QSFP28 vs QSFP56: Differences in Data Rate, PAM4 Signaling and 100G/200G Applications

1. QSFP28 and QSFP56: Two High-Speed Network Generations

QSFP28 stands for Quad Small Form-factor Pluggable 28. It is a four-lane transceiver format commonly used for 100 Gigabit Ethernet. A conventional 100G QSFP28 interface uses four electrical lanes operating at approximately 25Gb/s each.

QSFP56 is a later development of the QSFP family. It supports 200Gb/s-class connectivity using four approximately 50Gb/s PAM4 lanes. This doubles the aggregate bandwidth compared with a conventional 100G QSFP28 interface without doubling the number of lanes.

The main difference is the amount of data carried by each lane. QSFP28 typically uses NRZ signaling for its standard 100G electrical interface, whereas QSFP56 uses PAM4 to achieve a higher bit rate within the available signaling bandwidth.

2. QSFP28 vs QSFP56: Technical Specification Comparison

The lane rates above describe common Ethernet-oriented implementations. Exact signaling rates, encoding overhead, optical channel arrangements and FEC requirements depend on the applicable standard and module design.

3. Lane Rate: 4 × 25G vs 4 × 50G

Both QSFP28 and QSFP56 use four data lanes in each direction. Their bandwidth difference comes from the transmission rate of each lane.

A typical 100G QSFP28 Ethernet interface operates at approximately 25.78125Gb/s per electrical lane. Four lanes provide the signaling capacity for 100GbE.

A 200G QSFP56 interface uses four approximately 50Gb/s PAM4 lanes. The higher per-lane capacity allows the module to deliver twice the aggregate bandwidth through the same number of lanes.

This higher lane rate places greater demands on the host SerDes, PCB signal paths, connectors and transceiver electronics. The host platform must be designed to support the corresponding electrical signaling standard.

4. NRZ vs PAM4: The Main Signaling Difference

NRZ and PAM4 are two different electrical and optical modulation approaches used in high-speed communication systems.

NRZ uses two signal levels to represent one bit per symbol. It is widely used in 25G-per-lane electrical interfaces and conventional 100G QSFP28 architectures.

PAM4 uses four signal levels to represent two bits per symbol. This doubles the number of bits transmitted per symbol compared with NRZ at the same symbol rate. It enables approximately 50Gb/s-per-lane transmission without requiring the signaling frequency to double.

Because PAM4 has four signal levels, the voltage separation between adjacent levels is smaller than in a comparable NRZ signal. This makes the system more sensitive to noise, distortion and signal degradation.

Forward Error Correction (FEC) is therefore an important part of many 200G PAM4 link implementations. The required FEC type and configuration depend on the Ethernet or interconnect standard and must be matched across the link.

5. Physical Dimensions and Host Compatibility

QSFP28 and QSFP56 use a similar four-lane QSFP-family mechanical design. This helps maintain high port density in switches, routers, network adapters and data center equipment.

However, similar dimensions do not mean that every QSFP28 and QSFP56 module can operate in every QSFP port. Their electrical interfaces are designed for different lane rates and signaling technologies.

A QSFP56-capable host may support selected QSFP28 modules for backward-compatible 100G operation. This is useful when a network platform needs to connect to both 100G and 200G equipment.

Such compatibility depends on the host's cage, electrical interface, firmware, supported speed modes and transceiver compatibility list. It should never be assumed from physical fit alone.

6. Electrical Signal Integrity and FEC Requirements

Moving from 25G NRZ lanes to 50G PAM4 lanes increases the requirements for high-speed electrical design.

At the host level, the SerDes must generate and receive the appropriate signal format. PCB traces, connectors, package transitions and module electrical contacts must maintain adequate signal integrity across the higher-speed channel.

PAM4 also introduces tighter signal-level margins. Equalization, receiver performance and error correction become important factors in maintaining reliable data transmission.

FEC adds redundancy to transmitted data so that certain transmission errors can be corrected at the receiving end. In 200G PAM4 systems, the appropriate FEC mode is determined by the link standard and equipment implementation.

When connecting two devices, both ends must support compatible FEC settings. A mismatch can prevent a link from coming up or lead to degraded link performance, even when the optical modules and fiber appear to be correctly connected.

7. Optical Interfaces and Transmission Distance

QSFP28 and QSFP56 describe module form-factor and electrical performance families, not a single optical wavelength or transmission distance. Optical reach depends on the specific transceiver architecture, fiber type, wavelength, receiver sensitivity and link budget.

QSFP28 supports a broad range of 100G optical configurations, including parallel multimode, wavelength-multiplexed single-mode and single-lambda designs.

QSFP56 200G optical implementations also vary. A common short-reach configuration uses four 50G PAM4 optical lanes over multimode fiber, while other 200G designs may use different optical architectures.

These are representative configurations, not universal specifications for every product. The exact optical reach, fiber grade, connector type and FEC requirements must be checked against the selected module's datasheet.

8. Power Consumption and Thermal Management

QSFP56 supports twice the nominal aggregate bandwidth of QSFP28, but its power consumption cannot be determined from the form factor alone.

Module power depends on the optical reach, laser technology, DSP or retimer architecture, receiver design and internal signal-processing requirements. PAM4 implementations may require more advanced electronic processing than conventional NRZ designs, but the actual power difference varies by product.

For high-density 200G switches, total transceiver power and chassis airflow should be considered together. Network operators should use the maximum rated power and operating-temperature specifications of the selected product when planning system cooling.

9. Network Applications: 100G vs 200G Infrastructure

QSFP28 is widely used in 100GbE data center networks, enterprise aggregation layers, high-speed server connections and telecommunications equipment. Its four-lane architecture provides a practical balance between bandwidth, port density and optical reach.

QSFP56 is designed for 200G-class networking. It is suitable for high-performance computing, data center interconnects and network infrastructure requiring greater bandwidth per port.

QSFP56 is particularly relevant where 100G port bandwidth is becoming a constraint. It can increase the capacity of individual network links while retaining the four-lane QSFP form-factor approach.

10. C-LIGHT 100G QSFP28 and 200G QSFP56 Optical Transceivers

C-LIGHT provides 100G QSFP28 and 200G QSFP56 optical transceiver options for data center, enterprise, cloud computing and high-performance network deployments.

For a 100G-to-200G network upgrade, verify the switch or network adapter model, supported port speed, PAM4 capability, FEC mode, optical reach and fiber infrastructure before selecting the transceiver.

11. How to Select Between QSFP28 and QSFP56

The selection should begin with the bandwidth requirement and host platform capability. A 100G QSFP28 module is suitable for equipment designed around 25G-per-lane electrical signaling, while a 200G QSFP56 module requires a platform supporting the appropriate 50G PAM4 interface.

It is also important to distinguish QSFP56 from QSFP56-DD. QSFP56 uses four high-speed lanes for 200G-class connectivity, while QSFP56-DD uses a double-density electrical interface with eight lanes. They are different module and host-interface configurations and should not be treated as interchangeable.

12. Conclusion