Skip to content

Broadside-Coupled Differential Pair Calculator

Calculate differential and common-mode impedance for a broadside-coupled pair — two traces stacked on adjacent layers — for both shielded (between planes) and unshielded geometries.

Exact odd mode via virtual ground · even mode fitted to 2-D MoMAPI slug broadside-coupled-pair

Sign in to saveSave
Example
planeplaneεr 4.2W 150 µmt 35 µmd 200 µmh 365 µmh 365 µmb 1 mm

20 µm to 10 mm

Dielectric thickness between the two stacked traces

10 µm to 5 mm

Shielded geometry only — ignored when no planes are present

39.4 mil · 50 µm or more

1 oz · 5 µm or more

1 or more

Used only for the guided wavelength; the impedances do not depend on it in this model

1 MHz to 300 GHz

Formula

B. C. Wadell, "Transmission Line Design Handbook", Artech House 1991, §4.5 (broadside-coupled lines)
Zdiff=2Zodd,Zcommon=12ZevenZ_{diff} = 2 Z_{odd},\qquad Z_{common} = \tfrac{1}{2} Z_{even}
SymbolQuantityYour value
ZdiffDifferential impedance (Ω)84.8 Ω
ZoddOdd-mode impedance of one trace (Ω)42.4 Ω
WTrace width (mm)150 µm
dVertical separation between traces (mm)200 µm
bPlane-to-plane spacing (mm)1 mm
h(b − d − 2t)/2, each trace to its nearer plane (mm)365 µm

Assumptions

6 items
  1. 01Two identical traces stacked exactly over one another in one homogeneous dielectric, εeff = εr; quasi-static and lossless, with the frequency setting only λg = c/(f·√εr).
  2. 02Shielded: the pair is centred between planes b apart, each trace h = (b − d − 2t)/2 from its plane. The odd mode is exact by symmetry, each trace an offset stripline between its plane and a virtual ground d/2 away, and Zdiff = 2·Zodd; a stack-up with d + 2t ≥ b is refused.
  3. 03The even mode puts a magnetic wall at the midplane; its capacitance is the half-line value plus an excess fitted to a two-strip solver (2.2 percentage points on the coupling coefficient over W/b 0.05–0.8 and d/b 0.05–0.65), and Zcommon = Zeven/2.
  4. 04Unshielded: with no planes, each trace is a strip d/2 above the virtual ground in a homogeneous medium (Hammerstad’s form with εr), widened for copper by (t/π)(1 + ln(d/t)); the common mode has no return path and its impedance is reported as 0.
  5. 05The uncoupled Z₀ is the same trace with its partner removed, h from one plane and h + d + t from the other.
  6. 06Synthesis solves the trace width or the separation numerically within its range for the target Zdiff; the material picker sets εr to the laminate’s design Dk.

Worked example

100 Ω broadside pair between planes 1 mm apart

Two 0.15 mm traces are stacked on adjacent layers, centred between planes 1 mm apart in FR-4 (εr 4.4), with 17.5 µm copper. Solve the dielectric between them for Zdiff = 100 Ω.

StepExpressionResult
Separation for 100 Ω (solved)Zdiff(d) = 100 Ω; W = 0.15 mm, b = 1 mm0.28 mm
Each trace to its plane(1 − 0.28 − 2 × 0.0175) / 2342.5 µm
Odd-mode impedanceZ₀(W; h₁ = 0.3425, h₂ = d/2 = 0.28 / 2)50 Ω
Differential impedance2 × 50100 Ω
Common-mode impedanceZ_even / 247.4 Ω

How it works

A broadside-coupled pair stacks the two halves of a differential pair on adjacent layers, one directly above the other, instead of side by side on the same layer. It halves the routing width a pair consumes, which is why it dominates flex-circuit design and shows up on the densest rigid boards.

The analysis rests on one idea: symmetry. Drive a symmetric pair differentially and the plane exactly midway between the two traces is held at zero volts by symmetry alone. That virtual ground is an electric wall, and it turns a hard four-conductor problem into two independent single-line problems.

Shielded — between reference planes

With planes above and below, each trace in odd mode is an offset stripline: a virtual ground at d/2d/2 on one side, the real plane at h=(b−d−2t)/2h = (b - d - 2t)/2 on the other. The differential impedance is then simply twice that:

Zdiff=2ZoddZ_{diff} = 2 Z_{odd}

Even mode swaps the electric wall for a magnetic wall. No field crosses the midplane, so each trace is referenced only to its own plane — a half-stripline, whose impedance is twice the mirrored symmetric line. Common-mode impedance is half of that:

Zcommon=12ZevenZ_{common} = \tfrac{1}{2} Z_{even}

Unshielded — no planes at all

On a two-layer flex with no ground planes, the virtual ground is the only reference in the structure. Each trace becomes a strip at height d/2d/2 over a ground plane in a homogeneous dielectric, and Hammerstad's microstrip forms apply directly with εr\varepsilon_r in place of εr,eff\varepsilon_{r,eff} because there is no air–substrate boundary to average over.

Common mode in that geometry has no return path at all. Its impedance is not zero — it is undefined. Any common-mode current the pair picks up has nowhere to go except out into free space, which is why unshielded broadside pairs are efficient antennas and need common-mode chokes at both ends.

The real design risk

It is not impedance, it is registration. Differential impedance depends directly on the two traces sitting squarely above each other, and fabricators hold layer-to-layer registration far more loosely than they hold etch width. Lateral misalignment breaks the symmetry that the whole analysis depends on, and the broken symmetry converts differential signal into common mode — precisely the mode that radiates.

Practical tips

  • Ask your fabricator for the layer-to-layer registration specification before you pick trace widths, and size the traces so a worst-case shift is a small fraction of the width.
  • Broadside pairs are the natural choice on flex, where bend radius and width both matter and the two conductors bend together.
  • On unshielded flex, fit common-mode chokes at both ends. The geometry has no common-mode return and will radiate whatever imbalance it picks up.
  • Keep the trace separation smaller than the distance to the reference planes if you want strong coupling — that is what makes the pair immune to external noise.
  • Verify the model: differential impedance must be exactly twice the odd-mode impedance, and the odd mode must sit below the uncoupled single-line value.
  • Where a via transition breaks the stacking, keep the two vias as close together and as symmetric as possible — an asymmetric transition is a mode converter.

Common mistakes

  • Ignoring layer-to-layer registration tolerance. It is the dominant error source for broadside pairs and is typically far looser than etch tolerance — a misaligned pair converts differential signal into radiating common mode.
  • Assuming common-mode impedance exists on an unshielded pair. With no reference plane there is no common-mode return path at all; the quantity is undefined, not small.
  • Using edge-coupled formulas for a stacked pair. The coupling mechanism is completely different and the numbers are not transferable.
  • Placing a broadside pair on the two layers of a dual stripline without checking what else routes there. Anything running above or below the pair couples in as common mode.
  • Forgetting that both copper thicknesses eat into the plane spacing. On a thin shielded stack-up, two layers of 35 micron copper is a significant fraction of the available height.

Frequently asked questions

5 questions
When should I use a broadside pair instead of an edge-coupled pair?

When routing width is the binding constraint — dense rigid boards and almost all flex circuits. A broadside pair occupies roughly half the width of an edge-coupled pair with the same coupling, and on flex both conductors bend together rather than one taking a longer path around a bend.

What is the biggest risk with broadside pairs?

Layer-to-layer registration. The analysis assumes the two traces sit squarely above each other; any lateral shift breaks that symmetry and converts differential signal into common mode. Registration tolerance is typically much looser than etch tolerance, so it dominates the error budget.

Why is common-mode impedance undefined without reference planes?

Common mode means both conductors swing together, which requires a return path somewhere else in the system. With no plane there is no such path, so the impedance is genuinely undefined rather than zero. Any common-mode current couples into whatever is nearby and radiates.

How do I get 100 ohm differential from a broadside pair?

Target 50 ohm odd-mode impedance, since the differential impedance is exactly twice the odd-mode value. Then adjust trace width — wider traces lower the impedance — while keeping the vertical separation set by the available dielectric.

Can I use broadside pairs for USB, Ethernet, or PCIe?

Yes, and it is common on flex assemblies for exactly those interfaces. The requirements are the usual ones: hit the target differential impedance, keep the two traces stacked accurately, and manage the via transitions symmetrically at each end.

Shop Components

As an Amazon Associate we earn from qualifying purchases.

  • PCB Manufacturing (JLCPCB)Affordable PCB fabrication with controlled impedance optionsJLCPCB →
  • FR4 Copper Clad LaminateFR4 laminate sheets for custom PCB prototypingAmazon →
  • Thermal PasteThermal interface material for component heat managementAmazon →