On a Scalable Path for Multimode SDM Transmission

We investigate transceiver design and digital signal processing for spatially multiplexed transmission over multimode fibers. In conventional architectures, the full spatial domain of the transmission fiber has to be detected and processed such that the modal walk-off and mixture can be estimated and equalized. These architectures scale poorly with the number of modes supported, besides the sparsity of the fiber transfer matrix is not fully exploited. Instead, here we aim to employ selective mode vector launch and detection in order to minimize the number of optical front-ends required. In this case, an ideal basis for multiplexing is offered by principal modes, that to first order are frequency independent. We show that such mode vector basis can be used for full baud rate transmission over inter-data center distances despite limited coherence bandwidth and vulnerability to environmental-induced drift of the optical channel. It is shown that crosstalk at the receiver front-end can be significantly suppressed, critically reducing the number of coherent receiver front-ends to that of spatial tributaries aimed for data transmission – as opposed to the total number of fiber modes. Residual crosstalk can still be experienced due to environmental-induced channel drift and loss of orthogonality in presence of mode dependent loss. Multiple-input single-output digital signal processing is shown to be effective in this case, with the required equalizer array size scaling sub-linearly with the number of tributaries. A multimode fiber with 156 spatial and polarization modes and optimized for low modal dispersion is considered.