Signal Integrity and High Speed Digital High Speed PCB Design Informational

What is the effect of fiber weave on skew in a differential pair at high data rates?

Q: How do I specify spread glass to the PCB fabricator?

Include in your fabrication notes: "Signal layers shall use spread glass weave (e.g., 2116-spread or 1078-spread) to minimize fiber weave Dk variation." The fabricator will confirm availability and any cost premium. Some fabricators offer "homogeneous glass" or "flat glass" as their term for spread weave. Always confirm the specific glass style with your fabricator.

Q: Can I simulate fiber weave effects?

Full-wave 3D simulation of fiber weave is possible but computationally expensive (modeling individual glass bundles requires very fine mesh). More practical: use a statistical model with the known Dk variation range and estimate the worst-case skew. Many SI engineers treat fiber weave as a ±X ps skew contribution in the total skew budget without detailed simulation.

Q: Does fiber weave affect single-ended signals?

For single-ended signals: fiber weave causes a small propagation delay variation between different traces. This contributes to trace-to-trace skew in parallel buses (e.g., DDR memory). For SerDes (differential): the intra-pair skew is the primary concern. For matched-length buses: inter-trace skew due to fiber weave adds to the total bus skew budget.

What is the effect of fiber weave on skew in a differential pair at high data rates? The glass fiber weave in a PCB laminate creates localized Dk variations that can cause different propagation speeds for the two traces of a differential pair, resulting in intra-pair skew: (1) Root cause: FR-4 and low-loss laminates are composed of woven glass fiber bundles embedded in resin. The glass fiber has Dk ≈ 6.0-6.5, while the resin has Dk ≈ 3.0-3.5. If one trace of a differential pair sits predominantly over a glass fiber bundle and the other sits over resin: the two traces experience different effective Dk values. The trace over glass sees higher Dk → slower propagation speed. The trace over resin sees lower Dk → faster propagation speed. This creates intra-pair skew (timing difference between the P and N traces). (2) Magnitude: the Dk variation due to fiber weave can be up to ±5-10% of the nominal Dk. Propagation speed difference: up to 1-3% between the two traces. Skew: for a 6-inch differential pair: worst case ≈ 5-15 ps. At 10 Gbps (UI = 100 ps): 10 ps skew = 10% of UI (may be noticeable). At 25 Gbps (UI = 40 ps): 10 ps skew = 25% of UI (significant). At 56 Gbps PAM4 (UI = 35.7 ps): 10 ps skew = 28% of UI (critical). (3) Mitigation: route traces at an angle to the weave: rotate the differential pair routing by 5-10° relative to the fiber weave direction. This ensures both traces average over glass and resin equally. PCB fabrication standard: many fabricators offer this as a design option. Use tighter weave glass styles: standard weave (1080, 2116): coarse, Dk variation ±8-10%. Spread glass (flat or spread fiber): 2116-spread, 1078-spread: much more uniform Dk (variation < ±2%). Ultra-flat glass (NE glass): even flatter weave, lowest Dk variation. Cost: spread glass adds 10-30% to the material cost. Use alternating glass layers: alternate the weave orientation between adjacent dielectric layers. This averages the Dk variation over the via length.

Category: Signal Integrity and High Speed Digital

Updated: April 2026

Product Tie-In: PCB Materials, Connectors, Test Equipment

Fiber Weave Skew

Fiber weave skew is one of the most subtle and easily overlooked signal integrity issues, becoming significant at data rates above 25 Gbps.

Practical Design Rules

(1) For 10 Gbps and below: fiber weave skew is typically negligible. Standard glass styles (1080, 2116) are acceptable. (2) For 25 Gbps: use spread glass or route at an angle to the weave. Maximum acceptable intra-pair skew: 5-8 ps. (3) For 56-112 Gbps: spread glass is mandatory. Maximum acceptable intra-pair skew: 2-3 ps. Some designs use NE glass (no fiber weave variation) for the most critical layers. (4) Measurement: intra-pair skew can be measured with a TDR (Time Domain Reflectometry) or by measuring the differential-to-common mode conversion (Sdc21) on a VNA. High Sdc21 at a specific frequency indicates skew-induced mode conversion.

Fiber Weave Impact

Glass Dk ≈ 6.0, Resin Dk ≈ 3.0-3.5
Dk variation: ±5-10% (standard), ±2% (spread)
Skew: 5-15 ps over 6 inches (worst case)
25 Gbps: 10 ps = 25% of UI (significant)
Fix: spread glass, angled routing, NE glass

Common Questions

Frequently Asked Questions

How do I specify spread glass to the PCB fabricator?

Include in your fabrication notes: "Signal layers shall use spread glass weave (e.g., 2116-spread or 1078-spread) to minimize fiber weave Dk variation." The fabricator will confirm availability and any cost premium. Some fabricators offer "homogeneous glass" or "flat glass" as their term for spread weave. Always confirm the specific glass style with your fabricator.

Can I simulate fiber weave effects?

Full-wave 3D simulation of fiber weave is possible but computationally expensive (modeling individual glass bundles requires very fine mesh). More practical: use a statistical model with the known Dk variation range and estimate the worst-case skew. Many SI engineers treat fiber weave as a ±X ps skew contribution in the total skew budget without detailed simulation.

Does fiber weave affect single-ended signals?

For single-ended signals: fiber weave causes a small propagation delay variation between different traces. This contributes to trace-to-trace skew in parallel buses (e.g., DDR memory). For SerDes (differential): the intra-pair skew is the primary concern. For matched-length buses: inter-trace skew due to fiber weave adds to the total bus skew budget.

Keep Reading