What this calculator does
A coplanar waveguide (CPW) puts the signal conductor and its ground on the same layer. A centre strip of width is separated from two wide coplanar grounds by gaps of width , on a substrate of height . Adding a metal plane under the substrate gives grounded CPW (GCPW, also called conductor-backed CPW). GCPW is the usual choice on RF PCBs, because the backside plane provides mechanical support, heat sinking and a DC ground, and the coplanar grounds are stitched to it with vias.
The impedance is set mainly by the ratio , so a 50 Ω CPW can be made almost any width. That is why CPW is popular for probe pads, connector launches and mm-wave circuits. The calculator gives and both quasi-statically and at your frequency, the delay, and conductor and dielectric loss.
The models
All quasi-static results come from conformal mapping (Ghione & Naldi; Simons, ch. 2). With :
-
CPW on a substrate of finite height (air below). The substrate's share of the field is set by :
-
Grounded CPW. The backside plane adds a parallel microstrip-like capacitance with :
Metal thickness widens the strip by , narrows the gap, and pulls some field into the air between the conductor sidewalls (Gupta et al., eqs. 7.98–7.100).
Dispersion follows the Frankel et al. fit, in which rises toward as approaches the substrate's TE₀ surface-wave cutoff . The impedance scales as .
Losses. Conductor loss uses Ghione's closed form, derived from the incremental-inductance rule, which captures the current crowding at the gap edges. Dielectric loss uses the filling-factor formula shared with microstrip.
How to use solve-for
Click ⇄ Solve next to Gap to coplanar ground (or Centre conductor width) and enter a target impedance. A common workflow is to fix to match a connector pin or probe pitch, then solve for . You can also solve for the substrate height or back out from a measured impedance.
Worked example
GCPW on 0.508 mm (20 mil) RO4350B-type laminate, with , , mm, mm and 1 oz copper (m), at 10 GHz (the defaults):
- Quasi-static: Ω and .
- At 10 GHz: Ω, , delay 133.02 ps/in, mm.
- Losses: conductor 0.08164 dB/in and dielectric 0.1098 dB/in, so 0.1915 dB/in in total.
- The substrate TE₀ cutoff is 90.46 GHz.
To reach exactly 50 Ω at 10 GHz, solve for the gap: mm. Without the backside metal (CPW option), the same geometry gives Ω. Here , so the backside plane dominates the capacitance, and the calculator notes that the line behaves largely like a microstrip.
Accuracy and validity
| Quantity | Source | Accuracy |
|---|---|---|
| , () | Conformal mapping | Exact for infinitely wide grounds (CPW on a half-space, ); within ≈ 1 % for finite in practical ranges |
| Thickness correction | Gupta et al. | First-order. Good while ; the calculator warns when |
| Frankel et al. fit | Derived for CPW without a backside; an estimate for GCPW | |
| Ghione, incremental inductance | Needs ; ignores roughness and backside current | |
| Filling-factor formula | Exact to the extent is |
All results were cross-checked against scikit-rf's CPW media, which implements the same papers independently
(Qucs heritage). and agree to better than , both with and without the metal
backside. While checking, we found that scikit-rf 2.2.0 passes the elliptic modulus where SciPy expects the
parameter in its CPW conductor-loss formula. That makes its about 20 % low. After that
factor is corrected, the conductor and dielectric losses agree to better than 0.5 %.
Practical notes
- Grounds must be grounds. The formulas assume the coplanar grounds are wide (at least to each) and, for GCPW, tied to the backside with a via fence along both edges at a pitch well below . Without vias, the two coplanar grounds can be driven against each other (the slotline mode), and the GCPW structure can resonate as a patch antenna.
- GCPW versus microstrip. When is small compared with , most of the field goes to the backside plane, and the coplanar grounds only trim the impedance. That is fine, and it is the typical PCB situation. For a "true" coplanar line, make .
- Solder mask and plating sit right in the gap, where the field is strongest. They change more than they would on a microstrip. Expect a few ohms lower with solder mask in the gaps.
- Etching. Narrow gaps are hard to hold. For a 0.3 mm centre strip on 20 mil GCPW, a 25 µm change in around 0.15 mm moves by 3–4 Ω, so check your fabricator's minimum gap and tolerance before you commit to a geometry.
- Above the surface-wave cutoff the line leaks power into the substrate. Keep the operating frequency well below , which falls quickly with thicker or higher- substrates.