What this calculator does
A stripline is a flat conductor of width and thickness buried in a dielectric between two ground planes spaced apart. It is the standard inner-layer trace of a multilayer PCB. Because the dielectric surrounds the strip completely, the line carries a pure TEM wave. Its effective permittivity is simply , the delay is for any width, and the impedance depends only on the cross-section.
The calculator handles two geometries:
- Centred (symmetric) stripline, with the strip midway between the planes.
- Offset (asymmetric) stripline, with the strip at height above the lower plane and below the upper one. This is the usual case in real stack-ups, where the core and the prepreg have different thicknesses.
The outputs are , delay, capacitance and inductance per length, conductor and dielectric loss, and the cutoff of the first higher-order mode.
The models
Zero thickness, centred. Cohn's conformal mapping gives the exact result
where is the complete elliptic integral of the first kind.
Finite thickness. Wheeler's 1978 formula models a thick strip as a zero-thickness strip of increased width . Wheeler quotes 0.5 % accuracy for . On its own, Wheeler's formula at differs from Cohn's exact value by up to 0.5 %. To keep the result exact at and continuous as grows, the calculator applies Wheeler's thickness effect as a ratio: . Against a 2-D field solver, this cuts the error roughly in half compared with Wheeler alone (see the table below).
Offset strip. The image method (Wadell §3.6.3) treats the strip as two virtual centred striplines, one with plate spacing and one with , and adds their capacitances: . It is exact when the strip is centred. It always overestimates as the strip moves toward one plane, because the two halves' fringing fields interact. The calculator states how large that error is (see below).
Losses. Dielectric loss is exact for a homogeneous TEM line: . Conductor loss uses Wheeler's incremental-inductance rule, . The derivative is evaluated numerically by receding every metal surface by , so the same rule also covers the offset geometry. For the centred line it agrees with Pozar's closed form to within about 1–5 %.
Higher-order mode. The first TE mode, which has a field variation across the strip, cuts off at (Pozar §3.7). An offset stripline can also excite the parallel-plate mode between the planes at any frequency, which is why stripline cavities are via-fenced.
How to use solve-for
Click ⇄ Solve next to Strip width and enter a target impedance to design a trace. You can also solve for the plate spacing (useful when choosing a prepreg) or back out from a TDR measurement of or delay. In the offset geometry, is held fixed while solving for , so the extra thickness goes into the gap above the strip.
Worked example
RO4350B-type laminate with and . Plate spacing mm, mm, half-ounce copper (m), at 10 GHz, centred (the calculator defaults):
- Ω, delay 162.09 ps/in, pF/in and nH/in.
- Conductor loss 0.2064 dB/in and dielectric loss 0.1637 dB/in, so 0.3701 dB/in in total.
- Guided wavelength 15.670 mm. The TE-mode cutoff is 121.91 GHz, far above the operating frequency.
- Solving for exactly 50 Ω gives mm.
Now switch to Offset with mm, so the strip is 0.15 mm above the lower plane and 0.3325 mm below the upper one. drops to 44.453 Ω, and a 50 Ω line needs mm. At this position (strip centre at 32 % of from the nearer plane) the calculator notes that the image method reads about 0.5–2.5 % high, so check the final width with a field solver.
Accuracy and validity
The table compares the models with an independent 2-D finite-volume field solver (tooling/validation/laplace2d.py).
That solver agrees with Cohn's exact solution to better than 0.05 %.
| Case | Model error vs. field solver |
|---|---|
| Centred, | exact (Cohn) |
| Centred, , | −0.15 % … +0.03 % |
| Centred, –, – | −0.2 % … −0.6 % |
| Offset, strip near centre () | +0.2 % … +0.6 % |
| Offset, | +0.8 % (W/b = 2) … +2.3 % (narrow) |
| Offset, | +1.6 % (W/b = 2) … +5.5 % (narrow) |
| Offset, | +2.2 % (W/b = 2) … +11 % (W/b = 0.1) |
The calculator warns when the strip centre is within 25 % of of either plane (and notes the error band up to 40 %), and when . Conductor loss assumes smooth copper. Roughness adds 10–50 % at multi-GHz frequencies, depending on the foil.
Practical notes
- The dielectric is not uniform in a real board. Core and prepreg usually have different , and resin-rich prepreg around the trace lowers the local value. Use the stack-up's effective Dk, or a field solver for a mixed dielectric.
- Etch factor. Traces etch to a trapezoid. A common first-order fix is to enter the average of the top and bottom widths.
- Thickness matters more in thin stack-ups. At mm, going from ½ oz to 1 oz copper changes by several ohms.
- Via fences. In offset stripline and around discontinuities, stitch the two planes with vias spaced well under . Otherwise energy leaks into the parallel-plate mode.
- Loss budget. Inner-layer loss is usually dominated by the dielectric at high frequency (it grows as ), while conductor loss grows as . Low-loss laminates pay off quickly above a few GHz.