on radio reliability in factories
– Channel fading & multipath
– Channel coherence
– Radio performance
• In simulation
• On the wire
• Statistical: lost packets in factories
• Will channel diversity help?
7
Channel Fading
• Multipath effects
– Varies by position
– Varies by frequency
– Varies over time
• Overcome with diversity
– Path diversity
• Costs more routing nodes
– Frequency diversity
• Free with certain protocols
8
Fading Factors
Path Loss: L(dB) = 40 +10*N*LogD
L= path loss at 2.4GHz
D = distance in meters
N=Exponential Path Loss Factor N=2 for free space
Observed values of N: 1.3<N<3.7
Factors: Building Construction, Channel Obstructions
This equation represents a single static channel
L(dB) = 40 +10*Ne *LogD
Ne = Value based on installation type
Large Scale:L(dB) = 40 +10*Ne *LogD + Ll
Ll=RV based on loss from Obstructions (Walls, Doors, Ceilings)
Ll usually a log normal distribution – from installations of same type
Small Scale: L(dB) = 40 +10*Ne*LogD + Ls + Ll
Ls = RV based on measured Channel Characteristics
Both in time and space (Rayleigh, Rician, Nakagami, etc).
The better we characterize Ne the less variance in Ll
9
Frequency Diversity
In situ experiment
6
1
5
2
4
3
IEEE 802.5.4 Channels
10
Diversity Gains
NLOS Channel
Rayleigh Faded
Methods to Obtain Diversity Gain (Small Scale)
• Frequency Diversity (Greater than Coherence Bandwidth)
• Good Margin for fading and interference
• Path Diversity (Greater than 1 wave length)
• Good Margin for fading, some for interference
• Time Diversity (Greater than Coherence Time)
• Good Margin for interference, some for fading
We use all Three
Need to quantify Large Scale Diversity Gain
11
Power Level
• How much power is enough?
– 1 mW (0 dBm) is typical power level for 802.15.4
• Not enough for target 100 meter range
• Especially when no line of sight available
– Interference from Bluetooth and WiFi
• Need similar power level to be heard
• Supporting tests and simulations to be published
– Regulatory limits
• 36 dBm in US; 20 dBm in Europe (FH); 10 dBm some countries
– Transmitter energy consumption
• Transmitter percentage ~35% at 15 dBm; inflection point
• Tested at 15 dBm
– Seems about right; more field experience needed
12
Covering the Bases
Static
Multipath
Spatial Diversity
(Mesh Routing)
Frequency Diversity
(Channel Hopping)
Temporal Diversity
(Retries)
Transmit Power
(15 dBm)
Overall Risk
Coverage
Time Variant
Multipath
Static
Interference
Time Variant
Interference
(e.g. Microwave)
(Other Networks)




















• Assumptions
–
–
–
–
Spatial diversity supports a few paths
Frequency diversity supports many channels
Path diversity directs signal away from interferers
Interference is channel limited
13
Early Results
Summary
• Industrial protocols should support diversity
– Diversity gains >>10 dB
– Frequency diversity is “free”
– Some path diversity also provides redundancy
• Increase 2.4 GHz radio power to about 15 dBm
– Bluetooth/WiFi coexistence
– Hostile radio environment, line of sight often unavailable
– Range consistent with scale of industrial applications
• Likely range somewhat less than 100 meters
– Multihop architecture is necessary
• More testing needed in a wider variety of sites
14