Transmission Lines

CPW Ground Width

/see-pee-dubuh-yoo ground width/
Flanking either side of the signal trace in a coplanar waveguide, the ground planes have a finite lateral extent Wg that, together with the center-conductor width and slot gap, fixes the line's behavior. When each ground reaches roughly 3 to 5 gap-widths or one substrate thickness, the line approaches the semi-infinite-ground result and the characteristic impedance stops shifting. Narrow finite-ground designs (FGCPW) raise Z0 by 5 to 15 ohms and risk exciting the parasitic slotline mode if the two grounds are asymmetric or left untied. In conductor-backed CPW the same width drives substrate resonances unless the grounds are via-stitched.
Category: Transmission Lines
Asymptotic ground: ≥ 3g to 5g or 1h
FGCPW range: Wg ≈ 2g to 5g

Why Ground Width Governs CPW Behavior

A coplanar waveguide carries its signal on a center strip with return current splitting onto the two coplanar grounds beside it. Because the field is confined mainly to the slot gaps between the strip and the grounds, classical analysis treats the grounds as semi-infinite half planes. Real circuits cannot afford that area, so the grounds are truncated to a finite width Wg. As long as each ground extends far enough that the return-current density has decayed to a negligible value at its outer edge, truncation has no measurable effect; the conformal-mapping closed-form impedance still applies. The current density falls off roughly exponentially with distance from the slot, so the practical cutoff is about three to five slot-gap widths, or one substrate thickness h on thicker boards.

Below that threshold the line becomes finite-ground CPW (FGCPW). Squeezing the return current into a narrower conductor increases the series inductance per unit length and slightly reduces the shunt capacitance, so Z0 climbs. A 50 ohm line designed for semi-infinite grounds can drift to 55 to 65 ohms once Wg drops to one or two gap widths, which is why FGCPW geometries must be re-extracted in a full-wave or 2.5D solver rather than trusted to the textbook formula. FGCPW is nonetheless popular on GaAs MMICs because it conserves expensive die area and localizes the return path next to the signal.

Ground symmetry matters as much as ground width. The intended even (coplanar) mode keeps both grounds at equal potential. If the left and right widths differ, or if the grounds are not periodically tied together with air bridges or bond-wire straps, discontinuities can launch the odd slotline mode, which radiates, has a different phase velocity, and shows up as ripple in measured S-parameters. Keeping the grounds geometrically equal and bridging them at intervals shorter than one tenth of a guided wavelength preserves mode purity.

Impedance and Mode Onset Equations

Aspect ratio (semi-infinite ground):
k = a / b = (W/2) / (W/2 + g)  where W = center width, g = slot gap

Characteristic impedance (no backing):
Z0 = (30π / √εeff) × K(k′) / K(k)  εeff ≈ (εr + 1) / 2

Practical ground-width cutoff:
Wg ≥ max(3g to 5g,  h) → Z0 within ≈ 2% of semi-infinite value

Air-bridge / strap spacing for slotline-mode suppression:
sstrap < λg / 10  λg = c / (f × √εeff)

K() is the complete elliptic integral of the first kind, k′ = √(1 − k²), εr = substrate dielectric constant, h = substrate thickness, λg = guided wavelength. Example: alumina εr = 9.8, εeff ≈ 5.4, so λg at 40 GHz ≈ 3.2 mm and straps should sit closer than 320 μm.

Ground-Width Regimes Compared

Ground regimeWg (in slot gaps g)Z0 shift vs. idealDominant riskRequired mitigationTypical use
Semi-infinite (ideal)≥ 5g or ≥ hReference (0)Wasted die areaNoneDiscrete / large boards
Wide finite ground3g to 5g+0 to +3 ohmsEdge radiation at bendsSymmetric layoutGeneral MMIC routing
Narrow FGCPW1g to 3g+5 to +15 ohmsSlotline-mode couplingRe-extract; add strapsCompact GaAs/InP MMIC
Conductor-backed CPWWide top + backsideLowered by backingSubstrate / parallel-plate resonanceDense via fenceMultilayer modules
Asymmetric groundsUnequal left/rightMode-dependentOdd-mode excitationEqualize widths + bridgesAvoid in production
Common Questions

Frequently Asked Questions

How wide do the CPW ground planes need to be before impedance stops changing?

Once each ground extends about 3 to 5 slot gaps (g) beyond the gap edge, or roughly one substrate thickness h, Z0 sits within about 2% of the semi-infinite-ground value. A 50 ohm line on 254 μm alumina with a 40 μm center conductor and 20 μm gaps reaches that asymptote near Wg ≈ 100 to 150 μm. Narrower FGCPW geometries raise Z0 by 5 to 15 ohms and should be re-extracted in a 2.5D solver instead of trusting the closed-form result.

Why do unequal CPW ground widths excite the parasitic slotline mode?

Ideal CPW runs an even mode with both grounds at the same potential. Asymmetric ground widths, or grounds left untied, let discontinuities such as bends, tees, and bond-wire transitions launch the odd slotline mode, which radiates and shifts phase velocity, producing S-parameter ripple. Keep the two grounds geometrically equal and bridge them with air bridges or bond-wire straps spaced under λg/10 to short the odd mode while leaving the even mode untouched.

How does ground width interact with conductor-backed CPW and the substrate mode?

Adding a backside ground turns the line into CBCPW, where wide top grounds plus the backside plane form parallel-plate return paths that can resonate. If a top ground approaches a half guided wavelength, sharp transmission notches appear. Stitch the top grounds to the backside with a via fence pitched under λg/20 and keep the substrate thin. On 100 μm GaAs CBCPW at 40 GHz, unstitched grounds wider than roughly 600 to 800 μm can resonate in-band.

Millimeter-Wave Interconnects

Designing a CPW Feed Network?

RF Essentials builds coplanar-waveguide transitions and millimeter-wave assemblies with controlled ground geometry and via-stitched grounds for clean mode behavior through W-band. Talk to our engineers about your layout.

Get in Touch