BiDi SFP28 vs Duplex SFP28

The 25G generation inherited BiDi from its 10G predecessor, but the technical challenge is different. At 10G, NRZ signaling gives the optical engineer roughly 6 dB of link margin to work with, and the internal wavelength division multiplexer inside a BiDi module consumes 1 to 2 dB of that budget without threatening viability. At 25G, the same NRZ signaling operates at 2.5 times the lane rate, the link budget is tighter, the dispersion penalty is larger, and the tolerance for every additional dB of loss is reduced. BiDi SFP28 exists, and it works, but it does so under more constrained conditions than BiDi SFP+.

The choice between BiDi and duplex at 25G is therefore more consequential than it was at 10G. A BiDi SFP28 module uses one fiber with paired wavelengths; a duplex SFP28 module uses two fibers with the same wavelength in both directions. The fiber savings are identical—one fiber instead of two—but the cost of achieving them is higher at 25G, both in module premium and in link engineering effort.

This guide examines both transceiver types at 25G in depth: their architectures, wavelength plans, power budgets, reach classes, operational considerations, and the scenarios where each is the right choice. It builds on the principles established for 10G BiDi but addresses the specific constraints and opportunities that the 25G lane rate introduces.

1. What Is a Duplex SFP28?

A duplex SFP28 is the standard 25G optical transceiver form. It presents an LC duplex interface—two optical ports in a single connector body. One port is transmit, the other is receive. A duplex patch cord with two fibers connects the module at one end of the link to the module at the other end.

The 25G duplex module operates a single SerDes lane at approximately 25.78 Gbps using NRZ modulation with 64b/66b encoding. The transmit and receive wavelengths are typically identical in both directions. A 25GBASE-LR module transmits and receives at 1310 nm. A 25GBASE-ER module transmits and receives at 1550 nm. A 25GBASE-SR module transmits and receives at 850 nm.

Because the two directions are separated onto different fibers, there is no interference between them, no wavelength coordination is required, and every module of the same type is interchangeable with every other. This simplicity is duplex SFP28's defining characteristic and its primary operational advantage.

2. What Is a BiDi SFP28?

A BiDi SFP28 presents an LC simplex interface—a single optical port. That one fiber carries traffic in both directions, with the two directions separated by wavelength. The module contains an internal wavelength division multiplexer that combines the outgoing transmit wavelength with the incoming receive wavelength onto the single fiber, and a demultiplexer that separates the two wavelengths at the receiver.

Because the two directions use different wavelengths, a BiDi module is not symmetric. Each module has a transmit wavelength and a receive wavelength, and they are not the same. To form a working link, two BiDi modules must be paired such that the transmit wavelength of one matches the receive wavelength of the other. This is why BiDi modules are sold in matched pairs—commonly labeled "A" and "B"—and why a mismatched pair will fail to establish a link.

The line rate and modulation format are identical between the two. BiDi SFP28 does not use PAM4; it uses the same 25G NRZ lane as duplex SFP28. This is important because it means BiDi SFP28 inherits the signal integrity characteristics of 25G NRZ—the larger eye, the lower SNR requirement, the optional FEC—rather than the reduced noise margin of PAM4.

3. Wavelength Pairing: The Rule That Governs BiDi

The single most important operational difference between duplex and BiDi optics is that BiDi modules must be paired correctly. The transmit wavelength of the module at one end must equal the receive wavelength of the module at the other end, and vice versa. If this condition is not met, the link will not come up—the receiver will be looking for light at the wrong wavelength.

Standardized BiDi wavelength pairs exist for each reach class. The most common at 25G are summarized below.

The 1270/1330 nm pair is the standard for 25G BiDi at 10 km reach, inherited directly from the 10G BiDi wavelength plan. The module designated "BX-D" transmits at 1270 nm and receives at 1330 nm; the module designated "BX-U" transmits at 1330 nm and receives at 1270 nm. Together they form a matched pair.

The 1310/1490 nm pair is used for 40 km reach at 25G. This pair is also widely used in passive optical networks, where 1490 nm is the downstream wavelength and 1310 nm is the upstream wavelength. Because PON and BiDi share the same wavelength plan, there is potential for interference if the two systems share fiber.

Some 25G BiDi implementations use a 1290/1310 nm pair for 40 km reach, which places both wavelengths closer together in the O-band. This tighter spacing requires more precise wavelength control in the module but can simplify the WDM design in the fiber plant.

4. Fiber Savings: The Primary Advantage of BiDi

The reason BiDi optics exist is fiber conservation. A duplex 25G link requires two fibers; a BiDi 25G link requires one. For a single link, the difference is negligible. For a network with hundreds or thousands of links, the difference is substantial.

The savings matter most in three scenarios. First, in a congested conduit where additional fiber cannot be pulled without a construction project, BiDi doubles the capacity of the existing fiber. Second, in a leased fiber arrangement where the recurring cost is per strand or per fiber pair, BiDi halves the recurring cost for the same capacity. Third, in an access network or 5G transport deployment where the fiber to a site may be a single strand, BiDi is the only way to provide full-duplex service over that strand.

The savings also apply to patch cords, connectors, and patch panel ports. A BiDi deployment uses half the patch cords, half the connector pairs, and half the patch panel ports of an equivalent duplex deployment. At 25G, where the access layer port count is high and rack space is constrained, this density advantage is particularly valuable.

5. Reach and Power Budget at 25G

The link budget at 25G is tighter than at 10G, and this difference affects both duplex and BiDi deployments. The 25G NRZ lane operates at a higher baud rate, which increases the dispersion penalty and reduces the effective receiver sensitivity. Every dB of loss matters more at 25G than at 10G.

5.1 Duplex Reach Classes

Duplex 25G SFP28 modules are available in three standard reach classes: SR (850 nm multimode, up to 100 m over OM4), LR (1310 nm single-mode, up to 10 km), and ER (1550 nm single-mode, up to 30–40 km). Each class has a defined transmitter output power range and receiver sensitivity range, and the link must be engineered to satisfy the resulting power budget.

5.2 BiDi Reach Classes

BiDi 25G modules are typically specified for the same reach classes—10 km and 40 km—but the wavelength pair determines the actual loss budget. The 1270/1330 nm pair has slightly higher fiber attenuation than 1310 nm because 1270 nm is closer to the water-peak absorption region of standard single-mode fiber. This difference is small but measurable at 25G, where the margin is tighter.

The power budgets are comparable, but the BiDi module often has 0.5 to 1 dB less margin than the duplex equivalent. This is partly due to the internal WDM insertion loss and partly due to the slightly higher fiber attenuation at 1270 nm. In most deployments, this margin difference is not decisive—both module types can span 10 km comfortably—but in links that are already near the reach limit, the duplex module may provide the extra margin needed for reliable operation.

5.3 Internal WDM Insertion Loss

The internal WDM inside a BiDi module introduces 1 to 2 dB of insertion loss that a duplex module does not incur. This loss is accounted for in the module's specified output power and sensitivity, but it does mean that BiDi modules often require slightly higher transmitter output power or slightly better receiver sensitivity to compensate. At 25G, where the link budget is tighter than at 10G, this compensation is more demanding.

6. 25G NRZ: Why BiDi Avoids PAM4 Complexity

An important characteristic of both duplex and BiDi SFP28 is that they use NRZ modulation, not PAM4. This distinguishes the 25G generation from the 50G generation that followed, and it has direct implications for BiDi deployment.

NRZ uses two signal levels and encodes one bit per symbol. The eye diagram shows a single opening, the noise margin is large, and forward error correction is optional for short-reach links. This makes the signal more tolerant of the additional impairments that BiDi introduces—the internal WDM loss, the slightly higher fiber attenuation, and the wavelength-dependent dispersion.

PAM4, by contrast, uses four signal levels and encodes two bits per symbol. The eye diagram shows three openings, the noise margin is reduced by approximately 9.5 dB, and FEC is mandatory. A BiDi implementation using PAM4 would need to overcome both the PAM4 noise penalty and the BiDi WDM loss, which would make the link budget very difficult to close at 25G per lane.

This is why BiDi is more prevalent at 25G and below than at 50G and above. The NRZ modulation format provides the margin needed to accommodate the internal WDM, while PAM4 does not.

7. Cost Comparison

BiDi modules cost more than duplex modules of the same reach class, because they contain an internal WDM and require two different laser wavelengths rather than one. The premium at 25G is typically 25 to 45 percent per module.

The module premium is modest compared to the fiber savings. In a 10 km campus deployment where installing a new fiber pair costs thousands of dollars, the BiDi premium of $20 to $50 per module is trivial. In a leased fiber arrangement where the recurring cost is $100 per fiber pair per month, BiDi saves $1,200 per year per link—far more than the module premium.

At 25G, the cost comparison also includes the access-layer port count. A 25G BiDi deployment uses half the patch panel ports of an equivalent duplex deployment, which can mean the difference between fitting within the available rack space and requiring an additional panel. In high-density 25G access deployments, this density advantage has real cost implications.

8. Deployment Complexity and Troubleshooting

BiDi deployment is slightly more complex than duplex deployment, and this complexity is worth understanding before committing to a BiDi architecture.

8.1 Pairing Discipline

The most common BiDi deployment error is installing two modules of the same type at both ends of a link. Because BiDi modules look identical and may have similar part numbers, it is easy to accidentally deploy two "A" modules or two "B" modules. The link will not come up, and the cause may not be immediately obvious if the installer is not familiar with BiDi pairing rules.

Best practice is to keep BiDi pairs physically together—packaged as a pair, stored as a pair, and deployed as a pair—and to label each module clearly with its wavelength pair and A/B designation. Spare modules should be stocked as complete pairs, not as individual units, to avoid the risk of deploying a mismatched spare.

8.2 Troubleshooting Diagnostics

Duplex link troubleshooting follows a simple logic: check that light is present on the receive fiber, check the power level, and check the link status. BiDi troubleshooting adds a wavelength dimension: light may be present on the fiber but at the wrong wavelength, in which case the receiver will not detect it. A standard optical power meter calibrated at 1310 nm may read incorrectly or not at all when measuring a 1270 nm or 1330 nm signal.

To troubleshoot BiDi links correctly, the technician needs a power meter that supports the BiDi wavelengths and an understanding of the pairing rules. Digital diagnostic monitoring inside the module can provide some of this information, reporting transmit power, receive power, and in some cases the wavelength, but the module must be correctly paired for the DDM data to be meaningful.

8.3 Fiber Plant Considerations

BiDi deployment over a fiber plant that includes passive components—patch panels, splices, and adapters—is straightforward, because the fiber is transparent to wavelength. However, if the fiber plant includes any wavelength-selective components—CWDM or DWDM multiplexers, for example—the BiDi wavelengths must fall within the passband of those components. A BiDi link using 1270/1330 nm wavelengths cannot pass through a CWDM multiplexer designed for 1470–1610 nm, and the link will fail even though the fiber path is intact.

9. Where Duplex SFP28 Wins

Duplex SFP28 remains the default choice for most 25G deployments. Several factors favor it.

9.1 Fiber-Rich Environments

In data centers and campuses where fiber is abundant and cheap to install, the fiber savings of BiDi do not justify the module premium and the pairing complexity. Duplex modules are simpler, cheaper, and interchangeable, and the additional fiber is already available.

9.2 Standardized Deployments

Most enterprise, telecom, and data center networks are standardized on duplex transmission. The operational procedures, test equipment, staff training, and spare parts inventory are all built around duplex fiber pairs. Introducing BiDi into these environments requires retraining, new test equipment, and new sparing strategies. Unless fiber scarcity forces the change, most operators prefer to stay with duplex.

9.3 Interchangeability and Sparing

Duplex modules are fully interchangeable within the same reach class. Any 25GBASE-LR module works with any other 25GBASE-LR module, regardless of vendor, batch, or age. BiDi modules are not interchangeable—each must be paired with a module of the opposite type. This makes sparing more complex: a spare BiDi module is useless unless it is paired with a module of the opposite type, whereas a spare duplex module is always ready for deployment.

9.4 Multi-Fiber and Breakout Architectures

Duplex SFP28 is the foundation for the 100G breakout architecture. A 100G QSFP28 port breaks out into four 25G links using an MPO-to-LC breakout cable, and each 25G link is a duplex pair. BiDi is not compatible with this architecture because each BiDi link uses a single fiber, and the breakout cable is designed for duplex pairs. In deployments that rely on 100G-to-4×25G breakout, duplex SFP28 is the only option.

10. Where BiDi SFP28 Wins

BiDi SFP28 is the right choice when fiber is the binding constraint.

10.1 5G Fronthaul and Midhaul

In 5G transport, BiDi SFP28 is widely used in fronthaul and midhaul deployments where fiber to the radio site is limited. A single fiber can carry both the downstream and upstream fronthaul traffic using BiDi wavelengths, reducing the fiber count required to connect each radio unit. This is particularly valuable in dense urban deployments where fiber to each small cell site is expensive to provision.

The 25G NRZ modulation format is well-suited to 5G fronthaul, because it provides the link margin needed to accommodate the BiDi WDM loss while meeting the strict latency and synchronization requirements of the fronthaul interface.

10.2 Congested Conduit

In a campus or metro conduit that is already full, BiDi allows new 25G capacity to be added without pulling new cable. The existing fiber can be reused for new links at half the fiber count, effectively doubling the capacity of the conduit.

10.3 Leased Fiber

When fiber is leased under an Indefeasible Right of Use or a monthly recurring charge, the cost is typically per strand or per fiber pair. BiDi halves the number of strands required for a given capacity, directly halving the recurring cost. In this scenario, the BiDi module premium is recovered within months.

10.4 Campus Backbone with Limited Strand Count

In campuses where the backbone fiber was installed years ago with a modest strand count, BiDi allows the existing fiber to support more 25G links. This is particularly relevant for university campuses, enterprise campuses, and industrial sites where the fiber plant was designed for an earlier generation of networking and now must support more capacity.

10.5 Rapid Deployment

When a link must be deployed quickly and only one fiber is available, BiDi allows immediate service without waiting for additional fiber to be installed. The BiDi modules can be deployed on the available fiber while additional fiber is procured in parallel.

11. Comparison Summary

12. Selection Framework

13. Operational Best Practices for BiDi SFP28

If a deployment uses BiDi SFP28, several operational practices reduce the risk of pairing errors and simplify maintenance.

• Stock BiDi modules as complete pairs. Never separate an A module from its B counterpart in inventory. Order and store them as matched pairs so that a technician always has a working combination.

• Label modules clearly. Mark each module with its wavelength pair and A/B designation. Label the patch panel ports with the same information, and note the far-end pairing on the cable label.

• Use wavelength-aware test equipment. A standard 1310 nm power meter will not accurately measure a 1270 nm or 1330 nm signal. Use a power meter that supports the BiDi wavelengths, or use the module's own DDM readout.

• Document the wavelength plan. Maintain a record of which wavelength pair is used on which link, especially in mixed deployments where multiple BiDi pairs are in use across the same fiber plant.

• Train technicians on pairing rules. The most common BiDi failure is a mismatched pair. Brief training on the A/B pairing rule prevents most deployment errors.

• Check passband compatibility. Before deploying BiDi over a fiber plant that includes CWDM or DWDM components, verify that the BiDi wavelengths fall within the passband of those components.

• Verify link margin. At 25G, the BiDi module has slightly less margin than the duplex equivalent. For links near the reach limit, verify that the actual power budget closes with adequate margin before committing to BiDi.

14. Emerging Trends

14.1 BiDi at Higher Speeds

BiDi technology is expanding to 50G and beyond, but the technical challenges are greater. At 50G, the PAM4 modulation format reduces the noise margin, and the internal WDM loss becomes more difficult to accommodate. 50G BiDi SFP56 modules are in development but are less common than 25G BiDi, because the link budget is harder to close.

14.2 BiDi in 5G Transport

In 5G transport, BiDi SFP28 is the dominant interface for fronthaul and midhaul deployments where fiber to the radio site is limited. The combination of 25G NRZ signaling, single-fiber operation, and industrial-temperature operation makes it well-suited to the demanding requirements of 5G transport.

14.3 BiDi with WDM Overlay

BiDi and WDM can be combined to further increase fiber capacity. A BiDi link uses one fiber with two wavelengths; a CWDM or DWDM overlay can add additional wavelength pairs to the same fiber, multiplying the capacity. This hybrid approach is used in metro and access networks where fiber is scarce and capacity demand is high.

14.4 Tunable BiDi

Tunable BiDi modules, which can select their transmit and receive wavelengths in software, are emerging as a way to simplify wavelength management. Instead of stocking separate A and B modules, a single tunable module can be configured as either A or B, reducing inventory complexity. Tunable BiDi is still in the early stages of adoption but represents a promising direction for reducing operational complexity.

15.Conclusion

BiDi SFP28 and duplex SFP28 are two ways to build a 25G optical link. Duplex uses two fibers, one per direction, with the same wavelength in both directions. BiDi uses one fiber, with two different wavelengths multiplexed onto it—one for transmit and one for receive. The internal WDM that makes BiDi possible is what distinguishes the two, and it is why BiDi modules must be deployed in matched pairs.

The choice between them is governed primarily by fiber availability. In fiber-rich environments, duplex SFP28 is simpler, cheaper, and interchangeable, with no pairing rules and no wavelength coordination. In fiber-constrained environments—congested conduits, leased fiber routes, 5G fronthaul where fiber to the radio site is limited—BiDi SFP28 halves the fiber count per link, often justifying its module premium many times over.

The 25G generation adds a constraint that 10G did not have: a tighter link budget. The BiDi module's internal WDM loss and the slightly higher fiber attenuation at 1270 nm reduce the available margin by 0.5 to 1 dB compared to a duplex module. In most deployments, this is not decisive, but for links near the reach limit, the duplex module may provide the extra margin needed for reliable operation. The NRZ modulation format at 25G helps—it provides more margin than PAM4 would—but the budget is still tighter than at 10G.

As data rates increase and fiber demand grows, both technologies will continue to be deployed. Duplex remains the default for data center and enterprise applications, particularly those that rely on 100G-to-4×25G breakout architectures. BiDi remains the solution for the places where fiber is the constraint—and in those places, it is often the only way to add capacity without a construction project. Understanding both, and knowing which one fits a given link, is part of building a fiber plant that scales without waste.

16.Q&A