Experimental and Numerical Evaluation of CAZAC-type Training Sequences for MxM SDM-MIMO Channel Estimation

Spatial-division multiplexing (SDM) increases capacity by using advanced fibers, amplifiers, and switches. Powerful digital signal processing (DSP) is needed for SDM systems to equalize the multiple-input multiple-output (MIMO) transmission channel. Towards a commercialization of SDM systems in the medium term, training-based channel estimation and equalization can ensure more stable performance than blind adaptive algorithms. Furthermore, frequency domain equalization is expected to have lower computational complexity than time domain equalization for SDM applications with high modal dispersion and delay spread. However, as the number of orthogonal training sequences required in an MxM SDM-MIMO system is equal to the number M of parallel spatial paths/modes, the scalability of such an approach needs to be verified. To this end, this study experimentally and numerically compares cyclic shifted constant-amplitude zero-autocorrelation (CAZAC) training sequences (TS) with different number of repetitions, sequence lengths and scalings for channel estimation in an SDM-MIMO system. Validation experiments are performed using 32 GBd polarization-division multiplexed 16-ary quadrature amplitude modulated (PDM-16QAM) data signals in a 4x4 SDM-MIMO transmission testbed with two spatial paths to estimate the impact of the different TS configurations on the end-to-end system performance transmitting over a few mode fiber (FMF).