Channel Estimation for Gigabit Multi-user
MIMO-OFDM Systems
Franklin Mung’au
A Thesis Submitted in Partial Fulfillment
of the Requirements for the Degree of
Doctor of Philosophy
in
Engineering
cThe University of Hull
31st March, 2008
The University of Hull holds the copyright of this thesis. Any person(s)
intending to use a part or the whole of the materials in this thesis in a proposed
publication must seek copyright release from the Dean of the Graduate School.
Abstract
The fundamental detection problem in fading channels involves the correct estimation of
transmitted symbols at the receiver in the presence of Additive White Gaussian Noise
(AWGN). Detection can be considered when the receiver is assumed not to know the chan-
nel (non-coherent detection), or alternatively, when the random channel is tracked at the
receiver (coherent detection). It can be shown that for a given error probability, coherent
detection schemes require a Signal to Noise Ratio (SNR) that is 3dB less than the SNR re-
quired for non-coherent detection schemes. It is also known that the performance of coherent
detection schemes can be further improved using space-frequency diversity techniques, for
example, when multiple-input multiple-output (MIMO) antenna technologies are employed
in conjunction with Orthogonal Frequency Division Multiplexing (OFDM).
However, the superior performance promised by the MIMO-OFDM technology relies on
the availability of accurate Channel State Information (CSI) at the receiver. In the literature,
the Mean Square Error (MSE) performance of MIMO-OFDM CSI estimators is known to
be limited by the SNR. This thesis adopts a different view to estimator performance, by
evaluating the accuracy of CSI estimates as limited by the maximum delay spread of the
multipath channel. These considerations are particularly warranted for high data rate multi-
user MIMO-OFDM systems which deploy large numbers of transmit antennas at either end
of the wireless link. In fact, overloaded multi-user CSI estimation can be effectively studied
by considering the grouping together of the user antennas for the uplink while conversely,
considering a small number of antennas due to size constraints for the downlink. Therefore,
most of the work developed in this thesis is concerned with improving existing single-user
MIMO-OFDM CSI estimators but the results can be extended to multi-user system.
i
Acknowledgments
This thesis is dedicated to my parents, Jacob Mung’au and Mary Mung’au.
My dad, an Engineer and Mathematician, who has inspired me to achieve the
best results academically, but more than that, personally. My mom for putting
her career on hold to raise me and my sisters, and moulding us into the people
we are today. To both parents; I would not have achieved what I have today
were it not for your guidance, encouragement and faith in my abilities.
This thesis could also not have been done (on time at least!) were it not
for my wife, Becky, who has been very supportive in every way and also made
sure that I didn’t spend too much time on capoeira! It is also dedicated to my
two golden girls, Ayla and Leilani who are very special to me. Thanks to all
those York capoeira who would play me in the roda even when the stress was
evident and I was playing a bit too hard, axe, meu irmao.
Special thanks to Kai-Kit Wong, who introduced me to MIMO-OFDM and
the problem of channel estimation in overloaded systems. Thanks also to K.
Paulson, and Nick G. Riley for their contributions.
ii
Contents
List of symbols
1 An Introduction to MIMO-OFDM Systems
1.1 Predicting the Emerging and Future Wireless Communications
Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . .
1.3 The MIMO-OFDM System Model . . . . . . . . . . . . . . . . .
1
3
4
6
9
1.4 The MIMO-OFDM Air Interface
. . . . . . . . . . . . . . . . . 13
1.4.1
IQ Constellation Mapping . . . . . . . . . . . . . . . . . 13
1.4.2 Digital multitone/multi-carrier modulation . . . . . . . . 17
1.4.3 Maximum Likelihood Detection . . . . . . . . . . . . . . 19
1.5 The MIMO-OFDM Mapping/De-mapping Function . . . . . . . 23
1.5.1
Space-Frequency Coding . . . . . . . . . . . . . . . . . . 24
1.5.2
Spatial Multiplexing . . . . . . . . . . . . . . . . . . . . 27
1.6 Multi-user MIMO-OFDM . . . . . . . . . . . . . . . . . . . . . 29
1.7 Research Objectives of the Thesis . . . . . . . . . . . . . . . . . 31
2 The Wireless Channel
33
2.1 Multipath Propagation . . . . . . . . . . . . . . . . . . . . . . . 35
2.2 Tapped-Delay-Line System Model . . . . . . . . . . . . . . . . . 37
2.2.1
Statistical Model of a Multipath Channel . . . . . . . . . 41
2.2.2 Bandlimited transmission . . . . . . . . . . . . . . . . . 45
iii
2.2.3 Rayleigh Fading Channels . . . . . . . . . . . . . . . . . 48
2.3 Saleh-Valenzuela channel Model
. . . . . . . . . . . . . . . . . . 50
2.4 Correlation of the channel gain parameters of MIMO antennas . 55
2.5 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . 57
3 Coherent Detection for MIMO-OFDM Systems
61
3.1 OFDM Equalization . . . . . . . . . . . . . . . . . . . . . . . . 63
3.2 SISO-OFDM Channel Estimation . . . . . . . . . . . . . . . . . 68
3.2.1 One Dimensional Channel Estimation . . . . . . . . . . . 70
3.2.2 Two Dimensional Channel estimation . . . . . . . . . . . 73
3.3 MIMO-OFDM Channel Estimation . . . . . . . . . . . . . . . . 77
3.3.1 Least Squares Solution . . . . . . . . . . . . . . . . . . . 78
3.3.2 Time Domain LS Channel Estimation . . . . . . . . . . . 79
3.3.3 Performance of the Channel Estimator . . . . . . . . . . 84
3.4 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . 86
4 Reduced Parameter Channel Estimation
89
4.1 CSI Frequency Correlations
. . . . . . . . . . . . . . . . . . . . 90
4.1.1 OFDM Frequency Correlations
. . . . . . . . . . . . . . 91
4.1.2 Effects of Multipath on Frequency correlations . . . . . . 92
4.2 RP-CSI Basis Functions
. . . . . . . . . . . . . . . . . . . . . . 93
4.2.1 Wavelet Basis . . . . . . . . . . . . . . . . . . . . . . . . 94
4.2.2 Principal Component Analysis Basis
. . . . . . . . . . . 98
4.3 The Proposed Method . . . . . . . . . . . . . . . . . . . . . . . 102
4.3.1 OFDM Symbol based correlations . . . . . . . . . . . . . 102
4.3.2 OFDM sub-symbol based correlations . . . . . . . . . . . 108
4.4 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . 113
5 Reduced Parameter Channel State Information Analysis
116
5.1 The Lower Bound for MSE in Channel Estimate . . . . . . . . . 117
iv
5.2 RP-CSI Simulation Results . . . . . . . . . . . . . . . . . . . . . 121
5.2.1
Simulation Results for L = 4 . . . . . . . . . . . . . . . . 121
5.2.2
Simulation Results for L = 8 . . . . . . . . . . . . . . . . 124
5.2.3
Simulation Results for L = 16 . . . . . . . . . . . . . . . 126
5.3 OFDM Sub-symbol based MIMO-OFDM channel Estimation . . 128
5.3.1 Orthogonal Training Sequences for channel estimation . . 129
5.3.2 OFDM sub-symbol based MU-MIMO-OFDM channel
estimation . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5.4 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . 139
6 Time Varying Channels
142
6.1 Clarke’s Model
. . . . . . . . . . . . . . . . . . . . . . . . . . . 143
6.2 Slepian Basis Expansion . . . . . . . . . . . . . . . . . . . . . . 147
6.3 Kalman filter Tracking . . . . . . . . . . . . . . . . . . . . . . . 155
6.3.1 Deriving the Kalman Filter Process Model
. . . . . . . . 158
6.3.2 Deriving the Kalman Filter Measurement Model . . . . . 162
6.4 Conclusions & Future Work . . . . . . . . . . . . . . . . . . . . 165
A Eigen Decomposition of the Channel Covariance Matrix
178
B Power Spectral Density
C Channel gain frequency correlations
D WICOM-06 Conference Paper
181
184
186
D.0.1 OFDM Systems . . . . . . . . . . . . . . . . . . . . . . . 189
D.0.2 The Fading Channel
. . . . . . . . . . . . . . . . . . . . 189
D.0.3 The Overloaded Channel Estimation Problem . . . . . . 190
E MATLAB CODE
203
E.1 Saleh-Valenzuela Channel Model . . . . . . . . . . . . . . . . . . 203
v
E.2 Wiener Filter Implementation . . . . . . . . . . . . . . . . . . . 208
E.3 RP-CSI Estimator
. . . . . . . . . . . . . . . . . . . . . . . . . 218
E.4 Wavelet Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
E.5 Slepian/ Discrete Prolate Spheroidal Sequences
. . . . . . . . . 229
E.6 Orthogonal Training Sequence Training . . . . . . . . . . . . . . 230
E.7 Kalman Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
vi
List of Figures
1.1 A generic MIMO-OFDM communications system. Data flow in
the diagram is directed by the arrows, starting with the ”Input
bit sequence” at the transmitter, and ending at the ”Output bit
sequence” at the receiver. The wireless channel is the transmis-
sion medium.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.2 Constellation diagram of 4-QAM with Gray Coding . . . . . . . 15
1.3 Constellation diagram of rectangular 16-QAM with Gray Coding 16
1.4 Block Diagram representation of OFDM modulation. Because
the sub-carrier channels are orthogonal and separable at the
receiver, they are depicted as parallel channels. The symbols
and channel parameters are complex numbers representing the
separable I and Q components. The square blocks represent
complex variables. . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.5 Space-Frequency Alamouti Coding for a (2,2) MIMO-OFDM
system. The source OFDM symbol is mapped onto two OFDM
stacks which have space-frequency correlations.
. . . . . . . . . 25
vii