Evaluation of polynomial matrix SVD-based broadband MIMO equalization in an optical multi-mode testbed

Interference in frequency-selective multiple-input multiple-output (MIMO) systems can be removed by applying a block-transmission description based on spatio-temporal vector coding and using singular value decomposition (SVD). In this contribution a newly developed SVD algorithm for polynomial matrices (PMSVD) is analyzed and compared to the commonly used SVD-based MIMO equalization. Since the PMSVD processing results in independent single-input single-output layers exhibiting a frequency-selective characteristic, a simple zero forcing equalizer or an optimal Viterbi detector is applied in this work so as to remove the occurring inter-symbol interferences. The PMSVD orthogonalization is compared with the conventional SVD in terms of the achievable spectral efficiencies, showing that there is no loss applying PMSVD over SVD processing. In addition, the bit-error rate performances in different channel scenarios are evaluated and optimized by applying combined bit and power allocation schemes. One scenario involves the measured specific impulse responses of a (2×2) optical MIMO channel, consisting of a 1.4 km multi-mode fiber, optical fusion couplers for mode multiplexing and demultiplexing when operating at a wavelength of 1576 nm. The computer simulation results show that the PMSVD could be an alternative signal processing method compared to conventional SVD-based approaches in frequency-selective MIMO channels.