Direct Answer
Evaluate high-speed PCB signal integrity risk before fabrication by reviewing the stackup, material properties, via transitions, and design-for-manufacturing (DFM) rules against your target impedance, loss budget, and rise time. A structured pre-fabrication checklist catches reflection, crosstalk, and skew issues while the design is still editable, preventing costly respins and field failures.
---
Why Pre-Fabrication Signal Integrity Review Matters
Signal integrity failures are expensive to fix after fabrication. A board that passes electrical testing in simulation but fails in production often traces back to assumptions made during the layout phase—assumptions about material tolerances, copper roughness, or via behavior that the fabricator cannot compensate for without your input.
The pre-fabrication review is your last chance to align the design intent with the fabrication reality. At this stage, you can still adjust the stackup, change materials, add impedance control requirements, or modify via structures. Once Gerbers and drill files are released, changes mean new tooling, new lead time, and new cost.
The review should focus on five areas: stackup design, material selection, via transitions, DFM compliance, and test requirements. Each area maps to specific signal integrity risks that are measurable and actionable before fabrication begins.
---
Stackup Design: The Foundation of Impedance Control
The stackup determines the characteristic impedance of every trace on the board. If the stackup is wrong, no amount of routing discipline will fix the resulting reflections and signal degradation.
Dielectric Thickness and Spacing
For a microstrip trace, characteristic impedance depends on the trace width, copper thickness, and the distance to the reference plane below. For stripline, the distance to both reference planes matters. A common mistake is assuming the fabricator's standard prepreg thickness will produce the target impedance without verification.
Ask for the fabricator's impedance calculator or provide your own stackup with explicit dielectric thickness values. Specify the target impedance range (e.g., 50 Ω ± 10% for single-ended, 100 Ω ± 10% for differential) and require the fabricator to confirm they can hit that range with their available materials.
Copper Weight and Its Effect on Impedance
Copper weight affects impedance in two ways: thicker copper reduces the effective trace width for a given etch compensation, and it changes the dielectric thickness after lamination. A 1 oz (35 µm) copper trace behaves differently from a 2 oz (70 µm) trace at the same nominal width.
For high-speed signals, standardize on 1 oz or lighter copper unless current-carrying requirements demand more. If you need heavier copper for power planes, keep the signal layers on lighter copper and separate the power and signal stackup zones.
Reference Plane Continuity
Every high-speed signal trace needs a continuous reference plane directly adjacent to it. Gaps in the reference plane—caused by split planes, moats, or excessive clearance holes—force return current to detour, creating inductance and radiation.
Review the stackup to ensure each signal layer has an unbroken reference plane. If you must split a plane for different voltage domains, route high-speed signals over the solid portion and use stitching capacitors near the split to provide a return path.
---
Material Selection: Loss Tangent and Dielectric Constant
Material properties dominate high-frequency loss. Standard FR-4 has a loss tangent around 0.020 at 1 GHz, which becomes significant for traces longer than a few inches at data rates above 1 Gbps. For higher speeds, you need low-loss laminates.
Loss Tangent and Data Rate
The loss tangent determines how much energy the dielectric absorbs as the signal propagates. At 10 Gbps, a 10-inch trace on standard FR-4 can lose 6–8 dB, which may exceed your receiver's equalization capability. Low-loss materials like Megtron 6, Rogers 4000 series, or Isola I-Speed have loss tangents in the 0.002–0.010 range, reducing loss by 3–5×.
Match the material to your data rate and trace length. For short traces (< 3 inches) at moderate speeds (< 5 Gbps), FR-4 is usually sufficient. For long traces or high data rates, specify a low-loss laminate and verify the fabricator stocks it.
Dielectric Constant Tolerance
The dielectric constant (Dk) determines the propagation velocity and, together with trace geometry, the characteristic impedance. FR-4 has a Dk that varies from 4.2 to 4.8 depending on resin content and frequency. This variation shifts the impedance and the timing of signals.
For tight timing budgets, specify a laminate with a controlled Dk tolerance. Many high-speed laminates offer Dk tolerance of ±0.05 or better, compared to ±0.2 for standard FR-4. The fabricator should provide the Dk value they use for impedance calculations, and you should verify it matches your simulation assumptions.
Copper Roughness and Skin Effect
At high frequencies, current concentrates on the surface of the copper trace. Rough copper surfaces increase the effective resistance, adding loss that scales with frequency. Standard electrodeposited (ED) copper has a roughness of 2–5 µm RMS, which becomes significant above 10 GHz.
For high-speed designs, specify low-profile or rolled copper. The fabricator can also use smoother foil on the outer layers where microstrip traces run. This is a material selection decision that must be made before fabrication, not after.
---
Via Transitions: Stubs, Anti-Pads, and Return Paths
Vias are the most common source of signal integrity problems in high-speed designs. A via transition introduces a discontinuity in the transmission line—an impedance change that causes reflections, and a stub that resonates at certain frequencies.
Via Stub Length
A via stub is the unused portion of the via barrel below the signal layer. At frequencies where the stub length is a quarter wavelength, the stub acts as an open-circuit stub, creating a notch in the insertion loss response. For a 10 Gbps signal, a stub longer than about 100 mils (2.5 mm) can cause significant loss at the fundamental frequency.
Options to mitigate via stubs include back-drilling (removing the unused barrel), using blind or buried vias, or routing on layers that minimize the stub length. Back-drilling adds cost and requires the fabricator to have the capability. Blind and buried vias increase the layer count and fabrication complexity.
Anti-Pad Design
The anti-pad is the clearance hole in the reference plane around the via barrel. Its diameter determines the impedance of the via transition. Too small an anti-pad creates excess capacitance; too large creates excess inductance.
For a 50 Ω via transition, the anti-pad diameter typically needs to be 2.5–3.5× the via drill diameter, depending on the dielectric thickness and via pad size. Use a 3D field solver or the fabricator's via impedance calculator to verify the anti-pad size for your specific stackup.
Via Return Path
The return current must transition from one reference plane to another when the signal changes layers. If the via does not have a return path via adjacent to it, the return current takes a longer path, creating inductance and common-mode noise.
Place a ground via within 100 mils (2.5 mm) of each signal via that changes reference planes. For differential pairs, place the ground via symmetrically between the two signal vias. This is a layout rule that must be checked before fabrication, as adding vias after the fact is not possible.
---
DFM Rules: Trace Width, Spacing, and Clearance
Design-for-manufacturing rules directly affect signal integrity. The fabricator's minimum trace width and spacing determine what impedance values are achievable, and the clearance rules affect crosstalk and plane integrity.
Trace Width and Impedance Achievability
The fabricator's minimum trace width sets a lower bound on the impedance you can achieve for a given dielectric thickness. For a 50 Ω microstrip with 4 mil dielectric thickness, the trace width needs to be about 8 mils. If the fabricator's minimum is 4 mils, you have room to adjust. If the minimum is 6 mils, you need to increase the dielectric thickness or accept a lower impedance.
Review the fabricator's capability table before finalizing the stackup. Ask for their minimum trace width, minimum spacing, and the impedance range they can guarantee. If your target impedance falls outside their capability, adjust the stackup or the target.
Spacing and Crosstalk
Crosstalk between adjacent traces depends on the spacing relative to the dielectric thickness. The 3W rule—spacing traces at least 3× the trace width—is a rough guideline, but the actual requirement depends on the coupling length and the signal rise time.
For high-speed differential pairs, the spacing within the pair is set by the differential impedance target. The spacing between pairs should be at least 2× the pair spacing to limit crosstalk. Review the layout to ensure these spacing rules are met, and verify the fabricator can hold the required tolerances.
Clearance and Plane Integrity
Clearance rules for vias and through-hole components affect the reference plane integrity. Large clearance holes in the ground plane create gaps that disrupt return current paths. If a via must pass through a plane, the clearance hole creates an annular gap that can be a problem if the via is near a high-speed trace.
Coordinate with the fabricator on clearance values. The minimum annular ring requirement sets the via pad size, which in turn affects the clearance hole size. A larger pad allows a smaller clearance for a given drill size, preserving more plane copper.
---
Test Requirements: Impedance Coupons and Reports
The only way to verify that the fabricated board meets your signal integrity requirements is to test it. Impedance test coupons on the production panel allow the fabricator to measure the characteristic impedance of representative traces.
Coupon Design and Placement
The coupon should replicate the stackup and trace geometry of the actual design. It should include traces for each impedance target (e.g., 50 Ω single-ended, 100 Ω differential) and be placed on the production panel in an area that sees the same processing as the product.
Specify the coupon location and the number of coupons per panel. A common practice is one coupon per panel edge, but for high-volume production, you may want coupons at multiple locations to catch process variation across the panel.
Test Method and Acceptance Criteria
The fabricator typically uses time-domain reflectometry (TDR) to measure the impedance of the coupon traces. The acceptance criteria should match your design tolerance. If you specified 50 Ω ± 10%, the measured impedance must fall within 45–55 Ω.
Ask the fabricator to provide the impedance test report with the shipment. The report should include the measured impedance for each coupon trace, the test method, and the calibration data. Review the report before accepting the boards.
Additional Tests for High-Speed Signals
For very high-speed designs, impedance testing may not be sufficient. You may need to specify additional tests such as insertion loss measurement, return loss measurement, or eye diagram testing on a representative channel. These tests require specialized equipment and add cost, but they provide direct verification of the signal integrity performance.
Discuss the test requirements with the fabricator during the RFQ stage. Not all fabricators have the capability to perform high-frequency measurements, and those that do may have different test frequencies and methods.
---
Practical Example: A 10 Gbps SerDes Channel Review
Consider a design with a 10 Gbps SerDes channel running 8 inches (200 mm) from a connector to an ASIC. The target impedance is 100 Ω differential, and the loss budget is 10 dB at 5 GHz.
Stackup Review
The initial stackup uses standard FR-4 with 4 mil prepreg between the signal layer and the ground plane. The differential pair is routed with 5 mil trace width and 8 mil spacing. The fabricator's impedance calculator shows this produces 100 Ω differential impedance, but the loss tangent of 0.020 gives an estimated loss of 8 dB over 8 inches at 5 GHz—within the budget but with no margin.
Switching to a low-loss laminate with a loss tangent of 0.008 reduces the estimated loss to 3 dB, providing 7 dB of margin. The fabricator confirms they stock the material and can process it with the same stackup.
Via Transition Review
The SerDes channel changes layers once, using a through-hole via. The via stub length is 40 mils (1 mm), which is acceptable for 10 Gbps. The anti-pad diameter is 28 mils for a 12 mil drill, which the fabricator's calculator shows produces a 100 Ω differential via transition. A ground via is placed 50 mils from the signal via pair to provide the return path.
DFM Review
The fabricator's minimum trace width is 4 mils, so the 5 mil traces are within capability. The minimum spacing is 4 mils, so the 8 mil pair spacing and 15 mil pair-to-pair spacing are acceptable. The annular ring requirement is 3 mils, so the 12 mil via with a 18 mil pad meets the requirement.
Test Requirements
The RFQ specifies one impedance coupon per panel, with traces for 100 Ω differential and 50 Ω single-ended. The acceptance criteria are 100 Ω ± 10% and 50 Ω ± 10%. The fabricator will provide a TDR test report with each shipment.
This review catches the material issue before fabrication, preventing a board that would have marginal loss performance. The via and DFM checks confirm the design is manufacturable, and the test requirements ensure the fabricated boards meet the impedance targets.
---
Common Mistakes and How to Avoid Them
Assuming Standard FR-4 Is Sufficient
Standard FR-4 works for many designs, but not for all. If your data rate is above 5 Gbps or your traces are longer than a few inches, verify the loss budget with the actual material properties. The fabricator's standard material may have a different loss tangent than the datasheet value.
Ignoring Copper Roughness
Copper roughness adds loss that is not captured in a simple loss tangent calculation. If your design is sensitive to loss, specify low-profile copper and verify the fabricator can provide it. The difference can be 1–2 dB over a long trace.
Not Verifying the Fabricator's Impedance Calculator
Fabricators use different impedance calculators with different assumptions about material properties and etch factors. If you use your own calculator and the fabricator uses theirs, the results may differ. Provide your stackup and target impedance to the fabricator and ask them to confirm their calculated values.
Forgetting the Return Path for Vias
A signal via without a nearby ground via creates a large loop area and increases inductance. This is a common mistake in dense designs where ground vias are scarce. Check every signal via that changes reference planes and ensure a ground via is within 100 mils.
Not Specifying Test Requirements
If you do not specify impedance test coupons and acceptance criteria, the fabricator may not test at all. The boards may still work, but you have no verification that the impedance meets your target. Always specify test requirements in the RFQ.
---
When to Involve the Fabricator
Involve the fabricator early in the design process, not after the layout is complete. A good fabricator can provide stackup recommendations, impedance calculator results, and DFM feedback that saves time and money.
At minimum, contact the fabricator when you have a preliminary stackup and target impedance. Ask them to confirm the stackup is manufacturable, the impedance targets are achievable, and the materials are available. This conversation should happen before you finalize the layout.
For complex designs, consider a design review with the fabricator. Many fabricators offer this service, and it can catch issues that your internal review misses. The cost of a design review is small compared to the cost of a respin.
---
Checklist for Pre-Fabrication Signal Integrity Review
Use this checklist before releasing Gerbers to fabrication:
| Check Item | Verification Method | Risk Level |
|---|---|---|
| Stackup impedance targets | Fabricator impedance calculator | High |
| Dielectric thickness tolerance | Fabricator capability table | Medium |
| Material loss tangent | Datasheet vs. fabricator stock | High |
| Copper roughness | Material specification | Medium |
| Trace width vs. minimum | DFM check | High |
| Trace spacing vs. crosstalk | Layout review | Medium |
| Reference plane continuity | Layout review | High |
| Via stub length | Via calculator or field solver | High |
| Via anti-pad size | Via calculator or field solver | Medium |
| Ground via proximity | Layout review | High |
| Impedance coupon design | RFQ specification | Medium |
| Test method and acceptance | RFQ specification | Medium |
Work through each item and document the results. If any item is uncertain, contact the fabricator for clarification before releasing the design.
---
Related Considerations for Advanced Designs
For designs with very high data rates or dense routing, additional factors come into play. The principles in this article apply to most high-speed designs, but advanced designs may require more sophisticated analysis.
For guidance on evaluating advanced PCB technology risks, including material selection for higher frequencies and tighter tolerances, see our article on How to Evaluate Advanced PCB Technology Risk from High-Speed PCB and Integrity Trends. This covers the trade-offs between different laminate families and the impact of process variations on signal integrity.
If your design uses microvias or HDI structures, the via transition analysis becomes more complex. Microvias have different stub characteristics and require different anti-pad designs. Our guide on How to Evaluate HDI Microvia Aspect Ratio Risk Before PCB Fabrication explains the specific risks and verification methods for these structures.
For a deeper understanding of how stackup choices affect signal integrity, the article on PCB Stack-Up and Signal Integrity for High-Speed Boards provides a detailed comparison of different stackup configurations and their impact on impedance, loss, and crosstalk.
When the design moves to assembly, signal integrity issues can also arise from component placement and soldering. The article on How to Evaluate SMT Assembly Risk from PCB Fabrication and Sourcing Trends covers the assembly-side risks that interact with your signal integrity design.
For designs using advanced packaging or thermal processing, the interaction between thermal stress and signal integrity becomes relevant. Our guide on How to Evaluate SMT Assembly Risk from FOPLP and Thermal Processing Trends discusses how thermal profiles affect material properties and via reliability.
---
Working with a Manufacturing Partner
A reliable manufacturing partner can significantly reduce signal integrity risk. When evaluating a fabricator, ask about their experience with high-speed designs, their impedance testing capability, and their material inventory. A fabricator that stocks the low-loss materials you need and has the test equipment to verify impedance is a valuable partner.
At Omini, we work with fabricators that specialize in high-speed PCB production. We help our clients navigate the stackup design, material selection, and test requirement process. Our role is to bridge the gap between the design intent and the fabrication reality, ensuring that the boards you receive meet your signal integrity requirements.
The pre-fabrication review is not just a formality—it is a critical step in the design process. By following the checklist and involving the fabricator early, you can catch signal integrity issues before they become costly problems. The time spent on review is small compared to the cost of a respin or a field failure.