Fig. 6. (a) Performance (QdB) of 14 GBaud DP-16QAM transmissions using LDC, CSS DBP, LSS DBP, and BLSS DBP techniques over varying signal launch powers (dBm); (b) Performance (QdB) of CSS DBP, LSS DBP, and BLSS against the number of calculation steps/span at 13 dBm launch power.
The overall performance benefit of the BLSS DBP over other compensation techniques as a function of signal launch power is analyzed, as shown in Fig. 6(a). Observing the optimum launch power, indicating the NLT, the logarithmic step size-based DBP techniques demonstrate superior performance compared to the constant step size and linear dispersion compensation only. Using the BLSS DBP technique, performance benefits (∆Q) of 1.19 dB and 0.71 dB are obtained over CSS DBP and LSS DBP techniques, respectively. The results indicate that the BLSS DBP produces optimum step sizes for efficient DBP calculation. At higher powers above the NLT, the performance of DBP is degraded by accumulated stochastic ASE noise.
It is known that the effective implementation of DBP is significantly dependent on the number of calculation steps/span, . However, a high number of steps/spans comes at a high computational cost and are undesirable. Fig. 6(b) shows the accuracy of DBP calculation (in Q-factor) as a function of steps/span for different DBP implementations. It is observed that for N ≤ 10, BLSS DBP cancels dispersion and nonlinearity with higher accuracy than LSS DBP and CSS DBP. Specifically, the BLSS DBP reaches optimum performance with N = 10. In contrast, CSS DBP achieves optimum performance with N = 16, reducing the number of calculation steps by 6, corresponding to a reduction in complexity of 38% compared to the conventional CSS DBP. The constellation diagrams for N = 1, 2, 6, and 10 with BLSS DBP are shown in Fig. 7. At a high number of steps (N > 15), optimizing the step sizes becomes less important as the conventional DBP algorithm becomes equally accurate.