MIMO Channel Models

N. Czink, D. Seethaler, G. Matz,

M. J achan, and F. Hlawatsch
Institute of Communications and Radio-Frequency Engineering
Vienna University of Technology
INSTITUT FR
NACHRICHTENTECHNIK UND
HOCHFREQUENZTECHNIK
Outline
The wireless channel
Model categories Model categories
Propagation-based models
A l ti l d l Analytical models
Standardized models
Whi h d l i b t? Which model is best?
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 2
Outline of Part 3
The wireless channel
Multipath propagation Multipath propagation
The double-directional propagation channel
Clusters and dispersion p
Diversity
Model categories
Propagation-based models
Analytical models
Standardized models
Which model is best?
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 3
Multipath Propagation (1)
Interacting objects (scatterers) cause multipath propagation
Movement of Tx, Rx, and/or scatterers causes Doppler shifts , , pp
Each path has its reflectivity, delay, and Doppler shift
scatterer
Tx Rx
tt scatterer
object velocity object velocity
path weight
path delay
Doppler frequency
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 4
Doppler frequency
The Double-Directional Radio Channel
Multiple antennas enable spatial resolution:
path weight path weight
path delay
azimuth of dept.
scatterer
Tx Rx
azimuth of arrival
elevation of dept.
elevation of arrival scatterer
Additional properties:
Angles of departure (azimuth , elevation )
Angles of arrival (azimuth , elevation )
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 5
Multiple Antennas
MIMO channel:
bit stream
transmitter
(T )
receiver
(R )
decoded bit stream
(Tx) (Rx)
MIMO
h l

Input-output relation:
channel
ith h l i l t i with channel impulse response matrix
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 6
Multipath Clusters
Clusters lead to temporal and angular dispersion:
Tx
cluster
Rx azimuth spread
Tx
Tx
cluster
azimuth
spread
R l t
Rx
Rx cluster
spreads
Tx cluster
spreads
Global dispersion parameters:
rms delay spread
rms angular spreads
Global dispersion parameters
rms angular spreads
Cluster dispersion parameters:
cluster rms delay spread
Global dispersion parameters
are no meaningful description
of propagation environments!
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 7
cluster rms delay spread
cluster rms angular spread
Observing Clusters
Indoor scenario, NLOS, Laboratory environment
Rx fixed Tx is moved through the room Rx fixed, Tx is moved through the room
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 8
Diversity
Diversity is the presence of multiple copies of Diversity is the presence of multiple copies of
the transmitted signal
at different times (time diversity) at different times (time diversity),
at different delays or frequencies (delay/frequency diversity),
in space (spatial diversity),
where the different paths are fading independently.
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 9
Diversity Example
Demonstrating spatial diversity:
only weak eigenvalues fade
strong eigenvalues stay constant
Video of the environment
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 10
Source: Tokyo Institute of Technology
Outline
The wireless channel
Model categories Model categories
Propagation-based models
A l ti l d l Analytical models
Standardized models
Whi h d l i b t? Which model is best?
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 11
Why do we need channel modelling?
For MIMO deployment and network planning For MIMO deployment and network planning
site specific
For system design and test
site independent site independent
but not: one model serves all but not: one model serves all
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 12
MIMO Model Categories
Antenna configuration
Bandwidth
System parameters
Propagation-based models Analytical Models
deterministic:
Focus: physical wave propagation
correlation-based:
Focus: MIMO channel matrix
ray launching / tracing
stored measurements
geometry-based stochastic:
full correlation model
iid model
Kronecker model
g y
GSCM
COST 259
COST 273
Weichselberger model
propagation-motivated:
virtual channel representation p
finite scatterer model
Standardized models: - 3GPP SCM
IEEE 802 11n
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 13
- IEEE 802.11n
- IEEE 802.16 WiMAX
Outline
The wireless channel
Model categories Model categories
Propagation-based models
Deterministic
Propagation-based models
Deterministic
Geometry-based stochastic
Analytical models
deterministic:
Focus: physical wave propagation
Analytical models
Standardized models
Which model is best?
ray launching / tracing
stored measurements
geometry-based stochastic:
Which model is best ?
g y
GSCM
COST 259
COST 273
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 14
Propagation-Based Models
Modelling of all system parameters possible
Time-variance (moving Tx/Rx/scatterers) Time variance (moving Tx/Rx/scatterers)
Frequency selectivity
Spatial structure
Important for link-level simulations and hardware channel simulators
More art than science
Who decides on the scenario? Who decides on the scenario?
Canonical scenarios
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 15
Deterministic Models Ray Launching / Tracing
Model environment by geometry, place scatterers
Launch or follow rays from Tx to Rx
Tx Tx
Rx
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 16
Deterministic Models Stored Measurements
Conduct measurements
Store impulse responses Store impulse responses
Use impulse responses
in channel simulator
Playback Simulations
Question:
Which scenario
shall be used?
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 17
Geometry-Based Stochastic Channel Models (GSCM)
Model environment by scattering clusters placed in space
Within clusters, paths are modelled by statistic means
Use ray tracing to calculate channel response
LOS
NLOS
scatterer
Rx
scatterer
scatterer
Rx scatterer
scatterer
scatterer
scatterer
Tx
Tx
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 18
LOS: line-of-sight NLOS: non-line-of-sight
COST 259 Approach
Single-bounce model, no scatterers around BS
Fixed relationship between AOD, AOA, and delay
Well suited for smart antenna systems, but nature is
not single bounced
MS
f
near
scatterer
far
scatterer
BS
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 19
AOA: angle of arrival; AOD: angle of departure; BS/MS: base/mobile station
COST 273 Approach (1)
Model based on
clusters
single-interaction clusters
3 different
l t t cluster types
Clusters are Rx Clusters are
placed geometrically
and stochastically
Rx
Tx
l l l t
multiple interaction clusters
local cluster
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 20
multiple-interaction clusters
COST 273 Approach (2)
Local clusters around MS and/or BS may occur Local clusters around MS and/or BS may occur,
depending on the scenario
Any combination of delay ( ) and angles ( )
can be modelled, not limited to double scattering
All parameters are given per cluster;
there are no global spreads
Direct coupling between AOAs and AODs;
no Kroneckerstructure no Kronecker structure
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 21
Outline
The wireless channel
Model categories Model categories
Propagation-based models
A l ti l d l
Analytical models
Analytical models
Correlation-based
Propagation-motivated
correlation-based:
Focus: MIMO channel matrix
Propagation-motivated
Standardized models
Which model is best?
full correlation model
iid model
Kronecker model
Which model is best ?
Weichselberger model
propagation-motivated:
virtual channel representation p
finite scatterer model
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 22
Analytical Channel Models
Analytical channel models focus on modelling only the spatial structure
(up to now) ( p )
Narrow-band models
Delay spreads neglected
No Doppler shifts possible
Number of antennas is predetermined
Well suited for testing signal processing algorithms
Can be used in combination with propagation-based models
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 23
Analytical Channel Models Overview
Correlation-based models
Full-correlation model: m
o
r
e
Weichselberger model:
e


p
a
r
a
Kronecker model:
a
m
e
t
e
r
iid model (canonical model):
r
s
Propagation-motivated models
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 24
to be explained presently!
Channel Correlation Matrix
The channel correlation matrix
with
sufficiently characterizes the spatial structure
of the channel Size of :
with
of the channel. Size of :
Note: The operator stacks
the columns of a matrix
into a vector
Underlying assumption: Rayleigh fading channel
If this assumption is not fulfilled, all the following models
will inevitably fail!
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 25
distributed circular symmetric complex gaussian with mean and covariance
Correlation-Based Analytical Models
Full-correlation model
Very complex Very complex
Most accurate
Weichselberger model Weichselberger model
Good approximation
Good performance-complexity compromise
Kronecker model
Separates channel into Tx and Rx sides
Very limited validity
iid model
Most simple
No physical relevance
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 26
Full-Correlation Model
Synthetic channel realizations consistent with channel correlation
matrix can be generated by g y
with
where is an iid white Gaussian random vector
Can be interpreted as a noise coloring process: Can be interpreted as a noise-coloring process:
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 27
iid Model
All elements of the channel matrix are
complex Gaussian
independent identically distributed (iid) uncorrelated
Channel correlation matrix is modelled as
Ch l li ti b t d b Channel realizations can be generated by
with
Implications:
no spatial structure is modelled
f ( ) only valid for (very) rich scattering environments
BUT
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 28
cannot even be validated by measurements in indoor environments!
Kronecker Model Definition
Full-correlation matrix has too many parameters
treat correlation independently at Tx and Rx: treat correlation independently at Tx and Rx:
Transmit correlation matrix:
Receive correlation matrix: Receive correlation matrix:
Channel correlation matrix is modelled byy
Kronecker matrix product
Channel realizations can be generated by
with
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 29
Kronecker Model Implications
Kronecker model holds true only if channel can be separated into
Tx side and Rx side
Rx directions are independent of Tx directions
Only satisfied for few antennas or large antenna spacing
scattering environment
Tx Rx
local scatterers local scatterers
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 30
Weichselberger Model Definition
Relaxes assumptions of Kronecker model by using power coupling of
Tx and Rx eigenmodes, defined by
Rx eigenmodes Tx eigenmodes
Power coupling of eigenmodes is described by coupling matrix
Channel correlation matrix is modelled as
Channel realizations can be generated by
with
Channel realizations can be generated by
where is element wise square root of and
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 31
where is element-wise square-root of , and
element-wise matrix product (Hadamard product)
Weichselberger Model Parameters
What are eigenmodes and the coupling matrix ?
eigenmodes
scattering environment
Tx Rx
hot strong power
Coupling matrix
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 32
hot strong power
cold .. weak power
Outline
The wireless channel
Model categories Model categories
Propagation-based models
A l ti l d l Analytical models
Standardized models
3GPP SCM 3GPP SCM
IEEE 802.11n
IEEE 802.16 WiMAX (?) IEEE 802.16 WiMAX (?)
Which model is best?
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 33
3GPP Spatial Channel Model (SCM)
SCM is a geometry-based stochastic channel model, based on COST 259
1. Choose scenario
Suburban
macro
Urban
macro
Urban
micro
2. Determine user parameters
Angle spread
Angles of departure (paths)
Angle spread
Lognormal shadowing
Delay spread
Pathloss
Far scattering cluster
(urban macro)
Angles of departure (paths)
Angles of departure (subpaths)
Path delays
Average path powers
Pathloss
Orientation, Speed Vectors
Antenna gains
(urban macro)
Urban canyon
(urban macro)
Average path powers
Angles of arrival (paths)
Angles of arrival (subpaths)
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 34
3. Generate channel coefficients
Polarization
LOS (urban micro)
Options
IEEE 802.11n MIMO Channel Model Principles
Delay models (A-F)
1. Choose scenario
GSCM model
Cluster position is uniform in AOA/AOD
Cluster angular spreads are defined for
2. Physical model
different models
Add Doppler shift
Impose antenna structure 3. Analytical model
Use Kronecker model for each channel tap
Parameters are extracted from simulated
physical model
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 35
802.11n MIMO Channel Model Discrepancies
GSCM model (physical model used) is well-defined and reasonable
BUT
Kronecker model (analytical model used) is only valid for small number
of antennas and large antenna spacing
Tap-wise Kronecker model is even more inaccurate
Model is NOT WELL SUITED for indoor channels! Model is NOT WELL SUITED for indoor channels!
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 36
IEEE 802.16e WiMAX (?)
No MIMO model available at this point No MIMO model available at this point
Models already available for
pathloss
fading
d l d delay spread
Doppler spread
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 37
Outline
The wireless channel
Model categories Model categories
Propagation-based models
A l ti l d l Analytical models
Standardized models
Whi h d l i b t? Which model is best?
Propagation-based models
Analytical models Analytical models
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 38
Which Propagation-Based MIMO Model is Best ?
Physical (propagation-based) channel models
should be motivated by measurements should be motivated by measurements
should be able to model the behavior of different environments
Accuracy-complexity tradeoff
Challenges
Find one model that can be parameterized for many environments p y
Environments/scenarios must be parameterized accurately;
parameters should be extracted from extensive measurements
R d l it Reduce complexity
The best propagation-based model is not developed yet,
b t COST 273 l but COST 273 may come close.
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 39
Which Analytical MIMO Model is Best ?
Analytical channel models
should accurately reflect the spatial structure of the channel
should be validated by measurements
Accuracy-complexity tradeoff
Challenges
Optimal accuracy-complexity compromise
Extend analytical models to frequency-selective and
time-varying channels
The best analytical model?
There is no best one, but a most suitable one for specific
requirements requirements
When no specific requirements are in focus, the model should
reflect the spatial structure accurately
MIMO Tutorial Part 2: MIMO Channel Models Sept. 2005 40