Engineering

Link Design — putting every budget together to close a real link

This is the capstone. Every other engineering page on this site solved one budget in isolation — power, noise, dispersion, nonlinearity, capacity, protection. A real design has to pass all of them at once. This page is the method that stitches them into a single, ordered flow, walks a fully worked multi-span example that closes every gate, and gives you a scorecard you can reuse. The reach of a link is set by whichever gate fails first, so the whole job is to find that gate.

The one question this page answers

"I have two sites, a required service, and a fiber route between them — will a link close, and if so how far and how much can it carry?" The answer is not one calculation but a sequence of gates. If every gate passes with margin, the design closes. If any gate fails, that gate is the limiter, and you either fix that specific term or the reach stops there.

The governing principle — the weakest gate wins

A design has several independent gates: loss, OSNR, chromatic dispersion, PMD, nonlinear launch window, capacity/spectrum, and protection. Each has its own margin. The link closes only if every gate passes; the reach is set by the first gate that fails. Adding margin to a gate that already passes buys you nothing — you must find and address the limiting gate. Most of link design is bookkeeping to find which gate is tightest.

The design flow — seven gates in order

Run these in order. Each step names the page on this site where you learned to evaluate that gate — this page just sequences them.

THE DESIGN-CLOSURE FLOW 1. REQUIREMENTS ──► rate, distance, sites, protection, growth │ 2. FIBER / OPTICS ─► fiber type, channel/modulation, transponder class │ 3. LOSS BUDGET ────► Σ loss ≤ system budget, with margin [Testing & Loss Budget] │ [Link Engineering] 4. OSNR BUDGET ────► actual OSNR ≥ required + margin over spans [Link Engineering] │ 5. CD & PMD CHECK ─► residual dispersion & PMD within tolerance [Dispersion] │ 6. NONLINEAR / Pin ► launch power inside the bathtub window [Nonlinear Effects] │ 7. CAPACITY PLAN ──► channels/spectrum fit, with headroom [Capacity & Design] │ [ROADM & Flexgrid] 8. PROTECTION ─────► diversity + protection scheme meets SLA [Protection & OTN] │ ═══► MARGINS & CLOSURE: design margin + aging on every gate. Any gate FAIL → that gate is the limiter. All PASS → design closes.

The steps are numbered 1–8 but collapse to "requirements, choose components, then six budget gates." The order matters because early choices (fiber type, modulation) change every later gate — that is why fiber/optics selection sits before the budgets.

A fully worked multi-span example

Let's close a design end to end. Every number here is illustrative — chosen to teach the arithmetic, not lifted from an Ekinops spec. A real design uses the vendor planning tool with exact component data.

Requirements (step 1) and components (step 2)

Two sites, 4 amplified spans of 80 km each → ~320 km end to end.
Service: one 100G coherent wavelength, SD-FEC transponder.
Per-span loss: 22 dB. System loss budget (per span, amp-to-amp): 25 dB.
Amplifier NF: 5 dB. Planned per-channel launch power: +1 dBm.
Required OSNR (100G DP-QPSK, SD-FEC): 14 dB; reserve 4 dB margin → floor 18 dB.
Fiber CD coefficient ~17 ps/(nm·km); transponder tolerates a large residual because coherent DSP compensates CD electronically.
Growth target: room to reach 40 channels later.

GATE 3 — LOSS BUDGET (per span) span loss 22 dB ≤ 25 dB system budget → 3 dB margin → PASS (both directions checked; connectors/splices already inside the 22 dB) GATE 4 — OSNR BUDGET (from Link Engineering rule of thumb) single-span OSNR = 58 + Pin − NF − span_loss = 58 + 1 − 5 − 22 = 32 dB 4 identical spans: 32 − 10·log10(4) = 32 − 6 = 26 dB 26 dB actual ≥ 18 dB floor → 8 dB margin → PASS GATE 5 — CHROMATIC DISPERSION residual CD ≈ 17 ps/(nm·km) × 320 km ≈ 5440 ps/nm accumulated coherent DSP compensates this electronically, within tolerance → PASS PMD: low on modern fiber over 320 km, within tolerance → PASS GATE 6 — NONLINEAR / LAUNCH POWER +1 dBm per channel sits inside the planned bathtub window for this fiber and channel count → PASS (re-check at 40 ch — see below) GATE 7 — CAPACITY / SPECTRUM 1 channel today; plan reserves spectrum + amp gain for 40 → PASS GATE 8 — PROTECTION required scheme + route diversity provisioned → PASS ALL GATES PASS → DESIGN CLOSES with OSNR as the tightest budget (8 dB margin) and loss next (3 dB). (All values illustrative.)

Reading the result

Every gate passed, so the link closes. The tightest gate here is loss (3 dB margin), with OSNR comfortable at 8 dB — meaning if this route grew longer, loss per span would bite before OSNR. That tells you exactly where to spend effort if you later need more reach: reduce span loss (better splices, shorter spans) before touching anything else. Finding the tightest gate is the whole point of the exercise.

Now stress it — the growth case

The requirement said "room for 40 channels." Re-run the two gates that depend on channel count. OSNR per channel is roughly unchanged if per-channel launch power is held — but the nonlinear gate is not: XPM and FWM scale with channel count, so the launch window at 40 channels is lower than at 1 channel.

The gate that moves when you grow

A design that closes at 1 channel can fail the nonlinear gate at 40 channels if you keep the same hot per-channel launch power. Re-validate the launch plan for the fully loaded system, not just the channels lit on day one. This is why capacity headroom (Capacity & Design) and launch-power planning (Nonlinear Effects) are part of the same design, and why "just light another wavelength" is a design event, not a patch.

The design-closure scorecard

Capture every design as a scorecard. A design is "closed" only when every row is PASS with real margin. Keep the scorecard with the circuit record so the next engineer sees which gate was tightest.

#	Gate	Question	Where you learned it	Verdict
3	Loss budget	Σ loss ≤ system budget, both directions, with margin?	Testing & Loss Budget	PASS / FAIL
4	OSNR budget	Actual OSNR ≥ required + design + aging, over all spans?	Link Engineering	PASS / FAIL
5	CD & PMD	Residual dispersion and PMD within transponder tolerance?	Dispersion	PASS / FAIL
6	Nonlinear / Pin	Per-channel launch power inside the bathtub, at full channel load?	Nonlinear Effects	PASS / FAIL
7	Capacity	Channel plan / spectrum fits, with growth headroom?	Capacity & Design	PASS / FAIL
8	Protection	Route diversity + protection scheme meets the SLA?	Protection & OTN	PASS / FAIL
—	Closure	Every gate PASS with margin, and the tightest gate identified?	this page	CLOSED / NOT

Try it — the budget-closer simulator

The interactive tool below lets you adjust the design inputs — span count, span loss, launch power, noise figure, required OSNR and margins — and watch which gates pass and which gate becomes the limiter. Use it to build intuition for how one lever moves several gates at once. (All values it uses are illustrative teaching approximations, consistent with the rules of thumb on the linked pages — always confirm a live design against the Ekinops planning tool / BOM.)

1. A design passes loss with 6 dB margin and OSNR with 1 dB margin. Someone proposes improving the splices to cut loss further. Does that extend reach?

Barely, if at all — OSNR is the tightest gate (1 dB), not loss (6 dB). The reach is set by the OSNR gate, so effort should go there (lower NF, Raman, stronger FEC, fewer/lower-loss spans to help OSNR, or a 3R regen). Improving a gate that already passes comfortably does not move the limiter.

2. Given Pin = +1 dBm, NF = 5 dB, span loss = 22 dB, and a required-OSNR floor of 18 dB, does a 4-span link close on OSNR, and with how much margin?

Single-span OSNR = 58 + 1 − 5 − 22 = 32 dB. Four spans: 32 − 10·log10(4) = 32 − 6 = 26 dB. 26 ≥ 18, so it PASSES with 8 dB margin. (Illustrative rule of thumb.)

3. A link closes with 1 channel lit. Why must you re-run the nonlinear gate before growing to 40 channels?

XPM and FWM scale with channel count and density, so the safe per-channel launch window is lower at 40 channels than at 1. Keeping the same hot launch power can push the loaded system past the bathtub optimum and fail the nonlinear gate. Capacity growth requires re-validating the launch plan for the fully loaded system.

4. In the closure flow, why does fiber/optics selection come BEFORE the six budget gates?

Because those choices set the terms of every gate: fiber type sets loss and dispersion and FWM behavior; modulation/FEC sets required OSNR and dispersion tolerance; channel plan sets the nonlinear window and spectrum. Running budgets before choosing components would mean redoing them for every choice.

5. What does it mean for a design to be "closed," and what single fact should the scorecard always record?

Closed means every gate — loss, OSNR, CD/PMD, nonlinear, capacity, protection — passes with real margin (design + aging). The scorecard should always record the tightest gate, because that gate sets the reach and tells the next engineer where the link will fail first and where to spend effort to extend it.