10GBASE SFP+ Supported Distances & Fiber Types

 

The SR Story (and Why OM3/OM4 Actually Matters)

 

Everyone starts with 10GBASE-SR. It's cheap, it works, and 850nm VCSELs have been around forever. But here's where things get interesting-and where I've seen countless deployment mistakes.

The distance ratings you see on datasheets assume perfect conditions. 300 meters on OM3, 400 meters on OM4. Sure. But those numbers come from controlled lab environments with fresh fiber and clean connectors. Real world? You're dealing with patch panels that haven't been cleaned since 2015, bend radius violations hidden in cable trays, and that one splice someone did at 2 AM during a maintenance window.

OM1 and OM2 are still out there, by the way. Legacy campus buildings, older data centers that never got rewired. On OM1 (62.5μm core), you're looking at maybe 33 meters. OM2 gets you to 82 meters. Not great. The modal bandwidth just isn't there-500 MHz·km for OM2 versus 2000 MHz·km for OM3. The difference matters enormously at 10G speeds where modal dispersion becomes the limiting factor rather than attenuation.

 

LR and the Single-Mode Transition

10GBASE-LR operates at 1310nm over single-mode fiber. Ten kilometers. That's the spec. In practice, with good fiber and proper link budget planning, some deployments push it further-I've personally validated links at 12-13km with adequate margin, though this voids warranty territory and isn't something you'd document officially.

The jump from multimode to single-mode represents more than just a distance upgrade. You're moving from 50μm or 62.5μm cores down to 9μm. The alignment tolerances during termination become far more critical. Connectors matter more. The polish type matters-PC, UPC, APC all behave differently. For LR applications, you typically want UPC connectors; the flat polish works fine at 1310nm where back-reflection isn't as catastrophic as it becomes at 1550nm.

What nobody tells you: the fiber itself is actually cheaper per meter for single-mode. The cost disparity comes entirely from termination equipment and the transceivers themselves. An SR module runs maybe $30-50 from reputable third-party suppliers. LR? Triple that, minimum.

 

ER and ZR: When Distance Gets Serious

40 kilometers for 10GBASE-ER. 80+ kilometers for ZR. These are 1550nm optics, and they're a different beast entirely.

The power budgets are substantial-ER typically specifies +4 dBm launch power with receiver sensitivity around -15.8 dBm, giving you roughly 20dB to work with. ZR pushes this further with higher-powered lasers and more sensitive APD receivers. But power budget alone doesn't tell the whole story. At these distances, chromatic dispersion accumulates. The 1550nm window sits right where dispersion-shifted fiber was supposed to solve everything (it didn't, but that's a different rant about G.653 and why it's basically obsolete now for DWDM).

Standard G.652 single-mode has chromatic dispersion around 17 ps/(nm·km) at 1550nm. Over 80 kilometers, that adds up. ZR optics include electronic dispersion compensation to handle this, which is part of why they cost what they cost.

Honestly? If you're looking at ZR distances, you should probably be evaluating coherent optics or DWDM solutions anyway. The price premium for ZR has shrunk relative to entry-level coherent in recent years.

 

LRM: The Forgotten Standard

10GBASE-LRM exists. 220 meters over legacy multimode using 1310nm. It's designed for FDDI-grade installations-older buildings with OM1 fiber that can't be replaced economically.

I'm mentioning it for completeness. In 15 years of network engineering, I've deployed LRM exactly twice. Both times in university buildings from the 1980s where the conduit runs made pulling new fiber prohibitively expensive. The technology works through electronic dispersion compensation in the receive path, essentially cleaning up the modal mess that 1310nm transmission creates in multimode fiber.

If you have a choice, don't use LRM. Just budget for fiber replacement.

 

Link Budget: The Math Nobody Wants to Do

Here's a quick reality check formula:

Available margin = Tx power − Rx sensitivity − fiber attenuation − connector losses − splice losses − safety margin

 

For a typical LR deployment over 8km with four mated connector pairs:

 

 

The 0.35 dB/km figure is conservative for modern OS2 fiber at 1310nm. Some installers quote 0.4 dB/km to pad their numbers. G.652.D fiber typically measures around 0.32-0.34 dB/km when new.

At 1550nm (ER/ZR territory), attenuation drops to approximately 0.22 dB/km. This is why longer reaches are possible despite the dispersion challenges.

 

 

Quick Reference (Because Sometimes You Just Need the Numbers)

 

 

Compatibility and the Third-Party Question

Every major switching vendor-Cisco, Juniper, Arista, HPE-implements some form of transceiver authentication. Cisco's is the most aggressive; certain IOS versions will outright refuse to enable ports with non-TAA-compliant optics. Juniper tends to log warnings but function. Arista is generally permissive.

Third-party optics work fine in most cases. Companies like Finisar (now II-VI), Lumentum, and various white-label manufacturers produce the same silicon that ends up in OEM modules. The premium you pay for Cisco-branded optics is primarily logo and support assurance.

That said-and this matters for enterprise deployments-using third-party optics typically voids vendor support for link-level issues. If you open a TAC case for packet loss and they discover you're running FS.COM transceivers, the conversation becomes difficult.

DDM (Digital Diagnostic Monitoring) provides real-time telemetry regardless of vendor. Temperature, Tx power, Rx power, bias current, voltage. Every modern SFP+ supports it per SFF-8472. Use it. The data tells you when things are degrading before they fail.

 

BiDi, CWDM, and Other Variants

Worth mentioning: not all 10G SFP+ modules are simple point-to-point. BiDi transceivers use WDM within a single fiber strand-typically 1270nm/1330nm pairs for 10km BiDi-LR applications. Halves your fiber count. Useful when you're running low on dark fiber in existing conduit.

CWDM and DWDM SFP+ modules multiplex multiple 10G channels over one fiber pair. CWDM uses 20nm spacing (up to 18 channels practically), DWDM uses 0.8nm spacing (80+ channels). These aren't substitutes for your standard LR module-they're systems-level decisions that involve mux/demux equipment, wavelength planning, and usually a vendor conversation.

 

Final Thought

The specifications exist for a reason, but fiber networks remain stubbornly physical. Dust on an endface adds 0.5dB. A 15mm bend radius (spec says 30mm) introduces macro-bend loss you won't find on any datasheet. That fusion splice done in the field never quite matches factory pigtails.

Test everything. Trust nothing. Keep your cleaning kit stocked.