Optical Link Budget Explained: Will Enough Light Survive the Trip?

The short answer

What is an optical link budget, in simple terms? It is a check that enough light survives the trip. You take the weakest light the transmitter is guaranteed to send, subtract everything the fibre path will dim along the way — the fibre itself, every connector, every splice — and check that what is left is still brighter than the faintest light the receiver can read, with something spare for old age. Three subtractions, done on paper, before anyone climbs a ladder. If you remember one sentence, make it that one: a link budget is a wallet, not a wish — the path spends light, and the receiver sets the minimum balance.

Think of it as money, because the arithmetic behaves exactly like money. The optic pair hands you a fixed wallet: the difference between what the transmitter launches and what the receiver needs. Every element on the path is a bill. The fibre charges by the kilometre. Each connector pair charges a flat fee. Each splice charges a small one. At the end, whatever is left in the wallet is your margin — your savings against the day a connector gets dirty, a laser ages, or a repair adds a splice you did not plan. Spend more than the wallet holds and no amount of configuration will bring the link up, because the problem is physics, not settings.

dB and dBm, without the fear

These two units scare people off fibre maths, and they are genuinely simple. dBm is an amount: it says how much light, full stop. The scale is logarithmic, which sounds alarming but only means the steps are multiplicative — go down 10 dBm and you have a tenth of the light; go down 3 dBm and you have roughly half. A transmitter launching -4.3 dBm is shining about 0.37 milliwatts. A receiver with -10.6 dBm sensitivity can still read about 0.087 milliwatts. Small numbers, but light is cheap to make and receivers are good at squinting.

dB, without the m, is a difference — and differences are what make budgets easy, because differences in dB add and subtract like ordinary numbers even though the underlying light is multiplying and dividing. Fibre that loses 2 dB leaves about 63 percent of the light it was given; the point is not the percentage, it is that a 2 dB fibre loss followed by a 1 dB connector loss is a 3 dB total loss, every time. No calculator gymnastics, no exponents at the workbench. That is the entire trick of this article: convert everything to dB, add the losses, compare with the budget.

The three lines of every link budget

Line one — the budget. Power budget = minimum launch power minus receiver sensitivity. Both numbers come from the optic's datasheet, and you use the minimum launch, not the headline or typical figure, because the datasheet only promises the minimum. If one particular optic ships brighter, that is a bonus, not a plan.

Line two — the loss. Total channel loss = (attenuation in dB per km, times the distance) plus (connector pairs, times about 0.5 dB each) plus (splices, times about 0.1 dB each). Those per-item figures are planning values — the ones this site's calculator uses — and a real design checks them against the fibre and component specifications for the actual build.

Line three — the margin. Margin = budget minus loss. Also useful: the light that actually arrives is launch minus loss, and the maximum reach at a chosen safety target is (budget minus connector loss minus splice loss minus target margin), divided by the attenuation per km. Every worked number below is one of these three lines with real values in it.

One link, worked end to end

The job: join two buildings 8 kilometres apart with a 100-gigabit link, using 100G-LR4 optics — the QSFP28 long-reach class built for exactly this kind of run. The calculator on this site carries a preset for this optic pair: minimum launch power -4.3 dBm, receiver sensitivity -10.6 dBm. The fibre is OS2 single-mode, run at 1550 nm, where the calculator's attenuation figure is 0.25 dB per kilometre. The light crosses two mated connector pairs (one patch panel at each end) and two fusion splices along the route.

Line one: budget = -4.3 minus -10.6 = 6.30 dB. That is the whole wallet.

Line two: fibre loss = 8 km times 0.25 = 2.00 dB. Connector loss = 2 pairs times 0.5 = 1.00 dB. Splice loss = 2 times 0.1 = 0.20 dB. Total channel loss = 3.20 dB.

Line three: margin = 6.30 minus 3.20 = 3.10 dB. The light arrives at -4.3 minus 3.20 = -7.5 dBm — comfortably brighter than the -10.6 dBm the receiver needs, with 3.10 dB left in the wallet. Turned around, the maximum reach at a 3 dB safety target works out to (6.30 - 1.00 - 0.20 - 3.00) divided by 0.25 = 8.4 km, so this 8 km design sits just inside its own limit. Verdict: it passes — but notice how close the honest numbers land. There is no room in this design for sloppy extras, which is exactly what the next section is about.

Read the staircase left to right and you have understood every link budget ever written: a bright start, a step down for each thing the light passes through, a floor set by the receiver, and a gap — the margin — that decides whether the design is robust or lucky.

Margin is what the real world eats

Why insist on spare decibels at all? Because the budget you calculate describes the link on its best day: fresh connectors, a young laser, the fibre exactly as installed. Real links age. Lasers dim slowly over years. A fibre cut gets repaired with a new splice that was never in the design. A connector gets reseated by someone in a hurry and picks up a fingerprint. Each event spends margin. The common planning rule — and the default target in this site's calculator — is to keep about 3 dB spare on any run you cannot easily revisit. In the worked example, that rule is doing real work: add just two more patch-panel pairs (1.00 dB) and one repair splice (0.10 dB) and the loss grows to 4.30 dB, the margin falls to 2.00 dB, and the same link drops from pass to marginal. Stretch the route to 20 km instead and the fibre alone spends 5.00 dB, the total reaches 6.20 dB, and the margin is 0.10 dB — a link that lights on day one and becomes a mystery flapping ticket by winter.

What a power budget does not tell you

A passing budget proves the light arrives bright enough. It does not prove the signal arrives clean enough, and three neighbours of this calculation deserve an honest mention. On multimode fibre, reach is usually limited by modal bandwidth — the light smearing across modes — not by loss, so a multimode link can pass its power budget and still fail at speed. On long amplified DWDM spans, the binding budgets are different ones again: signal-to-noise ratio (OSNR), chromatic dispersion and polarisation effects, which need their own planner. And a budget is a prediction, not a measurement: an optical power meter tells you the actual loss end to end once the link exists, and an OTDR trace shows you where along the route the loss happens, event by event. Calculate first, measure after, and investigate any gap between the two — that gap is usually a dirty or damaged connector introducing itself.

One more honest corner: very short links have the opposite problem. A powerful long-reach optic across a short patch lead can arrive hotter than a receiver's maximum input allows. Datasheets list that maximum too. It rarely binds in practice, but a budget habit that only ever checks the floor will eventually meet a ceiling.

The mistakes that eat budgets

1. Budgeting with the typical launch power. The typical figure is what an average optic ships at; the minimum is what the vendor promises. Design to the promise.

2. Forgetting the patch panels. Engineers count the two transceiver faces and stop. The light also crosses every patch-panel pair at both ends — count every mated pair on the path, because at about 0.5 dB each, four pairs spend 2 dB before the fibre spends anything.

3. Counting connectors one at a time. The loss happens per mated pair, not per connector. A pair is one join; price joins.

4. Designing to zero margin. A budget that balances exactly has already failed — it just has not met its first repair splice yet.

5. Using one wavelength's attenuation for another. The same OS2 fibre costs more light per kilometre at 1310 nm than at 1550 nm in the calculator's table (0.40 versus 0.25 dB per km). The wavelength is part of the budget, not a footnote.

Try it yourself

Run the worked example with your own hands in the free optical link budget calculator: press the 100G-LR4 preset, choose OS2 single-mode at 1550 nm, set the distance to 8 km, connector pairs to 2 and splices to 2. You should see exactly the article's numbers — total channel loss 3.20 dB, power budget 6.30 dB, link margin 3.10 dB, max reach 8.40 km. Then break it the way the field would: push the distance to 20 km, or add two more connector pairs, and watch the verdict fall from pass to marginal. When your link stops being a single span and becomes a chain of amplified DWDM spans, the budget grows extra dimensions — noise, dispersion, polarisation — and the DWDM / 400ZR reach planner takes over from there, while the pluggable that carries those spans is explained in 400ZR Explained.