Downlink spatial division multiple-access (DL-SDMA)-assisted multiuser transmission (MUT) employing linear dispersion codes (LDCs) is proposed for space-time signaling, which attains a high diversity order and may support a high-rate data transmission. However, a potentially high complexity may be imposed by the signal detection when aiming to achieve optimum maximum-likelihood (ML) performance. Hence, a novel irregular sphere detection (IrSD) design is proposed to maintain an infinitesimally low bit error rate (BER) at the lowest possible iterative decoding complexity. Quantitatively, in the specific scenario, when six 4-quadratic-amplitude-modulated (QAM) data symbol streams were transmitted by the DL-SDMA system, the full-search-based ML detector had to evaluate all 4 ⁶ = 4096 possible decision candidates. By contrast, when searching through a fraction of this space, the proposed IrSD may reduce the complexity by three orders of magnitude, quantified in terms of the number of add-compare-select (ACS) arithmetic operations, namely, from the order of 10 ⁸ to about 10 ⁵ at E _b / N ₀ = 5 dB, and still approach the ML performance.