25G SFP28 Module Guide: How to Choose the Right One

A 25G SFP28 module is a compact, hot-pluggable optical transceiver that delivers 25 Gigabit Ethernet connectivity over a single lane. It shares the same physical form factor as SFP and SFP+ modules, which means existing cabling plant can often be reused when the host port already supports SFP28 operation. SFP28 modules are used across data centers, 5G fronthaul networks, enterprise campuses, and high-performance computing clusters where 10G bandwidth has become a bottleneck but 100G remains cost-prohibitive at the access layer.

 

What Does SFP28 Actually Mean?

"SFP" stands for Small Form-factor Pluggable - the industry-standard design for hot-swappable network modules that has been in use for over twenty years. The "28" refers to the electrical interface rate of up to 28 Gbps. In practice, the module is used for 25 Gigabit Ethernet (defined by IEEE standards 802.3by and 802.3cc) and 32G Fibre Channel applications. Why 28 Gbps instead of exactly 25? The extra capacity provides headroom for protocol overhead, error correction encoding, and clock recovery - housekeeping tasks that consume bandwidth beyond the nominal 25G data rate.

From a physical standpoint, an SFP28 transceiver looks identical to the older SFP+ module. Both share the same dimensions (8.5 × 13.4 × 56.5 mm) and use the same 20-pin electrical connector. This backward compatibility is a major reason why 25G adoption has moved so quickly in data center environments: network operators can plug an SFP28 module into an SFP28-capable port on a newer switch and immediately get 2.5 times the throughput they had with SFP+ - no new fiber runs, no rack reconfiguration needed.

 

 

SFP28 vs. SFP+: Why the Upgrade Path Matters

If your network is still running 10G SFP+ links at the server access layer, the upgrade case for 25G SFP28 is compelling. Each SFP28 port delivers 25 Gbps - 2.5 times more bandwidth than SFP+ - while drawing roughly the same power (typically 1.0 W to 1.5 W depending on reach). On a per-gigabit basis, that makes SFP28 significantly more energy-efficient, and the cost per gigabit has dropped steadily as production volumes have increased.

Backward compatibility deserves special attention here. If you plug an older SFP+ module into an SFP28 port, the port will automatically slow down to 10G and work normally. This makes gradual, step-by-step migrations practical: you can upgrade server NICs and top-of-rack switches to 25G-capable hardware now, and then swap in SFP28 optics over time as your workload demands grow. That flexibility is especially valuable for organizations running mixed-speed SFP transceiver environments where 10G and 25G coexist on the same chassis.

One important caveat that catches many engineers off-guard: the backward compatibility works in one direction only. SFP28 ports will accept older SFP+ modules and negotiate down to 10G - but SFP28 modules generally will not work in older SFP+ ports. Most SFP+ ports cap at 10 Gbps and will reject the faster module outright. Some switches also require you to manually set the port speed via command line before inserting a different-rate module; skip that step, and you could spend an hour troubleshooting a link that just needs one CLI command to fix.

 

 

 

 

 

Types of 25G SFP28 Modules and Their Specifications

Not all 25G SFP28 transceivers are interchangeable. The right module for your network depends on three factors: how far the signal needs to travel, what type of fiber cable you already have installed, and the environmental conditions of the deployment. Here is a breakdown of the primary variants, followed by a side-by-side comparison table for quick reference.

 

25G SFP28 SR (Short Reach)

The SR module is designed for short-distance connections - up to 100 meters. It uses an 850 nm laser paired with multimode fiber (the type of fiber cable most commonly pre-installed within data center buildings). You will need OM3 or OM4 grade multimode fiber with standard LC duplex connectors. Maximum reach is 70 meters on OM3 fiber and 100 meters on OM4, provided that RS-FEC (a form of error correction) is enabled on the switch port. Without FEC enabled, usable distance drops to roughly 30–50 meters depending on the condition of your fiber.

This is the workhorse module for connections inside a data center - server-to-switch links, storage connections within a rack row, and high-performance computing interconnects. SR modules are the lowest-cost optical SFP28 option and consume less than 1.0 W of power each, making them ideal for high-density deployments where dozens populate a single switch.

 

25G SFP28 LR (Long Reach)

LR modules are built for distances up to 10 kilometers. They use a 1310 nm laser and require single-mode fiber (OS2 grade) - a different fiber type than what SR modules use. Power consumption is slightly higher than SR, typically between 1.0 and 1.5 W per module.

Typical uses include connecting separate buildings on a campus, linking data center facilities across a metropolitan area, and 5G midhaul (the link between centralized and distributed network functions). The 10 km range covers the majority of enterprise and carrier inter-site distances without requiring signal amplifiers or repeaters, which keeps total link cost manageable. An important cost-saving tip: if your actual distance is 8 km, you do not need a 40 km-rated ER module - choosing the right reach for your real-world distance is one of the simplest ways to reduce optics spending without sacrificing reliability.

 

25G SFP28 ER (Extended Reach)

When your link distance extends to 40 kilometers, ER modules are the answer. These use a more powerful laser transmitter and a more sensitive receiver to maintain signal quality over longer fiber runs. They operate on single-mode fiber at 1310 nm and consume approximately 1.5–2.0 W. RS-FEC on the host port is required for full-distance operation.

Service providers connecting geographically separated facilities, enterprises linking distant campus sites, and 5G backhaul links spanning suburban or rural distances all rely on ER optics. While the per-module cost is higher than LR, the ability to cover 40 km without adding intermediate equipment often makes ER the more economical choice when total infrastructure cost is considered.

 

25G BiDi SFP28

Standard modules use two fiber strands - one to transmit, one to receive. Bidirectional (BiDi) modules reduce that to a single strand by transmitting and receiving on different wavelengths of light simultaneously. They must be deployed in matched pairs: a "Type A" module at one end and a "Type B" module at the other. Each type uses the reverse wavelength assignment (for example, Type A transmits at 1310 nm and receives at 1490 nm, while Type B does the opposite). BiDi SFP28 transceivers typically reach 10–15 km over single-mode fiber.

BiDi is especially valuable in situations where fiber strand count is limited, since it cuts the fiber requirement in half compared to standard two-fiber modules. This makes it a popular choice for 5G fronthaul deployments, where fiber duct space is often constrained between cell towers and baseband processing units. Network operators evaluating industrial-temperature optical transceiver options for outdoor radio cabinets frequently select BiDi SFP28 modules with extended operating ranges (-40°C to 85°C).

 

 

 

 

25G CWDM and DWDM SFP28

Wavelength-division multiplexing (WDM) modules allow multiple independent 25G channels to share a single fiber pair - effectively turning one fiber into several virtual fibers. CWDM modules use wavelengths spaced 20 nm apart and reach up to 10–40 km. DWDM modules pack channels more tightly, enabling even more simultaneous connections on the same fiber.

These modules are essential in 5G fronthaul networks where six or more 25G connections need to share a single fiber run between a central processing unit and multiple antenna sites. They also serve metro aggregation scenarios where installing new fiber is impractical or prohibitively expensive. Understanding network transceiver features like wavelength matching and modulation schemes is critical when specifying WDM optics to avoid interoperability failures.

A common and costly mistake with WDM deployments: purchasing two modules that appear compatible on paper - both single-mode, both rated for the same distance - only to discover the link refuses to come up. The culprit is often a wavelength mismatch (e.g., one module at 1310 nm and the other at 1550 nm). Unless you are deliberately using BiDi technology designed for different wavelengths, both ends of a link must use the same wavelength. When ordering CWDM or DWDM optics for multi-site deployments, always verify exact wavelength assignments per channel before placing the order.

 

 

 

 

Quick-Reference Comparison: Which Module Do You Need?

Side-by-side comparison of all major 25G SFP28 module types - use this table to match your link distance, fiber type, and deployment scenario to the right module before ordering.

 

Module Type	Max Distance	Fiber Type	Wavelength	Power Draw	Relative Cost	Best For
SR (Short Reach)	100 m (OM4) 70 m (OM3)	Multimode OM3 / OM4	850 nm	< 1.0 W	$ Lowest	Within the data center - server-to-switch, storage, HPC cluster links
LR (Long Reach)	10 km	Single-mode OS2	1310 nm	1.0 – 1.5 W	$$ Mid	Building-to-building campus links, metro DCI, 5G midhaul
ER (Extended Reach)	40 km	Single-mode OS2	1310 nm	1.5 – 2.0 W	$$$ Higher	Long-haul campus, carrier inter-office, 5G backhaul across rural / suburban spans
BiDi (Bidirectional)	10 – 15 km	Single-mode OS2, single strand	1270/1330 nm or 1310/1490 nm (paired A+B)	1.0 – 1.5 W	$$ Mid	Fiber-constrained sites - 5G fronthaul, limited duct space
CWDM	10 – 40 km	Single-mode OS2	1271 – 1331 nm (20 nm spacing)	1.0 – 1.5 W	$$$ Higher	Multiplexing several 25G channels on one fiber - 5G fronthaul, metro WDM
DWDM	40 km+	Single-mode OS2	C-band (dense spacing)	1.5 – 2.5 W	$$$$ Highest	Maximum channel density on existing fiber - carrier-grade metro networks

 

How to Read This Table (For Procurement Teams) Start with two questions:

(1) How far apart are the two ends of this link? That determines your distance requirement and narrows the "Module Type" column.

(2) What kind of fiber cable is already installed? Multimode (OM3/OM4) is common inside data center buildings; single-mode (OS2) is used for longer outdoor runs between buildings. Match those two answers to the table, and you will have your shortlist. If your distance is under 100 m and you have multimode fiber, SR is almost always the right choice. If you are connecting separate buildings up to 10 km apart on single-mode fiber, LR is the standard pick. Only move to ER, BiDi, or WDM modules when distance, fiber scarcity, or channel density specifically requires it.

 

 

Key Applications Driving 25G SFP28 Adoption

Three deployment scenarios account for the bulk of 25G SFP28 demand worldwide.

 

Data center server connectivity (spine-leaf architecture). The most common use case is connecting servers to their nearest switch. As server network cards have migrated from 10G to 25G - a shift that began around 2018–2019 and accelerated through 2024 - the switch port facing each server needs a matching SFP28 module. A typical leaf switch with 48 SFP28 downlink ports and 6 QSFP28 uplink ports aggregates 1.2 Tbps of server bandwidth into 600 Gbps of upstream capacity - a 2:1 ratio that works well for most cloud and enterprise workloads. Organizations running latency-sensitive applications like financial trading or real-time analytics often aim for 1:1 ratios, which drives higher-speed spine links even when the access layer stays at 25G. The 100G QSFP28 uplinks themselves use four 25G lanes internally, which means the 25G SFP28 transceiver is the fundamental building block of the entire network fabric.

 

5G fronthaul and midhaul. The architecture of 5G radio networks created massive demand for 25G optical links between centralized processing units and remote radio units at cell towers. The 25G SFP28 segment was the dominant optics category in the 5G market in 2024, driven by the rapid buildout of macro base stations worldwide. For outdoor deployments, the key challenges are environmental - wide temperature swings, precise timing requirements, and the physical constraints of remote cabinets with limited power and space. Industrial-rated SFP28 modules (-40°C to 85°C) with synchronization support are becoming the standard for these environments. When deploying across multiple cell sites with limited fiber availability, CWDM SFP28 modules can multiplex several 25G channels onto a single strand, significantly reducing fiber infrastructure costs. Selecting the right transceiver for telecom environments involves factors beyond reach and wavelength - understanding where different transceiver optical fiber types perform best helps prevent costly deployment mistakes.

 

Enterprise campus and storage networks. Organizations connecting multiple buildings within a campus typically need 1–10 km links at 25G - a natural fit for SFP28 LR modules over existing single-mode fiber. Storage area networks are another growth area: 32G Fibre Channel, which uses the same SFP28 physical interface, is replacing the older 16G standard in all-flash storage environments where performance demands have outgrown previous-generation connectivity.

 

 

How to Select the Right 25G SFP28 Module

Choosing a 25G SFP28 transceiver involves more than matching a distance number to a fiber type. Several practical factors determine whether a module will perform reliably in your specific environment.

 

Verify platform compatibility. SFP28 modules follow an industry-wide standard (the Multi-Source Agreement, or MSA), but some switch manufacturers implement firmware checks that reject modules not specifically coded for their brand. Reputable third-party transceiver suppliers program vendor-specific identification codes and test modules on actual switch hardware before shipping. The decision between buying original equipment manufacturer (OEM) optics and third-party compatible optics often comes down to a structured evaluation of SFP transceiver types, cost, and the vendor's validation process.

 

Confirm your fiber plant. Inserting a single-mode module (like LR) into a link that uses multimode fiber will not work - the two fiber types carry light in fundamentally different ways. Before ordering optics, verify: (1) the type of fiber installed (multimode vs. single-mode), (2) the connector polish type (UPC is standard for most SFP28 applications; APC connectors will cause problems), and (3) the total signal loss across the link including connectors and splices. On short single-mode runs, also check that received power won't exceed the module's maximum input - too much signal can damage the receiver, just as too little signal causes errors.

 

Account for FEC requirements. Most 25G SFP28 modules need forward error correction (FEC) enabled on the switch port to achieve their rated distance. FEC is a digital error-recovery mechanism built into the switch - think of it as a safety net that catches and corrects transmission errors before they affect your traffic. Without FEC, usable reach drops significantly. Confirm that your switch supports the required FEC type (RS-FEC is most common) and that both ends of the link are configured identically.

 

Evaluate power and thermal constraints. In a high-density leaf switch where all 48 SFP28 ports are populated, the combined power draw from modules alone can reach 50–70 W. Mixing SR and LR modules in the same switch creates uneven heat profiles that can affect neighboring ports. Always confirm that your switch's power budget and cooling design can handle fully loaded operation under worst-case temperature conditions before deploying at scale.

 

 

Compatibility with 100G QSFP28 Through Breakout

One of the most practical deployment patterns connects four SFP28 ports to a single QSFP28 port using a special "breakout" cable. Here is how it works: a 100G QSFP28 module actually carries four independent 25G lanes. A breakout cable splits those four lanes so that each one connects to a separate SFP28 port on a downstream switch or server. This can be done with direct-attach copper breakout cables (for distances under 5 meters), active optical breakout cables (up to 100 meters), or a fiber harness with separate optical modules at each end. The breakout approach maximizes port utilization on upstream switches and simplifies cabling in structured environments.

For organizations in the middle of a phased migration where older 10G switches and new 100G switches coexist, breakout cables can serve as a bridge - though this depends on the specific switch models and firmware involved. The 100G side runs at full 4×25G, and each lane connects to a downstream port. Whether that downstream port runs at 10G or 25G depends on the hardware. Always verify breakout compatibility in your switch documentation before assuming this topology will work as expected.

 

DAC and AOC Alternatives to Pluggable Optics

For very short connections - within a single rack or between neighboring racks - you do not necessarily need optical modules at all. 25G SFP28 Direct Attach Copper (DAC) cables are pre-terminated copper assemblies that plug directly into SFP28 ports at both ends. Passive DAC supports up to 3 meters at 25G; active DAC extends to roughly 10 meters. These one-piece cables are the lowest cost per link, eliminate the risk of dirty optical connectors, and draw less power than separate transceivers. Active Optical Cables (AOC) bridge the gap between DAC and pluggable optics, reaching up to 100 meters using fiber but without requiring you to buy separate transceivers for each end. Both options integrate naturally into environments where transceivers serve multiple roles across compute, storage, and network tiers.

 

 

Frequently Asked Questions