Engineering Summary – Stackup as a Skew Control Variable
Glass weave effect is often treated as a material or routing problem. In reality, stackup design has a significant influence on how much skew reaches your differential pairs. By placing high‑speed layers adjacent to solid reference planes, controlling dielectric thickness, selecting appropriate glass styles, and managing layer symmetry, you can reduce skew sensitivity by 30‑60% before routing even begins.
Key stackup strategies UltroNiu applies for high‑speed designs:
- Use stripline instead of microstrip for critical pairs – 30‑50% skew reduction
- Tighten dielectric thickness to ≤4 mil (100 µm) for ≥25 Gbps layers
- Pair with solid, uninterrupted reference planes – never route over splits
- Maintain symmetrical stackup with copper balance within 10%
- Combine spread glass prepreg (106, 1080‑SG) with tight stackup for 60‑70% total reduction
UltroNiu stackup review service: At quoting stage, we evaluate your stackup, reference plane placement, and dielectric thickness. We recommend prepreg styles and layer assignments to minimise skew before layout begins. Free test coupon with TDR data included for qualified designs.
How to mitigate glass weave effect through stackup design – PCB manufacturer guide
1. Why stackup design matters for glass weave
Most engineers look at glass weave effect as a problem of routing angle, glass style, or material selection. These ignore a fundamental fact: the electromagnetic field of a differential pair extends into the dielectric above and below. The stackup determines how tightly the field is confined, how much dielectric variation affects propagation delay, and whether the pair sees the same weave pattern on adjacent layers. A poorly designed stackup amplifies glass weave skew. A well‑designed stackup reduces its impact – even with the same glass style.
For high‑speed PCB designs, stackup is a first‑order control variable.
2. The physics: field confinement reduces skew sensitivity
Key principle: The more tightly the field is confined to the immediate vicinity of the trace, the less it is affected by dielectric inhomogeneity far from the trace.
| Stackup Feature | Effect on Field Confinement | Impact on Glass Weave Sensitivity |
|---|---|---|
| Signal layer far from reference plane (>5 mil) | Loose field, extends deep into dielectric | High sensitivity – weave variation fully sampled |
| Signal layer close to reference plane (2‑3 mil) | Tight field, concentrated near trace | Reduced sensitivity – weave impact averaged |
| Stripline (between two planes) vs microstrip | Very tight confinement | Lowest sensitivity |
3. Stackup strategies to mitigate glass weave
3.1 Use stripline instead of microstrip for critical pairs
| Structure | Field Distribution | Glass Weave Sensitivity | When to Use |
|---|---|---|---|
| Microstrip (outer layer) | Fields extend into air and prepreg – wide, asymmetric | Highest | Only when routing density forces outer layer |
| Stripline (inner layer, between two planes) | Fields tightly contained between two copper planes | Lowest (30‑50% reduction vs microstrip) | Preferred for high‑speed pairs ≥25 Gbps |
3.2 Tighten dielectric thickness to reference plane
| Dielectric Thickness (to reference) | Relative Skew Sensitivity | Manufacturing Feasibility |
|---|---|---|
| ≤3 mil (75 µm) | Baseline (lowest) | Feasible for HDI / build‑up layers |
| 4‑6 mil (100‑150 µm) | 30‑50% higher | Typical for standard multilayer |
| >6 mil (>150 µm) | 2‑3× higher | Avoid for high‑speed pairs |
For data rates ≥25 Gbps, assign high‑speed differential pairs to layers with ≤4 mil prepreg to the reference plane.
3.3 Pair with solid, uninterrupted reference planes
| Reference Plane Type | Impact on Skew | Recommendation |
|---|---|---|
| Solid ground plane, continuous | Best – field is stable | Required for high‑speed pairs |
| Power plane (with decoupling) | Acceptable if adjacent to solid ground | Use only when necessary, keep close coupling |
| Split plane under signal | Unpredictable, can double skew | Never route differential pairs over splits |
3.4 Use symmetrical stackups to avoid warpage‑induced skew
Warpage changes the physical distance between signal traces and reference planes unevenly. For multilayer PCB designs ≥25 Gbps, maintain copper balance within 10% across layers. Avoid unbalanced heavy copper planes on one side.
3.5 Spread high‑speed pairs across multiple layers strategically
Do not put all high‑speed pairs on the same layer with the same glass orientation. Distribute across layers with different prepreg styles, routing angles, or reference plane distances to prevent a single weave alignment anomaly from affecting all channels simultaneously.
4. Material + stackup co‑design
4.1 Choose prepreg styles that match your stackup goals
| Prepreg Glass Style | Typical Thickness (per ply) | Field Confinement | Skew Potential |
|---|---|---|---|
| 106 | 2‑3 mil (50‑75 µm) | Very tight – excellent for high‑speed | Lowest |
| 1080 | 3‑4 mil (75‑100 µm) | Tight – good for 25‑56 Gbps | Low |
| 2116 | 4‑6 mil (100‑150 µm) | Moderate – acceptable for ≤25 Gbps | Medium |
| 7628 | 8‑10 mil (200‑250 µm) | Loose – avoid for high‑speed | High |
4.2 Use spread glass prepreg for critical layers
| Material + Stackup | Typical Skew Reduction vs Standard FR‑4 + Loose Stackup |
|---|---|
| Standard 1080 + 6 mil reference distance | Baseline |
| Standard 1080 + 3 mil reference distance | 30‑40% reduction |
| Spread glass 1080 + 6 mil reference distance | 40‑50% reduction |
| Spread glass 1080 + 3 mil reference distance | 60‑70% reduction |
Spread glass and tight stackup are complementary. Use both for best results.
4.3 Avoid mixing very different glass styles in adjacent signal layers
For multi‑layer high‑speed regions, standardise on one glass style and one reference distance to prevent skew mismatch between channels.
5. Practical stackup examples
5.1 Stackup for 25‑56 Gbps (8 layers, cost‑optimized)
| Layer | Type | Prepreg | Thickness to ref plane | Notes |
|---|---|---|---|---|
| 1 | Signal (microstrip) | 1080 | 3 mil | High‑speed pairs – use angled routing |
| 2 | Ground | – | – | Solid plane |
| 3 | Signal (stripline) | 1080 | 3 mil each side | Preferred for critical pairs |
| 4 | Power | core | – | – |
| 5 | Ground | core | – | – |
| 6 | Signal (stripline) | 1080 | 3 mil each side | High‑speed |
| 7 | Ground | – | – | – |
| 8 | Signal | 1080 | 3 mil | Lower priority pairs |
5.2 Stackup for 56‑112 Gbps (12 layers, performance‑optimized)
| Layer | Type | Prepreg | Thickness to ref plane | Notes |
|---|---|---|---|---|
| 1 | Ground | – | – | Shield |
| 2 | Signal (stripline) | 106 | 2 mil each side | All critical pairs |
| 3 | Ground | – | – | Solid |
| 4 | Signal (stripline) | 106 | 2 mil each side | High‑speed |
| 5 | Power | core | – | – |
| 6 | Ground | core | – | – |
| 7 | Signal | core | – | Lower speed |
| 8 | Power | core | – | – |
| 9 | Ground | – | – | – |
| 10 | Signal (stripline) | 106 | 2 mil each side | High‑speed |
| 11 | Ground | – | – | – |
| 12 | Signal | 106 | 2 mil | Optional |
6. Stackup design checklist for skew mitigation
| Check Item | Target | Verification Method |
|---|---|---|
| High‑speed layers are stripline (not microstrip) | ≥25 Gbps | Stackup drawing |
| Dielectric thickness to reference plane | ≤4 mil (100 µm) for ≥25 Gbps | Prepreg stackup calculation |
| Reference plane is solid, no splits | Always | Layout review |
| Symmetrical stackup | Copper balance within 10% | Stackup calculation |
| Same glass style across adjacent high‑speed layers | Yes | Material list |
| Spread glass used for layers ≥56 Gbps | Yes | Material spec |
| Coupon includes multiple routing angles | 0°, 5°, 10°, 45° | Panel drawing |
7. How UltroNiu supports stackup‑based skew mitigation
UltroNiu integrates stackup design and glass weave mitigation from the earliest engineering phase.
- Stackup review: We evaluate your proposed layer structure, reference plane placement, and dielectric thickness to identify skew‑sensitive configurations.
- Prepreg selection: We recommend specific glass styles (106, 1080, spread glass) based on your data rate and skew budget.
- Symmetry control: Our stackups are engineered with copper balance to minimise warpage‑induced skew variation.
- Coupon validation: For high‑speed designs, we build coupons with multiple routing angles (0°, 5°, 10°, 45°) on production panels and provide TDR skew measurement data.
- Material substitution: If your chosen material is not optimal, we propose alternatives (e.g., spread glass versions of standard laminates) that reduce skew without changing your electrical parameters.
For designs at 56G, 112G, or phase‑sensitive applications (radar, aerospace), UltroNiu’s engineering team can design a stackup that minimises glass weave effect before routing begins – not just react to skew problems after fabrication.
For Rogers PCB and special PCB designs, we apply the same stackup principles.
References
- IPC‑TM‑650 2.5.5.17 – “Glass Weave Effect on Propagation Delay”
- DesignCon 2019 – “Glass Weave Skew: Measurement and Practical Mitigation”
- Isola “Spread Glass for High‑Speed Designs” application note
- Panasonic “Megtron‑SG” stackup guidelines
This article is part of UltroNiu’s engineering library for high‑speed PCB design. For project‑specific stackup recommendations, contact our application engineers.
Related Engineering Resources
Tags:
Related Articles
Related Products
6-Layer RO4350B/RO4450F Controlled-Impedance RF PCB — 49Ω Impedance Control (RF Power Stage, Driver Boards)
• Material: Rogers RO4350B + RO4450F • Layers: 6L • Key Tech: 49Ω Impedance Control
20-Layer High-Frequency PCB — Shengyi S7136H (RO4350B Alternative) Yin-Yang Copper · Multi-Group Impedance · Controlled Depth Slot
• Material: Shengyi S7136H • Layers: 20L • Key Tech: Yin-Yang Copper · Multi-Group Impedance · Controlled Depth Slot
2-Layer RO4350B High-Frequency RF PCB
• Material: Rogers RO4350B (TG280) • Layers: 2L • Key Tech: Controlled Impedance + Resin Plugged
Related Products/Solutions
Quick links


