| Summary: | In MIMO wireless communication systems when the transmitter is provided with some Channel State Information (CSI), then it is possible to apply a matrix transformation known as precoding to optimally allocate the power, thus resulting in high performance advantage in terms of capacity as well as the quality of the transmission compared to non precoded systems even with simple linear decoding algorithms at the receiver. However in practice perfect knowledge of the channel state information at the transmitter can not be guaranteed, hence precoding techniques based on partial CSI through limited feedback from the receiver are recently proposed and widely accepted in practical systems. These techniques assume invariant channels during transmission. Consequently, the optimal precoding matrix can be first found at the receiver during the training block, fed back to the transmitter and repeatedly used with the subsequent data blocks. However, for a time varying channel, the channel coefficients do not remain constant with time. Moreover, the feedback information is vulnerable to delay in the feedback channel, hence these methods become inefficient as the precoding matrix derived based on the current channel values becomes outdated. Therefore, the feedback information needs to be frequently updated. This thesis is mainly concerned with designing a feasible channel prediction technique to apply in time varying channels, in order to provide the transmitter with the optimal linear precoding matrix for subsequent data blocks, thus, eliminating the effect of the time varying channel and feedback delay. Firstly, the existing limited feedback unitary precoding algorithms are extensively investigated and a performance comparison among most of the proposed techniques is presented. The results are then used to establish the precoding selection criteria, size of the codebook and the linear receiver to be used.
|
|---|
