Adaptive MIMO Systems with High Spectral Efficiency

Adaptive MIMO Systems with High Spectral Efficiency

Zhendong Zhou (University of Sydney, Australia) and Branka Vucetic (University of Sydney, Australia)
DOI: 10.4018/978-1-59904-988-5.ch014

Abstract

This chapter introduces the adaptive modulation and coding (AMC) as a practical means of approaching the high spectral efficiency theoretically promised by multiple-input multiple-output (MIMO) systems. It investigates the AMC MIMO systems in a generic framework and gives a quantitative analysis of the multiplexing gain of these systems. The effects of imperfect channel state information (CSI) on the AMC MIMO systems are pointed out. In the context of imperfect CSI, a design of robust near-capacity AMC MIMO system is proposed and its good performance is verified by simulation results. The proposed adaptive system is compared with the non-adaptive MIMO system, which shows the adaptive system approaches the channel capacity closer.
Chapter Preview
Top

Background

MIMO Channel Model

Consider a point-to-point wireless communication channel equipped with 978-1-59904-988-5.ch014.m01 transmit and 978-1-59904-988-5.ch014.m02 receive antennas. This channel can be modeled by a multiple-input multiple-output (MIMO) system, which can be represented as a baseband channel matrix H of size 978-1-59904-988-5.ch014.m03. The entry of H at the j-th row and i-th column, 978-1-59904-988-5.ch014.m04, represents the complex channel coefficient relating to the i-th transmit antenna and the j-th receive antenna. We assume that the antennas are placed at an enough distance between each other, such that the channel link between each transmit-receive antenna pair experiences independent fading. The fading is modeled as flat (non-frequency selective), since frequency selective fading can be combated by the orthogonal frequency division multiplexing (OFDM) technique.

The fading distribution is assumed to be Rayleigh, or equivalently, complex Gaussian, which represents the worst case fading situation. Proper normalization is assumed such that each entry 978-1-59904-988-5.ch014.m05 can be modeled as a complex Gaussian random process with zero mean and unity variance (variance 0.5 for both real and imaginary parts). We denote this as 978-1-59904-988-5.ch014.m06, where 978-1-59904-988-5.ch014.m07 denotes complex Gaussian distribution with mean 978-1-59904-988-5.ch014.m08 and variance 978-1-59904-988-5.ch014.m09.

The input to the MIMO channel is represented by an 978-1-59904-988-5.ch014.m10 vector 978-1-59904-988-5.ch014.m11, where xi denotes the transmitted (modulated) symbol from the i-th transmit antenna. The output from the MIMO channel is represented by an 978-1-59904-988-5.ch014.m12 vector 978-1-59904-988-5.ch014.m13, where yj denotes the received symbol from the j-th receive antenna. The additive white Gaussian noise (AWGN) is represented by an 978-1-59904-988-5.ch014.m14 vector 978-1-59904-988-5.ch014.m15, where nj denotes the equivalent complex noise at the j-th receive antenna. The channel equation can be written as

978-1-59904-988-5.ch014.m16. (1)

Graphically, this can be illustrated by Figure 1.

Figure 1.

MIMO channel model

978-1-59904-988-5.ch014.f01

For convenience, we define

978-1-59904-988-5.ch014.m17
978-1-59904-988-5.ch014.m18
978-1-59904-988-5.ch014.m19.

We apply a singular value decomposition (SVD) to the matrix H and get

978-1-59904-988-5.ch014.m20
(2) where D is an 978-1-59904-988-5.ch014.m21 diagonal matrix. Its diagonal entries are the singular values of H, that is, the non-negative square roots of the eigenvalues of HHH. We denote the eigenvalues of HHH as 978-1-59904-988-5.ch014.m22, thus

978-1-59904-988-5.ch014.m23
978-1-59904-988-5.ch014.m24 and 978-1-59904-988-5.ch014.m25 are 978-1-59904-988-5.ch014.m26 and 978-1-59904-988-5.ch014.m27 unitary matrices with left and right singular vectors of H as their columns, respectively. 978-1-59904-988-5.ch014.m28 and 978-1-59904-988-5.ch014.m29 are also called the eigenvectors of HHH and HHH associated with 978-1-59904-988-5.ch014.m30, respectively. The eigenvalues and eigenvectors are related as
978-1-59904-988-5.ch014.m31
978-1-59904-988-5.ch014.m32, 978-1-59904-988-5.ch014.m33.

Substituting Eq. (2) into (1), we obtain

978-1-59904-988-5.ch014.m34
978-1-59904-988-5.ch014.m35
(3) where

978-1-59904-988-5.ch014.m36.

Eq. (3) readily shows a parallel channel model. To put it more explicitly, we can rewrite Eq. (3) in a component-wise form

978-1-59904-988-5.ch014.m37, 978-1-59904-988-5.ch014.m38(4)978-1-59904-988-5.ch014.m39, 978-1-59904-988-5.ch014.m40 and 978-1-59904-988-5.ch014.m41. (5)

Eq. (4) and (5) are further illustrated in Figure 2.

Figure 2.

Parallel equivalent subchannel model

978-1-59904-988-5.ch014.f02

This clearly shows that the MIMO channel has been converted into m parallel subchannels. The equivalent channel input and output are 978-1-59904-988-5.ch014.m42 and 978-1-59904-988-5.ch014.m43, respectively. The subchannel power gains are 978-1-59904-988-5.ch014.m44.

Complete Chapter List

Search this Book:
Reset