Propagation Delay Calculator — Network Latency From Distance

Background.

Propagation delay is the latency a network path adds simply because signals take time to cross distance, and it is the one component of latency that no amount of money can remove. Buying a faster link does not help: propagation delay is completely independent of data rate. A 10 Gbps fibre and a 10 Mbps fibre following the same route add exactly the same delay, because both carry signals at the same speed through the same glass. The only ways to reduce it are to shorten the route or to change the medium.

This calculator uses the propagation figures ITU-T publishes rather than guessed constants. Recommendation G.114, "One-way transmission time", contains a table of planning values for the delay of transmission elements, and it gives 5 µs per km for an optical fibre cable system with digital transmission, 4 µs per km for terrestrial coaxial cable or radio-relay, and 6 µs per km for a submarine coaxial cable system, the last of which explicitly allows for delay in repeaters and regenerators. Those three values are what the medium selector applies. The vacuum option uses the speed of light itself, 299,792,458 metres per second, which NIST marks as exact because the metre is defined from it.

One honest caveat about the fibre number. Five microseconds per kilometre corresponds to a velocity factor of 0.667 — signals travelling at two-thirds of light speed — which is slightly slower than the group index of a bare G.652 fibre would predict. That is deliberate on the ITU's part: it is a planning value, absorbing cable slack, splice trays and the small excess length coiled into every joint. G.114 itself flags the value as provisional and under study. If you want the propagation delay of a specific fibre from its published group index, choose the custom medium and enter one divided by that index.

The two thresholds the verdict is judged against also come from G.114. Section 4 recommends that a one-way delay of 400 ms should not be exceeded for general network planning, and the same section says that if delays can be kept below 150 ms, most applications — speech and non-speech alike — will experience essentially transparent interactivity. The default of 5,000 km is the ITU's own example: G.114 notes that for intercontinental routes of about 5,000 km or less, VoIP users are likely to experience mouth-to-ear delays under 150 ms. This calculator returns 25 ms one-way for that path, which is the propagation share of that budget; the rest goes to codecs, packetisation, jitter buffers and queueing, none of which this page models.

Use route length, not straight-line distance. Fibre follows roads, railways, rights of way and seabed contours, and a real route is commonly 20% to 50% longer than the great-circle distance between its endpoints. If you are estimating from a map, that discrepancy will dominate every other source of error on this page. And remember that propagation is only one term in end-to-end latency — serialisation, queueing, switching and processing all add on top, and on a congested path queueing frequently exceeds propagation by a wide margin.

What is propagation delay calculator?

Propagation delay is the time a signal takes to travel from one end of a link to the other, equal to path length divided by propagation velocity. It is one of the four classical components of network latency, alongside serialisation delay (the time to clock a packet's bits onto the wire, which does depend on data rate), queueing delay (time spent waiting in buffers) and processing delay (time spent in switching and routing logic). Propagation delay is distinguished from the other three by being fixed for a given route: it cannot be reduced by upgrading capacity, buying better hardware or reducing load. Signals in a transmission medium travel more slowly than light in vacuum, by a ratio called the velocity factor — the reciprocal of the medium's refractive or group index. For optical fibre, ITU-T G.114's network-planning value of 5 microseconds per kilometre corresponds to a velocity factor of about 0.667, so light in glass covers roughly two-thirds the distance per second that it would in vacuum. The practical consequence is a rule of thumb worth memorising: fibre adds about 5 milliseconds of one-way delay, or 10 milliseconds of round trip, for every 1,000 kilometres of cable route. Round-trip time is exactly twice the one-way figure for a symmetric path, and it is round-trip time that governs how a windowed protocol such as TCP behaves — which is why this page and the bandwidth-delay product calculator are used together.

How to use this calculator.

1. 01Enter the length of the cable route, not the straight-line distance between the endpoints. Real routes are typically 20–50% longer than the great-circle distance, and that difference will swamp everything else on this page.
2. 02Choose the unit. Miles and feet convert with NIST's exact factors, so no precision is lost.
3. 03Pick the medium. The fibre, coax and submarine-coax options apply ITU-T G.114's published planning values of 5, 4 and 6 µs per km respectively.
4. 04Use the vacuum option for free-space radio, microwave and satellite paths — air's refractive index is close enough to 1 that the difference is negligible at these scales.
5. 05Choose Custom and enter a velocity factor for a specific cable. For a fibre with a published group index n, the velocity factor is 1 ÷ n. Values above 1 are rejected, because nothing propagates faster than light.
6. 06Read the round-trip figure if you are feeding a bandwidth-delay product or a TCP window calculation; read the one-way figure if you are budgeting interactive voice or video against G.114's 150 ms and 400 ms thresholds.

The formula.

The core is one division: delay equals distance divided by speed. Everything else is bookkeeping about which speed and which units.

Distance normalises to kilometres first, using NIST's exact factors — a mile is 1,609.344 metres and a foot is 0.3048 metres, both defined exactly rather than measured, so no precision is lost in conversion.

Speed comes from the medium. For the three cable options this calculator stores ITU-T G.114's published microseconds-per-kilometre figures directly rather than deriving them from a velocity, because those integers — 5, 4 and 6 — are the quantities the Recommendation actually publishes, and round-tripping them through a velocity and back would introduce rounding that is not in the source. The velocity shown alongside is derived from them: one thousand million metres per second divided by the microseconds per kilometre. For vacuum the direction reverses, since there the exact quantity is the speed of light itself and the microseconds per kilometre is the derived figure, 10⁹ ÷ 299,792,458 = 3.3356409520 µs/km.

Multiply normalised distance by microseconds per kilometre and you have the one-way delay in microseconds; divide by a thousand for milliseconds; double it for the round trip. The round trip is exactly twice the one-way figure on a symmetric path, and asymmetric routing — where traffic returns by a different and often longer path than it took outbound — is one of the standard reasons a measured RTT exceeds twice the computed one-way delay.

The delay-per-1000-km output looks like a coincidence and is not. A thousand kilometres at x microseconds per kilometre is 1000x microseconds, which is x milliseconds. So the two figures share a numeral and differ only in unit, which is the arithmetic behind the fibre rule of thumb: 5 µs/km means 5 ms per 1,000 km, and therefore 10 ms of round trip per 1,000 km of route.

What this does not model. Serialisation delay, which is packet size divided by link rate and is the only latency component that does depend on bandwidth. Queueing delay, which under congestion routinely exceeds propagation delay and is the reason a link that looks fine on paper feels slow. Processing and switching delay at every intermediate hop. Optical-electrical conversion, regeneration and amplification along the route — G.114's 6 µs/km submarine figure explicitly folds repeater delay into the per-kilometre number, which is why it exceeds the terrestrial coax value. And for satellite paths, the delay is dominated by the altitude of the spacecraft rather than by ground distance; G.114 gives 260 ms for a geostationary hop at 36,000 km, propagation through space only, between earth stations.

A worked example.

Frequently asked questions.

References& sources.

1. [1]ITU-T Recommendation G.114 (05/2003), 'One-way transmission time', Table I.1 'Planning values for the delay of transmission elements'. Source of all three cable constants used by the medium selector: 'Terrestrial coaxial cable or radio-relay system: FDM and digital transmission — 4 µs/km'; 'Optical fibre cable system, digital transmission — 5 µs/km (Note 1)', where 'NOTE 1 – This value is provisional and is under study'; 'Submarine coaxial cable system — 6 µs/km. Allows for delay in repeaters and regenerators'; and the satellite entries '400 km altitude 12 ms, 14 000 km altitude 110 ms, 36 000 km altitude 260 ms — propagation through space only (between earth stations)'. Section 4 supplies the two thresholds the verdict is judged against: 'it is recommended to not exceed a one-way delay of 400 ms for general network planning', and that below 150 ms 'most applications, both speech and non-speech, will experience essentially transparent interactivity'.
2. [2]NIST, CODATA value for the speed of light in vacuum: 299 792 458 m s⁻¹, with standard uncertainty listed as '(exact)' because the metre is defined from it. Used by the vacuum medium and by the velocity-factor output.
3. [3]NIST Special Publication 811, Appendix B.9, 'Factors for units listed alphabetically'. Source of the exact length conversions used by the distance selector: mile (mi) to metre, 1.609 344 E+03, and foot (ft) to metre, 3.048 E−01, both printed in boldface, which the guide defines as indicating an exact factor.
4. [4]RFC 6349, 'Framework for TCP Throughput Testing' (IETF, 2011). Why round-trip propagation matters to throughput: it defines the bandwidth-delay product as 'the product of a data link's capacity (in bits per second) and its end-to-end delay (in seconds)' and works an example at a 25 ms round-trip time.
5. [5]RFC 5136, 'Defining Network Capacity' (IETF, 2008). Establishes the separation this page rests on — capacity is measured in bits per second over an interval, delay in seconds, and neither determines the other — and warns that 'literature often overloads the term bandwidth to refer to what we have described as capacity in this document'.