Short links are connector links
The intuition most people carry over from copper is that distance is the enemy, and on fibre inside a building that intuition is simply wrong. Work through the default numbers: 900 feet of fibre at 0.4 dB per kilometre is about 0.11 dB. Four connector pairs at 0.5 dB each is 2.0 dB. The glass is spending under five percent of the total, and about one eighteenth of what the connectors are spending.
The practical consequence is worth having in front of you when somebody proposes a change. Adding an intermediate patch panel to that link adds two mated pairs — one dB, at those figures — and one decibel of fibre at that attenuation is about two and a half kilometres of cable. The panel costs the optical equivalent of a run across a small town. Nobody would accept that framing if it were presented as extra cable, and it is exactly what a panel does.
It also tells you where to buy margin when a link is tight. Not by rerouting to shorten the cable, which achieves almost nothing on a short link, but by taking mated pairs out of the path.
Counting connector pairs is the usual mistake
Every mated pair in the path counts, and a pair is two ends coming together. A transceiver, a patch cord to a panel, a horizontal run across a building, a panel at the far end, a patch cord to the far transceiver: that is a pair where each patch cord meets each panel, and a pair where each panel meets the horizontal cable — four pairs, not two, and not one per panel. Counting panels rather than mated pairs is the error, it is always an undercount, and it always makes the calculated budget look better than the installed link tests.
Every figure on this page is yours
| Figure | Where it comes from |
|---|---|
| Attenuation in dB per kilometre | The cable data sheet, for that fibre at the wavelength you are operating |
| Loss per connector pair | The connector manufacturer, or better, your own test results on installed work |
| Loss per splice | The splice method and the person doing it — fusion and mechanical are not comparable |
| Transmitter output and receiver sensitivity | The transceiver data sheet, minimum output rather than typical |
| Receiver overload point | The transceiver data sheet, and it matters on short links |
| Margin to hold back | Your own judgement about ageing, repairs and future changes |
The defaults in those fields exist so the page produces output when it loads, and each one says so in its hint. They are not typical values, not recommendations and not standard figures, and none of them should survive contact with your actual data sheets. This page does not know which fibre type you have, which standard you are working to, or what your optics are.
A link can also have too much power
Loss budgets are usually discussed as a floor: enough power has to arrive. There is a ceiling too. A long-reach transceiver designed to push signal tens of kilometres, connected across a rack with a two metre jumper, can deliver more power than the receiver can handle. The result is not a clean failure — it is errors that look like a marginal link, which sends people looking for loss when the problem is the opposite. The overload field is optional here for that reason; it is worth filling in whenever the optics are longer-reach than the link.
Distance and dispersion limits are a separate matter again. Optics have a specified reach for a given data rate, and exceeding it can produce errors on a link that has ample power arriving. Loss arithmetic will not warn you about that, and this page makes no attempt to.
The number that actually settles it
Everything here is a design estimate made before anything is installed. What determines whether a link works is an optical loss test on the finished link, measured end to end at the wavelength in use. If the tested figure and the calculated figure disagree, the tested one is right. The usual reasons for a gap are a connector count that was too low, connectors that need cleaning, or splices that did not go as well as assumed.
Related pages
For the copper equivalent, the ethernet drop calculator works out cable, jacks, panels and whether the longest channel is inside a length limit you enter, and the PoE power budget calculator covers power sent down that cable and lost in it. Rack space for the panels either end is on the structured wiring panel planner. If the fibre is being pulled into a raceway, what that pull costs in tension is on the cable pull tension calculator — and fibre has its own maximum tension and minimum bend radius from the cable manufacturer, which that page will restate and will not judge.
The parts this page will not touch
Bare fibre is glass at a hundred and twenty five microns. The offcuts from cleaving are almost invisible, they go through skin readily, and they do not show up on an x-ray. They belong in a marked container from the moment they exist, and food and drink do not belong on the same bench. Separately, an active fibre end is an optical source you cannot see: nothing about the end of a lit fibre looks dangerous, which is the reason eye injuries from them happen at all. Do not look into a fibre end or a port. That is the hazard named; how any of it is handled is training and manufacturer instruction, not calculator content.
This page does not supply limits and cannot judge yours. Fill percentages, support and strap spacing, minimum bend radii, tray loading rules, cover depths and every other installation limit come from the code your jurisdiction has adopted and from the manufacturer of the product you actually bought, and the two do not always agree. The arithmetic here runs on figures you type in. Whether the result is acceptable is a conversation with a licensed electrician and with the inspector who signs the job off.
Questions people ask
How many connector pairs should I count on a link?
Every mated pair in the path, which is more than most people count. A typical run — transceiver, patch cord, panel, horizontal cable, panel, patch cord, transceiver — has four mated pairs: each patch cord meets a panel, and each panel meets the horizontal cable. Counting one per panel gives two and understates the loss by half. This is the single most common reason a calculated budget looks better than the installed link tests.
Why does the fibre length barely matter on my link?
Because inside a building it genuinely does not. At the figures in the default fields, 900 feet of fibre contributes about 0.11 dB while four connector pairs contribute 2.0 dB. One decibel of fibre at that attenuation is around two and a half kilometres. On a short link the budget is spent almost entirely at the joins, which is why adding a patch panel costs more optically than adding a great deal of cable.
What attenuation figure should I use?
The one on the data sheet for the fibre you have, at the wavelength you are operating. It varies substantially between fibre types and between wavelengths on the same fibre, so a general number would be misleading in either direction. The default in the field is a placeholder that exists so the page produces output, and the hint on the field says exactly that.
Does the calculator tell me whether my link will work?
No. It reports the computed loss beside the budget your own transceiver figures imply, and restates the difference without judging it. Whether a link works depends on things this page has no access to: the actual installed loss measured on the finished link, the distance and dispersion limits of your optics at your data rate, and how clean the connectors are on the day. The optical loss test on the installed link is the number that settles it, and where it disagrees with this arithmetic, it wins.
Can a fibre link have too much light?
Yes, and it is a real failure mode that gets misdiagnosed as a marginal link. Every receiver has an overload point as well as a sensitivity, and a long-reach transceiver connected across a short link can deliver more power than the receiver handles. The symptom is errors, which sends people hunting for loss when the problem is the reverse. Enter the overload figure from your transceiver data sheet and the page reports the arriving power beside it.