Training-Based and Blind Channel Estimation and Their Impact on MIMO System Performance

Xia Liu (2010). Training-Based and Blind Channel Estimation and Their Impact on MIMO System Performance PhD Thesis, School of Information Technology and Electrical Engineering (ITEE), The University of Queensland.

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
s4101784_PhD_finalabstract.pdf Thesis abstract application/pdf 17.37KB 1
s4101784_PhD_finalthesis.pdf Full thesis final version application/pdf 1.51MB 17
Author Xia Liu
Thesis Title Training-Based and Blind Channel Estimation and Their Impact on MIMO System Performance
School, Centre or Institute School of Information Technology and Electrical Engineering (ITEE)
Institution The University of Queensland
Publication date 2010-09
Thesis type PhD Thesis
Supervisor Marek E. Bialkowski
Vaughan Clarkson
Total pages 214
Total colour pages 50
Total black and white pages 164
Subjects 08 Information and Computing Sciences
Abstract/Summary In the last decade, the number of mobile and wireless communications users has dramatically increased. Wireless communication systems have experienced evolution of first generation (1G) and second generation (2G). Currently, the third generation (3G) wireless communication system is rapidly spreading all over the world providing high data rate wireless multimedia services. Yet, the pursuit for higher data rates, larger coverage and more spectral efficient mobile communication still goes on. Multiple Input Multiple Output (MIMO) systems have emerged and been nominated as a promising solution for future generation wireless communications systems due to their capability of enhancing channel capacity, spectral efficiency and coverage. These enhancements are possible in a rich scattering environment when array antennas are used at transmit and receive ends of a communication link. To realize these benefits, a MIMO system requires the channel state information (CSI) to be available at the receiver. It has to be known before any data can be decoded. In practice, CSI is obtained by channel estimation and as a result most of MIMO detection schemes rely on an accurate estimation of CSI at the receiver end. Therefore, channel estimation is extremely critical to the proper functioning of MIMO systems. However, as CSI can be determined only in an approximate manner, it is unclear how estimation errors affect MIMO system performance. Most of works focusing on the MIMO systems’ potential assume perfect CSI available at the receiver end while estimating capacity. This assumption is not true in practice and thus the existing gap on the relationship between an estimated CSI and MIMO channel capacity forms the main motivation for the work undertaken in this thesis. The MIMO channel estimation can be performed by sending training sequences which are known both to the transmitter and the receiver. This is the most popular method to estimate the MIMO channel. In majority of works which reported the training-based MIMO channel estimation methods, the channel coefficients in the channel matrix are assumed to have Gaussian identical independent distribution (i.i.d). This assumption is not true in practical scenarios because of a limited number of scattering objects and non-ideal operation of array antennas. The shortfall of training-based channel estimation is that the training sequence does not contain any information and thus it sacrifices a considerable amount of bandwidth. To save the bandwidth and increase the spectral efficiency, blind channel estimation and semi-blind channel estimation can be used to obtain CSI. For blind and semi-blind channel estimation, no training symbols or fewer training symbols are needed to estimate the channel. Also, the transmitter does not need to cooperate with the receiver. Most of these methods are based on second or higher order statistic models of the received signal. Their disadvantage is the increased computational complexity, gradual convergence, and scalar and phase ambiguities of channel matrix elements. The aim of this thesis is to investigate the performance of training-based channel estimation and assess its impact on MIMO channel capacity under realistic channel models; optimize training sequences when CSI is available to the transmitter and the receiver, and develop blind channel estimation algorithms with low complexity and without scalar or phase ambiguity. To realize these goals, the thesis introduces the information theory of MIMO system and shows how MIMO channel capacity is related to the properties to the MIMO channel matrix. Next, realistic channel models are described which include varying distributions of scattering objects and actual electrical properties of different configurations of antennas with varying elements spacing. These models are used to assess performances of training-based channel estimation methods including Least Square (LS), Scaled Least Square (SLS) and Minimum Mean Square Error (MMSE) methods when CSI is available only at the receiver side. It is shown that as the assumed channels properties divert from the i.i.d. case, because the scattering environment and antennas operation turn away from ideal conditions, spatial correlation affects both channel estimation and capacity. In order to have a better insight into the obtained results, spatial correlation is linked to physical parameters of the channel as well as to the mathematical properties of the channel matrix. It is shown that an increased spatial correlation helps to improve the channel estimation accuracy. However, the overall effect is the decreased channel capacity. This finding shows that the MIMO system does not have to rely on the perfect knowledge of CSI to achieve increased capacity. In the next step, considerations extend to the case when CSI is assumed to be available both at the receiver and the transmitter. This scenario creates an opportunity to optimize transmitted training sequences and thus improve channel estimation. As in the previous considerations, optimized training sequences are devised under the assumption of advanced channel models. This part of investigations also includes derivations of closed-form expressions of MIMO channel capacity and bit error rate (BER) performance by taking into account channel estimation accuracy. The obtained results show that the use of LS, SLS and MMSE methods has a different effect on MIMO BER performance, with LS offering the worst performance and MMSE giving the best BER performance. The final part of the thesis focuses on devising a new blind channel estimation algorithm which employs a simple coding scheme to avoid scalar and phase ambiguities for the MIMO channel matrix. Its validity is verified by extensive simulations followed by experiments on a MIMO test-bed employing a Field Programmable Gate Array (FPGA). It is shown that the proposed blind channel estimation algorithm is easy to implement in a DSP firmware due to its low computational complexity. Also it shows fast convergence. It offers good performance for fast fading channels in addition to slow fading channels. The work undertaken as part of this thesis has been published in several journals and refereed conference papers, which underline the originality and significance of the thesis contributions.
Keyword MIMO
Channel estimation
estimation error
channel model
channel capacity
spatial correlation
mutual coupling
Additional Notes 55, 60-63, 65-67, 69, 71, 73-77, 97-98, 101, 103-109, 112-114, 119-120, 128, 130, 132-133, 136-137, 150-151, 153, 160-161, 163-164, 178, 185, 187, 191, 193, 196, 199

Citation counts: Google Scholar Search Google Scholar
Created: Wed, 02 Mar 2011, 21:46:21 EST by Mr Xia Liu on behalf of Library - Information Access Service