Optimizing CSI Feedback for MU-MIMO: Tradeoffs in Channel Correlation, User Diversity and MU-MIMO Efficiency

Multi-user MIMO (MU-MIMO) is seen as one of the key technologies for 4G communications systems. A critical factor affecting the throughput gains provided by MU-MIMO is the feedback mechanism that allows the transmitting basestation to periodically receive channel state information (CSI) from mobiles. Sufficient feedback is necessary for efficient MU-MIMO operation. At the same time feedback itself is an overhead using resources that could otherwise be used for data transmission. It is therefore important to carefully consider how feedback is designed and used to support MU-MIMO. This work studies key design parameters in such a feedback process: how often in time and in frequency should mobiles perform feedback (given coherence in the channel), the number of bits of feedback (accuracy of CSI quantization), and the number of mobiles to provide feedback. Within a bound on feedback resources these parameters can determine how best to use such resources to optimize MU-MIMO spectral efficiency. The results show that spectral efficiency is often maximized if mobiles quantize their channels very accurately but perform feedback at moderate intervals in time, such that CSI inaccuracy is primarily determined by the variation of the channel in time and frequency rather than being dominated by quantization errors.