The 400ZR Reach Math Nobody Shows You

Open any 400ZR datasheet and you'll find a reach number. 120 km. Sometimes 80 km. Sometimes 140 km on the more aggressive specs. That number is a clean-link ceiling. It is not the distance you should be designing your network to.

The difference between the datasheet ceiling and what you can actually deploy is the difference between a one-time lab demo and a network that will still link in three years on the second connector you didn't clean. The full budget rarely shows up on the vendor slide. Here it is.

The components of the OSNR budget

The link budget for a coherent ZR link is governed by Optical Signal-to-Noise Ratio at the receiver. The OIF spec calls out the minimum OSNR required for the FEC to decode at the target BER. That's the threshold. Everything you add or subtract above it is your margin.

Six things eat into your margin:

Span loss. The dominant term. A typical G.652 fiber loses 0.2 dB per kilometer. 100 km gives you 20 dB of pure span loss before anything else.

Connector and splice loss. Every connector is 0.3 to 0.5 dB on a good day. A real deployment has six to twelve connectors per link (patch panels at both ends, intermediate distribution frames, the test point nobody documented). Add 3 to 5 dB.

Amplifier noise figure. EDFAs add ASE noise. Each amp pass adds noise figure (4 to 6 dB typical), and the OSNR degrades cumulatively. Three amplifier hops on a 240 km link easily costs you 4 to 5 dB of OSNR margin.

Fiber type and chromatic dispersion compensation. G.652 is standard. G.655 changes the dispersion budget. If you don't know which one is in your patch panel, you don't know your reach.

Nonlinearity at high launch power. Push the launch power too high to overcome span loss and you start clipping the modulation. SPM and XPM eat the constellation. The optimal launch power is lower than you'd intuitively guess.

Polarisation-dependent loss. Aged fiber, splices, and connectors all contribute. 1 to 2 dB on a real deployment is normal. The DSP compensates but the cost shows up in margin.

The clean-link assumption

Vendor datasheets assume new, single-section, factory-tested fiber with two connectors and no splices. They also assume the link is exactly at the launch power the spec optimised for, with no operational headroom.

Your link is not that link. Your link has six connectors of varying age, two repair splices from the last cable break, and an amplifier you inherited from the previous generation that runs slightly hot. The datasheet number was correct. It was just for somebody else's link.

What the real margin looks like

Take a 100 km metro DCI link, two amplifier hops, eight connectors, a single OADM passthrough, G.652 fiber, normal operational state.

Span loss: 100 km × 0.2 dB/km = 20 dB.
Connector loss: 8 × 0.4 dB = 3.2 dB.
OADM passthrough: ~2.5 dB.
Amplifier noise contribution: ~3 dB OSNR penalty.
PDL: ~1.5 dB.
Aging margin (3-year deployment): ~2 dB.

That's a roughly 32 dB total budget for a link the datasheet would call 100 km clean. The 400ZR module's launch power and FEC margin handle it. Just barely.

Now run the same calculation at 130 km. The span loss alone is 26 dB. Add the rest, you're at 38 dB. The module's launch power is fixed. The FEC margin tightens to the point where one extra connector — say, a patch panel you added in year two — pushes you under threshold.

This is why "400ZR up to 120 km" is true on the datasheet and "400ZR comfortably to 80 km, plan to 100 km, push 120 km only on clean fiber you control end to end" is true in practice.

The questions to actually ask

When you spec a 400ZR link, ignore the marketing reach number. Ask these instead.

What's the launch power, in dBm? What's the receive OSNR threshold for the FEC class? What's the manufacturer's recommended margin for aging? How does the module behave at low input power — does the DSP go into a degraded state cleanly, or does it flap?

The answers won't match the front-page datasheet number. That's the point.

When 400ZR+ is the honest answer

400ZR+ exists because the OIF community knew this. The plus variants raise launch power, sharpen the FEC, and trade a bit of power consumption for real margin. If your link plan needs more than about 100 km on real-world fiber with normal connector counts, 400ZR+ is not a luxury upgrade. It's the version that matches your budget.

The vendor sales motion will sell you 400ZR for the price-point story. The right engineering answer for half the metro DCI builds I see is 400ZR+. Build the budget honestly first, then pick the module.