Performance and Complexity Analysis of Terahertz-Band MIMO Detection

Abstract

Achieving terabit-per-second (Tbps) data rates in terahertz (THz)-band communications requires bridging the complexity gap in baseband transceiver design. This work addresses the signal processing challenges associated with data detection in THz-band multiple-input multiple-output (MIMO) systems. We begin by analyzing the trade-offs between performance and complexity across various detection schemes and THz channel models, demonstrating significant complexity reduction by leveraging spatial parallelism across subspaces of correlated, typically ill-conditioned THz MIMO channels. We also derive accurate theoretical bounds on the detection error probability by incorporating THz-specific channel distributions and accounting for mismatches introduced by subspace decomposition. In addition, we propose a variation of subspace detectors that combines channel-matrix sorting, QR decomposition, and puncturing. Furthermore, under wideband THz UM-MIMO systems, we introduce a channel-matrix reuse strategy that minimizes exhaustive computations while maintaining reliable detection performance within a coherence bandwidth. Simulations over accurate THz channels show that the proposed efficient spatial parallelization schemes yield multi-dB performance gains, while the proposed reuse strategy offers significant computational savings with minimal performance degradation.