Quick answer: Multimode distance is the reach at which modal dispersion closes the eye at the receiver, so it drops sharply as speed rises. Typical published limits are 550 m at 1G on OM3 and OM4, 300 m and 400 m at 10G, and 100 m or 150 m at 40G and 100G with parallel optics. OM1 and OM2 stop being useful at 10G.
Multimode Fiber Distance: OM1 to OM5 Limits by Speed
A network designer inherits a 10G backbone built on OM2 and asks whether the new 300 m run to the warehouse will work. It will not, and the reason has nothing to do with the connectors, the patch cords or the switch port. Multimode reach is set by how much the launch light spreads out in time before it reaches the receiver, which depends on the core profile and the light source. Two cables printed 50/125 can have very different reach at the same speed. Reading a datasheet means reading the bandwidth column, not the core size.
Why the Limit Comes From the Fiber, Not the Connector
A connector adds a fixed loss of a few tenths of a dB and, if it is dirty or badly polished, a reflection. Those matter for link budget but they do not set the published distance. The distance limit is where the pulse has spread so much that adjacent symbols overlap at the receiver, which is a property of the fiber and the transmitter. That is why the same OM4 cable reaches 400 m at 10G and 150 m at 100G, and why changing patch cords moves neither number. A clean, well terminated link on marginal fiber still fails at its published distance. That is also why a link can pass a certification test on a short reel and fail in the building, where the route adds panels, splices and temperature swings.
Modal Dispersion and Effective Modal Bandwidth
Light launched into a multimode core travels in several guided paths, called modes, and each path has a slightly different physical length. A short pulse therefore arrives spread out in time, and the spread grows with distance until symbols overlap. The fiber is rated by effective modal bandwidth, measured with a controlled launch that mimics a real VCSEL. Higher bandwidth means less spread, which means a longer reach at the same signaling rate. That number is the only one in the datasheet that predicts distance. Core diameter and jacket color tell you the class, not the bandwidth. One consequence is that two fibers with the same core size and the same jacket color can differ by a factor of two in reach at 10G.
VCSEL Launch and Differential Mode Delay
An 850 nm VCSEL does not fill the whole core evenly. It launches a patch of light, and if that patch lands where the core profile is imperfect, a few modes arrive late and dominate the spread. That effect is differential mode delay, and it is why a fiber tested with an overfilled launch can look better than it behaves with a real transceiver. OM3 and OM4 are specified with a restricted launch for exactly this reason, and the test method is part of the standard.
Reading the Bandwidth Column
Datasheets list bandwidth in MHz*km at 850 nm and 1300 nm, plus a laser launch bandwidth for laser optimized grades. For 10G and above, use the laser launch figure. A grade rated at 500 MHz*km under overfilled launch may carry a much lower laser launch number, and the laser number is what the standard uses to set reach. If only one figure appears, ask which launch condition it was measured under before you trust the reach.
Published Reach by Speed and Class
The figures in the table are the ones in the standards, and they assume a clean link with a small number of mated connectors. Real projects should plan inside them rather than at the edge. OM1 and OM2 were designed for LED sources at 1G and run out of bandwidth quickly at 10G. OM3 is the practical floor for 10G in a building backbone. OM4 adds margin at 10G and is the common choice for 40G and 100G inside a data hall. OM5 extends the same fiber across a wider wavelength range for short wave division multiplexing. Design to the standard reach, then confirm the loss budget with the optics you intend to order, because vendor figures sometimes differ from the table.
1G and 10G Reach
At 1G, 1000BASE-SX reaches 550 m on any 50/125 grade and about 275 m on 62.5/125, so class barely matters. At 10G the picture changes: 10GBASE-SR reaches 300 m on OM3, 400 m on OM4 and 82 m on OM2. That single step is why most buildings installed after 2010 use OM3 or OM4 even where 1G was the only requirement at the time.
40G and 100G Parallel Optics
40GBASE-SR4 and 100GBASE-SR4 do not send one lane down one fiber. They send four lanes in parallel over an MPO connector, so reach is set by the worst lane and by the MPO loss. Published reach is 100 m on OM3 and 150 m on OM4 for both 40G and 100G. A twelve fiber MPO trunk carries four 10G or 25G lanes, which is where the port count and the connector loss budget come from.
OM5 and Wideband Multiplexing
OM5 is a 50/125 laser optimized fiber with specified bandwidth across 850 nm to 953 nm rather than only at 850 nm. That range allows short wave division multiplexing, which puts four wavelengths on one pair and reaches 100G over a duplex instead of an MPO trunk. Reach is typically 150 m, and the gain is cable density rather than raw distance. The fiber is backward compatible, but the optics must support it.
When Single Mode Becomes the Cheaper Answer
Past roughly 500 m the installed cost comparison usually flips. Multimode optics are cheaper per port, but the saving is tens of dollars per port while fiber, pathway and labor dominate the project. Single mode reach is limited by the optics you buy rather than the glass, so a 2 km run needs no special cable. Once a route needs 100G beyond 150 m, or needs headroom for 400G later, single mode with duplex or parallel single mode optics is usually the lower total cost. Price both before you commit a backbone to multimode. The optics price gap narrows every year, so a backbone expected to serve 15 years should not be locked to a class that tops out at 150 m.
Typical published multimode reach by class and speed
| OM class | Core/cladding | 1G | 10G | 40G | 100G |
|---|---|---|---|---|---|
| OM1 | 62.5/125 | 275 m | 33 m | Not rated | Not rated |
| OM2 | 50/125 | 550 m | 82 m | Not rated | Not rated |
| OM3 | 50/125 laser | 550 m | 300 m | 100 m | 100 m |
| OM4 | 50/125 laser | 550 m | 400 m | 150 m | 150 m |
| OM5 | 50/125 wideband | 550 m | 400 m | 150 m | 150 m |
Real-World Application Scenarios
A hospital campus links three buildings at 10G over 250 m, and OM4 is specified because it sits inside the 400 m rating with margin for two extra patch panels. A factory floor runs 40G between a cabinet and a line controller at 80 m, where an OM4 MPO trunk is cheaper than single mode optics at both ends. A data hall with 100G spine links at 120 m uses OM4 MPO trunks for the same reason. In each case the class came from the speed and the distance, not from what was already on the shelf.
Related Guides on This Site
Class choice drives reach more than any other single decision, so start with the OM1 OM2 OM3 OM4 comparison guide before you price a run. If the choice is between the two laser optimized grades, the difference between OM3 and OM4 guide shows where the extra 100 m at 10G comes from. Wideband fiber is covered in the OM5 wideband guide, and parallel optics in the 100G OM4 data center guide.
Frequently Asked Questions (FAQ)
Q: What is the maximum distance for multimode fiber?
A: It depends on speed, not on a single number. At 1G, 550 m is typical on 50 micron fiber. At 10G it is 400 m on OM4, and at 40G and 100G it is 150 m on OM4 with parallel optics. The same cable carries different distances at different signaling rates, which is why speed has to be part of the question.
Q: Can I run 10G over OM2 fiber?
A: Yes, but only to about 82 m, which is short for a building backbone. OM2 was designed for LED launch at 1G and its effective modal bandwidth is low. Running 10G over an existing OM2 link often works in a short rack to rack run and fails on a longer one, so measure before you assume it will hold.
Q: Why is OM4 limited to 150 m at 100G?
A: 100G over multimode uses parallel lanes, typically four lanes at 25G, and each lane has its own dispersion limit. The published 150 m is set by the worst lane plus the connector loss budget of the MPO trunk. It is a standard reach rather than a physical wall, but planning beyond it leaves no margin for contamination.
Q: Does multimode fiber work with single mode optics?
A: No. A single mode transceiver launches a small spot into a 9 micron core, and a multimode core is far too wide, so most of the light is lost at the launch and the link will not close. The reverse also fails. Choose the transceiver for the fiber class, and never mix the two on one link.
Q: How much margin should I leave under the distance limit?
A: Keep the designed length at about 80 percent of the published reach, and count every mated connector in the loss budget. Leave additional margin where the route has many patch panels or where the cable may be spliced later. A link designed exactly at the limit usually fails after the first retermination.