Enter for the full results page · Esc to close
Light in glass covers about 204,000 km per second — roughly 4.9 microseconds per kilometre. For trading, storage replication and anything with a distance limit written into the protocol, that number decides the design.
Where the number comes from. The speed of light in vacuum is 299,792 km/s. Divide by the fibre's group index — about 1.4682 for standard single-mode — and you get roughly 204,190 km/s in glass, or 4.897 microseconds per kilometre. That is the floor; nothing you buy makes light go faster.
Route length, not map distance. Fibre follows ducts, railway lines and road verges, and it has slack coils at every joint. A route that measures 100 km on a map is commonly 120–140 km of installed fibre. If latency matters to you, get the actual OTDR length from the carrier rather than estimating.
Active delay adds up. Every switch, transponder, amplifier and regenerator adds its own delay. Cut-through switching is under a microsecond; store-and-forward, FEC and OTN mapping can each add several. On a long route with several amplifier huts, the active delay can rival a good fraction of the fibre delay.
About 490 microseconds one way, or 980 microseconds round trip, in standard single-mode. That is why synchronous storage replication is usually capped around 100 km — beyond that the round trip starts to dominate application response time.
Because light travels through air, not glass. The effective index is close to 1.0, so the delay drops from about 4.9 µs/km to nearer 3.4 µs/km — roughly 30% faster. For latency-sensitive trading routes that difference is worth a great deal of money; for everything else it is not yet worth the cost.
Propagation delay, no — light travels at the same speed whether you are sending 1G or 800G. Serialisation delay does change: pushing a 1500-byte frame onto a 1G link takes 12 µs, onto a 100G link 0.12 µs. Over a long span propagation dominates; inside a rack, serialisation and switching dominate.
How long a given volume of data takes across a link, with protocol overhead and efficiency factored in
Open tool →Total channel loss, power budget and pass/fail margin for single-mode and multimode links
Open tool →Pulse spread from Δt = D · L · Δλ, and the distance at which dispersion — not loss — ends the link
Open tool →Transceivers, MPO assemblies, WDM and patch cords — built and tested to the figures you just calculated.