Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Estimation and Capacity of Channels in
Smart Antenna Wireless Communication Systems
Murat Torlak
Final Oral Examination
Dept. of Electrical & Computer Eng.
The University of Texas at Austin
Dr. Guanghan Xu Dr. Brian L. Evans
Dr. Edward J. Powers Dr. Mircea D. Driga
Dr. Wolfhard Vogel Dr. Srivinas Bettadpur
1
MOTIVATION
Communication Anywhere-Anytime-Anytype
Increasing Demand for Wireless Services
Unique Problems compared to Wireline communications
2
SMART ANTENNA SYSTEMS
Wireless systems broadcast signals across wide areas to individual users
Most of energy is wasted while little reaches
the intended user
Spatial processing characterizes the RF environment
through antenna arrays and innovative digital signal processing sotware
Additional dimension of functionality
3
ADDED DIMENSION IN WIRELESS COMMUNICATION
Space division compelements existing air interfaces
Frequency Frequency Frequency
Tim
e
Tim
e Tim
e
Tim
e
FDMA TDMA CDMA
NEW DIMENSION--SPACE
Tim
e
Tim
e
Tim
e
Spac
e
Spac
e
Frequency Frequency Frequency
Spac
e
SPACE DIVISION MULTIPLE ACCESS (SDMA)
AMPS GSM IS-95
CochannelInterference
CochannelInterference
CochannelInterference
toward intended users – away from unintended users
4
WIDEBAND CHANNEL MODEL
The symbol period, , is comparable to multipath spread,
max
Channel Impulse Responseh(t)
time
ampl
itude
T (Symbol Period)
Single user case with paths to receiver
Vector channel model with an antenna array
5
UPLINK PROCESSING FOR WIRELESS BASESTATIONS
Objective: Combine the intended user signals while suppressing the
co-channel signals.
How? Estimate the wireless channel parameters ( ) based on the
data received
s (n)
s (n)
s (n)1
2
h (n)
h (n)
h (n)
P
2
1
P
x(n)
Channels BaseStationSignals
User
6
CODE DIVISION MULTIPLE ACCESS (CDMA)
Uses available bandwidth efficiently
Distinguishes users by codewords
Modulate &Code
Spreading Code Noise & OtherCDMA Signals
Decode&
Demod
DespreadingCode
Channel&
Delay
Synchronous CDMA All signals must be synchronized
Narrowband CDMA
RAKE receiver coherently resolves multipath fading
Wideband CDMA (W-CDMA)
Emerging third-generation standards
Estimate (vector) FIR channel parameters for symbol estimation
7
CDMA SYSTEMS WITH PERIODIC SPREADING
Single receiver with Users
Blindly estimate
Assume knowledge of spreading codes
Assume knowledge of propagation delays .
8
DATA FORMULATION
Baseband signal, single receiver, P users:
denotes the user index
is the number of users
are the information symbols
is the symbol duration
is a signature waveform
is the spreading code of user
is the code length
are the channels
9
MATRIX FORM
Represent the data vector of samples with a symbol period in a matrix form
.
.
....
.
.
.
Stack successive samples of the signal in vector form
.
.
.
.
.
.
is defined as the smoothing factor.
10
INPUT/OUTPUT STRUCTURE
shows the algebraic relation between the input and output
.
.
....
. . .. . .
.
.
....
.
.
....
Signature waveform matrix has rich structure
.
.
....
. . .. . .
.
.
.
.
.
....
.
.
....
. . ....
11
MULTI-USER CHANNEL ESTIMATION
Estimate signature vectors determine the channel vectors
Data model when including additive white noise
Perform subspace decomposition on by a singular value decomposition
(SVD)
Vectors in span the signal subspace, .
Vectors in span the noise subspace.
12
MULTI-USER CHANNEL ESTIMATION
Estimate from without knowledge of
Orthogonality between noise and signal subspace
Using yields
is a Toeplitz matrix of spreading codes
is in the null space of
13
PROPOSED ALGORITHM FOR A SINGLE RECEIVER
1. Construct data matrix
2. Apply SVD to , or eigenvalue decomposition (EVD) of , to obtain
orthogonal subspace .
3. For each user, estimate the channel vector
4. Reconstruct signature vectors and
Exploits fact that each signature vector is a linear function of a unique
spreading code [Torlak & Xu, IEEE-TSP Jan. 1997]
14
CAPACITY INCREASE USING ANTENNA ARRAY
Problem: Existing systems are not designed to accommodate more than
(spreading gain) users, i.e. P
Goal: Adjust the algorithm to manage an overloaded system (P )
Proposed algorithm in an overloaded system breaks down since the
dimension of the orthogonal subspace reduces
Additional orthogonal vectors are required to guarantee more equations
than unknowns.
Solution: spatially oversample by means of multiple receivers
15
CAPACITY INCREASE USING ANTENNA ARRAY
Assume receivers at the base-station
......
and subspace space relation between and still hold
Orthogonal vectors in increases from to
Signal space remains fixed while noise space increases
Extend our proposed algorithm to handle
16
A-CDMA SYSTEM SIMULATION
Average bit error rate (BER) for different receivers with estimated and
actual channels
Spreading gain=16, M=1, SNR=10 dB, and 9 users
0 1 2 3 4 5 6 7 8 9 1010
10
10
10
10
SNR [dB]
BE
R
__ :actual channels
M=1, P=9
0 1 2 3 4
0Channel response, real part
Time [T]0 1 2 3 4
0
0.05Channel response, imaginary part
Time [T]
0 1
0
0.5
1
0 0.1
0
0.1
RAKE Receiver
17
A-CDMA SYSTEM SIMULATION
Average BER for overloaded A-CDMA systems
Spreading gain=16, M=2, SNR=10 dB, and 18 users
0 1 2 3 4 5 6 7 8 9 10
10
10
10
10
10
SNR [dB]
BE
R
___:actual channels
M=2, P=18
0 1
0
0.5
0 1
0
0.5
0 0.1
0
0.1
user #11:RAKE Recceiver
0 0.1
0
0.05
0.1
user #18:RAKE Receiver
18
CDMA WITH APERIODIC SPREADING SEQUENCES
Aperiodic spreading sequences
distribute signal spectrum uniformly
have inherent interference averaging capabilities
are beneficial to the soft capacity in IS-95 CDMA systems
Subspace-based methods require periodic spreading sequences
Current methods in CDMA sytems with aperiodic spreading sequences
1-D RAKE receivers
2-D RAKE receivers (with an antenna array)
Principal Component (PC) Algorithm
Proposed a new iterative blind channel estimation method
Channel estimation at chip level
Joint block equalization at symbol level
19
DATA MODEL
CDMA system with users
is the chip period
is the noise vector
is the chip delay index assumed to be known.
Wideband channel model
Source symbols are drawn from a finite alphabet
Pseudo-noise (PN) spreading codes
20
STEP 1: ESTIMATION OF CHANNELS GIVEN TRANSMITTED SYMBOLS
Construct the data matrix of the signal
...
is constructed from the estimated input symbols and the code sequence for
th user
Solution for the above equation
Most of is known due to the known PN spreading sequence.
21
STEP 2: ESTIMATION OF SYMBOLS GIVEN CHANNELS
Stack the spatial data samples so that the data matrix
...
are to be estimated
Estimated wideband channel parameters and spreading codes form
. . .
blocks
...
22
DEFINITION OF AND SOLUTION FOR
is the shifted blocks of the codes
...
...
...
...
.
.
.
...
...
.
.
.
...
.
.
.
columns
Each block covers one symbol period and has Toeplitz structure
Use to solve for , then
23
ALGORITHM SUMMARY
and allow us to use both frameworks to exploit the
discrete-alphabet property of CDMA signals and knowledge of spreading
codes.
Adopt Iterative Least Squares with Projection (ILSP) algorithm originally
developed by Talwar, Viberg, and Paulraj for TDMA systems.
Update iteratively, which updates , and under the constraint that
the information symbols are from finite alphabets.
Continue to iterate until or converge.
24
ALGORITHM OUTLINE
1. Randomly choose and set
2.
(a) where
...
(b) Construct with the channel estimates and PN sequences.
(c) is estimated through
(d) Project to closest discrete values.
3. Continue until .
25
SIMULATION OF CDMA SYSTEM WITH APERIODIC SPREADING
Signal constellation for 1-D RAKE, 2-D RAKE, and the proposed method
for M=1 and M=2 cases
Spreading gain=16, SNR=15 dB, and
4 5 6 7 8 9 1010
10
100
Number of Users
MS
E
2D RAKE
M=1
M=2
M=1
M=2
Proposed Method
1D RAKE
-2 0 2
-2
0
2
1D-RAKE with PC-MMSE-1 0 1
-0.5
0
0.5
Proposed Method
P=8,M=1
-1 0 1
-0.5
0
0.5
1
Proposed Method
P=13,M=2
0 5-4
-2
0
2
2D-RAKE with PC-MMSE
26
CONTRIBUTIONS AND OTHER RESEARCH
Uplink processing for W-CDMA with smart antennas
Blind multi-user channel estimation
CDMA with aperiodic spreading codes
Downlink processing for multi-transmitter diversity
Designing weight vectors for maximum capacity
Smart Antenna Testbed development
Statistical vector channel modeling
Real-time signal processing
TDMA with smart antennas
Fast source separation algorithms for Digital Wireless Applications
27
FUTURE DIRECTIONS
1G and 2G are optimized for Voice Communications (FDMA, TDMA,
CDMA) at 8-13 kbps
3G technology must deliver data at 144 kbps-2 Mbps
– Coordinated system concept
– Wideband-CDMA : The Universal Platform
Fourth-generation systems
– Multimedia/Telemedicine/Internet access
– Accommodating mixed traffic
Wideband testbed development with antenna arrays
– Channel propagation studies
– Experimental validations of algorithms
28