Abstract-In wireless communication systems, signal space diversity techniques are usually adopted to combat channel fading by exploiting time diversity, frequency diversity, spatial diversity or a combination of them. Most existing schemes to achieve signal space diversity are based on linear constellation spreading. In this paper, we propose a novel nonlinear signal space diversity technique based on maximum distance separable (MDS) codes. The new technique provides a design flexibility for almost any number of diversity channels and desired diversity orders. We also propose a simple and suboptimal diversity channel selection (DCS) decoding for our new scheme. DCS decoding can greatly reduce the decoding complexity at a cost of marginal performance loss relative to the optimal detection while keeping the diversity order. Simulation results show that with the same throughput but a lower decoding and implementation complexity, our scheme can have superior performance than the optimal linear spreading schemes over either independent fading or additive white Gaussian noise (AWGN) channels.