Modeling Mutual Coupling and OFDM System with ...ece.drexel.edu/telecomm/Talks/kirsch.pdf · OFDM System with Computational Electromagnetics ... Laboratory Telecommunication Seminar

Modeling Mutual Coupling and OFDM System with Computational Electromagnetics

Nicholas J. KirschDrexel University Wireless Systems

Laboratory

Telecommunication SeminarOctober 15, 2004

Introduction

MIMO CommunicationsMutual Coupling and Local ScatteringMIMO-OFDMMobile Ad-hoc NetworkingConclusions

MIMO Communications

Multiple transmitting antennas to multiple receiving antennas

Modeling MIMO Channels

Effects of scatterers in the near fieldIncorporate effects of OFDMExtend simulations to experimental test bed

Mutual Coupling and Local Scattering

MotivationDetermine the degree of importance of including mutual coupling (MC)Accurately model vehicular mobile ad hoc networks (MANETs) which include local scattering (LS)

Hybrid Computational Electromagnetic Simulation

Near fieldMethod of Moments (NEC, FEKO)

Far fieldRay tracing (FASANT)

Inclusion of mutual coupling and local scattering

Run simulation with isotropic radiation patterns to determine the electric fields and geometric solutions for multipath rays : Eθ(θ) and Eφ(φ)

Inclusion cont’d

Inclusion cont’d

Create antenna e-field weight signal components with appropriate radiation pattern value (which includes MC and LS)Sum L the appropriate multipath Ez rays for every pairwise combination of transmit and receive antennas to create the channel matrix H

Inclusion cont’d

( ) ( ) ( ) ( )ϕϕϕϕ ϕ nmmn

weight RETE =),(

( ) ( ) ( ) ( )θθθθ θ nmmn

weight RETE =),(

∑−

=

=1

0

),(,,

L

l

mnlzmn Eh

[ ]MNmnh ×

= ,H

Method of AnalysisSpatial Multiplexing (SM) Capacity - to evaluate spectral efficiency due to multipath richness (without path loss)Beamforming (BF) Capacity –used to compare the effectiveness of antennas in vehicular MANET

SM and BF capacities are used to evaluate which signaling technique is best in a MANET

( )

+=

)(log2 H

HHIHα

ρM

CH

N

( )

+=

)()(

1log2max

2 HHH

αρλ

B

21)(FM

HH =α

Angular spread

Angular Spread – used to understand the multipath richness

20

211FF

−=Λ

∫ −=π

θ θθ2

0

)( depF jkk

Simulations

Inclusion of MC at both ends of the link

4 x 4 MIMO System, ULA

Performance analysis of a vehicular MANET

4x4 MIMO, antennas arranged on each side of vehicle

Radiation patterns2.4 GHz dipole antenna in a λ/2 ULA2.4 GHz patch antenna on the front of a bumper of a vehicleRadiation patterns due to near field effects are shown

Results: Inclusion of MC at both ends of the link

Inclusion of MC at both ends of the link results in greater capacity in both the LOS and NLOS cases.

0.3240Philadelphia - LOS

0.6960Philadelphia - NLOS

Angle SpreadCase

Results: Performance evaluation of a vehicular MANET

Due to the array geometry, some transmitting antennas may not communicate with certain receiving antennas even in a multipath rich environmentA rank deficient channel will have lower spatial multiplexing capacity, but higher beamforming capacity

MIMO-OFDM

Current research on MIMO-OFDM based on statistical channel modelsStatistical channel model: either insufficiently wideband or not suited for the physical environment of interestRay-tracing simulation: site-specific information

Electromagnetic Ray TracingComputational electromagnetic-based simulation technique;ERT is computationally complex and demanding;ERT has been applied to narrowband system so far.

Channel Response Extrapolation

AssumptionsSubcarrier frequencies are not so far apart that the geometric ray solutions remain the same across all subcarriersThe number and type of multipath signal components remain the same from one subcarrier to another

Channel Response Extrapolation

FASANT computes the electric field of each ray

The response Ez,n at carrier fn can be computed by extrapolation

where

mm

fcdj

zdjk

zmz eEeEEπ2

,

−− ==

fcdj

mz

fcdj

znz eEeEE n ∆−−==

ππ 2

,

2

,

mn fff −=∆

SISO Channel Model

Time-varying channel

For a finite bandwidth receiver, some of the rays cannot be distinguished in time. Group them to be a cluster The time-clustered channel is

∑ −

= −= 10

2)(, )(),(

)(Ni i

tfjimz

ideEt ττδτα π

∑ == lp

iimzl tEt

1)(

, )()(β

∑ −

= −= 10 )()(),( L

l ll tt ττδβτα

MIMO Channel Model and Capacity

Input Output relationFrequency response of wideband channel matrix at kth subcarrier

If CSI is not available at the transmitter

If CSI is known at the transmitter: use water-filling to maximize mutual information.

∑−

=−=

1

0)()(

L

ll lnn xHy

∑ −

=−=

1

0/2/2 )( L

lNlkj

lNkj ee ππ HH

( ) ( )∑∑−

=

−

=

+==

1

0

/2/222

1

0detlog11 N

k

NkjHNkj

ntMr

N

kk ee

NMP

NI

NI ππ

σHHI

Simulated Capacity in Ad Hoc network

Center frequency = 2.4GHzNumber of carriers = 64Channel spacing = 312.5KHz

1 2 3

4 5 6

7 8 9

Street

Street

50m 20m

50m

20m

Simulated Capacity

Capacities fluctuate at different subcarriers

0 10 20 30 40 50 608

9

10

11

12

13

14

15

16Capacity vs Subcarrier(Node 7 to Node 3)

Subcarrier

Cap

acity

(bit/

sec/

Hz)

Equal power SNR=15dBEqual power SNR=20dB1D Waterfilling SNR=15dB1D Waterfilling SNR=20dB

0 10 20 30 40 50 606

7

8

9

10

11

12

13

14

15

16Capacity vs Subcarrier(Node 9 to Node 2)

Subcarrier

Cap

acity

(bit/

sec/

Hz)

Equal power SNR=15dBEqual power SNR=20dB1D Waterfilling SNR=15dB1D Waterfilling SNR=20dB

Eigenvalues vs Subcarriers

The larger the eigenvalue an eigenmode has, the more power flows to that modeFrequency diversity achieved

0 10 20 30 40 50 60 702

2.5

3

3.5

4Dominant eigenvalues vs Subcarrier(Node 9 to Node 2)

Subcarrier

Eig

enva

lue

#1

0 10 20 30 40 50 600

0.5

1

1.5

2

Subcarrier

Eig

enva

lue

#2

0 10 20 30 40 50 60 702

2.5

3

3.5Dominant eigenvalue vs Subcarrier(Node 7 to Node 3)

Subcarrier

Eig

enva

lue

#1

0 10 20 30 40 50 60 700

0.5

1

1.5

2

Subcarrier

Eig

enva

lue

#2

Mobile Ad hoc Network

Ad-hoc network MIMO CommunicationsSoftware Defined RadioModular Networking LayerWide bandwidth (10MHz Baseband)

Node infrastructure overview

Summary

Demonstrated importance of including MC and LS for design of ad hoc network nodes making use of antenna arraysDeveloped simulation methodology for including these practical electromagnetic effects

Summary cont’d

Fast way to generate channel responses in OFDM system with CEM.This technique is used for simulating MIMO-OFDM Ad Hoc network, which shows both spatial and frequency diversity are achieved.Combine ray tracing channel model with correlation (stochastic) channel model to provide site-specific information as well as numerous channel realizations – Semi-stochastic spatial-temporal channel model.