Title:
|
Blind equalisation schemes for dispersive stationary and mobile channels
|
In this thesis the problem of blind equalisation of stationary and fading mobile channels is investigated. Blind equalisers achieve a bandwidth economy by eliminating the training overhead, often invoked in trained channel equalisation. The family of blind equalisers found in the literature is critically appraised and the achievable performance is evaluated by computer simulations. We focus our attention on the set of per-survivor processing based techniques as well as on the Bussgang techniques. A specific application is also considered in the context of digital video broadcasting. An extension of the well-known constant modulus algorithm, namely the DFE-CMA [1,2], is proposed for equalising channels having a long inverse impulse response. This algorithm is characterised in terms of its performance and convergence in comparison to the conventional constant modulus algorithm. It is found that the proposed algorithm outperforms the conventional constant modulus algorithm for transmissions over channels exhibiting impulse responses having equal-weight taps. The extension of blind equalisation using channel decoding assisted feedback is also considered in the thesis. The corresponding modified per-survivor processing algorithm is iteratively invoked in order to provide improved joint data detection and channel estimation, by utilising the enhanced-reliability feedback of the channel decoder. The application of the M-algorithm is also considered in order to reduce the complexity of this blind turbo equalisation algorithm for transmissions over channels having a high number of CIR taps. The performance of this algorithm is evaluated by computer simulations and it is found that it improves the performance of non-iterative per-survivor processing, owing to the employment of the channel decoding feedback. Finally, different channel coding techniques were invoked and benchmarked in conjunction with the above blind turbo-PSP equaliser. Specifically, a range of convolutional coding, turbo convolutional coding, trellis-coded modulation and turbo trellis-coded modulation assisted blind turbo-PSP schemes were comparatively studied.
|