Transmit Precoding for the Multiple Antenna Broadcast Channel

Manish Airy, Sandeep Bhadra, Robert W. Heath Jr., Sanjay Shakkottai

Abstract: In this paper we compare the following two methods of transmit preceding for the multiple antenna broadcast channel: vector perturbation applied to channel inversion (also termed zero forcing or ZF) precoding and scalar Tomlinson-Harashima (TH) precoding applied to sum-rate achieving transmit precoding. Our results indicate that vector perturbation applied to channel inversion preceding can significantly reduce power enhancement and yields the full diversity afforded by the channel to each user. Scalar TH-modulo reduction significantly reduces the power enhancement for precoding based on sum-rate criterion. The solution to vector perturbation applied to ZF precoding requires the solution to an integer optimization problem which is exponentially complex, or an approximation to the integer optimization problem which requires the Lenstra-Lenstra-Lovasz algorithm of polynomial complexity. Instead we propose a simpler solution (an approximation) to the vector perturbation problem based on the Rayleigh-Ritz theorem (R.A. Horn and C.R. Johnson, 1985). This approximate solution achieves the same diversity order as the optimal vector perturbation technique, but suffers a small coding loss. This solution is of polynomial complexity order. Further, a small increase in complexity with a "sphere"-based search around this solution yields significantly better performance. Since this vector perturbation is required to be done at the symbol rate, the lower complexity of the proposed algorithm is valuable in practice