For the downlink of direct sequence code-division multiple access (DS-CDMA) systems, transmitter (Tx)-based multiple access interference (MAI) cancellation techniques, termed multiuser precoding, and Tx-based diversity techniques can significantly increase system capacity while remaining low complexity at mobile stations (MS). We have proposed novel linear and nonlinear decorrelating precoding algorithms. Hybrid transmitter designs were developed to combine MAI cancellation and diversity techniques. We have also analyzed the important duality between Tx-based multiuser precoding and receiver (Rx)-based multiuser detection.
The class of decorrelating precoding techniques is simple, efficient and satisfies the minimum mean square error (MMSE) criterion. The nonlinear decorrelating Tomlinson-Harashima precoding (THP) inherently outperforms linear MAI cancellation methods. For frequency-selective fading channels, we developed the THP with Pre-RAKE combiner (PreRAKETHP) and Multipath Decorrelating THP (MDTHP) designs. While the PreRAKETHP is the optimal THP design by both ZF and MMSE criteria, it is computationally complex because the MAI cancellation filters depend on the instantaneous channel gain coefficients and therefore need to be updated frequently. In the MDTHP, the precoding filter is independent of the channel state information (CSI). The MDTHP is simpler than PreRAKETHP at the expense of moderate performance loss.
For Pulse Amplitude Modulated (PAM) and Quadrature Amplitude Modulated (QAM) systems with small constellation index, the practical performance of THP is degraded due to the side effects of modulo operation. In contrast, linear precoding is not influenced by modulation and is simpler to implement. We have developed several linear precoding techniques. The PreRAKE Linear Decorrelating Precoding (PreRAKELDP) is the optimal ZF and MMSE precoding method. The Multipath Decorrelating Precoding (MDP) provides a simpler but suboptimal scheme. It is shown that the PreRAKELDP and MDP outperform the existing linear precoding techniques with similar complexity. The system performance is further improved by employing multiple antennas in transmitter design.
Both multiuser precoding and transmit diversity techniques require the knowledge of CSI at transmitter. For rapidly varying channels, the long range channel prediction (LRP) provides the accurate CSI in time. We have demonstrated that the LRP method enables the proposed Tx-based techniques for practical systems.
