Low-Complexity Adaptive MIMO Equalizer With Hybrid-Precision Tap Bit-Width Optimization for Hardware-Efficient 200G Coherent PON

The adaptive multiple-input multiple-output (MIMO) equalizer deployed in coherent receivers can compensate for channel impairments in passive optical networks (PONs). The polarization rotation, chromatic dispersion, and device bandwidth limitation are effectively mitigated. However, the implementation of electronic adaptive MIMO equalizers in application-specific integrated circuits (ASICs) becomes increasingly challenging due to the substantial complexity with higher signal rates. To overcome this problem, we propose a hybrid-precision fixed-point quantization strategy that reduces the total tap bit-width, thereby lowering the hardware computational complexity. The genetic algorithm (GA) is employed to determine the optimal bit-width configuration. Moreover, for taps with minor magnitudes, redundant high-order bits are removed through left-shift operations to reduce the bit-width further. The simplified architecture is adapted to field-programmable gate array (FPGA) deployment. In the FPGA implementation, employing shared feedback modules for multi-channel tap coefficient updates can eliminate the structural redundancy. Finally, the proposed scheme is experimentally demonstrated in a C-band band-limited 200-Gbps coherent PON system. A 57.20% average reduction in resource utilization, accompanied by 0.60-dB and 0.62-dB power budget penalties in X and Y polarizations, is achieved when the signal-to-noise ratio (SNR) loss threshold for bit-width optimization is 0.4 dB after removing redundant bits. The optimized equalizer is employed in the 50-channel deployment to achieve a real-time rate of 100 Gb/s. The power budgets in X and Y polarizations, at 33.73 dB and 33.63 dB, respectively, are attainable with a power consumption of 38.2 W.