Short answer: A fiber array unit (FAU) converts a multi-core fiber or a chip-scale optical interface into a fan-out of standard single-fiber or MPO-connectorized paths. Choosing the right FAU depends on more than channel count: you must match the chip interface pitch, the fiber type and mode field, the polishing and insertion-loss budget, and the mechanical tolerances of the module. Start by defining the optical interface on the transceiver or photonic chip, then select fiber parameters and connectorization that keep loss low and the assembly stable.
What Is a Fiber Array Unit (FAU)?
A fiber array unit is a precision optical component that aligns a row or grid of optical fibers to a photonic chip, a planar waveguide, or a multi-fiber connector. It is the bridge between the sub-micron world of silicon photonics and the practical world of patch cords and MPO trunks. FAUs appear in co-packaged optics (CPO) modules, external laser sources for pluggable transceivers, fiber-optic sensors, and high-density test equipment, where a single module must fan out to dozens of individual fibers without introducing excessive loss.
Because the alignment tolerances are measured in fractions of a micron, FAU design is a combination of glass, mechanical, and connector engineering. The result is a component that looks simple but behaves like a precision instrument — which is exactly why selection criteria such as fiber parameters and chip interfaces matter as much as the number of channels.
Why Channel Count Is Only the Starting Point
Channel count tells you how many fibers the FAU carries, but it says nothing about whether the unit will actually perform in your optical system. Two FAUs with the same channel count can differ dramatically in insertion loss, return loss, crosstalk, and reliability depending on the fiber parameters, the polishing quality, and the alignment between the fiber array and the chip interface.
- Channel count defines capacity, not performance. A 16-channel FAU can still fail a link budget if the pitch mismatch or mode-field mismatch adds loss at the interface.
- Chip interfaces dictate geometry. Silicon photonics die often use a specific fiber pitch (for example 127 µm or 250 µm), which the FAU must match precisely; a mismatch produces systematic alignment loss.
- Fiber parameters determine optical behavior. Core size, numerical aperture, and mode-field diameter must be matched to the waveguide mode on the chip and to the downstream fiber plant.
- Mechanical stability decides real-world yield. Thermal cycling, vibration, and connector mating stress can shift alignment over time; a mechanically sound FAU keeps performance stable in the field.
For a broader view of how interface-level components fit into high-density systems, see our guide to optical interconnect adoption and CPO challenges and the article on OXC architecture in AI data center networks.
Matching the Chip Interface
The optical interface on the chip is the single most important input to FAU selection. Photonic integrated circuits (PICs) expose light through edge couplers or grating couplers, and each type has a different alignment strategy:
| Interface type | Typical pitch | FAU alignment strategy | Main risk |
|---|---|---|---|
| Edge coupler (silicon photonics) | 127 µm / 250 µm | Fiber array facet aligned to waveguide facet | Mode-field mismatch, sub-micron X/Y offset |
| Grating coupler | Variable (often 127 µm) | Angled polished fiber array, controlled fiber angle | Angle error, polarization sensitivity |
| Lensed fiber interface | Single or multi-fiber | Lens on chip or in ferrule; array focuses light | Working distance tolerance |
| MPO-to-chip fan-out | MPO-defined (e.g., 12F) | FAU terminates to MPO connector for external trunking | Connector mating loss and polarity |
The pitch tolerance between the fiber array and the chip determines how much of the available coupling efficiency is captured. If the FAU fiber positions drift by even one micron, coupling loss can increase noticeably, which is why FAU suppliers specify position accuracy in the datasheet. Always ask for the position error budget and verify it against your system’s total loss allocation.
Fiber Parameters That Matter in an FAU
Once the interface geometry is fixed, the fiber itself must match the optical design. The most important parameters are:
- Core type and mode-field diameter: single-mode fiber (typically G.652 or G.657) is required for silicon photonics edge coupling; multi-mode fiber is rarely used at the chip interface because the waveguide modes are single-mode.
- Numerical aperture (NA): the fiber NA should accept the divergence of the light exiting the waveguide; a fiber with too low an NA clips the beam and adds loss.
- Insertion loss (IL): the sum of fiber-to-chip coupling loss and internal FAU loss; target values depend on the link budget, often 0.5–1.0 dB per interface for production modules.
- Return loss (RL): reflections at the interface can destabilize laser sources; high RL (low reflection) is required for coherent and high-power applications.
- Polishing quality: the fiber array facet must be polished to optical quality (for example, an 8° angled polish to suppress back-reflection) with controlled surface defects.
- Bend and thermal behavior: the fiber jacket and adhesive system must survive the module’s operating temperature range without shifting alignment.
These parameters interact with the rest of the fiber plant. For example, if the FAU terminates to an MPO trunk, the connector endface must meet the same physical-contact standards described in our article on physical contact in fiber connector ferrules, and the overall link budget must follow the same logic as our fiber optic attenuation guide.
FAU Selection Checklist
| Selection step | What to verify | Typical question to ask the supplier |
|---|---|---|
| 1. Define the chip interface | Coupler type, pitch, mode field | What is the recommended FAU pitch tolerance for this PIC? |
| 2. Choose fiber type | SM vs MM, G.652 vs G.657, coating | Is the fiber bend-insensitive and matched to my wavelength? |
| 3. Set the loss budget | IL/RL per interface, channel uniformity | What are the max IL and channel-to-channel uniformity? |
| 4. Specify connectorization | MPO pin count, polarity, ferrule grade | Is the FAU connectorized to MPO or bare fiber pigtails? |
| 5. Validate reliability | Temperature range, thermal cycling, vibration | What reliability data do you have for thermal cycling? |
| 6. Plan installation | Cable management, bend radius, mating cycles | What bend radius and mating cycle limits apply? |
For the connector side of the equation, review the MPO vs MTP connector guide and the fiber polarity methods guide before finalizing the fan-out architecture.
Common FAU Mistakes and How to Avoid Them
- Ignoring pitch tolerance: ordering a standard FAU without confirming the chip’s fiber pitch creates systematic loss. Verify the datasheet against the PIC layout.
- Using the wrong fiber type: a multi-mode FAU cannot serve a single-mode chip interface. Match the core and mode field.
- Overlooking return loss: high back-reflection can detune lasers and degrade coherent links. Specify an angled polish or an index-matching solution.
- Forgetting connector standards: the fan-out connector must match the plant: MPO-12, MPO-16, or LC. See the fiber patch cord selection guide for the options.
- Skipping reliability testing: a lab-perfect FAU that drifts under temperature will fail in the field. Require thermal and vibration data.
Density decisions also interact with the FAU choice. High-density patch panels, VSFF connectivity, and port-density strategies from LC to MMC all determine how the FAU fan-out is routed and managed in the rack.
Frequently Asked Questions
What does FAU stand for in optical networking?
FAU stands for fiber array unit, a precision component that aligns multiple optical fibers to a photonic chip, a planar waveguide, or a multi-fiber connector. It is commonly used in silicon photonics modules, co-packaged optics, and fiber-optic sensing systems.
Is channel count the most important FAU specification?
No. Channel count defines capacity, but insertion loss, pitch accuracy, fiber type, return loss, and mechanical stability determine whether the FAU works in the system. A higher-channel-count FAU with poor alignment can be worse than a lower-count unit that matches the interface perfectly.
What is the difference between an FAU and a fan-out cable?
A fan-out cable typically splits a multi-fiber trunk into individual patch cords using standard connectors, while an FAU terminates a fiber array to a chip or waveguide interface with sub-micron alignment. FAUs are the chip-side precision component; fan-out cables handle the plant-side breakout.
How do I calculate the FAU loss budget?
Start from the transceiver or laser power budget, subtract the fiber plant loss, connector losses, and required margin, and the remainder is the maximum loss allowed at the chip-to-FAU interface. Compare that with the FAU datasheet IL, and remember to include temperature margin. Our attenuation and link budget guide walks through the math.
To see how these precision components fit into a complete optical solution, browse the ADTEK fiber optic product series and the pre-terminated trunk solutions page.
Fiber parameter standards for FAU design follow the IEC 60793 series (IEC), and single-mode fiber dimensions follow the ITU-T G.652/G.657 recommendations (ITU-T). The Optical Internetworking Forum provides related co-packaged optics specifications (OIF).