Cognitive Radio Research

1
Cognitive Radio Research

• Danijela Cabric
• Berkeley Wireless Research Center
• University of California at Berkeley
• Summer Retreat 2006

2
Evolution of Commercial Wireless Systems
Unlicensed (SS)
Unlicensed (mm-wave)
Unlicensed (underlay)
Overlay ? (opportunistic)
Regulation
Licensed
Multiple access CDMA cellular CS-OFDM - WLAN
Pt-2-Pt WLAN 60 GHz
Ad-hoc UWB - PAN
? Cognitive radio
Broadcast Pt-2-Pt
System type
medium
wide
ultra-wide
dynamically allocated
Bandwidth
narrow
1998
2002
2004
Time
1986
3
Spectrum Utilization Today
PSD
0 1 2 3
4 5 6 GHz

• Measurements show that there is wide range of
  spectrum utilizations
• across 6 GHz of spectrum

4
Looking Forward
Spectrum utilization ()
N gtgt K
100
N K
N of systems K of unused bands
N ltlt K
10
Time
No scarcity N ltlt K No temporal and spatial
variations Early stage of cognitive radio
networks
Medium scarcity N K Small temporal and spatial
variations More than one cognitive radio network
Significant scarcity N gtgt K Significant temporal
and spatial variations Multiple competing
cognitive radio networks
5
CORVUS Approach (Cognitive Radio approach for
usage of Virtually Unlicensed Spectrum)
Primary User Frequency Bands
Active Primary Users
Frequency
Time
Sub-Channel
Secondary User Channels
Non-frequency specific cognitive radio system
architecture
Proposed unique PHY and Link Layer system
functions
6
Cognitive Radio Communication Stack
Universal Control Channel
Data Transfer Channel
Group Control Channel
Link Layer
Link Management
MAC
Control MAC
Spectrum Allocation
Control MAC
PHY Layer
Spectrum Sensing
Wideband Data Transmission
Control PHY
7
Spectrum Sensing
Key challenge detecting weak primary user signals
Typical case SNR -10dB to -40 dB
Network cooperation
Digital Processing
Analog Processing
A/D
Spectrum Sensing is a cross-layer
functionality
8
Sensing Radio Architectures
9
Wideband Sensing Radio
wideband antenna
A/D
AGC
LNA
RF Filter
High speed A/D converter
Huge dynamic range
Multi-GHz A/D -gt Nyquist sampling High A/D
resolution (gt 12 bits)
Challenging specifications
Frequency RF MEMS filter bank Time Active
cancellation Spatial Filtering using multiple
antennas
Dynamic range reduction
10
Time Domain Interference Cancellation

• Mixed signal approach
• Flexibility offered by adaptive digital signal
  processing
• Feed forward architecture with 2 stage low
  resolution A/D conversion to achieve overall high
  resolution 2M2N ltlt 2MN

Yang, Brodersen
11
Spatial Filtering Approach
Phased antenna array

• Create beam that suppress strong primary signals
  
• Potentially enhance sensitivity in weak signal
  direction

Poon, Tse, Brodersen
12
Interference Suppression
Spectrum map Spatial vs. frequency view
1. Frequency analysis through wideband FFT
enabled by high speed A/D 2. Spatial analysis
through beam sweeping 3. Beam coefficient set to
reduce the dynamic range
Goal Equalize the Spectrum
map
13
An Example
Before dynamic range reduction

• FFT N128 points
• 4 antennas, 8 sweeps
• Avg. SNR 10 dB per sub-carrier
• 2 strong PUs
• ?145 P140dB k100 bin
• ?270 P230dB k50 bin
• Other signals random DoA
• Constraint max power10 dB

After dynamic range reduction
Beam that reduces dynamic range
14
Signal Processing for Spectrum Sensing
15
Energy Detector
Signal type sinewave or modulated data
Metric Probability of detection and false alarm
for given SNR
Resource Number of samples N needed for sensing
Matched filter N1/SNR
Theoretical result
Energy detector N1/SNR2
Sahai, Tandra
16
Testbed Setup
Joint work with Artem Tkachenko
17
Probability of detection vs. Sensing time
Sinewave signal
QPSK signal
1024 FFT averaged 200 (3.2 ms) to 52,000 (0.83
s) times
Threshold set to meet the Pfa10 over 1000
experiments
Noise power in 1 FFT bin is -103 dBm gt SNR
-7 to -25 dB
18
Sensing time for fixed Pd and Pfa
QPSK signal
Sinewave signal
Pfa10 and Pd60
FFT is coherent for sinewave, averaging is
non-coherent, thus N1/SNR1.5
SNRwall set by noise uncertainty (0.5 dB) -128
dBm for sinewave -110 dBm for
QPSK
19
Network Sensing

• Fading channel
• Multipath small scale
• Shadowing large scale
• Cooperation gain
• Pd and Pfa increase with n
• Maximized if independent
• Spatial correlation
• Function of the environment
• Limits the gain
• Threshold rules
• Noise uncertainty
• Interference from other CRs

Primary Tx
Primary Rx
p1
p1
20
Experimental Setup
Indoor wireless experiments inside BWRC
54 location on a 2m by 2m grid
21
Cooperation gains
4MHz QPSK signal (-30dBm)
Sinewave signal (-40dBm)
System prob. of detection and false alarm
monotonically increase with n
Pd63 and Pfa10 and n 5 QD99 and QF10
QD1-(1-Pd)n
QF1-(1- Pfa)n
Gains are higher for QPSK due to frequency
diversity
22
Threshold rule
Fixed threshold is sub-optimal, reduces the gain
by 15-25
Each radio must estimate local noise and
interference gt sensing overhead
23
Spatial correlation
Cooperation gain improves with distance
24
More powerful detectors
25
Theoretical Performance
26
Robustness Complexity of Feature Detectors

• Interference from secondary
• Strong signals in adjacent bands
• Receiver nonlinearity, Cyclostationary noise,
  Phase noise
• Coherence time of the channel
• Implementation complexity O(N2NlogN)

27
Wideband Transmission
Dynamic Spectrum Allocation
PU present
PU absent
CR1
CR3
CR2
f1
fN
Spectrum pool

• What access scheme can assign
• ANY sub-channel ?? ANY CR user
• If we restrict one user per sub-channel
• Orthogonal Frequency Division Multiple Access
  (OFDMA)

28
OFDMA
29
Interference to Primary Users

• Observations from the point of PU system

Primary user receiver is designed for this case
Primary user PU

• PU is neither OFDMA nor synchronized with CR
  system ? no orthogonality!
• CR Tx signal decays slowly 13.6dB first side lobe

Adjacent Primary PU1
Adjacent Primary PU2
Co-channel primary PU3
30
Interference Suppression Techniques
Guard bands
Power Control
Windowing
Cancellation Carriers
31
Transmitter Architecture
Guard Bands, Power control, Cancellation Carriers
d(1)
g(1)
x(1)
Windowing, Filtering
d(2)
g(2)
x(2)
IFFT
Data Source
S/P
P/S
Cyclic Prefix
D/A
.
d(N)
g(N)
x(N)
Optimization problem
Given - Interference limit I -
Bandwidth B - BER requirement
Maximize - Rate - Power
efficiency
32
Conclusions

• Opportunistic spectrum use requires innovation
  in
• Radio front-end architectures (active
  cancellation, spatial filtering, and RF MEMs
  filter banks)
• Signal processing for weak signal detection
  (pilot, energy, and cyclostationary feature
  detectors)
• Network cooperation (robust thresholds and
  combining)
• Wideband transmission (OFDMA optimized through
  guard bands, power allocation, cancellation
  carriers , or windowing)