Dual Stripline Impedance Calculator
Calculate impedance for both signal layers of a dual stripline (two routing layers sharing one plane pair), plus the broadside crosstalk penalty when traces overlap.
Cohn offset stripline per layer · broadside odd/even couplingAPI slug dual-stripline
Formula
S. B. Cohn, “Characteristic Impedance of the Shielded-Strip Transmission Line”, IRE Trans. MTT-2, 1954 (each layer, with the off-centre fringing fitted to rftools.io’s 2-D method-of-moments solver); H. A. Wheeler, IEEE Trans. MTT-26, 1978 (thickness factor); IPC-2141A; E. Bogatin, "Signal and Power Integrity — Simplified", 3rd ed., ch. 11
| Symbol | Quantity | Your value |
|---|---|---|
| Ĉₙ | Capacitance per unit length of layer n, normalised to ε₀εr | — |
| C₀(W, b) | Exact capacitance of a zero-thickness strip centred between planes b apart (Cohn): 4K(k)/K(k′), k = tanh(πW/2b) | — |
| ΔCoff | Edge fringing that superposing the two halves misses when the trace is off centre, fitted to a 2-D method-of-moments solver (as in the asymmetric-stripline calculator) | — |
| Ct(W, h) | Capacitance of the centred strip of thickness t that a half with gap h mirrors into, its planes 2h + t apart | — |
| Ft | Wheeler’s (1978) copper-thickness factor Z(W, b, 0)/Z(W, b, t) at b = 2h + t, stated to 0.5 %; 1 when t = 0 | — |
| η₀ | Impedance of free space, µ₀·c = 376.730 Ω (Ω) | — |
| b | Plane-to-plane spacing (mm) | 800 µm |
| h₁ | Layer 1 to upper plane (mm) | 200 µm |
| h₂ | Layer 2 to lower plane (mm) | 200 µm |
| s | Separation between the two signal layers (mm) | 330 µm |
| kb | Broadside backward-crosstalk coefficient | 18.9 % |
Assumptions
6 items
- 01Each signal layer is an offset stripline: its own gap to the nearer plane, and s + t plus the other layer’s gap to the far plane, computed with the asymmetric-stripline model: Cohn’s exact centred strip, Wheeler’s thickness factor and a fitted off-centre correction, which holds to 0.75 % over W/b 0.05–1.2 and a near gap of at least 0.05 of the layer’s dielectric and is flagged outside it.
- 02The other signal layer is left out of each layer’s Z₀, which holds where the two layers route orthogonally.
- 03Layer 2’s gap is what remains, h₂ = b − h₁ − s − 2t; a stack-up that leaves h₂ ≤ 0 is refused.
- 04The broadside coupling (Zeven − Zodd)/(Zeven + Zodd) is for traces directly over each other with the pair centred between the planes: it uses b, s and t only, so unequal h₁ and h₂ do not change it. An off-centre pair couples less, so past |h₁ − h₂| = 0.2·b a warning calls the figure an upper bound, and another says it reads low once s + 2t exceeds half of b.
- 05One homogeneous dielectric, εeff = εr, quasi-static and lossless; the frequency sets only λg = c/(f·√εr).
- 06Synthesis solves the trace width for either layer’s Z₀, or layer 1’s offset, numerically within its range; the material picker sets εr to the laminate’s design Dk.
Worked example
Two routing layers between planes 0.8 mm apart
A dual stripline in FR-4 (εr 4.4) has planes 0.8 mm apart, the upper signal layer 0.2 mm below the upper plane, 0.3 mm between the two signal layers and 35 µm copper. Solve the width that gives the upper layer 50 Ω, then check the lower layer and the coupling where traces overlap.
| Step | Expression | Result |
|---|---|---|
| Width for 50 Ω on the upper layer (solved) | Z₀₁(W) = 50 Ω | 0.2113 mm |
| Lower layer to the lower plane | 0.8 − 0.2 − 0.3 − 2 × 0.035 | 230 µm |
| Upper layer impedance | Z₀(W; h₁ = 0.2, h₂ = 0.3 + 0.035 + 0.23) | 50 Ω |
| Lower layer impedance, same width | Z₀(W; h₁ = 0.23, h₂ = 0.3 + 0.035 + 0.2) | 52.4 Ω |
| Coupling where the traces overlap | (Z_even − Z_odd) / (Z_even + Z_odd) × 100; h = (0.8 − 0.3 − 2 × 0.035) / 2 | 24.9 % |
How it works
Dual stripline puts two routing layers between one pair of reference planes. It is a cost play: you get two signal layers for the price of one plane pair, which on a dense board can save two copper layers and a lamination cycle. The price you pay is coupling between the two signal layers, and it has to be managed by routing discipline rather than by geometry.
Each signal layer on its own is simply an offset stripline. Layer 1 sits from the upper plane and from the lower one; layer 2 is the mirror. Both are solved with Cohn's parallel-half construction:
Here is the layer and the other one. with is Cohn's exact capacitance of a zero-thickness strip centred between planes apart, restores the edge fringing that superposing the two halves misses when the trace is off centre (fitted to a 2-D method-of-moments solver), and is Wheeler's (1978) copper-thickness factor, 1 at , applied to each half's centred strip with its planes apart; .
The other signal layer is deliberately not treated as a reference. It is a floating conductor, not a plane, and it carries its own signal. That assumption is exactly why the two layers must route orthogonally — the moment two traces run parallel over each other, the model stops applying and you have a broadside-coupled pair instead.
Why overlap is so damaging
When two traces on adjacent signal layers do run over each other, they face each other across a thin dielectric with no plane in between. That is the tightest coupling geometry available on a PCB. Decomposing into modes: odd mode places an electric wall midway between the traces, even mode places a magnetic wall there and leaves each trace referenced only to its own plane. The backward-crosstalk coefficient follows from the split:
The pair is solved centred between the planes, each trace from its plane, so the figure does not change with . An off-centre pair couples less: past the figure is an upper bound, up to about 10 % above a 2-D field solution at that offset, and the calculator warns.
For a typical dual-stripline geometry this lands around 20%, and climbs past 50% when the two signal layers sit closer to each other than to their planes — several times worse than the edge coupling between neighbouring traces on the same layer. That is not a number you design around; it is a number you avoid by routing one layer north-south and the other east-west.
Impedance asymmetry
Unless the stack-up is symmetric, the two layers will not have the same impedance. Whichever layer sits closer to its plane is the lower-impedance one. If both need to hit the same target, use different trace widths per layer and tell the fabricator both values.
Practical tips
- Route the two layers orthogonally without exception. Enforce it with a design rule, not with reviewer discipline.
- Give each layer its own trace width if the stack-up is asymmetric, and put both widths on the fabrication drawing.
- Keep the two signal layers as far apart as the stack-up allows and each layer as close as possible to its own plane — that is the geometry that minimises coupling.
- If you must run two nets parallel across layers, make sure they are unrelated and slow. Never do it with a clock or a strobe.
- Where the budget allows a plane between the two signal layers, take it. Dual stripline is a cost compromise, not a performance choice.
- Check the reported coupling coefficient before committing to the stack-up. Above about 10%, the orthogonality rule stops being advice and becomes a hard requirement.
Common mistakes
- Routing both signal layers in the same direction. Overlapping broadside traces typically couple at around 20%, and far more when the layers are close together — no amount of spacing within a layer fixes it.
- Assuming both layers have the same impedance. They only match if the stack-up is symmetric; otherwise the layer nearer its plane is lower.
- Treating the other signal layer as a reference plane in the impedance model. It is a floating conductor carrying its own signal, and modelling it as ground gives an optimistically low impedance.
- Putting a differential pair on one dual-stripline layer and a fast single-ended clock on the other, directly above it. The clock couples into the pair as common mode, which the receiver cannot reject.
- Forgetting to add the two copper thicknesses when deriving the second layer offset from the plane spacing — a 0.07 mm error on a 0.2 mm gap is a 35% error in the wrong place.
Frequently asked questions
5 questions
What is dual stripline?
Two signal routing layers placed between the same pair of reference planes, with no plane in between them. It saves copper layers and cost compared with giving each signal layer its own plane pair.
Do the two layers really have to be orthogonal?
In practice, yes. Traces facing each other across a thin dielectric with no intervening plane are the most tightly coupled structure on a PCB — commonly around 20% backward crosstalk, and far more on tight stack-ups. Routing one layer horizontally and the other vertically reduces the coupled length at each crossing to almost nothing.
Why do my two layers show different impedance?
Because they sit at different distances from their nearer plane. The layer closer to its plane has the lower impedance. Fix it with different trace widths per layer rather than by trying to force a symmetric stack-up.
Can I run differential pairs on dual stripline?
Yes, provided the orthogonality rule holds and nothing fast runs directly over the pair on the adjacent layer. Broadside coupling from an aggressor above appears as common mode on the pair, which the receiver's rejection cannot remove.
Is dual stripline worth the trouble?
On dense, cost-sensitive designs, yes — two layers saved is real money at volume. On high-speed designs with tight crosstalk budgets, put a plane between the signal layers instead and accept the extra cost.
Shop Components
As an Amazon Associate we earn from qualifying purchases.