Study On Linear Precoding Techniques Based On Iterative Detection

In modern communication systems, certain channel state information at thetransmitter (CSIT) can be obtained through reciprocity or feedback. Linear precoding(LP) techniques can improve the system performance by properly pre-processing thetransmitted signal based on the available CSIT. This strategy is theoreticallycapacity-achieving and practically can be implemented through simple linear transformwithout changing the channel code and modulation.Most existing work on LP techniques is focused on either information theoreticalanalysis or uncoded systems. Direct applications of these designs in practically codedsystems usually lead to a poor performance. This paper considers the LP techniquesbased on iterative detection, which can utilize the information provided by the channelcoding through the feedback in iteration and optimize the precoder using the evolutionanalysis. Based on the above considerations, this paper has studied the following issues.The precoder design in multiple input multiple output (MIMO) orthogonalfrequency division multiplexing (OFDM) systems with complete and perfect CSIT isfirst considered. In this case, the channel can be fully diagonalized and the design isproved to be a convex optimization problem that can be easily solved using the standconvex optimization tools. Numerical results demonstrate that the proposed design canutilize the CSIT efficiently and improve the system performance significantly. Forexample, compared with the conventional design based on non-iterative detection, theproposed design shows a performance gain of about5.0dB.The paper next studies the precoder design in systems with incomplete CSIT. Insystems with delay constrained traffic, CSIT is obtained slot by slot. In this case, theprecoder is designed based on the incomplete CSIT that has been obtained in each slot.A search algorithm based on the quantization of received power is developed tooptimize the precoder design, and a hybrid power allocation algorithm is proposed toconstruct a low-complexity suboptimal precoder. Both designs obtain significantperformance improvement that is predicted by the information theoretical analysis. In asystem with delay constraint4slots, the optimized and suboptimal precoders obtain aperformance gain of about6.1dB and5.8dB, respectively. Due to the fact that the conditions of the channel reciprocity cannot be fullysatisfied or the delay between the CSIT acquisition and utilization, the CSIT obtainedby the systems are usually imperfect. The precoder design in OFDM systems withimperfect CSIT is studied. A simple but efficient approximate water-filling precoder isproposed, which improves the system performance by4.1dB at most. Based onevolution analysis, the precoder is further optimized in ergodic and quasi-static channels,which provide an additional performance gain.In MIMO systems, the uncertainty in the CSIT affects not only the channel gainbut also the channel directions. A permutation matrix is induced in the precoder design,which helps to make the coded symbols undergo the nearly equal channel conditions.The equation for evolution analysis is derived. Based on the evolution analysis, theprecoder is optimized in ergodic channels, and a suboptimal but simple average powergain (APG) precoder is proposed in quasi-static channels. Significant performance gainsare obtained by both designs. The simulation results demonstrate that a performancegain of6.5dB or so can be obtained by the proposed scheme in a cooperative cellularnetwork.