Design of linear dispersion codes: some asymptotic guidelines and their implementation

The family of linear dispersion (LD) codes is a diverse set of space-time block codes that subsumes several standard designs. In this paper, we provide a design technique for LD codes that generates codes which enable large capacities and perform well when decoded with a standard suboptimal detector. Our design technique is motivated by the observation that for an independent Rayleigh fading channel, as the number of transmit antennas grows, LD codes with a certain orthogonal structure simultaneously approach a maximized upper bound on the capacity, and a minimized lower bound on a certain mean square error performance measure. Using this asymptotic result as a guide, we impose the orthogonal structure on finite sized codes and optimize the resulting system. Imposing this structure significantly simplifies the design procedure, and as we demonstrate via simulation, leads to codes that perform well in practice. Using insight generated by our study of the asymptotic properties of LD codes, we also propose a row interleaving scheme which is shown to result in significant performance enhancement.