DR vs ER

DR and ER sit at nearly opposite ends of the single-mode optical reach spectrum. DR covers the last few hundred meters inside a data hall. ER spans tens of kilometers across a metro area. The distance gap between them is roughly sixty-fold, and that gap drives every other difference: the wavelength plan, the optical engine, the connector, the power budget, the amplification strategy, and the cost profile.

DR was defined for the dense, short-reach fabric inside a data center. It uses parallel single-mode lanes, each with its own laser and photodiode, to deliver 400G or 800G over eight or sixteen fibers at 500 meters. ER was defined for metro and regional transport, where a signal must cross a city without regeneration. It concentrates multiple wavelengths onto a single fiber pair and uses wavelengths in the 1550 nm window that can be amplified by erbium-doped fiber amplifiers when the link exceeds the reach of a single span.

The two interfaces rarely compete for the same link. DR connects a leaf switch to a spine switch in the same building. ER connects that spine switch to a router in a central office across town. They appear in the same network, but at different layers and with different fiber plants. Understanding what separates them clarifies where each belongs.

1. The Reach Gap and What It Forces

DR is rated for 500 meters. ER is rated for 30 to 40 kilometers. That 60:1 ratio is not a minor specification difference—it reshapes the entire module design.

At 500 meters, the fiber loss at 1310 nm is under 0.2 dB. The power budget can be modest, the transmitter can be simple, and the receiver does not need extraordinary sensitivity. At 30 kilometers, the fiber loss alone consumes 6 to 10 dB depending on the wavelength window. The module must launch more power, the receiver must detect weaker signals, and the link must be engineered with dispersion and optical signal-to-noise ratio in mind.

This is why DR and ER use different wavelength windows. DR operates at 1310 nm, where chromatic dispersion in standard single-mode fiber is near zero and no compensation is needed. ER operates at 1550 nm for the longest reaches, where fiber attenuation is at its minimum and erbium-doped fiber amplifiers can extend the span. The trade-off is higher chromatic dispersion at 1550 nm, which requires either dispersion compensation or careful link engineering.

2. Lane Architecture: Parallel Fibers vs Wavelength Multiplexing

DR uses parallel single-mode lanes. A 400GBASE-DR4 module has four lanes, each with its own laser and photodiode, and each lane occupies its own fiber. Eight fibers carry the four transmit and four receive lanes. An 800GBASE-DR8 module doubles that to eight lanes and sixteen fibers.

ER takes the opposite approach. A 400GBASE-ER4 module uses four wavelengths multiplexed onto a single fiber pair. Two fibers carry the entire 400G link—one for transmit, one for receive. The four wavelengths are combined by a LAN-WDM multiplexer inside the module and separated at the far end. The fiber count drops from eight to two, a fourfold reduction.

This architectural difference determines what fiber plant each interface requires. DR needs MPO trunk cables with enough strands to carry every lane. ER needs a duplex LC pair and a patch panel that can route it. In a building with limited conduit space, ER's fiber efficiency is a structural advantage. In a data hall where MPO trunks are already deployed for SR and DR, DR's parallel lanes fit the existing cabling pattern.

3. Wavelength Plan: 1310 nm O-Band vs 1550 nm C-Band

DR keeps all lanes at 1310 nm. There is no wavelength multiplexing, so no wavelength plan is needed. The module is a simple parallel array: every lane transmits on the same wavelength, and the receiver filters nothing. This simplicity reduces cost and eliminates wavelength-dependent impairments.

ER4 uses four LAN-WDM wavelengths around 1310 nm for shorter ER reaches, or wavelengths in the 1550 nm window for the longest reaches. The 1550 nm window is chosen because it has the lowest fiber attenuation and because it falls within the gain band of erbium-doped fiber amplifiers. When an ER link exceeds the reach of a single span, an EDFA can be inserted to boost the signal without electrical regeneration. DR has no equivalent amplification option at 1310 nm.

The wavelength plan also affects coexistence. DR's single-wavelength parallel lanes do not interfere with other services because each lane has its own fiber. ER4's four wavelengths share a fiber pair, so the fiber plant must be free of other signals in those wavelength slots unless a WDM coexistence element is used.

4. Optical Power Budget and Link Engineering

The DR power budget is sized for 500 meters. A typical DR4 link needs roughly 6 to 8 dB of budget, most of which is consumed by connector loss and margin rather than fiber attenuation. The link is short enough that dispersion and nonlinear effects are not limiting factors.

An ER link at 30 to 40 kilometers needs 12 to 20 dB of budget, depending on the wavelength. At 1550 nm, fiber attenuation is around 0.2 dB/km, so 40 kilometers consumes 8 dB. At 1310 nm, attenuation is around 0.35 dB/km, so 40 kilometers consumes 14 dB. The remaining budget covers connectors, splices, dispersion penalty, and aging margin.

This larger budget is why ER modules launch more power and have more sensitive receivers. It is also why ER links are often engineered with an optical budget calculation that accounts for every connector and splice in the path. A DR link inside a rack rarely needs that level of scrutiny.

5. Connector and Fiber Plant Implications

DR's MPO connector is designed for parallel lanes. It presents 8 or 16 fibers in a single ferrule and mates with a single push. The fiber plant that supports DR is built around MPO trunk cables, MPO patch panels, and MPO-to-LC breakout cassettes. This is the same infrastructure used by SR, so a data hall that already supports SR can often support DR with the same cabling.

ER's duplex LC connector is the standard interface for single-fiber-pair transport. It mates with the same LC patch cords and LC patch panels used by LR, FR, and PON. In a campus or metro environment where the fiber plant is built around duplex LC, ER drops in without new connector types or breakout hardware.

The practical consequence is that DR and ER belong to different fiber plant ecosystems. DR lives in the MPO world of the data hall. ER lives in the LC world of the campus and metro. Converting from one to the other requires a change in patch panels and cabling, not just a module swap.

6. Power Consumption and Thermal Load

DR modules consume less power than ER modules of the same data rate. The 1310 nm parallel lasers in DR are simpler to drive, and the absence of a wavelength multiplexer reduces the optical path loss. An 800GBASE-DR8 module typically draws 14 to 16 watts. An 800GBASE-ER8 module draws 18 to 22 watts, depending on the wavelength plan and the transmitter design.

The additional power in ER comes from several sources: higher laser output power, temperature control for wavelength stability, the LAN-WDM multiplexer, and a more sensitive receiver with a higher-gain transimpedance amplifier. In a switch chassis with 32 or 64 ports, this power difference accumulates and affects the thermal design and cooling requirements.

7. Where DR Fits

DR lives inside the data hall. Its 500-meter reach covers the distance from a leaf switch in one rack to a spine switch in another rack, or from a GPU node to a leaf switch in an adjacent row. In AI clusters, DR is the standard interface for GPU-to-leaf connections where the distance exceeds what multimode fiber can support but stays within a single building.

The DR module's MPO connector and parallel lane architecture match the cabling pattern of modern data halls. An MPO trunk cable can carry multiple DR links, and the breakout cassettes that convert MPO to LC are already in place for SR. DR extends that infrastructure to single-mode fiber without changing the physical cabling model.

8. Where ER Fits

ER lives outside the data hall. Its 30 to 40 kilometer reach covers metro DCI between data centers in the same city, regional links between a data center and a carrier point of presence, and long campus backbones that span more than a few kilometers. ER is also used in carrier access networks where a single link must reach a central office without intermediate regeneration.

ER's duplex LC connector and LAN-WDM wavelength plan match the fiber plant of metro and regional networks. The fiber is single-mode, the patch panels are LC, and the links are engineered with optical budget calculations that account for every connector and splice. ER modules are designed to fit that environment, not the MPO-heavy world of the data hall.

9. Cost Profile and Total Cost of Ownership

DR modules cost less per port than ER modules. The 1310 nm parallel lasers are cheaper to manufacture, the module has no wavelength multiplexer, and the packaging is simpler. In a data hall with hundreds or thousands of DR links, the module cost advantage is significant.

ER modules cost more because of the higher-power lasers, the LAN-WDM multiplexer, the temperature control, and the more sensitive receiver. But ER uses only two fibers per link, while DR uses eight or sixteen. In a metro environment where fiber is scarce or expensive to lease, ER's fiber efficiency can offset its higher module cost. The total cost of ownership depends on the cost of fiber relative to the cost of modules.

In a data hall where MPO trunk cables are already installed and fiber is not a constraint, DR is the lower-cost option. In a metro network where every fiber pair is leased or newly installed, ER's two-fiber link is more economical despite the higher module price.

10. Standard and Ecosystem

DR is defined in IEEE 802.3bs (400GBASE-DR4) and 802.3cd (100GBASE-DR). ER is defined in IEEE 802.3ba (100GBASE-ER4), 802.3bs (400GBASE-ER8), and earlier specifications for 10GBASE-ER. The two interfaces belong to the same IEEE family but were developed for different network segments.

11. Comparison Summary

12. Choosing Between DR and ER

The choice between DR and ER is determined by the link distance and the fiber plant, not by a performance comparison. If the link is inside a data hall and under 500 meters, DR is the appropriate interface. If the link crosses a campus or metro area and exceeds 500 meters, ER is required.

The decision framework follows the physical layout of the network:

• Distance under 500 meters: DR. The parallel lane architecture and MPO connector match the data hall cabling pattern, and the lower module cost is advantageous at high port counts.

• Distance between 500 meters and 2 kilometers: FR or LR. DR does not reach, and ER is over-specified for this range.

• Distance between 2 and 10 kilometers: LR or ER. LR is the standard 10 km interface; ER provides additional margin.

• Distance between 10 and 40 kilometers: ER. This is the range where ER's 1550 nm option and amplification capability become relevant.

There is no overlap where DR and ER compete directly. The 500-meter to 30-kilometer gap is covered by FR, LR, and intermediate reach classes. DR and ER occupy distinct segments of the reach spectrum, and the choice between them is determined by where the link falls.

13. Common Misconceptions

• "DR and ER are interchangeable if the fiber is single-mode." False. DR uses parallel lanes with an MPO connector and reaches 500 meters. ER uses LAN-WDM with a duplex LC connector and reaches 30 to 40 kilometers. The fiber count, connector, and wavelength plan are different, and the two cannot be swapped on the same link.

• "ER is just a longer version of DR." False. ER is not an extended DR. It uses a different optical architecture, a different connector, and often a different wavelength window. The two interfaces were designed for different network segments and different fiber plants.

• "DR can be amplified to reach ER distances." False. DR operates at 1310 nm, outside the gain band of erbium-doped fiber amplifiers. There is no practical amplification option for DR at 1310 nm, and the parallel lane architecture does not support the wavelength multiplexing needed for amplified transport.

• "ER is always more expensive per gigabit." Not necessarily. ER modules cost more per port, but ER uses two fibers per link versus eight or sixteen for DR. In a fiber-constrained environment where fiber is leased or newly installed, ER's fiber efficiency can make it the lower-cost option on a total cost of ownership basis.

• "DR replaces ER in modern data centers." False. DR serves in-building links under 500 meters. ER serves metro and regional links up to 40 kilometers. The two interfaces operate at different layers of the network and do not compete.

14. Summary

DR and ER are two IEEE-defined optical reach classes that occupy opposite ends of the single-mode reach spectrum. DR reaches 500 meters over parallel single-mode lanes with an MPO connector, delivering 400G or 800G inside a data hall. ER reaches 30 to 40 kilometers over LAN-WDM wavelengths with a duplex LC connector, delivering 100G, 400G, or 800G across a metro or regional network.

The reach gap drives every other difference. DR uses 1310 nm across all lanes, requires no amplification, and fits the MPO cabling pattern of the data hall. ER uses LAN-WDM or 1550 nm wavelengths, supports EDFA amplification at 1550 nm, and fits the duplex LC cabling pattern of the campus and metro. DR modules are cheaper and consume less power. ER modules cost more but use only two fibers per link.

DR and ER do not compete. They serve different network layers, different fiber plants, and different distance ranges. The choice between them is determined by the physical layout of the network, not by a performance comparison. A data center fabric uses DR for in-building links. A metro network uses ER for inter-building and inter-city links. Both interfaces coexist in the same end-to-end network, each doing the job it was designed for.

15. Q&A