Block Diagonal Inversion Precoding for MIMO Broadcast Channels

The sum capacity of a broadcast MIMO channel can be achieved sub-optimally using linear precoding techniques. Block diagonalization (BD) is a linear precoding technique that achieves near to the sum capacity with low complexity by nullifying the inter-user interference. However, one limitation of BD is that the equivalent channel after block diagonalization has to be communicated to the users using additional pilot symbols or a limited feedforward link. The former method increases the system overhead and the latter suffers from quantization error. Transmitter zero forcing may be an alternative but it has a high power enhancement. Besides, transmitter zero forcing lacks flexibility for use in multi-antenna users. In this paper, a new linear precoding technique called Block Diagonalization Inversion (BDI) is proposed. The proposed scheme avoids the need to communicate the equivalent channel by using a combination of block diagonalization and partial inversion of the equivalent channel at the transmitter. BDI enables per-stream power alloca- tion and adaptive modulation like BD; besides, BDI has a lower power enhancement than transmitter zero forcing. It is shown using simulations that BDI achieves a sum rate performance in-between an ideal BD and transmitter zero forcing. When a limited feedforward link is used with BD, it is demonstrated that BDI provides higher sum rate than BD.