At the heart of every WDM subsystem is an optical filtering architecture that combines and separates wavelength channels with high precision. Among the most widely used solutions are Z-Block and Thin Film Filter (TFF) Prism technologies. Although both perform the same fundamental function, they differ in optical design, manufacturing processes, and application characteristics.

This article provides an overview of how these two technologies work and the key factors to consider when selecting a WDM optical solution.

What Is Z-Block Technology?

Block is a highly integrated free-space optical assembly designed for wavelength multiplexing and demultiplexing. It integrates multiple precision optical components—including thin-film filters, mirrors, collimating lenses, and fiber arrays—within a single compact optical module.

The name “Z-Block” comes from the zigzag optical path that light follows as it passes through multiple reflective and transmissive optical elements inside the assembly.

By integrating these components into one precisely aligned structure, Z-Block technology delivers consistent optical performance while reducing the number of discrete optical assemblies required.

How Does Z-Block Work?

During demultiplexing, the incoming optical signal enters the Z-Block through a collimating lens, which converts the divergent beam into parallel light before it reaches a sequence of thin-film filters.

Each filter extracts one wavelength channel while allowing the remaining channels to continue through the optical path. Mirrors redirect the separated wavelengths toward their designated output ports until all channels have been demultiplexed.

When operating as a multiplexer, the process is simply reversed, combining multiple wavelength channels into a single output fiber.

This integrated optical path helps maintain low insertion loss and stable wavelength performance across multiple channels.

Advantages and Limitations of Z-Block

Key advantages include:

• Highly integrated architecture
• Low insertion loss
• Excellent wavelength consistency
• Compact package design
• High mechanical stability

Limitations

• Complex optical alignment during manufacturing
• Higher production cost compared with simpler assemblies
• Limited flexibility after final assembly
• Tight manufacturing tolerances

What Is TFF Prism Technology?

Thin Film Filter (TFF) Prism technology is another well-established optical filtering architecture widely used in CWDM and DWDM systems. Rather than integrating all optical elements into a single assembly, it combines precision thin-film filters with prism optics to separate or combine optical wavelengths.

Each thin-film filter is coated with multiple dielectric layers designed to selectively reflect or transmit specific wavelengths while the prism guides the optical path between filtering stages.

TFF Prism technology has been widely deployed in CWDM and DWDM applications due to its mature manufacturing process and proven reliability.

How Does TFF Prism Work?

As the optical signal enters the prism assembly, it is directed toward a series of thin-film filters.

Each filter is designed for a specific wavelength channel. Depending on its optical coating, a particular wavelength is either reflected toward an output port or transmitted to the next filtering stage.

This sequential filtering process continues until every wavelength has been separated or combined.

Because each optical element performs an individual filtering function, TFF Prism assemblies offer excellent wavelength selectivity while supporting flexible channel configurations.

Advantages and Limitations of TFF Prism

Key advantages include：

• Mature and proven technology
• High wavelength selectivity
• Excellent channel isolation
• Flexible optical configurations
• Reliable long-term performance

Limitations

• Larger optical assembly compared with integrated designs
• More individual optical components
• Cumulative insertion loss may increase as channel count grows
• Assembly complexity increases with higher channel densities

Z-Block vs. TFF Prism

Although both technologies are designed for wavelength multiplexing and demultiplexing, each offers distinct advantages depending on the application.

Neither technology is universally better than the other. The optimal choice depends on system requirements such as channel count, insertion loss, package size, manufacturing complexity, and overall network design.

Conclusion

Z-Block and TFF Prism technologies both play important roles in modern WDM systems. While they share the same objective of efficiently multiplexing and demultiplexing optical wavelengths, they achieve this through different optical architectures and manufacturing approaches.

Understanding the characteristics of each technology allows network designers and system integrators to select the most appropriate solution based on performance requirements, scalability, and application needs.

As WDM technologies continue to advance, both Z-Block and TFF Prism architectures will remain essential optical building blocks, supporting higher-capacity networks with reliable wavelength multiplexing and demultiplexing performance.

FIBERSTAMP Insight

From CWDM and DWDM to O-Band TFF MUX/DEMUX and coherent optical solutions, FIBERSTAMP focuses on delivering precision-engineered WDM subsystems designed for stable optical performance and long-term reliability.

The post Inside the Optics: Understanding Z-Block and TFF Prism Technologies in WDM Systems first appeared on FIBERSTAMP.

As AI and data center interconnects accelerate toward 800G and 1.6T, short-reach optical connectivity is no longer a question of power optimization alone. It has evolved into a system-level engineering challenge, encompassing port density, thermal design, link stability, and large-scale deployability.

Amid growing industry discussions around “de-DSP” architectures, LPO has been pushed toward its theoretical limits, while traditional DSP-based solutions continue to serve as the foundation for reliability. LRO, however, has yet to establish a stable position in large-scale deployments. Its reliance on TX-only DSP architectures increasingly exposes constraints in chip reusability, ecosystem scalability, and long-term sustainability.

By contrast, HYBRID (semi-DSP) architectures take a more pragmatic approach. By reusing mature full-duplex DSP platforms and reducing DSP utilization rather than redefining DSP architectures, HYBRID achieves a more realistic balance across power efficiency, latency, signal quality, and engineering controllability.

From a system engineering perspective, this article argues that AI and data center short-reach interconnects are entering a phase where architectural momentum is shifting from LRO toward HYBRID-based design methodologies. The following sections examine the technical and industrial factors driving this transition.

2. Revisiting the Core Architectures: DSP, LPO, and LRO

2.1 DSP-Based Modules: The Anchor of Performance

Architecture:

Full DSP processing on both TX and RX paths.

Strengths:

• Strongest signal processing capabilities (equalization, CDR, FEC, nonlinear compensation)
• Supports medium-, long-, and ultra-long-reach links
• Mature ecosystem and standardized interfaces for plug-and-play deployment

Limitations:

• High power consumption (typically >14–16 W at 800G)
• Increased latency
• Highest cost

Typical Applications:

Metro networks, backbone networks, DCI, and mission-critical links where reliability is paramount.

2.2 LPO: Maximum Energy Efficiency

Architecture:

No DSP inside the module; all signal processing is handled by the host SerDes.

Strengths:

• Lowest power consumption (30–50% lower than DSP modules)
• Ultra-low latency
• Lowest module BOM cost

Limitations:

• Requires exceptional host SerDes performance and channel consistency
• Limited reach (typically ≤100 m)
• Complex system tuning; ecosystem still maturing

Typical Applications:

Ultra-short-reach links within or between AI racks.

2.3 LRO: A Conceptual Compromise

Architecture:

DSP retained on TX; linear reception on RX.

Strengths:

• Lower power than full DSP
• Longer reach than LPO

Challenges:

• Fragmented TX-only DSP variants
• Poor chip reusability, weak economies of scale
• Ecosystem adoption remains limited

3. HYBRID: The Semi-DSP, Engineering-Ready Approach

3.1 What HYBRID Means

HYBRID is not simply LRO. It represents a system-level reallocation of DSP resources:

• Only a portion of TX/RX channels pass through DSP
• Remaining channels use a linear architecture
• DSP may be applied selectively to either TX or RX

Conceptually:

HYBRID ≈ (LRO + LTO) / 2

3.2 Key Differences: HYBRID vs. LRO

Core Insight:

HYBRID reuses mature duplex DSPs and reduces channel usage, avoiding the low-volume, high-risk TX-only DSPs used in LRO. Early validation shows certain linear-receive links in HYBRID even outperform LRO, reinforcing its real-world advantage.

4. HYBRID: Advantages and Practical Considerations

Key Advantages:

Power Efficiency: 20–30% lower than full DSP

Ultra-Low Latency: DSP usage halved, latency comparable to LRO

Predictable Signal Quality:

MMF 50 m: PRE-FEC BER E-7 to E-8

SMF 500 m: PRE-FEC BER E-10

Cost Optimization: ~20% reduction vs. traditional DSP modules

Supports Higher Density: Enables 16-channel / 3.2T pluggable modules

Challenges:

• Linear RX channels require slightly higher host SI tuning
• System-level co-optimization is needed; not plug-and-play
• Large-scale adoption is in early stages and requires close collaboration

Compared with LPO and LRO, HYBRID offers more controllable engineering risk.

5. Parallel Coexistence Is Inevitable

6. Conclusion: HYBRID as the Most Practical Middle Ground

LPO represents the idealized extreme for efficiency

DSP remains the foundation for reliability

LRO offers conceptual compromise but faces scalability limits

HYBRID emerges as the deployable, scalable intermediate solution

HYBRID leverages mature duplex DSPs, avoiding the specialized TX-only variants of LRO. This gives it superior performance, efficiency, cost-effectiveness, ecosystem readiness, supply-chain stability, and real-world deployability.

Looking forward, LPO, HYBRID, and DSP will coexist, forming the technological foundation for AI and next-generation data center interconnects.

The post Why HYBRID Architectures Outperform LRO in Real-World Systems—and Are Reshaping Short-Reach Interconnects in AI Data Centers first appeared on FIBERSTAMP.

At the heart of this transition are three core components: optical transceivers, optical DSPs (oDSPs), and switch ASICs. How these elements are partitioned, optimized, and integrated defines the technical and economic trade-offs between LPO and CPO.

I. Core Components of Optical Interconnect Systems

1. Optical Transceivers: The Electro-Optical Bridge

Optical transceivers enable bidirectional conversion between electrical and optical signals and are indispensable in modern data centers and telecom networks.

Key functions include:

Electro-Optical Conversion

Laser sources modulate electrical signals into light at the transmitter, while photodetectors convert optical signals back to electrical form at the receiver.

Rate and Reach Adaptation

Supporting industry-standard form factors such as QSFP-DD and OSFP, transceivers scale from 100G to 1.6T and address distances from short-reach (50 m) to 2 km interconnects.

Signal Conditioning

Conventional pluggable optics rely heavily on onboard DSPs for equalization, forward error correction (FEC), and dispersion compensation—often accounting for nearly half of the module’s total power consumption at 400G and above.

2. Optical DSPs (oDSPs): The Intelligence Engine

Optical DSPs represent the most complex and highest-value electrical components inside transceivers, typically contributing 20–30% of the BOM cost.

Their core roles include:

Advanced Modulation and Demodulation

In data center environments, PAM4 DSPs enable higher per-lane data rates (50G/100G) while compensating for signal distortion. In long-haul networks, coherent DSPs employ advanced modulation formats to maximize reach and sensitivity.

Signal Regeneration and Error Correction

Digital processing such as FEC restores degraded signals and ensures transmission reliability over extended distances.

Power Consumption Constraints

At 800G, oDSPs alone can consume 6–8 W, making them the dominant contributor to module power dissipation and a key bottleneck for further scaling.

3. Switch ASICs: The Data Center Traffic Hub

Switch ASICs form the backbone of data center networking, responsible for ultra-high-speed packet forwarding and port connectivity.

Key capabilities include:

High-Speed SerDes Connectivity

Supporting 112G and future 224G SerDes, ASICs enable massive port density and low-latency interconnects between servers, accelerators, and storage.

Integrated Signal Processing

Modern ASICs increasingly incorporate PAM4 interfaces, clock recovery, and basic equalization to preserve signal integrity.

Expanded Role in LPO Architectures

In LPO systems, switch ASICs take on part of the signal conditioning traditionally handled by DSPs, enabling simplified, lower-power optical modules.

II. LPO and CPO: Technology Paths and Industry Impact

1. LPO: Driving Efficiency Within the Pluggable Ecosystem

Linear-drive Pluggable Optics (LPO) represent an evolutionary step that optimizes today’s pluggable form factors for power and cost efficiency.

Key Technical Characteristics

DSP-Free Architecture

LPO replaces traditional DSPs with high-linearity Driver and TIA components, eliminating CDR and complex digital processing. This significantly reduces power consumption, latency, and system cost, particularly in 800G applications.

Operational Compatibility

Retaining QSFP-DD and OSFP formats, LPO preserves hot-swappability and operational flexibility—critical for large-scale deployment and maintenance.

Standards Alignment

LPO is built around OIF CEI-112G-Linear-PAM4 specifications. While 800G adoption is underway, future 224G SerDes validation remains a key milestone.

Industry Impact

Power and Cooling Savings

In dense AI racks, LPO can deliver substantial energy savings, reducing both electrical and cooling costs at the system level.

Supply Chain Diversification

By minimizing DSP dependency, LPO accelerates innovation in Driver/TIA silicon and broadens supplier participation.

Application Boundaries

Because signal compensation shifts upstream to the switch ASIC, LPO is best suited for short-reach, homogeneous environments, such as AI training clusters.

2. CPO: Unlocking the Next Level of Bandwidth Density

Co-packaged Optics (CPO) represent a more disruptive architectural shift—bringing optical engines physically closer to, or directly integrated with, the switch ASIC.

Technical Evolution

From Near-Package to Co-Package

Transitioning from co-board optics to true CPO shortens electrical traces from centimeters to millimeters, delivering 30–50% power reduction and improved signal integrity.

Advanced Packaging Architectures

CPO implementations range from 2.5D integration to chiplet-based and future 3D stacking approaches, enabling deeper convergence of silicon photonics and switching silicon.

Silicon Photonics at the Core

High-density, scalable silicon photonics is foundational to CPO, with industry projections indicating it will dominate optical integration in the coming decade.

Industry Impact

Extreme Performance Scaling

CPO architectures enable system bandwidths beyond 51.2T with ultra-low latency, directly addressing the demands of next-generation AI and accelerated computing fabrics.

Ecosystem and O&M Challenges

Early CPO deployments rely on proprietary designs and face higher operational complexity, including limited field replaceability.

Clear Market Segmentation

CPO primarily targets scale-up AI networks, while scale-out data center architectures continue to rely on pluggable optics.

III. Technology Outlook: Competition, Coexistence, and Convergence

1. Multi-Path Evolution

LPO as a Mid-Term Mainstream Solution

Between 2025 and 2027, LPO is expected to see rapid adoption in AI clusters and mid-scale data centers, driven by its balance of efficiency and deployability.

CPO as a Long-Term Strategic Direction

Beyond 2030, CPO is likely to gain traction in hyperscale environments as packaging, silicon photonics, and open ecosystems mature.

DSP Remains Essential

DSP-based optics will continue to dominate long-reach and heterogeneous network scenarios, evolving toward more power-optimized designs.

2. Synergistic Innovation

Silicon Photonics as a Common Enabler

Advances in silicon photonics benefit both LPO—through lower-cost, high-performance optics—and CPO—through dense, integrated optical engines.

Packaging Breakthroughs

3D integration, TSVs, and advanced thermal solutions will be critical to unlocking CPO’s full potential.

Standards and Ecosystem Development

LPO benefits from ongoing IEEE and OIF standardization, while CPO’s future depends on broader industry collaboration and open interfaces.

3. Supply Chain Transformation

Chip Vendors

Traditional DSP leaders continue to innovate, while system players integrate optics more tightly with switching silicon to enhance platform differentiation.

Optical Module Providers

Vendors such as FIBERSTAMP are actively investing in LPO and future CPO-ready technologies, balancing near-term deployment with long-term architectural shifts.

Manufacturing and Foundries

Expanded silicon photonics capacity and advanced packaging capabilities are becoming strategic enablers for next-generation optical interconnects.

IV. Conclusion

The rise of LPO and CPO marks a decisive transition from DSP-centric pluggable optics toward more power-efficient, tightly integrated interconnect architectures. LPO offers a practical, near-term path to reduced power and cost, while CPO represents the long-term vision for ultra-high-bandwidth, AI-driven data centers.

Rather than a zero-sum competition, LPO and CPO will coexist and evolve synergistically, each serving distinct application layers within the data center. For the industry, success will depend on advances in silicon photonics, packaging, and—most critically—open standards and ecosystem collaboration.

At FIBERSTAMP, we believe this multi-path evolution will define the next decade of optical networking—driving smarter, more efficient, and more scalable connectivity for AI and cloud infrastructure.

The post LPO and CPO: A Strategic Turning Point and Parallel Evolution in Optical Interconnect Architectures first appeared on FIBERSTAMP.

1. Dispersion

Dispersion occurs when light of different wavelengths travels at varying speeds through a medium, causing pulse broadening and signal distortion. This phenomenon can be classified into several types:

Modal Dispersion – Common in multimode fibers, where multiple propagation paths exist. Light traveling near the fiber axis arrives sooner than light reflecting off the edges, resulting in pulse spreading. For example, a 10 Gbps signal transmitted over more than 300 meters in multimode fiber can suffer such severe modal dispersion that individual bits (“0”s and “1”s) become indistinguishable.

Material Dispersion – Arises from the wavelength-dependency of a fiber’s refractive index (typically silica). Light comprising multiple spectral components—such as that from a wide-linewidth laser—experiences differential speeds even within the same mode, leading to smearing of the signal.

Waveguide Dispersion – Results from the fiber’s structural geometry (core/cladding refractive index difference, core diameter). Part of the light’s energy extends into the cladding, causing wavelength-dependent propagation constants and thus dispersion.

Polarization Mode Dispersion (PMD) – A specialized form of dispersion. Fiber imperfections, such as core ellipticity or uneven mechanical stress, cause orthogonal polarization components (e.g., horizontal vs. vertical) to travel at different speeds, introducing timing discrepancies even at the same wavelength.

2. Polarization

Polarization defines the orientation of the light wave’s electric field relative to its propagation direction. Common polarization states include:

Linear Polarization – The electric field oscillates along a fixed direction (e.g., horizontal or vertical).

Circular Polarization – The field rotates uniformly while maintaining constant amplitude.

Elliptical Polarization – The field direction and amplitude both vary.

In optical fibers, polarization states can be altered by manufacturing defects (e.g., core ellipticity) or external stressors (e.g., bending or compression). These changes can cause:

Polarization-Dependent Loss (PDL) – Different polarization orientations incur varying attenuation in optical components (e.g., filters, couplers), leading to power fluctuations in the signal.

Polarization Mode Dispersion (PMD) – As noted above, timing differences between orthogonal polarization components lead to pulse broadening—especially problematic at higher data rates.

Why Dispersion and Polarization Are Critical for Optical Transceivers

Optical communication relies on precise transmission of digital pulses (“0”s and “1”s). Dispersion and polarization-related distortions can compromise signal integrity, raise the bit error rate, or even cause link failure.

Pulse Distortion from Dispersion

Higher bit rates (e.g., ≥ 10 Gbps) and longer fiber spans (e.g., > 10 km) exacerbate inter-symbol interference, where overlapping pulses become indistinguishable. For instance, in single-mode fiber at 1550 nm, dispersion is around 17 ps/(nm·km). If using a 1 nm linewidth laser over 100 km, dispersion-induced pulse spreading reaches ~1700 ps—far exceeding the 100 ps pulse duration of a 10 Gbps signal.

Instability Due to Polarization Effects

PDL can induce signal power swings of 1–3 dB depending on polarization, potentially lowering the signal below receiver sensitivity thresholds over long distances.

PMD becomes increasingly troublesome at speeds ≥ 40 Gbps, where even 50 km links can suffer > 50 ps of broadening. When combined with chromatic dispersion, this severely degrades signal quality.

Mitigating Dispersion and Polarization: Design and Operational Strategies

A. Design Phase Optimization

Dispersion Control

Use narrow-linewidth lasers (< 0.1 nm) to mitigate material dispersion. For long-reach systems, prefer external modulation lasers (EMLs) over directly modulated lasers (DMLs) to avoid additional chirp-induced dispersion.

Integrate dispersion-compensating elements—like chirped fiber Bragg gratings (CFBG) or dispersion-compensating fiber (DCF); e.g., a 100 Gbps long-haul module may include a CFBG delivering –1000 ps/nm of compensation.

Match fiber types to distance: for short multimode runs (< 300 m) using OM3/OM4, optimize the graded-index profile; for longer spans, use single-mode fiber (SMF), and consider dispersion-shifted or non-zero dispersion-shifted fiber (DSF or NZ-DSF) tailored to the operating wavelength.

Polarization Control

Choose components with low PDL (< 0.5 dB), ensuring polarization-insensitive performance (e.g., lenses, filters, isolators).

Implement adaptive PMD compensation circuitry in high-speed modules (≥ 100 Gbps) to correct delays between polarization components, aiming to keep PMD-induced broadening to < 10% of the pulse period (i.e., < 25 ps for 40 Gbps).

Employ polarization-diverse reception (such as with PDM-QPSK detectors) to capture both polarization states and mitigate mismatch losses.

B. Application and Maintenance Best Practices

Link Planning

For short data center connections (< 100 m), multimode fiber suffices (e.g., OM4 supports 100 Gbps to about 150 m), beyond which SMF becomes necessary. For long-haul links (> 10 km), use SMF and pre-calculate dispersion budgets, deploying compensation modules (e.g., DCF) strategically, such as every ~100 km.

Avoid sharp bending or compression—bends tighter than 30 mm in radius can increase stress and aggravate PMD. Ensure proper routing and strain relief.

Monitoring and Maintenance

Regularly assess dispersion (via OTDR or dispersion analyzers) and PMD (using PMD testers). Maintain dispersion < 1600 ps/nm for 10 Gbps links and PMD < 0.5 ps/√km for 40 Gbps links.

For ultra-high-speed (e.g., 400 Gbps) or ultra-long-haul (> 800 km) systems, deploy coherent optical communication systems with real-time digital signal processing (DSP) to compensate for both dispersion and PMD—currently the industry standard for such applications.

Conclusion

Dispersion and polarization are inherent physical challenges in optical communications that threaten transmission fidelity, speed, and reach. Through thoughtful transceiver design—including component selection and compensation strategies—and diligent application-level planning, monitoring, and maintenance, these impairments can be effectively managed to support high-speed, long-distance optical links.

The post Dispersion and Polarization in Optical Communications first appeared on FIBERSTAMP.

Three OSFP Packaging Options：

Finned-Top OSFP

Top fins improve airflow and cooling

13.00 mm height, air-cooled switches compatible

Ideal for traditional rack-mounted switches

Example: 800G OSFP SR8 VCSEL 850nm 100m MMF MPO Optical Transceiver

Close-Top OSFP

Smooth top with ventilation holes, internal heat structure

13.00 mm height, fully compatible with Finned-Top

Optimized for airflow in air-cooled systems

Example: SiPh 800G OSFP DR8 1310nm 500m SMF MPO Optical Transceiver

Flat-Top OSFP (OSFP-RHS)

Flat, no fins or openings

Slim 9.50 mm height, fits liquid-cooling plates or external heat sinks

Perfect for liquid-cooled servers, high-density GPU racks, or OEM systems

Example: 400G OSFP-RHS SR4 VCSEL 850nm 100m MMF MPO Optical Transceiver

Application Scenarios (Illustrated with NVIDIA Use Cases)

Finned Top Finned Top

Ideal for air-cooled OSFP switches, such as standard Ethernet switches relying on chassis fan airflow for cooling.

Finned Top Flat Top

Suited for hybrid cooling setups, like NVIDIA DGX H100 Cedar systems connecting to air-cooled switches. The DGX rack uses liquid cooling internally, while switches remain air-cooled—requiring a Finned-to-Flat configuration to bridge different thermal environments.

Flat Top Flat Top

Designed for fully liquid-cooled OEM systems or back-to-back adapter testing. Here, air cooling is insufficient, so modules must make direct contact with cold plates or heat sinks to ensure optimal thermal performance and reliability.

Why Thermal Design Matters

With increasing compute demands, data centers are moving from traditional air-cooling to liquid-cooling and hybrid setups. Choosing the right OSFP package ensures:

Maximum airflow utilization in air-cooled systems

Efficient heat transfer in liquid-cooled or high-density setups

Flexibility with mixed-use or future upgrades

FIBERSTAMP’s OSFP modules provide the right thermal solution for your performance needs, ensuring reliable connectivity and higher energy efficiency.

The post Choosing the Right OSFP: Balancing Performance and Thermal Innovation first appeared on FIBERSTAMP.

Historically, optical transceivers have been optimized around bandwidth and reach. But as system-level latency becomes a profit driver and performance bottleneck, particularly in nanosecond-sensitive environments, deterministic low latency is emerging as the new design frontier.

The Trade-Off of FEC: Error Resilience vs. Latency

Most 25G/100G Ethernet optical links rely on Forward Error Correction (FEC)—such as RS-FEC—to reduce the Bit Error Rate (BER) below 1E-12. While effective in protecting against signal degradation, these schemes introduce significant latency, typically between 200–250 nanoseconds per link.

This added latency is acceptable in general-purpose data networks. However, it becomes a critical limitation in real-time systems like:

• Ultra-low-latency financial trading
• AI workloads sensitive to cache and interconnect delays
• Closed-loop edge control systems and TSN environments

In these scenarios, every nanosecond matters—and FEC processing becomes a performance liability.

FIBERSTAMP’s Zero-FEC 25G SFP28 SR: Designed for Real-Time Demands

To address this bottleneck, FIBERSTAMP has developed a FEC-free 25G SFP28 SR ultra-low-latency optical transceiver, purpose-built for environments where latency is non-negotiable.

Key Highlights:

Zero-FEC Architecture

Eliminates encoding/decoding latency entirely, removing up to 250ns from the data path.

Outstanding BER Performance

Despite FEC removal, typical BER remains below 1E‑15, with <1E-12 guaranteed under standard operating conditions.

Thermal Stability for Edge and Co-location

Passes stress testing at 70°C for 30+ minutes, ensuring reliable deployment in dense, high-power environments.

Clean Signal Integrity at 25.78 Gbps

Integrated laser driver and limiting amplifier support low jitter and strong eye openings, even without DSP processing.

Plug-and-Play Compatibility: No Network Redesign Needed

FIBERSTAMP’s ultra-low-latency module is validated with mainstream switch and NIC platforms, offering full plug-and-play interoperability. This ensures fast integration without rearchitecting the entire network—an essential advantage in:

• Regulated environments like finance
• Edge deployments where engineering resources are limited
• Any use case where latency optimization must not introduce operational complexity

Why This Matters: Deterministic Latency Is Now a Business Metric

In the trading world, latency Is Currency. For AI and edge systems, latency determines feasibility. Across all industries, predictable infrastructure is quickly becoming a strategic differentiator.

By eliminating FEC overhead while maintaining signal quality, FIBERSTAMP’s zero-FEC modules are setting a new standard for optical interconnects—one that favors speed, stability, and simplicity.

Conclusion: Building the Future of Predictable Infrastructure

As system architectures reach physical limits, latency becomes the decisive variable. Bandwidth alone is no longer sufficient. The need for deterministic, low-latency, and thermally robust optical interconnects is reshaping design priorities across finance, AI, and edge computing.

FIBERSTAMP is proud to lead this shift—providing the building blocks for next-generation low-latency networks through high-performance, zero-FEC optical modules.

Explore FIBERSTAMP’s Zero-FEC Product Line →

https://www.fiberstamp.com/200g-qsfp-dd-sr8.html

https://www.fiberstamp.com/100g-qsfp28-esr4.ht

https://www.fiberstamp.com/25g-sfp28-sr.html

The post Ultra-Low-Latency Optics Are Reinventing Financial Trading Networks first appeared on FIBERSTAMP.

We believe that not all networks are based on common infrastructure purposes. The diversification of network purposes is the ecological foundation for network diversification. Precisely because of this diversification, we cannot blindly accept the theory that networks based on a 100G PAM4 underlying architecture are infallible. Further refinements to this theory are inevitable.

Generally speaking, for applications like high-frequency trading and edge networks that require prompt responses, a more stable network underlayer is necessary. This underlayer does not necessarily rely on 100G PAM4 technology, as the stability of such networks is best achieved on the signal foundation of NRZ and 50G PAM4. If our goal is to achieve higher signal sensitivity and eliminate noise interference, NRZ and 50G PAM4 serve as the optimal bases for network practices. NRZ has an advantage over PAM4 primarily in signal-to-noise ratio (SNR), which is the foundation for eliminating DSP. As we all know, DSP introduces power consumption and significant network latency. Although both 50G PAM4 and 100G PAM4 belong to the PAM4 technology paradigm, their impacts on networks are strikingly different, with vastly different system SNR and power consumption implications. Technologies similar to LPO, which are currently trusted, can indeed form the basis for the next-generation network architecture akin to 50G PAM4. Attempts at this network have actually given us glimpses of the facts.

For high-frequency trading and edge networks, FIBERSTAMP offers two evolutionary blueprints for technological practices. One is a network built on 400G QSFP-DD PSM8 optical transceivers, and the other is a network built on 400G QSFP-DD 2×FR4. These two networks can fulfill the purposes of high-frequency trading centers and edge networks, and they possess long-term effectiveness. If people must evolve towards 800G, we can still form a network architecture based on 16×50G PAM4. However, we also believe that both 50G PAM4 and 100G PAM4-based linear LPO and CPO technologies are goals for this network. The main dividing point remains: you must create a sufficiently stable, low-latency, and reliable foundation for this network. There is no need to pursue industry hotspots at the expense of blurring the practical functions of the network.

It is now believed that NV provides the most advanced network hardware for data center computing power, such as pluggable networks based on 200G SERDES/1.6T. However, this network faces challenges such as thermal power consumption and signal integrity. The industry has various technical concepts for improving this network, which are currently being tested and deployed. However, technologies like 3nm DSP, LPO, LRO, or immersion liquid cooling all require a certain amount of time for trial and error. Therefore, generally speaking, CPO may be a better solution. However, the premature definition of CPO makes this vision difficult to achieve in the short term. It is evident that NPO technology is the best practical blueprint for CPO implementation. Personally, I believe that data centers for computing power can continue to innovate along the path of 100G PAM4 without getting caught up in extending or evolving single-channel baud rates to 200G or 400G. Increasing baud rates reduces the system’s SNR and thermoelectric stability. I believe there are limits to deploying higher baud rate networks, and this model can be scientifically refuted, requiring reconsideration. Mainly, people must dispel the misconception that higher baud rates equate to technological advancement.

FIBERSTAMP has a rich product line for AI&DC 800G computing power networks. We are currently establishing an innovative and differentiated product line for computing power networks starting from 800G and 1.6T. Products like FIBERSTAMP’s 800G CWDM8/LR8 will serve as a supplementary architecture for 800G networks. For 1.6T, GIGALIGHT will begin a new product line layout and will release it by Q4 2025 at the latest.

The stagnation in 100G PAM4 technology research and development has led to difficulties in the innovation and evolution of metropolitan telecommunication networks. It is hard to believe that vast telecommunication networks in different locations can be built on the current underlying foundation of 100G PAM4 Ethernet. Future telecommunication networks must be based on three assumptions: first, building an infrastructure based on space wavelength division optics and a 50G network underlayer; second, developing a new technological platform for the next generation of 100G PAM4 DSP that is suitable for telecommunication networks; and third, establishing the next generation of high-speed networks entirely on the descent of various wavelength coherent technologies. These network architectures are not mutually exclusive, and each network can accommodate these three technologies. The main argument is that at different network nodes, corresponding technologies based on correct understanding are needed. However, this “correct understanding” must be based on the talent, preferences, and understanding of network rights of the network architects.

Based on the above understanding, FIBERSTAMP will build the next generation of telecommunication network products within its own technological cognition. Indeed, this is a historical period dominated by computing power networks. However, people should not forget that it is the innovation in telecommunication networks that brings about the foundation for economic and industrial inclusiveness.

As for the most popular data center networks currently, since their concept is entirely focused on the general purpose of “data storage and retrieval,” this purpose is primarily based on simplicity and low cost. Therefore, their evolution is mainly driven by the interest in technological discoveries, the advanced deployment goals of the company’s future business, and assessments of its own economic strength and capabilities. Since the increase in general network baud rates cannot bring any additional benefits, we previously believed that 200G networks represented the optimal balance between cost and technology, and we still believe this today. Investments in 400G, 800G, and 1.6T general networks are mainly aimed at establishing a hierarchical structure within the network. If not driven by the demand for a hierarchical structure, the deployment of higher baud rate networks is solely motivated by the advancement of silicon photonics technology.

In the field of general data center architectures, FIBERSTAMP has a comprehensive product line of silicon photonics, as well as a rich product line based on copper cable interconnection, VCSEL-based AOC, and EML technology. Due to the solidification of architectures, these products mainly compete on cost in the industry.

The post The Basic Dialogue Between Fiberstamp And Network Architecture first appeared on FIBERSTAMP.

Figure 1 – The relationship between the transmission loss and optical band

    The O-band wavelength ranges from 1260nm to 1360nm. The signal distortion caused by light dispersion in this band is minimal and the loss is low, which is an early optical communication band. Hence the name O-band (O-band), where O means “Orignal”.

O-band In Perfect Availability For 5G Front-hual

    The unique advantage of the O-band is that it is in the low dispersion region, and under NRZ modulation, as the rate increases, the sensitivity to dispersion will also increase. “The advantages of O-band, low dispersion and low cost, can provide guarantees for 5G front-hual transmission performance.”

    Currently, the O-band LWDM applied to 5G front-haul adopts a mixed rate WDM system of 25Gb/s and 10Gb/s. 25Gb/s request O-band DWDM, 12-channel BiDi, 800GHz interval, using DML+PIN+TEC; The 10Gb/s request CWDM, 12-channel BiDi, 20nm interval, using DML+PIN. In terms of standardization, 25Gb/s LWDM meets the ITU-T standard definition of DWDM.

Figure 2 – Hybrid transmission with 25Gb/s and 10Gb/s

O-BAND DWDM Non-coherent Transmission Solution

    Fiberstamp’s 100G PAM4 DWDM O-BAND solution is designed for customers who require multiple 100G connections over medium length distances. 100G QSFP28 DWDM1 O-BAND Silicon optical series adopts 150Ghz O-BAND DWDM Grid, adopts single-wave 100G PAM4 silicon optical modulation technology platform, adopts Duplex LC interface, provides 16 wavelengths optional. The total bandwidth of 1600G is realized on the dual-fiber single-mode fiber, which meets the 5G fron-haul transmission and DCI interconnection, and 30km transmission can be achieved under the support of SOA optical amplifier (10km transmission without external SOA), without the need for dispersion compensation DCM module. The power consumption less than 3.5W. Compared with coherent DWDM, this scheme has the advantages of low cost, low delay, low power consumption and easy deployment.

Figure 3 – 1600G Transmission Solution with O-BAND 100G QSFP28 PSM DWDM1

Color X 100G QSFP28 DWDM1 O-BAND 10km

• 100G PAM4 silicon optical MZ modulation technology
• The electrical port side adopts 4X25G NRZ with built-in FEC-KP4
• DWDM O-BAND 150GHZ, 16 channels
• can be transmitted for 10km
• with an external SOA, it can transmit 30km in a single span segment
• Power consumption is expected to be less than 3.5W
• for 5G pretransmission and DCI interconnection

    Fiberstamp’s 100G QSFP28 PSM DWDM4 for Non-Coherent long-distance Subsystem. These modules, built upon a 4x25G NRZ quad-carrier DWDM solution, offer options in C-band and O-band spectra, effectively spanning distances exceeding 80km. Here we will introduce the 100G QSFP28 PSM DWDM4 O-band optical transceivers.

   In a dual-fiber configuration, the module delivers a total network bandwidth of 400G. It can be directly inserted into a 100G QSFP28 switch port, eliminating the need for traditional DWDM optical layer equipment and dispersion compensation modules.

Figure 4 – 400G Transmission Solution with O-BAND 100G QSFP28 PSM DWDM4

Color ZR+ 100G QSFP28 PSM DWDM4 O-BAND 

• 4x 26Gbps NRZ per channel O-band 150GHz DML DWDM transmitter 16CH wavelengths and PIN photo detector array receiver
• Supports 4 channels of bidirectional 100G or unidirectional 100G business access.
• Offers multiple customer-side business interfaces: 100G Base-SR4/CWDM4/LR4/PSM4.
• Line side supports 2 channels of coherent CFP2 optical modules at 200G.
• Enables DWDM transmission and wavelength conversion.
• Utilizes SNMP-based unified network management platform.
• Features CDR function for optimizing output and monitoring DDM signals.
• Supports software-based port shutdown.
• Up to 80km reach for G.652 SMF with external Mux/Demux,daul SOA
• Up to 10km reach for G.652 SMF without external SOA
• Low power consumption <5W
• Single male MPO-12(APC 8-degree) connector receptacle

O-BAND Next-generation Coherent Communication Solution

    The specialist of transmission technology put forward a hypothesis that “cloud and hyperscaler have promoted the PAM4 industry in the past ten years, and the AI cluster may incubate the coherent 20km scene in the next ten years”. There is also an articel about coherent from Marvel, it says that AI has higher bandwidth requirements than traditional cloud servers, and as data centers’ demand for bandwidth grows, they will need to scale data centers in multiple dimensions, including building larger data centers and adding longer internal connections. The O-Band Coherent has been in discussion for years but has not yet to be deployed commercially on a large scale.

   The O-band Coherent technology platform uses low-dispersion O-band wavelengths, so there is no need to design complex dispersion compensation algorithms like C-band. However, for the short distance typical interconnection of 500m to 10km in the data center, it is more economical to build optical networks using O-band. Fixed wavelength lasers operate at around 1310nm, providing a significant advantage in terms of cost, as dispersion is not a major issue. Since the O-band does not need to mitigate CD and DGD, the complexity of optical DSP can be simplified. It will reduce the technical complexity of DSP and reduce the power consumption and latency. Using coherence technology in combination with O-band optics can extend transmission distance, increase bandwidth, and reduce the cost of optical transceivers. O-band may be also able to provide the performance required to replace standard coherent technologies in campus environments to save energy.

    In the next few years, as the demand for AI rises, it may gradually enter the market. Coherent technology is traditionally used to connect geographically distant data centers, but for the typical interconnection distance of 500m to 10km in a data center, it is more economical to build an optical network using O-Band, which can extend transmission distance, improve bandwidth, and reduce the cost of optical transceivers.

Figure 4 – 400G/800G Coherent Optical Transceivers

O-Band Coherent: An Idea Whose Time Is (Nearly) Here

    Over the past 20 years, the data rate of optical technology has increased by a factor of 1,000 while the power consumption per bit has decreased by a factor of 100, an amazing trajectory that has in many ways paved the way for cloud computing, mobile Internet and streaming media.

    AI represents the next turning point in bandwidth demand. Servers powered by AI accelerators and Gpus have greater bandwidth requirements than typical cloud servers: just seven high-end Gpus are enough to exhaust switches that can typically handle 500 cloud dual-processor servers. Equally important, the demand for AI services, as well as higher value AI services such as medical imaging or predictive maintenance, will further drive the need for more bandwidth. The AI market is expected to reach $407 billion by 2027.

    O-band coherent or coherent streamlining – a technology that has been discussed at conferences for years but has yet to be deployed commercially in a meaningful way – may begin to penetrate the market in the coming years, helping cloud service providers address some of these challenges.

    Coherent technology has traditionally been used to connect geographically distant data centers with connections spanning more than 100km. In terms of optics, DWDM technology using tunable lasers in the C-band is a must for long-distance transmission to save fiber costs and maintain performance. When operating at C-band, coherent DSP are needed to mitigate optical obstacles such as dispersion (CD) and differential group delay (DGD). Essentially, the complexity of coherent DSPS and the need to use tunable lasers and complex coherent light receivers drive up costs, making them unsuitable for large-scale deployment.

    However, for the short typical interconnections of 500m to 10km in the data center, it is more economical to build optical networks using O-band. Fixed wavelength lasers operate at around 1310nm, providing a significant advantage in terms of cost, as dispersion is not a major issue. Since the O-band does not need to mitigate CD and DGD, the complexity of optical DSPS can be simplified. Using coherence technology in combination with O-band optics can extend transmission distance, increase bandwidth, and reduce the cost of optical modules. O-band may also be able to provide the performance required to replace standard coherent technologies in campus environments to save energy.

Figure 4 – Coherent DSP Technology

    Nothing stands still. Ten plus years ago, the industry standard was NRZ and PAM4 was an experimental idea. Now, PAM4 is pervasive and people have begun to build out the technology and use cases for O-band coherent. While we anticipate PAM4 technology will continue to evolve, the changing needs of data centers will create openings for O-band coherent. Although it’s difficult to predict how extensively O-band coherent will be used, we can predict that data centers will get larger and that bandwidth will continue to climb. Thus, even if it gets deployed in limited use cases at first, the potential opportunities will expand.

The post Why To Use O-BAND In Optical Communication? first appeared on FIBERSTAMP.

With the rapid development of high computing power applications such as AI and machine learning, diversified businesses such as autonomous vehicles, big data streaming, and ChatGPT emerge one after another. As an infrastructure to support applications such as AI and machine learning, compared with CPU (central processing unit) computing clusters, GPU (Graphics Processing Unit) computing clusters reuquire higher performance for network transmission. The new requirements of the computing center require new network equipment to match the development of computing power, and these networks are either based on low latency and lossless high-speed IB networks, or very large high performance Ethernet architectures. The network bandwidth has increased to the 400G and 800G switch architectures based on the PAM4-112G SerDes (serial/detunneling serializer). Currently, FIBERSTAMP has introduced a full range of high-performance 400G and 800G silicon optical transceivers to the AI computing network market to meet the needs of NVIDIA IB networks. NVIDIA, as the industry leader in 2024, has launched the PAM4-224G SERDES network card and switch.

Figure 1 – Peak Computing Power and Interconnect Broadband Trends

Figure 2 – Trends of SERDES electrical interface

Silicon photonic data centers are mainly based on silicon optical data center switches and silicon optical transceivers.

Compared with other technologies, silicon optical chips have the following advantages: good performance of signal integrity SI transmission, high reliability, excellent performance of optical waveguide transmission, high integration, low cost, long-term technological evolution and etc.

1.Good Transmission Performance, High Reliability, Low Power Operation

The modulation mode of silicon optical technology module determines its better ability to overcome SI signal fading. The 100G PAM4 signal transmits high-speed electrical signals based on PCB (Printed Circuit Board), which is a great challenge for SI, and silicon light can better overcome on-board noise because of its modulation method, and can tolerate higher operating temperatures because of its temperature insensitive characteristics. Based on silicon optical platform and linear direct drive mode such as LPO/TRO/CPO, the system power consumption can be greatly reduced. This also reduces the total input cost. CPO (co-package optic) is a new unpluggable optical technology route to on-chip integrated package, which is highly challenging, but it can improve the total bandwidth capacity of exchange and transmission under the premise of the same space size. Because its main technical policy is to keep the photoelectric chip close, it can greatly reduce the energy demand and signal loss. At the same time, the volume of the switch and server is also reduced, and the space saved can further carry out heat dissipation treatment such as liquid cooling of the optical module, thus further reducing power consumption. At present, there are samples of 1.6T pluggable optical transceivers, and it is difficult to pluggable in terms of power consumption, crosstalk and optical modulator size with 3.2T. It is believed in optical industry that 1.6Tbps is the watershed of CPO and pluggable optical modules, and CPO technology and products are ready to be developed.

Figure 3 – 51.2T system power reduction from pluggable to CPO

Cisco: Thanks to CPO I/O innovations, product power consumption can be reduced by up to 50%, resulting in a total fixed system power reduction of up to 25% to 30%.

2.Good Waveguide Characteristics, High Integration and Low Cost Advantages

Silicon is transparent to the 1.1-1.6μm communication band, has excellent waveguide transmission characteristics, and high refractive index can form a large refractive index difference. Silicon photonics technology has higher integration and more embedded functions, which can improve the integration of the chip. The computing power center is generally parallel computation, which requires higher baud rate and more channels. Silicon photonic chips have the advantage of large-scale multi-channel integration, which greatly improves the feasibility of silicon optical technology to the next generation of technologies such as optical I/O. This is unmatched by wearing traditional techniques. Because the manufacturing process of silicon optical chips does not require a low-nanometer process, the cost of large-scale production is objectively reduced.

Figure 4 – Silicon photonic chips

In March 2024, Marvell showed the industry’s first 3.2Tb/s silicon optical CPO core product prototype, based on silicon optical and 3D photoelectric hybrid packaging technology, to achieve 32×200Gb/s photoelectric chip performance for AI cluster and cloud data center network switch interconnection.

3.Long-term Technology Evolution

The AI computing power center needs a greater computing power breakthrough, and the GPU computing power cluster is formed through networking, and the computing power cluster needs more optical interconnection components. AI network puts forward higher requirements for optical modules: higher bandwidth, low power consumption and high reliability of the product, high performance and low error code to meet the requirements of lossless network, improve KP-FEC margin, reduce KP-FEC error correction time, and achieve low latency.

Figure 5 – Requirements for optical transceivers in AI scenarios

In the future generation of 1.6T networks, silicon photonics technology shows strong toughness, and only silicon photonics technology can complete the evolution from pluggable optical modules to next-generation networks. A relatively advanced NPO (Near-packaged optics) /CPO (Co-packaged optics) ecology based on silicon photonics has begun to take shape. In the future, CPO technology will become the mainstream business scheme. In OFC 2024, industry leader MARVELL demonstrated TRO DSP technology, and Broadcom showed Bailly, the industry’s first 51.2Tbps CPO silicon optical Ethernet switch with a single optical engine speed of 6.4T. In addition, we also see that many on-chip optical interconnection technologies represented by CPO are expanding to a wider range of optical I/O, and the future is expected to achieve GPU-to-GPU interconnection within the server and expand new application scenarios of optical interconnection.

Figure 6 – Optical transceivers package evolution

The post Why to Choose Silicon Photonics in Data Center? first appeared on FIBERSTAMP.

Fat-Tree Data Center Architecture

With the wide application of AI large model training in various industries, traditional networks cannot meet the bandwidth and delay requirements of large model cluster training. Large-model distributed training requires communication between Gpus, and its traffic patterns are different from traditional cloud computing, which increases east-west traffic in AI/ML data centers. Short and high volumes of AI data lead to reduced network latency and training performance in traditional network architectures. Therefore, in order to meet the short-term and high-volume data processing needs, the emergence of Fat-Tree network is inevitable.

In the traditional tree network topology, the bandwidth is aggregated layer by layer, and the network bandwidth at the bottom of the tree is much smaller than the total bandwidth of all leaf nodes. In contrast, a Fat-Tree looks like a real tree, with thicker branches near the roots. As a result, the network bandwidth gradually increases from leaf to root, improving network efficiency and speeding up the training process. This is the basic premise of the Fat-Tree architecture to enable a non-blocking network.

Data Center Network Evolution

As the complexity of data center applications continues to increase, so does the need for network speed. From 1G, 10G and 25G in the past to 100G in widespread use today, the rate of data center network upgrading and evolution is accelerating. However, in the face of large-scale AI workloads, 400G and 800G transfer rates have become the next key process in the evolution of data center networks.

AI data centers drive the development of 400G/800G optical transceivers.

Large-scale data processing requirements

The training and reasoning of AI algorithms require large data sets, so data centers must be able to efficiently handle the transfer of large amounts of data. The advent of 800G optical modules, which provide greater bandwidth, helps solve this problem. An upgraded data center network architecture typically consists of two tiers, extending from switches to servers, with 400G as the underlying layer. Therefore, upgrading to 800G will also drive demand growth for 400G.

Real-time demand

In some AI application scenarios, the need for real-time data processing is critical. For example, in autonomous driving systems, the massive amounts of data generated by sensors need to be transmitted and processed quickly, and optimizing system latency becomes a key factor in ensuring timely response. The introduction of high-speed optical modules quickly meets these real-time requirements by reducing the latency of data transmission and processing, thereby improving the responsiveness of the system.

Multitasking concurrency

Modern AI data centers often need to handle multiple tasks simultaneously, including activities such as image recognition and natural language processing. Support for this multitasking workload can be enhanced with high-speed 800G/400G optical modules.

The 400G/800G optical module market has broad prospects

At present, the demand for 400G and 800G optical modules has not yet seen significant growth, but it is expected to increase significantly in 2024, driven by the growing demand for AI computing. According to Dell’Oro, the demand for 400G optical modules will increase in 2024. The increasing demand for high rate data transmission driven by AI, big data and cloud computing is expected to accelerate the growth of the 800G optical module market. This trend highlights the bright future of the 800G/400G optical module market, which will gradually increase in response to the changing needs of advanced computing applications.

400G/800G optical module solution for typical data centers

This diagram shows the solution for upgrading to 800G data centers. The QDD-FR4-400G optical module forms a high-bandwidth link between the MSN4410-WS2FC switch in the backbone layer and the high-performance 800G switch in the core layer, and runs at a 400G interface rate.

Because these optical modules are packaged in a high-density QSFP-DD package, they can be deployed in high-density configurations. This increases transmission capacity and provides greater bandwidth rates. In addition, by employing PAM4 modulation and retiming technologies, these optical modules achieve faster data transfer rates while significantly reducing latency and improving overall system performance.

800G/400G optical module new era

With the growing demand for faster and more efficient data transmission, the era of 800G/400G optical modules has fully arrived. Favored for their outstanding bandwidth capabilities, advances in LPO technology, and economic benefits, these optical modules are expected to transform the AI field and redefine the data center. Using high-speed optical modules, fully developing and training AI is no longer just an idea.

The post AI data center network architecture requirements: 400/800G optical Transceivers first appeared on FIBERSTAMP.