Filter bank based channel equalization in broadband wireless single-carrier systems

Abstract

Channel equalization is a very important anti-multipath technique in broadband communication systems and it has received much attention during the era of digital communications. Frequency selective fading arises whenever the bandwidth of the transmitted signals is comparable to, or larger than the channel delay spread. In the absence of any suitable signal processing in the receiver, this leads to significant distortion of the signal due to intersymbol interference, which is a major barrier to high-speed digital transmission over wireless channels. This thesis considers frequency-domain signal processing techniques to combat intersymbol interference effects in the context of single-carrier broadband wireless transmission. Meanwhile, it has been recognized that filter bank transforms with high frequency selectivity can offer many advantages over the current discrete Fourier transform based approaches for frequency-domain processing. The main objective of this thesis is to establish a novel single-carrier frequency-domain equalization model utilizing perfect reconstruction, orthogonal, complex modulated filter banks.

An introduction to modulated filter banks and common channel equalization techniques is first given. The main research work presented in this thesis can be separated into two topic areas: frequency-domain channel equalization and combined equalization/decoding schemes for coded transmission. First, compared to the discrete Fourier transform approach, the important property of filter bank based equalization is that the channel subband response is not flat anymore. The subband equalizer responses are designed to cope with the channel response within each subband, by utilizing a low-complexity frequency sampling based approach. This is in contrast with the discrete Fourier transform approaches where channel equalization is done with a single complex multiplier per subband. One merit of using the filter bank approach is the absence of cyclic prefix preceding the data block, improving the data rate accordingly. Furthermore, this scheme can be used for any communications waveform and it exhibits improved tolerance against narrowband interference. The same filter banks can be used to provide a significant part of the channel filtering, thus relieving the receiver front-end complexity and leading to a very flexible receiver structure.

Second, in the case of coded transmission, the optimal way for equalization/decoding is to use the maximum a posteriori probability equalizer. The problematic issue of such an optimal method would be the high calculation complexity involved, especially when high-order modulation is applied and long channel delay spread may be encountered. This motivates to develop low-complexity solutions for the equalization/decoding loop. A brief introduction of well-known turbo equalization is given in this thesis and two low-complexity equalization/decoding methods are developed. Our approach is based on the decision feedback equalization concept utilizing the noise prediction model and decoding in the feedback loop.