Introduction to Smart Antenna Techniques and Algorithms RAWCON `99 Smart Antenna Workshop

1
Introduction to Smart AntennaTechniques and
AlgorithmsRAWCON 99 Smart Antenna Workshop

• Adrian Boukalov
• Helsinki University of Technology
• Communications Laboratory
• Otakaari 5 A, FIN-02150 Espoo, Finland
• e-mail adrian.boukalov_at_hut.fi
• fax/ voice int. 358-9 - 4512359/17

2
Overview

• Introduction
• Motivation for smart antennas (review)
• System integration of smart antenna (SA)
  technology
• Cellular network components influenced by smart
  antenna technology
• Spatial Processing
• Classification by reference type
• Receiver structures and algorithms
• Space-Time Processing
• Space-Time CDMA receivers
• SA integration into macrocell and microcell
  environment

3
"Spatial Processing remains as the most
promising,if not the last frontier, in the
evolution of multiple access systems"


Andrew Viterbi
4
Smart Antenna TechnologyMotivation
Link level improvements System improvements
- Interference cancellation on the up and down
links - SNR improvement due to antenna gain
- Multipath mitigation
capacity coverage Quality of service (QoS), bit
rate, mobility rate
5
Smart Antenna Technology Benfactors
Network capacity, coverage, filling
dead spots, fewer BSs, higher QoS, new
services...-gt revenues
New market for more advanced BSs, flexible
radio network control...
Higher QoS, more reliable, secure
communication, new services, longer battery
life...
Operator
OEM
User
6
System Integration of Smart Antennas
Different types of environments
Propagation Maps
Network Planning
Network Infrastructure (BS position, system
parameters, fixed network topology)

- offered traffic spatial distribution,
services. -mobility
Radio channel, interference environment, mobility
, services
Receiver, antennas parameters
Expected cells load variations Layers
structure ...
Smart Ant. Tech.
Radio Network Management
Radio Interface -Receiver structure
algorithms
Protocols, dynamics
C
U
DSP tech.
Air Interface
7
Cellular Network Components Influencedby Smart
Antenna Technology
Network Planning - Capacity, coverage,
interference planning - Joint fixed and radio
network optimization, planning -
System upgrade, economical issues
1G- analog systems 2G- digital systems 2.5G-
digitalpacket .. (GPRS,..) 3G - W-CDMA 4G-
cellular gigabit WLAN
Radio Interface Receiver structure, Tx, Rx
algorithms - Spatial proc. - Time domain proc. -
Coding - Detection - Diversity - ..
Radio Network Management
DSP tech. SW Radio
Services -gt MS location
Network control - R.resource management - call
control
Cell control - admission control - broadcast
channel control - handover control -
macro-diversity control
1G
2G
Air Interface - Multiple access - Duplexing -
Modulation - Framing - Availability of pilots
Link level control - Power Control - Quality
Control - Tracking
2.5G
3G
4G
8
Smart Antenna Receivers Many choices!
- Switched beam, adaptive algorithms.. - Side
reference information available (spatial
reference, reference signal, signal structure
and their combinations) for spatial processing
- Narrowband , broadband (CDMA) - Optimization
method (if any) maximum likelihood-ML, minimum
mean square error- MSE, minimum variance-MV,
maximum a-posteriori probability -MAP - Domains
-gt Space-only, space-time, space-frequency -
Amount and type of channel knowledge available -
Combination of space/space-time processing with
other technologies (diversity, interference
cancellation, channel coding, space-time coding
) - Smart antennas at the mobile
9
Spatial Processing Approaches
- Sectorization - Macro-diversity with
Combining maximum ratio combining - MRC
optimum combining -OC,.. Prefiltering/Coding
Space -Time Coding V-BLAST - Beamforming
(BF) Switched-beam Smart Antenna Adaptive
beamforming These approaches can be/should be
combined/mixed
together !
Sectorization
Macro-diversity
Switched-beam ant.
Adaptive BF
10
Spatial Processing Integration with Air
Interface
Antennas elements geometry, numbers of elements -
M.
Radio Transmission Technologies
MS
Internetworking
Physical Channel Definition, Multi- plexing
Multiple Access Technology
Frame Structure
Duplexing Technology
RF- Channel parameters
Channel Coding
Modulation Technology
Source Coding
Availability of the training signal Frame length
- T
Mapping control, traffic channels
FDD TDD
Modulation type CM... Finite Alphabet Linearity
FDMA CDMA
Combination with Space Processing
Bandwidth-B Carrier frequency fo
UL-gtDL link
Wide/narrow band SA rec, BF, AoA est
Blind methods SSBF, ST

Ref. Signal based BF, S-T
11
BF/OC Techniques Classified by Reference Type
Data-independent beamforming
- Spatial reference based beamforming,
Direction of arrival based beamforming
(DoABF) - Reference signal based/time reference
beamforming (TRB) and/or optimum combining (OC)
- Signal structure (temporal /spectral) based
beamforming, SSBF/property restored beamforming
Statistically optimum beamforming
12
Adaptive algorithms
Tracking in time
Data independent BF
AoA estimation..
AoA(s)tracking ML, ...
- DoABF - TRB and/or OC - SSBF
Calibration
Ref. multiplexed with des. signal or
reconstr. from detected symbol
Adaptive Alg. DMI, LS (LMS, RLS),non-linear
Synchronization
Constant Modulus (CMA), FA,...
CM-LMS
Statistically Optimum BF
13
Possible SA receivers realizations
Parameters that can be optimized
Data, BER
SINR
Time Ref. post det.
CIR
Time Ref.
Demod.
Detection
RF
IF
RF-BF
IF- BF
BB- BF/OC
14
Direction of Arrival Based Beamformers (DoABF )
- require angle of arrival (AoA) estimation -
sensitive to AoA estimation errors, calibration
problem - estimates output power at the output
or eigen-decomposition of correlation matrix -
problem with coherent multipath - Angular
spread (As) to array resolution (A) ratio should
be low - FDD applications
Array Processor
Array Output
15
DoABF Theory in brief
?
16
AoA estimation methods
1.Conventional techniques poor
angular resolution limited by aperture, search
of peaks in spatial spectrum - MV (some
degrees of freedom spent on interference
cancellation, improved resolution)
2.Based on statistical model of signal and
noise (optimal) - ML, MLM - data samples lt-gt
AoA - Block ML(comb. with Eigen decomposition)
joint pdf of sampled data needed, very
computationally extensive, can work well in
low SNR (or number of signal samples is small)
work well in correlated signal conditions,
number of sources should be known, non-linear
multi-dimensional optimisation (coincides
with LS estimator when assumption about noise do
not hold)

array data input matrix spatial signature
matrix signal wave form matrix noise matrix
17
AoA estimation methods (contd)
U(t)As(t)n(t)
3. Based on the model of the received signal
vector high resolution methods , fail in coherent
multipath (suboptimal, BB only ) - MUSIC,
WSF - ESPRIT subarraying (relaxed computational
and calibration requirements) Supplementa
ry techniques N sources, R- correlation matrix
estimation DOA estimation under coherent
conditions Spatial smoothing, multi-dimensional
MUSIC, ILSP-CMA, (integrated approach)

signal subspace noise subspace
Rss-signal correlation matrix
18
Time-Reference Signal Based Beamformers and/or
optimal combiner (TRB)
- requires reference signal or the replica
correlated with desired signal - based on
Wiener solution (MSE) - reference signal
multiplexed with desired signal or
reconstructed signal obtained from detected
symbols (detection and BF are interdependent
but attractive for tracking) - better for
varying radio channel - diversity - more
processing extensive methods - receiver is
simpler at expense of spectral efficiency
- synchronization problem - Delay spread (Ds) to
frame length (T) ratio should be low - TDD
applications
LS Beamformer
X1(t)
1
W1
Array output
2
X2(t)
W2
y(t)
N
Xn(t)
Wn
Control algorithm
Ref.
Error
-


Signal processor
Adaptive processor
19
TRB Theory in brief
MMSE
wR-1p
Ref.
Error Signal
d(t)
LS
Wk(AHA)-1 AHdk
20
Signal Structure Based Beamforming (SSBF)
- Does not require reference signal, thus
increased spectral efficiency - constant
modulus (CM)property of phase modulated
signals, - finite alphabet (FA) property of
digitaly modulated signals , - spectral
coherence restoral SCORE (only information
needed - bit rate) - Useful method for
tracking between references - Convergence
properties ? - Methods based on partial
information are usually non-linear -
Performance from robustness point of view
similar to reference signal based methods
BF (W)
CMA
21
Improvements available using spatial processing
- Improvement in SNR due to beamforming array
gain. (improved coverage. ) - Reduced ISI.
(depends on angular spread of multipath) -
Enhanced spatial diversity. - Interference
cancellation. In Tx and Rx. Capacity. These
goals may be conflicting. Need balancing to
achieve synergy with propagation environment,
offered traffic, infrastructure.
22
SNR maximization due to antenna gain
Beamforming
1/M
23
Co-Channel Interference (CCI) Cancellation
Beamforming
Combining
M-1 interferers cancellation. independent of the
environment
M-1
24
Diversity (Angle- and Space-) Gain
M
M
25
ISI Cancellation
M-1
M-1 delayed signals cancellation (M-1)/2 symbol
due to delay spread
26
Optimal Spatial Algorithms
Beamforming
Multi-path
BS
Interfering MS 2
MS 1
Path with ISI, uncorrelated paths
27
Optimal S-T Algorithms

Spatial domain processing Temporal domain
processing
28
Degrees of freedom number of SA elements
- Number of SA elements (M) can be considered as
a resource, i.e., degrees of freedom which
can be spent for SNR, CCI, diversity,
ISI, either separately or jointly (optimum) - M
determines spatial selectivity of SA
29
Beamforming Methods
Data independent beamforming (conventional beamf
ormer -CBF,..) Optimum BF - Based on the cost
function maximization/minimization (max
SINR,) - Based on statistical estimation ML
(likelihood function) Squared function based
MSE (Reference ) -Adaptive algorithms - Least
Square (LS), Maximum A-posteriori Probability
(MAP),
( for example, GSLC,)
30
Optimization Criteria
- Based on cost function maximization/minimization
(max SINR,)-gt difficult to obtain - Based
on Statistical Estimation ML (Likelihood
function)-gt treats interference as temporally and
spatially white Gaussian. Balance effect of
noise. MSE (Reference )-gt more attractive in
presence correlated CCI. -gt More efficient in
interference dominant environment. Do not balance
effect of noise
31
Spatial processing Summary
DoABF - better perform in environments with low
angular spread - require AoA estimation and
calibration - well suit for FDD applications -
macrocell environment - CDMA AoA estimation and
beamforming TRB or/and OC - well perform in
environments with high angular spread - require
reference signal (spectrum efficiency),
synchronization - well suit for TDD (micro/pico
cells), FDD is more problematic micro and
picocell - more robust methods in changing
environment (adaptive algorithms)can be/should
be combined with blind methods
32
Space-Time (S-T) Processing
- Space domain processing Efficient CCI
mitigation Space Diversity ISI mitigation
depends on angular spread of multipath and M
and cannot be very efficient - Time domain
processing Very limited against CCI Time/path
div., ISI mitigation - S-T Processing
Simultaneous operations in Time and Space
domains can combine strength of the both -
Multi-User-S-T Processing
Channel
ST-MLSE
Vector VA
Sk

Training
ST-MMSE
yk
Sk
Demod.
W

ST-MMSE/MLSE
STF W
Scalar VA MLSE
33
Space-Time (S-T) processing techniques
Decoupled S-T processing Joint S-T
processing Path diversity BF Combining Single
user MU Narrowband Wideband Up-link Down-link
34
Relations between spreads and relative quantities
of interests for different types of cells.
Location of scatters
at MS at BS Remote
Spread Type
Critical system parameter
Macrocell
Doppler spread Dp fd fo(v/c) Delay spread Ds Ang
ular spread As
B
fd/B
MS motion
T
Micro cell
Ds/T
Array Resolution A1/M As/A
static
35
Macrocell and Microcell Channel Response

Remote scatters
1800
1800
Scatters local to BS
-1800
0
1
0
20
Delay (microsec)
Delay (microsec)
Scatters local to MS
Macrocell
Microcell
As
Dp
Ds
As?
After A.Paulraj
36
Space-Time MLSE and MMSE
S-T
MLSE CCI statistic needed Delay
spread (Ds) cant handle large Ds Doppler
spread difficult to handle Blind
methods more problematic MU
S-T MLSE - optimum
S-T MMSE not needed (Rxx) less
problematic can handle by channel tracking
applicable S-T MMSE (V-BLAST)
37
TDMA Rx Structures (Ch. Knowledge lt-gt Optimality)
S-DIV CCI ISI T-DIV X X X X
X X X X X X
X X X X X X X
MU-MLSE
H1 H2
H1 RS-T
ST-MMSE-MLSE
S-MMSE-MLSE
H1 RS
Decreasing Channel Knowledge
MMSE
H1
ANT-HOP
Nil
After A. Paulraj
38
CDMA Rx Structures (Ch. Knowledge lt-gt
Optimality)
S-DIV T-DIV MUI X X X
X X X X
X X X X
X
ST-MU
H1 H2
H1 RS-T
ST-MMSE
ST-RAKE
H1 RS
Decreasing Channel Knowledge
BF-RAKE
H1
ANT-HOP
Nil
After A. Paulraj
39

Space-Time channel estimation
Underlying channel/signals structures
Tracking of fast varying channel
Channel Estimation methods
Reference
40
Space Domain Only and Space-Time SA Algorithms
41
Space-Time CDMA receivers
- Non- coherent combining (equal gain
diversity combining improves SNR, but CCI
cancellation not possible.) - Coherent combining
Beamforming- RAKE (1D, 2D) Reference signal
based beamformer - RAKE DoABF - RAKE (max.
SINR, ML, ..) SSBF- RAKE Combing - RAKE OC,
IRC,.. - Multi-user ST (MU-ST-MMSE,
MU-ST-MLSE) - Space -frequency RAKE (RS-F)
joint, and decoupled
42
ST MU-MLSE
- computational complexity linear to the number
of users - same degree of the near-far resistance
and error rate performance as optimum MU
receiver - require knowledge of the all users
channels - optimum in Gaussian noise only
43
SA integration into macrocell and microcell
environments
44
TDMA Tx Structures
S-DIV CCI T-DIV X X
X X X X
X X X
MRC-Adaptive ST - Nulling
H1 H2
MRC-Adaptive nulling
H1 RS-T
CBF
Decreasing Channel Knowledge
R11
ANT-HOP
After A. Paulraj
45
CDMA Tx Structures
S-DIV T-DIV MUI X X
X X
X X
HST
ST-MRC
HS
S-MRC
CBF
Decreasing Channel Knowledge
Rss
ST-Coding
After A. Paulraj
46
Summary
- Spatial structure based algorithms can work in
higher Doppler spread but are affected by
angular spreads - Temporal structure based
algorithms can better handle delay spread, but
higher speed can be problem - Single and
multi-user combination may be needed - Training
signal lt---gt receiver complexity trade-off -
Environment (spreading) lt--gt receiver and
algorithmic complexity, (how models
corresponds to reality)
47
Summary (contd)
Best solutions Combine trade-offs between -
Beamforming lt---gt combining - Algorithms
(MLlt---gt MSE) , subspace - Optimum lt---gt Data
independent approaches - Base band beamforming
lt---gt RF/or IF beamforming - Combination with
other methods like multi-user detection (MUD),
diversity, ST coding, adaptive modems Air
interfaces should be not only friendly for
S-T processing but flexible / adaptive to be able
to exploit advantages of spatial processing in
variable environments
48
Glossary