Influence of Channel Estimation on Wireless Communication System Performance
Many wireless communication systems today employ receivers that use channel estimation as an integral part of the receiver chain. Two sides of this problem are studied in this thesis: Impact of channel estimation on low complexity receivers for multi antenna and CDMA systems, and the performance of systems based on superimposed pilots and systems with peak power constraints.
First the connection between iterative least squares methods, relying on the finite alphabet properties of the transmitted signal, and the Expectation Maximization (EM) method is explored. A modification to the EM-algorithm known as SAGE, is proposed to be used in order to reduce the computational complexity and increase the performance. These algorithms are extended to cope with coded signals. By utilizing the extended distance between different codewords, significant performance gain can be seen. If a turbo code is employed for error protection, the complexity can be reduced even further by exploiting the iterative nature of the decoder. The last chapter in the first part is devoted to channel parameter estimation in a slotted DS-CDMA system. A near-far resistant estimator based on a large sample approximation to the maximum likelihood estimator is presented. The performance is evaluated by means of numerical simulations, and the estimator is also shown to be asymptotically efficient.
In the second part the performance of training based and peak power limited systems is evaluated by means of channel capacity. The performance of MIMO-systems and systems with frequency selective channels are studied when the training is done in the form of SuperImposed Pilots (SIP). The parameters of the SIP scheme are optimized based on the mutual information, and the advantage of using superimposed pilots over conventional pilots is demonstrated for short coherence times. One issue with SIP is that the accuracy of the channel estimate is decreased. A method based on data aided channel estimation is proposed to mitigate this problem, and the proposed algorithm is found to perform close to the noncoherent channel capacity. Finally, the capacity achieving distribution for a Rayleigh fading system with peak power constraint is found. It is shown that a uniform phase and a discrete amplitude distribution is the optimal signalling scheme.