MIMO SVD-based multiplexing with imperfect channel knowledge

In narrowband multiple-input multiple-output (MIMO) communication systems with perfectly known channel state information (CSI), the singular value decomposition (SVD) is commonly used to decompose the MIMO channel into independent single-input single-output subchannels. In theory, optimal interference-free data multiplexing may then be carried out using the subchannel power levels provided by the well-known waterfilling solution. In practice, however, when finite codebooks are used and perfect CSI is unavailable, adaptations to power levels and bit-loading schemes are often needed to maintain reasonable performance. In this paper, we use expressions for the per-subchannel signal-to-interference- and-noise ratio for the imperfect CSI case to derive approximately optimal subchannel power levels and thresholds on the amount of CSI imperfections and noise tolerable in SVD-based multiplexing systems using M-ary quadrature amplitude modulation. Numerical simulations demonstrate the usefulness of the derived expressions.