OM4 or OS2: which to buy

The short answer, by distance

Choose by the longest link you will ever build on that cable, then by the transceivers you plan to run. Inside a row or a room, OM4 usually does the job and costs less to light. Between rooms, floors, buildings or anything past about 100 metres at modern speeds, OS2 is the safer purchase, because its reach does not fall as the speed rises. The two are not interchangeable once installed: a multimode transceiver cannot use single-mode fibre, and a single-mode transceiver cannot use multimode fibre in any standard design.

The sections below give the numbers behind this table, then a worked example, a cost view, and a checklist. For a plain introduction to the two families, read single-mode vs multimode fibre first. This page assumes you already know the core size difference and focuses on the purchase decision. Related questions are answered in the Radix fibre guides.

Reach in the IEEE standards

The IEEE 802.3 Ethernet standards give a maximum channel length for each optical interface. Multimode reach shrinks as speed grows, and the grades of fibre make a difference. Single-mode reach is set by the optics, not the cable, so a given single-mode interface works on any OS2 length up to its limit.

Read the table by column. Going from 10G to 100G cuts the OM4 reach from 400 m to 100 m, while the single-mode columns stay at kilometre scale. That pattern is why designers who expect an upgrade path choose single-mode for backbone and long cross-connects, and keep multimode for short in-row links. The details of the grades are in OM3 vs OM4 vs OM5, and the reach tables for newer interfaces appear in the guide to 400G SR8, DR4 and FR4 cabling.

Figures in the table are the standard maximum lengths for a channel that meets the loss limit of the standard. A channel with too many connectors can fail well inside the length limit, as the next section shows.

Loss limits matter more than length

Multimode links have a tight loss allowance. The 10GBASE-SR channel is allowed about 2.6 dB on OM3 at 300 m and 2.9 dB on OM4 at 400 m. A 10GBASE-LR channel on single-mode has a larger allowance of about 6.2 dB. Cable attenuation is 3.5 dB/km for multimode at 850 nm and 0.4 dB/km for single-mode in the ISO/IEC 11801 OS2 grade, so on short links the connectors, not the fibre, use most of the budget.

Consider a multimode channel of 150 m with two mated connector pairs at 0.5 dB each (assumed values). Cable loss is 0.15 km x 3.5 dB/km = 0.53 dB, connectors add 1.0 dB, and the total is 1.53 dB. That passes a 2.9 dB limit with 1.37 dB spare. Add two more connector pairs, for a patch panel at each end, and the total is 3.53 dB, which fails the same limit at a length well inside the 400 m maximum. Single-mode has no such problem on the same layout: 0.15 km x 0.4 dB/km = 0.06 dB, plus 2.0 dB of connectors is 2.06 dB, against a 6.2 dB allowance.

So the higher bandwidth of OM4 does not reduce the penalty for each connector, and a structured cabling design with many patch points has more room on OS2. Use the fibre loss budget calculator with your own connector and splice counts, and see the loss budget guide for the method.

Worked example: two links, two answers

Assume a facility with two requirements, with all figures chosen for illustration.

• Link A: a leaf switch to a server row, 40 m by route, 25G now and 100G planned in six years.
• Link B: the main hall to a building across the yard, 460 m by route, 10G now and 25G or faster later.

For link A, both fibres work, and the choice comes down to the optics, the existing switch ports and the budget. For link B, OM4 fails on length before any other factor is considered. The point of the exercise is that one project often needs both, and the selection is made link by link, not site by site. Keeping a short table like the one above in the design file prevents the habit of buying one fibre type for everything.

Cost: where the money moves

Fibre cost has three parts: the cable, the connectors and the transceivers. The cable and cord cost difference between OM4 and OS2 is small compared with the transceiver and the labour. What differs is the direction of the transceiver cost gap, which has narrowed over successive speeds.

• 10G and 25G. Multimode short-reach modules have usually been cheaper than their single-mode equivalents, so a short OM4 link costs less to light.
• 100G and 400G. Parallel multimode optics (SR4, SR8) need 8 to 16 fibres and multi-fibre connectors, while single-mode DR4 needs 8. The optics cost gap is smaller and, depending on volumes, can reverse. The 12-fibre cabling for parallel optics is described in 400G and 800G parallel optics cabling.
• Cable life. Cabling is commonly planned for fifteen years or more, while switches are refreshed every five to seven. A fibre type chosen for today's switch has to serve at least two later generations.

Check current quotes for the optics you intend to use, since market prices change. A cable decision that looks cheaper by a few percent on the cord can cost far more if the link has to be re-pulled in five years. The guide to fibre optic cable price lists what makes up the price of an assembly.

Do not mix them

OM4 and OS2 patch cords look similar and their connectors fit the same adapters. Mating the wrong type causes loss that is easy to miss on a short link.

The colour code helps. Single-mode cord is normally yellow, and multimode OM3 and OM4 are normally aqua, with violet used by some makers for OM4. Connector colours vary, and the jacket and the printed text are better guides than the connector alone. Details are in the pages on multimode fibre colours and the single-mode yellow jacket and green connector. Label each end with the fibre grade as well.

A decision sequence for your purchase

1. Measure every route, not a typical one. The longest link in the group decides the fibre type for the whole run, unless you split the order.
2. List the transceiver or interface at each end and its reach on each fibre grade, from the IEEE table or from the module data sheet.
3. Count the connector pairs on the longest and the most patched link, and run the loss budget with realistic per-pair values.
4. Ask what speed the link needs in five and in ten years. If it is likely to go beyond 25G over more than 100 m, choose OS2.
5. Check port types on the equipment. Some switches only offer multimode ports at the speeds you need.
6. For mixed sites, order OM4 for in-row links and OS2 for the backbone, and label both clearly.
7. Write the polish for single-mode (APC or UPC) and the gender and polarity for any MPO. See the pages on UPC vs APC and MPO polarity.
8. Put the quantities, lengths and grades in an RFQ using the format in how to write an RFQ for fibre assemblies.

Radix makes single-mode and multimode patch assemblies to the customer's drawing in Pune, with quotes by form, phone or WhatsApp. The LC to LC OM4 duplex and LC to LC OS2 duplex pages describe the two cord types side by side. Radix tests each cord before dispatch.

When each choice is wrong

• OM4 is wrong for a backbone you plan to keep through several upgrades. The reach falls with each speed step and the pull cost is much higher than the cord cost.
• OM4 is wrong for any run past 400 m. There is no standard multimode channel for it at 10G and above.
• OS2 is a poor fit where the equipment is multimode-only. Old switches, some video gear and a few industrial devices only have multimode ports. Check the port before ordering.
• OS2 is wrong where dirty connectors are the main risk and nobody cleans them. A speck covers more of a 9 micron core than of a 50 micron one. Where single-mode is needed anyway, plan cleaning as part of the work.
• Either is wrong if the choice was made per supplier, not per link. A cheap bulk order of one grade for all sites tends to come back as a re-order.

If your project is a mix and you are unsure, send the route lengths and equipment list to Radix and ask for a cord set for each case. The full list of cord families is on the fibre optic patch cords page.