Loading…
Solve width, gap, or both at once for a differential target on edge-coupled microstrip, stripline or asymmetric stripline, then read the odd-mode, even-mode and common-mode impedances. The exact pair solve drives both conductors, and the landscape plots the width-gap surface so the sensitivity to each dimension is visible.
Direct answer
A 100 Ω differential pair is not two 50 Ω traces. At a practical gap the odd-mode impedance of each line is below 50 Ω, so each trace has to be narrower than an isolated 50 Ω line. On 0.2 mm FR-4 a 100 Ω edge-coupled stripline pair lands near 0.05 mm wide with a 0.14 mm gap, which is fine-line work; the same laminate at 90 Ω is nearer 0.08 mm wide with a 0.2 mm gap. So the pair model has to solve width and gap together, and the exact solve matters at those dimensions.
Calculator
Pick the pair cross-section and the target, then choose what the solver should hold: width, gap, or both at once along a pairing rule such as width = half the gap. The exact button re-solves the pair with the 2-D field solver, and the landscape plots the whole width-gap surface so the sensitivity is visible, not inferred.
Model and validity range. The calibrated coupled-line fit for edge-coupled microstrip and stripline, a harmonic-mean closed form for an offset stripline pair, and the exact odd/even 2-D Laplace solve with finite copper, soldermask and two independent Dk values. Width and gap are solved together on a one-parameter curve through the rules, and the landscape evaluates the same model cell by cell. The pair solve holds for a uniform cross-section; a BGA break-out, a via field or a connector launch needs a 3-D model.
Direction
Pairing rule
One target and two unknowns leave a choice, so the solver holds a second condition. The pairing rule is that condition: it decides which geometry on the solution curve the panel reports.
Target and geometry
Design rules (optional)
Blank derives it from the copper thickness.
Blank derives it from the copper thickness.
Blank derives it from the dielectric height.
Blank derives it from the dielectric height.
The solver never leaves these rules. When the target cannot be reached inside them it reports the closest geometry it is allowed to build and names the rule that blocked it, instead of printing a width that cannot be etched.
Material and frequency
Differential pair impedance
Odd, even and common mode from the pair geometry, with the exact 2-D solve one click away.
80.00 Ω differential at 0.066 mm / 0.166 mm
80.00 Ω differential at 0.066 mm / 0.166 mm
Closed-form coupling fit. Run the exact pair solve to capture the gap, the mask and an asymmetric cavity.
0.00% from the 80.0 Ω target.
One line driven against the other.
Both lines driven together.
Half the per-trace even-mode impedance.
Geometric mean of the odd and even impedances.
(Zeven - Zodd) / (Zeven + Zodd).
Sets the differential-mode velocity.
Sets the common-mode velocity.
Run the exact solve for the mode delay difference.
2.6 mil.
6.5 mil edge to edge.
| Design | Value |
|---|---|
| Pairing rule | width = 0.40 × gap |
| Achieved | 80.00 Ω (+0.00 Ω) |
| Width range | 0.050 – 1.600 mm |
| Gap range | 0.050 – 1.600 mm |
| Model evaluations | 5 |
| Solver | closed form |
This target needs fine-line geometry
What this does not cover
Impedance landscape
Every cell is a full forward solve of this cross-section. The white band is the set of geometries within ±2 Ω of the target, so its width is the etch margin the pair is buying. Hover a cell to read it; click or press Enter to adopt its width and gap.
Hover or use the arrow keys to read a cell; Enter adopts its geometry.
White outlines mark the cells within ±2.0 Ω of the 80.0 Ω target. Every cell is a full closed-form solve, not an interpolation.
Free, no account. An engineer reviews these inputs before the fixed fee is issued.
Results are engineering estimates, not design verification. Confirm them against the fabrication drawing, a TDR coupon or a 3-D extraction before release. Read the tool terms.
A wrong number, a rejected in-range value or a missing model? Send it to the engineer.
Every result above traces to an equation below. Read the equations and their validity limits.
Equations
Each block gives the equation the calculator evaluates, every symbol with its unit, and the range over which the model holds, with what it cannot see stated alongside. The button in the corner copies the LaTeX source.
The differential signal drives the two lines against each other, so it sees the two odd-mode impedances in series.
Odd mode is the superposition of two equal and opposite drives. Each line sees the other as a return path, so the pair presents twice the impedance of one line in that mode.
This is why a 100 ohm pair is not two 50 ohm lines. The odd-mode impedance of each line is about 50 ohm only when the coupling is negligible, which takes roughly three to four dielectric heights of separation.
Model limits
The relation is exact for a symmetric pair. A pair with unequal widths or unequal reference heights has two different odd-mode impedances, and the single 2 × Z formula no longer describes the differential mode.
When both lines are driven together, the pair behaves as one conductor with twice the width but also twice the return path.
Even mode drives both lines at the same potential. The mutual capacitance between them is not excited, so each line sees its own capacitance to the reference plane and the pair behaves more like an isolated line.
The common-mode impedance is half the per-trace even-mode impedance because the two lines are in parallel for the common-mode current. A common-mode choke or a cable shield sees this number.
Model limits
The formula assumes a symmetric pair driven exactly in phase. A real common-mode component can be driven by a skew, a ground bounce or an asymmetric driver, and the local impedance then varies along the route.
The geometric mean of the two mode impedances is the impedance a TDR sees when it looks at one line of the pair.
The geometric mean follows from the capacitance split: odd mode adds mutual capacitance, even mode subtracts it, and the isolated line is the product.
The number is useful because a TDR on a single line of a pair, with the other line terminated or floating, reports something between the odd and even values. The geometric mean is the reference.
Model limits
The relation is a definition for a symmetric pair. A TDR measurement depends on how the other line is terminated, so a measured single-ended number is not automatically this value.
One number between zero and one that describes how strongly the two lines couple, independent of the target impedance.
A pair with no coupling has equal odd and even impedances and k = 0. A pair whose gap is very small has an even impedance far above the odd one and k approaches 1.
The coefficient is useful because it separates the coupling from the impedance level. Two pairs can both be 100 ohm differential and have very different coupling, and therefore very different mode delay, crosstalk and gap sensitivity.
Model limits
The coefficient is a lumped description of a distributed pair. It does not capture the frequency dependence of the coupling, the effect of a soldermask or the local change at a break-out.
Because the two modes see different effective permittivities, a common-mode component arrives at a different time from the differential component.
The odd and even solves produce two effective permittivities and therefore two delays per millimetre. Their difference is the mode skew.
A homogeneous stripline pair has almost no mode skew. A microstrip pair, an asymmetric stripline pair or a covered pair does. The skew grows with route length, so a long backplane pair is where it matters.
Model limits
The formula is a uniform-line result. A pair that changes layer, passes a via or opens its gap at a break-out has a local mode conversion that the average delay difference does not describe.
Definition
A single trace has one impedance. A pair has two modes: odd, where the lines are driven against each other, and even, where they are driven together. The differential impedance is twice the odd-mode impedance, the common-mode impedance is half the even-mode impedance, and the single-ended impedance of one line is the geometric mean of the two.
The coupling is what makes the pair more than two traces. A tight gap lowers the odd-mode impedance and raises the even-mode impedance at the same time. A layout that sets the width from a single-ended table and then picks a gap will land on the target by accident at best.
Geometry
One differential target and two geometry variables leave a choice, and a calculator that silently picks one of them is hiding the decision. This one names it: hold the width, hold the gap, or solve both at once by walking a single curve through the plane. Width = half the gap (the usual 100 ohm geometry on FR-4), equal features, a fixed pitch, the widest permitted gap or the narrowest permitted width.
The curve matters because it is what makes the answer repeatable. Along width = 0.5 x gap, opening the gap also widens the trace, so the two effects partly cancel and the impedance moves slowly with the scale; hold the width instead and the gap alone moves the answer, quickly. The pairing rule says which of those sensitivities the design is buying.
A tight pair is not automatically better. It lowers the odd-mode impedance, raises the even-mode impedance and increases the mode delay difference. It also makes the gap harder to etch and more sensitive to the fabricator's process. Use the widest gap that meets the target and the routing density, and let the design rules bound what the solver may return.
The pair model shares its cross-section with the Impedance Calculator and its delay with the Propagation Delay. Solve the pair and the single-ended line on the same stackup so the layer change is not a surprise.
Sensitivity
A single solved pair says what impedance one geometry gives. It does not say how far that answer moves when the etcher takes a micron off the trace or the press-out raises the dielectric. The landscape answers that: width across, gap up, and a full forward solve in every cell, so the target becomes a band rather than a point.
Read the band, not the cell. A wide band means the target survives the process window. A band one cell wide means the geometry is on a cliff and the fabricator's tolerance will decide the impedance. Put dielectric height on one axis and the same plot shows how much of the answer the stackup owns, which is usually more than the trace width.
The exact landscape runs the 2-D pair solve once per cell across a pool of workers, so the picture is the same model the panel prints. The closed-form grid is instant and is what the panel draws before the exact run; clicking a cell adopts its width and gap so the readout and the inputs cannot disagree.
The same sensitivity question, one line at a time, is what a TDR coupon answers. Our SI/PI verification service work reads both, and the Controlled Impedance PCB covers how to put the tolerance window on the fabrication drawing.
Consequence
In a microstrip pair the odd and even modes see different effective permittivities. That means they travel at different speeds, so a pure common-mode component arrives at a different time from the differential component. The difference is small per millimetre and adds up over a long route.
A stripline pair is homogeneous and its mode delays are nearly equal. An asymmetric stripline pair or a covered microstrip pair is not, which is why the exact solve reports the two delays separately. Length matching on the differential mode alone does not fix a mode delay that the common-mode component sees.
For PCIe, USB4 and backplane channels, the mode delay and the reference-plane continuity are part of the same budget. Our high-speed PCB layout service, backplane and high-layer layout service and SI/PI verification service work check both, and PCB layout pricing prices the layout once the pair rules are fixed.
FAQ
Send the stackup, the interface list and the differential pair geometry inputs. An engineer re-solves them against the real layout before the fixed fee is issued.