400G Midhaul Upgrade on a Long DWDM Span
An operator is upgrading a 5G midhaul network that connects 40 distributed units to three central units over 20–60 km spans. The existing DWDM line system carries 100G wavelengths with 50 GHz spacing, 24 dB span loss, and 22 dB OSNR. To support 400G per wavelength, the vendor proposes 64QAM with 75 GHz spacing and coherent optics, claiming a 3 dB OSNR margin. After a trial, the 60 km span shows 15.5 dB OSNR with pre-FEC BER 8×10^-3, and latency occasionally spikes from 250 µs to 2 ms. The SLA requires 1 ms one-way midhaul latency, 99.999% availability, and PTP timing within 1.5 µs. The operator can add an amplifier site or reduce to 200G. You must recommend a design that meets the SLA and timing without unlimited cost.
1. Assess whether 400G at 64QAM is viable on the 60 km span and explain the OSNR/latency symptoms.
Read the worked answer
400G at 64QAM has a high OSNR requirement. The trial’s 15.5 dB OSNR is well below what a robust 64QAM 400G channel needs, especially with only a claimed 3 dB margin and 24 dB span loss. The pre-FEC BER of 8×10^-3 indicates the receiver is operating near or beyond its correction limit; occasional latency spikes to 2 ms are consistent with FEC stress, queueing, or protection/retry behavior rather than pure fiber delay.
The 60 km fiber itself contributes only about 300 µs one-way at 5 µs/km, so a 2 ms spike is not propagation delay. It suggests the link is not stable at 400G. I would not accept 64QAM 400G on this span without adding OSNR margin. Either add an amplifier site, reduce to 200G with a lower-order modulation, or use 400G only on short spans where OSNR is proven.
2. Compare adding an amplifier site with falling back to 200G for SLA and timing.
Read the worked answer
An amplifier site improves OSNR and can make 400G viable, but it adds cost, power, space, and a new failure domain. It also adds amplified spontaneous emission noise and requires careful gain setting; it does not automatically guarantee PTP asymmetry or latency. For 40 DUs, 200G per wavelength is likely sufficient and has a lower OSNR requirement, so it can run on existing spans without civil works.
My recommendation is a mixed design: 400G on short spans with adequate OSNR margin, 200G on the 60 km span, and no amplifier unless capacity forecasts prove 200G insufficient. This meets the 1 ms latency budget because the fiber delay is unchanged; the main latency risk is congestion and protection switching, which a stable 200G link reduces. Verify PTP with boundary clocks and measure asymmetry; an amplifier would not fix timing if asymmetry comes from mapping or protection.
3. What protection and timing design would you verify before accepting the midhaul?
Read the worked answer
Verify protection at the optical/OTN layer with a tested switchover time inside the 99.999% availability target, and confirm that protection does not create a latency spike above 1 ms. Use redundant paths that are physically diverse where possible; a single amplifier site must not become an unprotected common point. Check that pre-FEC BER and OSNR margins remain acceptable under worst-case aging and temperature.
For timing, require G.8275.1 PTP with SyncE or boundary clocks as appropriate, and measure time error at the far end under normal and protection-switch conditions. Confirm that the PTP path is symmetric and that 200G/400G mapping does not introduce variable delay. Acceptance should include a 24-hour soak with latency, BER, and time-error captures, not just a one-off power reading.