
By the Sanoc Engineering Team — last reviewed and validated against IEEE 802.3, ITU-T and TIA fiber standards in 2026. Every distance, wavelength and loss-budget figure below has been cross-checked against modules our New Taipei City factory builds, programs (Sanoc FreeCode) and bench-tests before shipment.
Quick answer: Single-mode SFPs use 9 µm fiber with 1310/1550 nm lasers for 10–80 km links; multimode SFPs use 50 µm fiber with 850 nm VCSELs for links up to 300–550 m depending on data rate (OM3/OM4). Use multimode inside a room or building, single-mode between buildings or sites — the modules and fiber types are not interchangeable.
“Should I use a single-mode or multimode SFP?” is the single most common question our applications engineers field every week — and the wrong answer either burns budget (paying for single-mode optics on a 30-metre run) or, far worse, produces a link that won’t come up at all (driving multimode optics across a 10 km span). This guide is the reference we wish every network engineer had on the bench: a deep, standards-anchored, field-tested walkthrough of single-mode vs multimode SFP transceivers — the fiber types, core sizes, wavelengths, lasers, connectors, reach tables, and a repeatable decision process you can apply to any port.
We manufacture optical transceivers — SFP, SFP+, SFP28, QSFP+ and QSFP28 — that are compatible with Cisco, HPE, Juniper, Arista, MikroTik and Ubiquiti switches, so the figures here are not copied from a datasheet PDF; they come from optics we program and measure ourselves. Where a real-world reading differs from the headline spec, we say so. Let’s get into it.
1. Single-Mode vs Multimode SFP at a Glance
Before the deep dive, here is the comparison most engineers actually need pinned above their desk. An SFP fiber optic module is a hot-pluggable transceiver that converts electrical signals to light and back; whether it is single-mode or multimode is determined by the fiber it is designed to launch into and the laser inside it, not by the SFP cage or the connector shape.
The headline comparison table
| Attribute | Multimode SFP (MMF) | Single-Mode SFP (SMF) |
|---|---|---|
| Fiber core diameter | 50 µm (OM3/OM4/OM5) or 62.5 µm (OM1) | ~9 µm (OS1/OS2) |
| Fiber standard | ITU-T G.651 / TIA-492 (OM1–OM5) | ITU-T G.652 / G.657 (OS2) |
| Typical jacket colour | Aqua (OM3/OM4), lime (OM5), orange (OM1/OM2) | Yellow |
| Laser / light source | 850 nm VCSEL (short reach) | 1310 nm or 1550 nm DFB/EML |
| Typical reach | 30 m – 550 m (10G); up to ~2 km at 1G | 2 km – 120 km |
| Common IEEE names | 1000BASE-SX, 10GBASE-SR | 1000BASE-LX/EX/ZX, 10GBASE-LR/ER/ZR |
| Connector | LC duplex (most), SC (legacy) | LC duplex (most), SC (legacy) |
| Relative module cost | Lower for short reach | Higher per module, far cheaper fiber per km |
| Best for | In-rack, in-row, data-centre, building risers | Campus backbone, metro, inter-building, long haul |
The one-sentence rule of thumb
If the link is inside a building or data centre and under ~400 m, reach for a multimode SFP (SX/SR) and aqua OM3/OM4 fiber. The moment you cross buildings, leave a campus, or exceed a few hundred metres, switch to a single-mode SFP (LX/LR and beyond) on yellow OS2 fiber. Everything in the rest of this article is the engineering detail that turns that rule of thumb into a defensible design.
The three independent facts every engineer confuses
Most of the confusion around sfp single mode vs multimode comes from collapsing three separate properties into one word. Keep them apart and the topic becomes simple:
- Fiber mode — single-mode (9 µm) or multimode (50/62.5 µm). Decides reach class and jacket colour.
- Wavelength / laser — 850 nm VCSEL, 1310 nm DFB, or 1550 nm DFB/EML. Must match the fiber mode.
- Connector — LC, SC or MPO. A purely mechanical fit that is independent of the first two.
An “SFP LC single-mode 1310 nm LR” module is one sentence stating all three facts. Once you read SFP specs that way, picking the right part is mechanical.
Why people get this wrong
Two myths cause most failures we troubleshoot. First, “an LC connector means single-mode” — false; both fiber modes use LC. Second, “more distance is always safer” — also false, because single-mode optics launched into multimode fiber (or vice versa) produce huge modal/coupling penalties and unstable links. Matching module wavelength to fiber type is non-negotiable, and we’ll show you exactly how to verify the match before you patch the cable.
2. Fiber Core Sizes: 9 µm vs 50/62.5 µm
The deepest reason single-mode and multimode behave differently lives at the glass. A fiber is a core surrounded by a lower-index cladding; light is trapped by total internal reflection. The core diameter decides how many spatial paths (modes) the light can travel — and that, in turn, decides bandwidth and reach.
Single-mode fiber (SMF): the 9-micron highway
Single-mode fiber per ITU-T G.652 (and bend-tolerant G.657) has a core of roughly 8.3–9.5 µm, with a standard 125 µm cladding. At 1310 nm and 1550 nm only one spatial mode propagates, so there is essentially no modal dispersion. The penalty is a tiny core: aligning a 9 µm core demands precise connectors and clean end-faces, which is why single-mode patch cords and splices are less forgiving of dirt. The reward is enormous reach — tens of kilometres on a single SFP — because the light stays coherent. OS2 (the modern, low-water-peak grade) is the fiber you want for any new single-mode build; OS1 is the older indoor-tight-buffered grade.
Multimode fiber (MMF): the wide, short road
Multimode fiber per ITU-T G.651.1 and graded by TIA-492 (OM1–OM5) uses a 50 µm core (OM2/OM3/OM4/OM5) or a legacy 62.5 µm core (OM1). The large core makes coupling cheap and tolerant — great for a low-cost 850 nm VCSEL — but it lets many modes propagate, and they arrive at slightly different times (modal dispersion). That spreads the pulse and caps the distance-bandwidth product. Higher OM grades have tighter index profiles (“effective modal bandwidth”, EMB) that push the usable distance up.
OM1 / OM2 / OM3 / OM4 / OM5 — the multimode grades
| Grade | Core | Jacket colour | EMB @850 nm | 10GBASE-SR reach | Notes |
|---|---|---|---|---|---|
| OM1 | 62.5 µm | Orange | ~200 MHz·km | ~33 m | Legacy; avoid for 10G+ |
| OM2 | 50 µm | Orange | ~500 MHz·km | ~82 m | Legacy 1G plant |
| OM3 | 50 µm | Aqua | 2000 MHz·km | 300 m | Laser-optimised, data-centre baseline |
| OM4 | 50 µm | Aqua (or violet) | 4700 MHz·km | 400 m | Preferred for 10/25/40/100G short reach |
| OM5 | 50 µm | Lime green | 4700 MHz·km | 400 m+ | Wideband (SWDM) ready |
Two practical takeaways. (1) If your building still has OM1 orange fiber, a 10GBASE-SR SFP will only reach ~33 m on it — plan single-mode for anything longer. (2) For new short-reach installs, OM4 is the sweet spot; OM3 is acceptable; OM5 only buys you something if you intend to run SWDM optics. Our 1G and 10G multimode modules are validated against OM3/OM4 plant — see the 1000BASE SFP transceivers and 10GBASE SFP+ transceivers for the SX/SR variants.
Why the core size dictates the physics — modal dispersion explained
It is worth understanding why a wider core costs you distance, because it explains every reach number in this article. In a 50 µm multimode core, light can travel straight down the axis or bounce at shallow angles off the core-cladding boundary. Each of those paths — each “mode” — has a different physical length, so a single launched pulse arrives smeared across time. That smearing is modal (intermodal) dispersion, and it grows with distance until adjacent bits overlap and the receiver can no longer distinguish them. Graded-index multimode fiber fights this by varying the refractive index across the core so off-axis modes (which travel further) also travel faster, partially re-aligning their arrival. The quality of that index profile is exactly what the “effective modal bandwidth” (EMB, in MHz·km) rating measures — and why OM4’s 4700 MHz·km reaches further than OM3’s 2000 MHz·km at the same speed.
In a 9 µm single-mode core there is simply not enough room for higher-order modes to propagate at 1310/1550 nm, so only the fundamental mode survives. With modal dispersion eliminated, the distance limit moves out to chromatic dispersion and pure attenuation — both far gentler — which is why single-mode reach is measured in tens of kilometres rather than hundreds of metres. This single fact, core size, is the root cause of the entire single-mode-vs-multimode trade-off.
Bend radius, OS2 and a note on fiber handling
One often-overlooked difference: single-mode’s small core and long wavelength make it more sensitive to macrobends — a tight loop behind a patch panel can add measurable loss at 1550 nm. This is why G.657 bend-insensitive single-mode fiber exists for tight in-building routing, and why you should keep single-mode patch cords above their rated bend radius (typically 10–30 mm). Multimode is comparatively forgiving of bends but unforgiving of grade mismatches and dirty connectors. Knowing which failure mode each fiber is prone to lets you troubleshoot faster: suspect bends and end-faces on single-mode, suspect grade and modal bandwidth on multimode.
3. Wavelengths & Lasers: 850 nm VCSEL vs 1310/1550 nm DFB
The transmitter optic inside the SFP is as decisive as the fiber. Wavelength, laser type and the resulting spectral width set both the cost and the maximum reach. This is where “single-mode” and “multimode” become physical, not just labels.
850 nm VCSEL — the multimode workhorse
Short-reach multimode SFPs (SX, SR) use an 850 nm VCSEL (Vertical-Cavity Surface-Emitting Laser). VCSELs are cheap to make, run cool, and couple efficiently into a 50 µm core — which is exactly why multimode short reach is so economical inside a rack or row. The trade-off is that 850 nm light suffers more attenuation per kilometre (~3 dB/km) and the VCSEL’s broader spectrum interacts with modal dispersion, capping reach in the hundreds of metres.
1310 nm DFB — the single-mode standard reach
Single-mode “long reach” optics (LX, LR) use a 1310 nm DFB (Distributed-Feedback) laser. At 1310 nm, G.652 fiber has its zero-dispersion window and moderate attenuation (~0.33 dB/km), so a DFB-driven LR module covers 10 km comfortably with margin to spare. DFB lasers have a very narrow spectral width, which is what makes long, low-dispersion links possible.
1550 nm DFB/EML — extended and long haul
For ER (40 km), ZR (80 km) and DWDM optics, modules move to 1550 nm, where SMF attenuation bottoms out near 0.22 dB/km — the lowest-loss window in standard glass. At 1550 nm, chromatic dispersion becomes the limiting factor over very long spans, so high-end ZR/DWDM optics use externally-modulated lasers (EML) and tighter line-width control. ITU-T G.694.1 defines the DWDM grid these wavelengths sit on.
VCSEL vs DFB — side by side
| Parameter | 850 nm VCSEL | 1310 nm DFB | 1550 nm DFB/EML |
|---|---|---|---|
| Fiber mode | Multimode | Single-mode | Single-mode |
| Spectral width | ~0.45 nm (RMS, wide) | < 1 nm | < 0.1 nm (narrow) |
| Fiber attenuation | ~3.0 dB/km | ~0.33 dB/km | ~0.22 dB/km |
| Typical reach | ≤ 400 m (10G) | 10–25 km | 40–120 km |
| Relative cost | Lowest | Medium | Highest |
| Example modules | 1000BASE-SX, 10GBASE-SR | 1000BASE-LX, 10GBASE-LR | 1000BASE-ZX, 10GBASE-ER/ZR |
Field note: this is why you cannot simply swap a multimode SFP for a single-mode one to “go further.” An 850 nm VCSEL launched into 9 µm single-mode fiber couples poorly, and a 1310 nm DFB launched into 50 µm multimode fiber triggers DMD (differential mode delay) penalties unless you use a conditioning launch cord. Match the laser to the fiber every time.
Spectral width, chromatic dispersion and why ZR optics are special
There is a second, subtler reason laser choice matters: spectral width. No laser emits a single perfect colour; it emits a band of wavelengths, and different wavelengths travel at slightly different speeds in glass (chromatic dispersion). A wide-spectrum source over a long span smears pulses just as modal dispersion does. VCSELs have relatively wide spectra — fine over hundreds of metres, hopeless over kilometres. DFB lasers are spectrally narrow, and the EML (externally-modulated laser) sources in ZR/DWDM optics are narrower still, which is precisely what lets a single strand carry 10G across 80 km. When you see a premium attached to an ER or ZR SFP, you are paying for that laser quality and its tight wavelength control — not marketing.
How wavelength choice interacts with the power budget
Wavelength does not just set reach class; it sets your loss-per-kilometre, which feeds straight into Section 8’s power budget. At 850 nm you lose ~3 dB every kilometre — brutal, but irrelevant over 300 m. At 1310 nm you lose ~0.33 dB/km; at 1550 nm only ~0.22 dB/km. That is why long-haul optics moved to 1550 nm: it is the lowest-loss window in standard silica fiber. When you budget a link, always pair the wavelength with its attenuation coefficient — using the wrong figure (e.g., 1310 nm loss for a 1550 nm link) is a common and costly arithmetic error.
4. Reach: SR/LR/ER/ZR and SX/LX/EX/ZX Distance Tables
Engineers live and die by reach. Below are the IEEE-defined naming conventions and the distances you can actually expect — both the standard maximum and what we typically measure on clean, properly-budgeted plant in our own lab. Real readings often beat the spec on short runs (because the spec assumes worst-case fiber and connectors), but never design to the measured number — always design to the standard with margin.
The naming code, decoded
The suffix tells you everything: S = short (multimode 850 nm), L = long (single-mode 1310 nm), E = extended (single-mode 1550 nm ~40 km), Z = “ZR” extended (single-mode 1550 nm ~80 km). At 1G the family is SX/LX/EX/ZX; at 10G it becomes SR/LR/ER/ZR (the “R” denotes the 64b/66b coded PHY). The letters map directly onto the laser/fiber pairings from Section 3.
1 Gigabit SFP reach (1000BASE) — IEEE 802.3z
| Module | Wavelength | Fiber | Spec max reach | Typical measured (clean plant) |
|---|---|---|---|---|
| 1000BASE-SX | 850 nm | OM3/OM4 MMF | 550 m (OM3 ~700 m+) | ~600–800 m on OM4 with margin |
| 1000BASE-LX | 1310 nm | OS2 SMF | 10 km | 10–15 km with full budget |
| 1000BASE-EX | 1310/1550 nm | OS2 SMF | 40 km | ~40 km |
| 1000BASE-ZX | 1550 nm | OS2 SMF | 80 km | 70–80 km depending on fiber loss |
10 Gigabit SFP+ reach (10GBASE) — IEEE 802.3ae
| Module | Wavelength | Fiber | Spec max reach | Typical measured (clean plant) |
|---|---|---|---|---|
| 10GBASE-SR | 850 nm | OM3 / OM4 MMF | 300 m (OM3) / 400 m (OM4) | up to 400 m on OM4 |
| 10GBASE-LRM | 1310 nm | OM1/OM2/OM3 MMF | 220 m | ~220 m (legacy FDDI plant) |
| 10GBASE-LR | 1310 nm | OS2 SMF | 10 km | 10–15 km with margin |
| 10GBASE-ER | 1550 nm | OS2 SMF | 40 km | ~40 km |
| 10GBASE-ZR | 1550 nm | OS2 SMF | 80 km | 70–80 km depending on fiber loss |
Beyond 10G the same logic scales to 25G SFP28 transceivers (25GBASE-SR/LR) and 40G QSFP+ transceivers (40GBASE-SR4/LR4), where short reach uses parallel MPO multimode and long reach uses duplex LC single-mode. The single-mode/multimode decision you make at SFP level is identical at every speed.
Why measured can beat spec — and why you still design to spec
IEEE reach figures assume worst-case transmitter, worst-case fiber attenuation and a connector-loss allowance. On a fresh OM4 run with clean LC end-faces, an SR link may light at 450 m on the bench. That is not a licence to deploy at 450 m — aging, microbends, additional patches and temperature will eat that headroom. Treat measured numbers as confirmation of margin, not as your design distance.
5. SFP Connectors: LC / SC / MPO — Connector ≠ Fiber Mode
This is the section that clears up more confusion than any other. The SFP connector is the mechanical interface where the patch cord meets the module — and it tells you almost nothing about whether the link is single-mode or multimode. Let’s separate the two ideas cleanly.
LC — the dominant SFP fiber connector
The vast majority of modern SFP and SFP+ optics use an LC duplex connector (two 1.25 mm ferrules in a small clip — one Tx, one Rx). “SFP LC” simply means the module accepts LC patch cords; an LC SFP can be single-mode or multimode depending on the optic and fiber. When engineers search “sfp lc” they almost always just want to confirm their patch-cord polarity and ferrule type — and the answer is: use a duplex LC-LC cord, single-mode (yellow, OS2) for LR/ER/ZR optics, multimode (aqua, OM3/OM4) for SR/SX optics.
SC and other legacy connectors
Older 1000BASE-SX/LX modules and some industrial GBICs used the larger SC (square, push-pull, 2.5 mm ferrule) connector. SC still appears on legacy plant and some 1G single-mode links, but new SFP designs have standardised on LC for density. Again, SC vs LC is purely mechanical — it does not imply a fiber mode.
MPO/MTP — when one SFP isn’t enough
Parallel optics (40GBASE-SR4, 100GBASE-SR4) use an MPO/MTP ribbon connector carrying 8 or 12 fibers. You won’t see MPO on a standard single-lane SFP/SFP+, but you will on QSFP modules — and the same mode rules apply (SR4 = multimode MPO, PSM4/DR = single-mode MPO).
The crucial point: connector vs mode vs wavelength
| Property | What it controls | What it does NOT tell you |
|---|---|---|
| Connector (LC/SC/MPO) | Mechanical fit, ferrule size, polarity | Whether the link is single- or multimode |
| Fiber mode (SMF/MMF) | Core size, reach class, jacket colour | The connector shape |
| Wavelength/laser (SX/LR/ER…) | Reach, fiber type it must launch into | The connector shape |
So the answer to “sfp lx — what connector?” is: LX is a single-mode 1310 nm 1G optic, and it almost always terminates in a duplex LC (older units in SC), running over yellow OS2 fiber. The “LX” tells you the reach and fiber; “LC” tells you the plug. They are independent facts.
6. BiDi & Single-Fiber SFP: Halving Your Fiber Count
Most SFP links use two fibers — one to transmit, one to receive. BiDi (bidirectional) SFPs let you run a full-duplex link over a single fiber strand, which is gold when fiber is scarce or leased per-strand.
How BiDi works — WDM in a module
A BiDi SFP integrates a WDM filter so that one end transmits at, say, 1310 nm and receives at 1490 nm, while the far end does the reverse. They are sold as matched A/B pairs — you must order one of each, never two of the same. Because both directions share one strand, BiDi is almost always a single-mode technology (LC simplex connector).
When to choose BiDi
- Leased dark fiber priced per strand — BiDi halves your recurring cost.
- Existing single-strand runs you cannot re-pull (e.g., into a remote cabinet).
- FTTx / access aggregation where strand count is the constraint.
BiDi caveats
Each direction uses a different wavelength, so the per-direction power budget differs slightly — always check both the 1310-Tx and 1490-Tx side budgets. And because pairs are matched, keep clear labelling: a mismatched pair (two “A” units) simply will not link. For multi-strand parallel needs at higher speed, DAC cables or AOC cables are often the cheaper in-rack alternative to optics entirely.
BiDi vs CWDM/DWDM — when to escalate further
BiDi solves the “one strand, one link” problem. When you need many links over one strand, you move up to CWDM (coarse WDM, ~20 nm channel spacing, up to 18 channels per ITU-T G.694.2) or DWDM (dense WDM, the 0.8/0.4 nm grid of G.694.1, dozens of channels). Both are single-mode technologies that stack multiple coloured SFPs onto one fiber through a passive mux. The selection logic extends naturally: BiDi for a single duplex link on one strand; CWDM for a handful of links on a budget; DWDM for high channel counts and metro/long-haul scale. All of them inherit the single-mode power-budget rules in Section 8 — you simply add the mux/demux insertion loss to the total.
7. How to Tell If Your SFP Is Single-Mode or Multimode
Here is the hands-on, first-party knowledge our engineers use daily. You have an unlabelled SFP in your palm — how do you know its mode without a datasheet? Use these checks in order; any one is usually conclusive.
Check 1 — the bale-clasp / housing colour code
The single fastest tell is the colour of the bale clasp (the latch handle) and sometimes the pull-tab:
- Black or beige bale → almost always multimode (SX/SR, 850 nm).
- Blue bale → single-mode 1310 nm (LX/LR).
- Yellow bale → single-mode 1550 nm (EX/ER/ZX/ZR).
- Purple/violet bale → often BiDi or CWDM/DWDM single-mode.
This colour convention is widely (though not universally) followed; treat it as a strong hint, then confirm with Check 2 or 3.
Check 2 — read the part-number/label markings
Look at the laser-etched label. The wavelength and reach are almost always printed: “850nm 550m” or “SX” → multimode; “1310nm 10km” or “LR/LX” → single-mode; “1550nm 80km” or “ZR/ZX” → single-mode long haul. The IEEE suffix (Section 4) is the definitive answer once you can read it.
Check 3 — read DDM/DOM with a CLI (the engineer’s method)
Every compliant SFP supports Digital Diagnostics Monitoring (DDM/DOM) per SFF-8472. Plug it into a managed switch and read the optics:
- Cisco:
show interface transceiver detail - Arista/Juniper:
show interfaces transceiver/show interfaces diagnostics optics
The output reports the wavelength (850/1310/1550 nm), the vendor/part number, and live Tx/Rx optical power in dBm. Wavelength alone tells you the mode: 850 nm = multimode; 1310/1550 nm = single-mode. The live Tx/Rx dBm values are also exactly what you need for the power-budget check in Section 8.
Check 4 — inspect the fiber, not just the SFP
If the SFP is already cabled, the patch-cord jacket colour is a final cross-check: yellow = single-mode, aqua/orange/lime = multimode. If the SFP says 850 nm but the cord is yellow single-mode, you have found a mismatch — stop and fix it before chasing “link errors.”
First-hand tip: when in doubt, we measure
Every module we ship is bench-read for wavelength, Tx power and extinction ratio, and we offer free EEPROM coding (Sanoc FreeCode) so the host switch reads the correct vendor and reach data. If you want certainty on an unknown unit, the cleanest path is a DDM read on a switch port or a request for a free, factory-tested sample so you can compare against a known-good reference.
8. Optical Power Budget & Link Loss
Choosing single-mode vs multimode is only half the job; the link must close — meaning the received power has to stay above the receiver’s sensitivity after all the loss along the way. This is the optical power budget, and it is where most “it won’t link” tickets are actually resolved.
The power budget equation
The link margin is simply:
Link Margin (dB) = Tx Power (dBm) − Rx Sensitivity (dBm) − Total Link Loss (dB)
where Total Link Loss = (fiber attenuation × distance) + (connector loss × number of connectors) + (splice loss × number of splices). You want a comfortably positive margin — we recommend keeping at least 3 dB of headroom for aging and repair.
A worked single-mode example (10GBASE-LR, 10 km)
Typical 10GBASE-LR figures: Tx power −8.2 dBm (min), Rx sensitivity −14.4 dBm.
- Power budget = −8.2 − (−14.4) = 6.2 dB available.
- Fiber loss @1310 nm: 0.33 dB/km × 10 km = 3.3 dB.
- Connectors: 4 mated pairs × 0.5 dB = 2.0 dB.
- Splices: 2 × 0.1 dB = 0.2 dB.
- Total loss = 3.3 + 2.0 + 0.2 = 5.5 dB.
- Margin = 6.2 − 5.5 = 0.7 dB → too tight! Reduce connectors, clean end-faces, or move to an ER optic.
This is the classic trap: a 10 km LR link “should” work, but four dirty connectors quietly eat the budget. The fix is usually housekeeping, not a bigger optic.
A worked multimode example (10GBASE-SR, 300 m on OM3)
- SR power budget ≈ 2.4 dB at 300 m on OM3 (the limit here is modal bandwidth, not just attenuation).
- Fiber loss @850 nm: 3.0 dB/km × 0.3 km = 0.9 dB.
- Connectors: 2 × 0.5 dB = 1.0 dB → total 1.9 dB, leaving ~0.5 dB. Move to OM4 to regain margin.
Do the math without a calculator — use our dBm tool
Converting between mW and dBm by hand is error-prone, so we built a free converter you can keep open while you budget a link. Plug in your DDM Tx/Rx readings and the tool returns the dBm/mW pair instantly: mW ⇄ dBm conversion calculator. For a full step-by-step on assembling the budget — including connector and splice allowances and how to read margin against receiver sensitivity — see our optical power budget & link loss guide, and reach for the dBm calculator any time a DDM value needs converting. Used together, these two resources turn the power-budget step from guesswork into a 30-second check.
9. Selection Decision Tree: Picking the Right SFP
Here is the repeatable process. Answer four questions in order and the right module class falls out.
Step 1 — How far is the link?
- < 400 m, in-building / data centre → multimode SR/SX on OM3/OM4. Cheapest, lowest power.
- 400 m – 10 km → single-mode LR/LX on OS2. The campus/metro default.
- 10 – 40 km → single-mode ER/EX (1550 nm).
- 40 – 80 km → single-mode ZR/ZX (1550 nm).
- > 80 km or many channels → CWDM/DWDM single-mode + amplification.
Step 2 — What fiber already exists in the ground?
Survey before you buy. If the building is full of yellow OS2, you are committed to single-mode optics regardless of distance. If it is aqua OM3/OM4, multimode short reach is on the table. Orange OM1 existing plant severely limits 10G — plan single-mode or re-cable.
Step 3 — How many strands do you have?
Plenty of strands → standard duplex optics. Single strand or per-strand leasing → BiDi single-fiber SFP (matched A/B pair). In-rack only → consider DAC/AOC and skip optics altogether.
Step 4 — Will the budget close?
Run the Section 8 math with real connector counts. If margin is under 3 dB, either reduce connectors/splices, clean end-faces, or step up one reach class. Verify against the receiver sensitivity, not just the headline distance.
Decision summary table
| Scenario | Recommended SFP | Fiber | Connector |
|---|---|---|---|
| Top-of-rack / in-row, <100 m | 10GBASE-SR (or DAC/AOC) | OM4 MMF | LC duplex |
| Data-centre spine, ≤400 m | 10GBASE-SR | OM4 MMF | LC duplex |
| Building-to-building campus, 1–10 km | 10GBASE-LR / 1000BASE-LX | OS2 SMF | LC duplex |
| Metro, 40 km | 10GBASE-ER / 1000BASE-EX | OS2 SMF | LC duplex |
| Long haul, 80 km | 10GBASE-ZR / 1000BASE-ZX | OS2 SMF | LC duplex |
| Single-strand run | BiDi SFP (A/B pair) | OS2 SMF | LC simplex |
For deeper specification reference on the families above, our complete SFP transceiver guide covers form factors and coding, and the QSFP / QSFP28 guide extends the same single-mode/multimode logic to 40G and 100G.
10. Common Mistakes & Field Troubleshooting
Even with the right module, links fail for predictable reasons. Here are the ones we see most.
Mode mismatch (the #1 cause)
A single-mode SFP on multimode fiber, or vice versa. Symptom: link flaps, very high Rx loss, or no link at all. Fix: confirm the SFP wavelength via DDM and the cord jacket colour; they must agree. A useful sanity rule: if your DDM read shows 850 nm but the patch cord is yellow, or 1310/1550 nm but the cord is aqua, you have a mode mismatch — fix it before changing anything else.
Reach class too aggressive for the budget
Specifying LR (10 km) for a 12 km run, or relying on the headline distance without counting connectors. Symptom: link comes up but flaps under temperature, or shows Rx power hovering near sensitivity. Fix: run the real power budget (Section 8), count every mated connector and splice, and step up to ER if margin is under 3 dB. Conversely, do not over-spec: an ER optic slammed into a 500 m link can overload the receiver — some long-reach optics need an inline attenuator on short spans.
Dirty or damaged end-faces
A single contaminated LC ferrule can add 1–3 dB. On a budget-tight long single-mode run, that is the difference between up and down. Always inspect and clean before blaming the optic.
Wrong OM grade for the speed
Running 10GBASE-SR over OM1/OM2 and expecting 300 m. The grade caps you (OM1 ≈ 33 m). Check the jacket and EMB before committing.
BiDi pair mismatch
Two identical BiDi units will never link — they need opposite Tx/Rx wavelengths. Order and label as A/B pairs.
Frequently Asked Questions
What is the difference between a single-mode and multimode SFP?
A multimode SFP uses an 850 nm VCSEL and launches into large-core (50/62.5 µm) multimode fiber for short reach (typically under 400 m at 10G). A single-mode SFP uses a 1310 nm or 1550 nm DFB laser and launches into small-core (~9 µm) single-mode fiber for long reach (2 km to 120 km). The mode is set by the laser and the fiber, not by the SFP cage or connector.
Can I use a single-mode SFP with multimode fiber (or vice versa)?
No — not reliably. Mixing modes causes large coupling and modal-delay penalties, producing unstable links or no link at all. Always match the SFP’s wavelength to the fiber: 850 nm with multimode (aqua/orange), 1310/1550 nm with single-mode (yellow). The only exception is 10GBASE-LRM with a conditioning launch cord on legacy multimode.
Does an LC connector mean the SFP is single-mode?
No. The LC connector is purely mechanical and is used by both single-mode and multimode SFPs. To determine the mode, read the module’s wavelength via DDM (850 nm = multimode; 1310/1550 nm = single-mode) or the printed reach/IEEE suffix, and check the fiber jacket colour.
How do I tell if my SFP is single-mode or multimode?
Use the bale-clasp colour (black/beige = multimode, blue = 1310 nm single-mode, yellow = 1550 nm single-mode), read the printed label (SX/SR = multimode, LX/LR/ER/ZR = single-mode), or read DDM/DOM on a managed switch to see the exact wavelength. Cross-check with the patch-cord jacket colour: yellow = single-mode, aqua/orange/lime = multimode.
What fiber does a 1000BASE-SX SFP use, and how far does it reach?
1000BASE-SX is an 850 nm multimode optic. On OM3 it reaches roughly 700 m+ and on OM4 even further; on legacy OM1/OM2 the reach drops to ~220–550 m. Use aqua OM3/OM4 fiber with a duplex LC connector for best results.
What fiber does a 10GBASE-SR SFP module need?
10GBASE-SR is an 850 nm multimode module. It reaches 300 m on OM3 and 400 m on OM4 — but only about 33 m on legacy OM1. For 10G short reach always specify OM3 or, preferably, OM4 multimode fiber with clean LC end-faces.
What is the difference between SR/LR/ER/ZR and SX/LX/EX/ZX?
They are the same reach hierarchy at different speeds. SX/LX/EX/ZX are 1G (IEEE 802.3z) and SR/LR/ER/ZR are 10G (IEEE 802.3ae). In both, S = short multimode 850 nm, L = long single-mode 1310 nm (~10 km), E = extended 1550 nm (~40 km), and Z/ZR = 1550 nm long haul (~80 km).
How do I calculate the optical power budget for an SFP link?
Use Link Margin = Tx Power − Rx Sensitivity − Total Link Loss, where total loss = fiber attenuation × distance + connector loss × connectors + splice loss × splices. Keep at least 3 dB of margin. Read your live Tx/Rx values via DDM and convert mW to dBm with our free mW ⇄ dBm calculator; the power budget guide walks through a full worked example.
Are single-mode SFPs more expensive than multimode?
Per module, single-mode optics usually cost more because DFB lasers are pricier than 850 nm VCSELs. However, single-mode fiber itself is cheaper per kilometre and future-proofs you for higher speeds, so for any run over a few hundred metres single-mode is typically the better total-cost choice. For short in-rack links, multimode (or even DAC/AOC) is the economical pick.
Can I get a free, factory-tested SFP sample to verify compatibility?
Yes. Sanoc manufactures Cisco/HPE/Juniper/Arista/MikroTik/Ubiquiti-compatible single-mode and multimode SFP/SFP+/SFP28 modules with free EEPROM coding (Sanoc FreeCode), bench-measured wavelength and power, and a 1-year warranty. Request a free sample and we’ll send a factory-tested unit you can DDM-read and qualify on your own switch before you order.
Conclusion: Match the Laser to the Fiber, Then Budget the Link
The single-mode vs multimode SFP decision is not complicated once you separate the three independent facts: the fiber mode (9 µm SMF vs 50/62.5 µm MMF), the laser/wavelength (850 nm VCSEL vs 1310/1550 nm DFB), and the connector (LC/SC/MPO — which tells you nothing about mode). Use distance to pick the reach class, survey the installed fiber, confirm the strand count, and finally prove the power budget closes with 3 dB of margin. Do that and your links come up first time, every time.
As a Taiwan-based manufacturer with our own New Taipei City factory, 2025 Taiwan Excellence recognition, and every module bench-tested before it ships, we’re glad to help you specify the exact single-mode or multimode optic your network needs. Browse our 1000BASE SFP, 10GBASE SFP+, 25G SFP28 and 40G QSFP+ ranges, keep the dBm calculator handy for every link budget, and request a free, factory-tested sample to verify compatibility on your own hardware — zero risk, full support from our engineering team.
This article provides technical explanations and selection guidance. Check the exact module part number, full equipment model, software version, port mode, far-end device, and fiber conditions together. Use the applicable product and equipment documentation for specifications and support scope.
Sanoc provides a 1-year warranty. Warranty, DOA, returns, and after-sales terms follow the formal quotation, order terms, or applicable warranty policy. Consult the equipment manufacturer’s terms for the scope of the switch warranty.
Have the equipment model, module part number, and requirements ready when using the contact page.
Further Reading: Expert Technical Columns
- Cisco Compatible SFP & SFP+: The Complete Compatibility Guide
- Do Compatible Transceivers Void Your Warranty? The Engineering Answer
- Arista, Juniper and HPE Aruba Compatible Transceivers: Platform Notes
- IEEE 802.3 and the MSA: What Transceiver Standards Actually Guarantee
- The 400G to 800G Data Center Transition: What IT Leaders Should Plan For
- AI Networking and the Optical Interconnect Surge: A Strategic View
- My SFP Link Won’t Come Up — A Field Troubleshooting Guide
- Inside the Sanoc QA Lab: How We Bench-Test Every Batch
- Why Taiwan Optical Manufacturing Matters for Your Supply Chain