Stackup and routing decide signal integrity in a multilayer PCB

Signal integrity describes whether a signal reaches its receiver in a form the circuit can read correctly. In a multilayer PCB, a signal must arrive with the intended timing and amplitude, and its waveform must stay free of obvious overshoot, undershoot, ringing, or edge degradation. When impedance, the reference plane, or the dielectric material along the path is wrong, the waveform at the receiver drifts from what was intended, and signal integrity suffers.

When signal integrity fails in a multilayer PCB, the usual signs are a closed eye, a rising bit error rate, and shrinking timing margin. As data rates climb, signal edges shrink to a few hundred picoseconds, and effects that were once negligible, trace inductance, via stubs, and dielectric loss, start to matter. At best they narrow the setup and hold window; at worst the receiver reads a high level as a low one, or the other way around. These problems often surface only during prototype debugging, when a board change is expensive, so they need to be handled by rule during design.

Signal integrity in a multilayer PCB comes from the combined effect of stackup, impedance, return path, and material properties, and it is hard to trace back to any single trace. The stackup fixes which reference plane sits next to each signal layer, dielectric thickness and constant then set impedance, and routing governs crosstalk and loss. This article follows that chain and explains how each variable affects signal quality and what can be done about it in practice.

Where signal integrity problems in a multilayer PCB come from

Engineers group signal integrity problems in a multilayer PCB into a few recognizable categories, each with a clear cause. Reflection comes from an impedance discontinuity; a change in trace width, a via, a connector, or poor termination sends part of the signal energy back toward the source, where it adds to overshoot and ringing at the receiver. Impedance mismatch means the trace characteristic impedance differs from the source or load, which besides reflection lowers the transfer rate and speeds up attenuation, especially on high-speed differential pairs.

Crosstalk is electromagnetic coupling between adjacent traces, either capacitive or inductive, and the current direction in the aggressor line plus the edge rate of the signal both change the coupling strength. A broken return path is easy to overlook; once the ground or power plane has a gap, return current is forced to detour, bringing noise and electromagnetic interference. Transmission loss covers both dielectric absorption and conductor loss, and it grows with frequency. High-speed signals generally fall in the 50 MHz to 3 GHz range, and once frequencies reach into it, all of these effects act on the traces of a multilayer PCB together.

How impedance control in a multilayer PCB maps to stackup parameters

Impedance control starts with a target. Single-ended traces are usually set to 50 Ω or 75 Ω, differential traces commonly to 90 Ω or 100 Ω, and the differential termination resistor is typically 100 Ω to match the line. Once the target is set, stackup parameters are what achieve it: the distance from trace to reference plane, trace width, copper thickness, dielectric thickness, and dielectric constant together determine characteristic impedance.

The common microstrip approximation is Z0 = 87 / sqrt(εr + 1.41) × ln(5.98H / (0.8W + T)), and the stripline approximation is Z0 = 60 / sqrt(εr) × ln(4H / (0.67 × (0.8W + T))), where H is the distance from trace to reference plane, W is trace width, and T is copper thickness. The formulas show the direction in which each parameter pushes impedance, while an impedance calculator resolves the actual value against the board material data.

Tolerance is an easy part of impedance control to underestimate. Fabrication commonly quotes ±10%, which for a 50 Ω trace means the real value can land anywhere between 45 Ω and 55 Ω. In a channel running above 10 Gbps, that much deviation is enough to close the eye. High-speed links therefore ask the fabricator to hold tolerance to ±5% or even ±3%, at the cost of higher build expense. The impedance rule has to match the actual board material and lamination structure; setting a target in software without confirming stackup parameters with the fabricator leaves the rule as nothing more than a number on a page.

The choice of trace structure also serves impedance and electromagnetic compatibility. A microstrip runs on an outer layer with one reference plane beside it. It is simple, easy to fabricate and probe, but part of its field sits in air, so radiation is relatively high. A stripline runs on an inner layer with a reference plane above and below, which keeps the field inside the dielectric and gives better shielding and isolation, at the price of somewhat more complex fabrication and slightly slower propagation. High-speed signals sensitive to interference go on striplines first, while signals that need frequent rework tend to use microstrips.

How reference planes and return paths set crosstalk in a multilayer PCB

Crosstalk follows rules that can be stated plainly. Far-end crosstalk is usually larger than near-end crosstalk, and coupling grows with the aggressor frequency and edge rate. Engineers treat 100 MHz as a threshold where suppression should begin; even at lower frequencies, a device with very sharp edges needs careful routing. Trace spacing and parallel run length are the two most direct variables. Moving the center-to-center spacing of two traces from 5 mil to 10 mil reduces crosstalk noticeably, while extending the parallel run from 1.3 inches to 2.6 inches increases it sharply.

Stripline crosstalk is lower than microstrip crosstalk, and for the same routing structure, doubling the dielectric thickness also raises crosstalk markedly. The corresponding practices are to open up trace spacing, shorten parallel runs, route adjacent signal layers orthogonally, and follow the 3W rule of keeping spacing at least three times the trace width.

The return path is the common factor behind both crosstalk and electromagnetic interference. Return current at high frequency follows the path of least inductance, concentrating on the reference plane directly beneath the signal trace; the least-resistance path familiar from low-frequency circuits does not apply here. When the reference plane stays intact, the return loop stays small and radiation drops.

Once a signal crosses a split in the reference plane, the return current detours, the loop grows, and both crosstalk and radiation rise. When a signal changes layer, its return must change too, so a ground via is placed within roughly 0.5 mm of the layer-change via to give the return current a path. Where a split truly cannot be avoided, a stitching capacitor of 100 pF to 1 nF can carry the high-frequency return.

multilayer PCB

Where transmission loss in a multilayer PCB comes from

Conductor loss is largely a skin effect. As frequency rises, current crowds into a thin layer at the copper surface, and the skin depth at 1 GHz is about 2 μm, shrinking further as frequency climbs. Once copper surface roughness approaches or exceeds that depth, the current path lengthens and loss increases. In comparative tests, low-roughness copper foil improved insertion loss by roughly 9% at 18 GHz over conventional foil, with surface roughness the main difference. High-speed signal layers therefore specify low-profile foil to cut conductor loss at the source.

Dielectric loss depends on the dielectric constant and loss tangent of the board material. Ordinary FR-4 has a dielectric constant around 4.2 to 4.8 and a loss tangent near 0.02, while polytetrafluoroethylene based high-frequency laminates can go as low as 2.2 to 3.0 in dielectric constant and reach a loss tangent on the order of 0.0037. A lower and more stable dielectric constant means faster propagation and less absorption.

Solder mask on outer layers also adds loss, because its dielectric constant is higher than air; once the mask covers a trace, field lines that used to pass through air pass through the mask instead, raising insertion loss. High-speed traces therefore favor low-loss mask inks with controlled thickness. Material choice should match signal rate, using low-loss stock only on the layers that need it and conventional material elsewhere, which balances loss against cost.

How termination and layer changes affect signal integrity in a multilayer PCB

Termination absorbs energy arriving at the end of a line so it does not reflect back to the source. Source series termination adds a resistor at the driver output so the combined impedance matches the line, which suits point-to-point links. Parallel termination places a resistor at the load matched to the line impedance; it works directly but consumes DC power.

Thevenin termination uses two resistors as a divider, drawing less current from the supply. Series RC termination inserts a capacitor to block low-frequency content, keeping the DC load off the driver. A differential pair places a matching resistor across the receiver inputs, typically 100 Ω to match the differential impedance, positioned close to the receiver pins so no stub is introduced.

Vias associated with layer changes add impedance discontinuities and extra inductance. Blind vias reduce through-hole stubs, and back-drilling can remove the remaining stub on high-frequency signals. Routing corners with two 45 degree segments or an arc rather than a right angle reduces the reflection caused by capacitance change at the corner. Termination and via treatment for critical nets should be settled with simulation rather than copied from experience.

Design checks for signal integrity in a multilayer PCB

  • At stackup stage, confirm every high-speed signal layer has an adjacent intact reference plane, and check the return vias at layer changes.
  • Confirm impedance targets together with the fabricator’s material and lamination parameters, and write ±5% or tighter into the requirement for high-speed links.
  • Route critical signals as stripline between two planes, and separate sensitive signals from noisy ones on different layers.
  • Route adjacent signal layers orthogonally, keep spacing to the 3W rule, and shorten parallel runs.
  • Move a high-speed signal that would cross a split onto a layer with an intact reference plane, or add a stitching capacitor at the crossing.
  • Select copper roughness, solder mask, and laminate loss tangent to match the signal rate, and run insertion loss simulation.
  • Choose termination by interface type, and verify 100 Ω matching and intra-pair skew on differential pairs.
  • Run time-domain reflectometry and eye diagram simulation before fabrication, and check waveforms with a fixture and oscilloscope where practical.

Common questions about signal integrity in a multilayer PCB

Why is impedance tolerance so strict in high-speed design? A conventional ±10% tolerance on a 50 Ω trace means a range from 45 Ω to 55 Ω. That matters little at low speed, but in a channel above 10 Gbps, the reflection caused by impedance deviation shows up directly in the eye. The faster the rate, the more the tolerance needs to tighten to ±5% or ±3%, with the fabricator’s process capability to back it up.

How should microstrip and stripline be chosen in a multilayer PCB? A microstrip runs on an outer layer, which makes it easy to fabricate and rework, but its radiation is relatively high. A stripline sits between two reference planes, giving better shielding and isolation with slightly slower propagation. Signals sensitive to interference go on stripline first, while signals that need frequent rework can use microstrip; both require an intact adjacent reference plane.

Are signal integrity problems solved mainly by simulation or by routing experience? By both. Crosstalk, return paths, and impedance follow clear rules that translate directly into routing constraints, and following them removes most of the risk up front. The interaction of edge rate, vias, and termination is hard to judge from experience alone, so time-domain reflectometry and eye simulation anticipate it ahead of time, and a high-bandwidth oscilloscope verifies the real prototype.

The order for controlling signal integrity in a multilayer PCB is to settle the stackup first so every high-speed signal layer has an intact reference plane, then map impedance targets, trace geometry, and dielectric thickness onto the fabricator’s real process, and finally use simulation to check termination, vias, and crosstalk. Doing these steps before routing keeps signal problems out of the prototype stage.

Scroll to Top