pptx - Stanford University Networking Seminar

Report
Interference Centric Wireless Networks
Sachin Katti
Assistant Professor
EE&CS, Stanford University
Interference is Everywhere
WiFi
Zigbee
How to maximize throughput in the
Bluetooth
presence of interference?
Current Approach to Interference
• Fears & avoids interference at all costs
• Impacts all aspects of wireless design
– Radios are half duplex
– MAC protocols try to schedule one link at a time
– Coexisting networks use different channels if possible
– …….
Current Approach Cannot Scale
Dense and chaotic wireless deployments
 Interference is unavoidable
– Hidden terminals cause collisions
– Coexisting networks interfere with each other
– Legacy interferers (e.g. microwave) …..
Moreover,
Limited spectrum + interference avoidance design
 Achievable capacity is fundamentally limited
This Talk
Fundamental rethink:
Exploit interference instead of avoiding it
High-Level Approach
• Infer interference structure
• Exploit structure to better decode interfered
packets and increase throughput
Exploiting Interference in All Contexts
• Exploiting In-Link Interference
– Full Duplex Radios (Mobicom 10,11)
• Exploiting In-Network Interference
– Rateless & Collision Resilient PHY (Sigcomm11)
• Exploiting Cross-Network Interference
– Detecting Degrees of Freedom (Sigcomm11)
Exploiting In-Link Interference:
Full Duplex Radios
Jain et al, “Practical Real Time Full Duplex Wireless”
Mobicom 2010, 2011
“It is generally not possible for radios to receive and transmit
on the same frequency band because of the interference that
results. Thus, bidirectional systems must separate the uplink
and downlink channels into orthogonal signaling dimensions,
typically using time or frequency dimensions.”
- Andrea Goldsmith, “Wireless Communications,”
Cambridge Press, 2005.
In-Link Interference  Half Duplex Radios
TX
RX
TX RX
Self-interference is millions to billions (60-90dB)
stronger than received signal
In-Link Interference  Half Duplex Radios
Analog Self Interference
Analog Received
Signal
ADC
max
Digital Self Interference
Digital Received
Signal
Tx
Rx
- max
Self-interference drowns out received signal
Our Approach
1. Infer interference structure
– Easy, we know what we are transmitting!
2. Exploit knowledge of interference structure
to subtract and decode
First Attempt: Antenna Cancellation
TX1
RX
d
TX2
d + λ/2
• Signal null at RX antenna
• ~30dB self-interference cancellation
Bringing It Together
RX
Antenna
Cancellation
Hardware
Cancellation
TX Signal
QHX220
RF
ADC
Digital
Cancellation
TX Samples
-
Baseband
+
∑
Clean RX samples
Our Prototype
Antenna
Cancellation
Digital
Interference
Cancellation
Hardware
Cancellation
Antenna Cancellation: Performance
TX1
Only TX1 Active
TX2
Antenna Cancellation: Performance
TX1
TX2
Only TX1 Active
Only TX2 Active
Antenna Cancellation: Performance
TX1
TX2
Only TX1 Active
Null
Position
Both TX1 &
TX2 Active
Only TX2 Active
Antenna Cancellation: Performance
TX1
TX2
Only TX1 Active
Null
Position
Both TX1 &
TX2 Active
~25-30dB
Only TX2 Active
Bandwidth Constraint
A λ/2 offset is precise for one frequency
TX1
RX
d
fc
TX2
d + λ/2
Bandwidth Constraint
A λ/2 offset is precise for one frequency
not for the whole bandwidth
TX1
RX
d
fc -B
fc
fc+B
TX2
d + λ/2
Bandwidth Constraint
A λ/2 offset is precise for one frequency
not for the whole bandwidth
TX1
RX
TX2
d1 + λ-B/2
d1
TX1
RX
d
fc -B
fc
fc+B
TX1
TX2
d + λ/2
RX
d2
TX2
d2 + λ+B/2
Bandwidth Constraint
A λ/2 offset is precise for one frequency
not for the whole bandwidth
TX1
RX
TX2
d1 + λ-B/2
d1
TX1
RX
d
fc -B
fc
fc+B
TX1
TX2
d + λ/2
RX
d2
TX2
d2 + λ+B/2
WiFi (2.4G, 20MHz) => ~0.26mm precision error
Bandwidth Constraint
Bandwidth Constraint
•
•
•
WiFi (2.4GHz, 20MHz): Max 47dB reduction
Bandwidth⬆ => Cancellation⬇
Carrier Frequency⬆ => Cancellation⬆
First prototype gives 1.84x throughput gain with two
radios compared to half-duplex with a single radio.
Limitation 1: Need 3 antennas
Limitation 2: Bandwidth constrained (802.15.4 works)
Limitation 3: Doesn’t adapt to environment
Our Approach
1. Infer interference structure
– Easy, we know what we are transmitting!
2. Exploit knowledge of interference structure
to subtract and decode
Poor Man’s Subtraction
Cancellation using Phase Offset
SelfInterference
∑
Cancellation
Signal
Cancellation using Phase Offset
SelfInterference
∑
Cancellation
Signal
SelfInterference
∑
Cancellation
Signal
Frequency dependent, narrowband
Cancellation using Signal Inversion
SelfInterference
∑
Cancellation
Signal
SelfInterference
∑
Cancellation
Signal
Frequency and bandwidth independent
Second Design:
Balanced to Unbalanced Conversion
+Xt/2
Xt
-Xt/2
BALUN
Traditional Design
aT
T
TX Frontend
R
R+aT
RX Frontend
1. Invert the Signal
aT
+T
-T
R
R+aT
2T
balun
TX Frontend
RX Frontend
2. Subtract Signal
aT
+T
-T
R
R+aT
Σ
R+aT-T
2T
balun
TX Frontend
RX Frontend
3. Match Signals
aT
+T
-T
R
attenuator and
delay line
R+aT
Σ
v
-vT
R+aT-vT
2T
balun
TX Frontend
RX Frontend
Can Receive If v = a!
aT
+T
-T
R
attenuator and
delay line
R+aT
Σ
v
-vT
R+aT-aT
2T
balun
TX Frontend
RX Frontend
Signal Inversion Cancellation: Wideband Evaluation
• Measure wideband cancellation
• Wired experiments
• 240MHz chirp at 2.4GHz to measure response
TX
Xt
+Xt/2
+Xt/2
∑
-Xt/2
Signal Inversion
Cancellation Setup
RX
TX
Xt
λ/2
Delay
RF
Signal
Splitter
∑
+Xt/2
Phase Offset
Cancellation Setup
RX
Higher is
better
Lower is
better
Higher is
better
Lower is
better
~50dB cancellation at 20MHz bandwidth with balun vs
~38dB with phase offset cancellation.
Significant improvement in wideband cancellation
Other advantages
RX
TX
Attenuator and
Delay Line
+Xt/2
∑
-Xt/2
Xt
TX Frontend
•
•
RX Frontend
From 3 antennas per node to 2 antennas
Parameters adjustable with changing conditions
Adaptive RF Cancellation
RX
TX
RF Reference
Attenuation
& Delay
Σ
Balun RF Cancellation
Wireless
Receiver
Wireless
Transmitter
TX Signal Path
RX Signal Path
• Need to match self-interference power and delay
• Can’t use digital samples: saturated ADC
Adaptive RF Cancellation
RX
TX
RF Reference
Attenuation
& Delay
Σ
RSSI
Balun RF Cancellation
Wireless
Receiver
Wireless
Transmitter
TX Signal Path
RX Signal Path
• Need to match self-interference power and delay
• Can’t use digital samples: saturated ADC
RSSI : Received Signal Strength Indicator
Adaptive RF Cancellation
RX
TX
RF Reference
Attenuation
& Delay
Σ
Control
Feedback
RSSI
Balun RF Cancellation
Wireless
Receiver
Wireless
Transmitter
TX Signal Path
RX Signal Path
• Need to match self-interference power and delay
• Can’t use digital samples: saturated ADC
Use RSSI as an indicator of self-interference
Adaptive RF Cancellation
RX
TX
RF Reference
Attenuation
& Delay
Σ
Control
Feedback
RSSI
Balun RF Cancellation
Wireless
Receiver
Wireless
Transmitter
TX Signal Path
RX Signal Path
Objective: Minimize received power
Control variables: Delay and Attenuation
Adaptive RF Cancellation
Objective: Minimize received power
Control variables: Delay and Attenuation
➔ Simple gradient descent approach to optimize
Bringing It All Together
RX
TX
RF Reference
Control
Feedback
Attenuation
& Delay
Σ
RSSI
Balun RF Cancellation
Baseband ➔ RF
RF ➔ Baseband
DAC
ADC
Digital Interference Cancellation
Digital Interference
Reference
Encoder
TX Signal Path
FIR filter
∆
Channel
Estimate
Decoder
RX Signal Path
Performance
•
•
WiFi full-duplex: with reasonable antenna separation
Not enough for cellular full-duplex: need 20dB more
Full Duplex Implications
• Breaks a fundamental assumption in wireless
• Could eliminate the need for paired spectrum
• Impacts higher layer design
– Reduce control overhead (Radunovic et al)
• Other applications
– Security & Privacy (Gollakota et al)
• Many more …..
Exploiting In-Network Interference
Rateless & Collision-Resilient Codes
Gudipati, Katti “Strider: Automatic Rate Adaptation”
SIGCOMM 2011
In-Network Interference  Collisions
Carrier sense failure  Packet collisions and loss
Current Approach: Conservative backoff, RTS/CTS
Our Approach: Infer Interference Structure
• Current approach:
– Measure channel SINR and pick modulation, coding rate
– If channel SINR < decoding threshold, decoding fails
– Collision  SINR < decoding threshold
• Key insight: Novel rateless codes for wireless
• Rateless code  no need to know SINR in advance,
automatically achieves optimal throughput
Our Approach: Infer Interference Structure
• Key technique: Novel rateless codes
• P1 acts as interference to P2 and vice versa
Decode P1
P1
P1
P2
1. Use rateless code to decode P1
 Infer interference
Our Approach: Exploit Interference Structure
• Key insight: Exploit rateless code to decode one
packet, subtract it and decode next packet
Decode P1
P1
P1
P2
_
Subtract
interference
__
1. Use rateless code to decode P1
 Infer interference
2. Subtract P1 from received signal
and decode P2
Exploiting Cross-Network Interference
Detecting Degrees of Freedom
Hong, Katti “DOF: A Local Wireless Information Plane”
SIGCOMM 2011
Cross-Network Interference  Coexistence
WiFi
Zigbee
How to maximize throughput in the presence of
Bluetooth cross-network interference?
Our Approach: Infer Interference Structure
DOF infers coexisting interference structure
1. The protocol types operating in the local vicinity
2. The spectrum occupancy of each type
Heart Monitor
0
°
0
°
AoA
AoA
180°
WiFi AP
180°
3. The spatial directions of each type
2.3
2.3
GHz
GHz Freq
2.5
GHz
Smart Transmitter
Smart Receiver
Microwave
2.3
2.3
GHz
GHz
Freq
2.5
2.5
GHz
GHz
Key Insight
“Man-made” signals  hidden repeating patterns that
are unique and necessary for operation
CP
Data
CP
Data
CP
Data
…………………….
Repeating Patterns in WiFi OFDM signals
Time
Repeating Patterns in Zigbee signals
Leverage unique patterns to infer 1) type, 2)
spectral occupancy, and 3) spatial directions
Our Approach: Exploit Interference Structure
Exploit interference structure knowledge
• Policy 0 – Only use unoccupied spectrum
• Policy 1 – Use unoccupied spectrum + mw oven spectrum
• Policy 2 – Use unoccupied spectrum + mw oven
spectrum + compete for WiFi spectrum
WiFi
AoA
PSD
Frequency
2.3
GHz
Freq
2.5
GHz
Smart Rx
AoA
Smart Tx
Heart Monitor
(ZigBee Based)
2.3
GHz
Microwave
Freq
2.5
GHz
To Conclude
Future: dense, chaotic and limited spectrum
Interference is the dominant determinant of
future wireless network capacity
• Point to point link speeds are close to Shannon
Our approach: Fundamental rethink of wireless to
manage and exploit interference
• Increase concurrency  Increase network capacity

similar documents