RF PCB Design: Impedance, Grounding and Layout Practice

An RF board is not simply a board with a radio on it. The distinction is whether a conductor is treated as a wire carrying a current or as a transmission line carrying a wave, and that depends on the ratio of the physical length to the wavelength. Once that ratio becomes significant, the geometry of the trace and the properties of the material it sits on determine whether the signal arrives intact.

What Counts as an RF Board

The conventional dividing line between a lumped element model and a distributed model is a conductor whose length is five percent or more of the wavelength in the material. Below that, the voltage along the trace can be treated as uniform and ordinary circuit rules apply. Above it, the phase of the signal varies along the trace and the trace has to be designed as a transmission line.

Expressed in frequency, and assuming traces on a board are rarely longer than half a metre, that puts the lower end of practical RF work at roughly 30 megahertz. The upper end is a matter of convention, since above about three gigahertz the work is usually called microwave design, and practical limits on component spacing become the binding constraint before the frequency itself does. Treating boards that carry analogue signals between about 30 megahertz and 6 gigahertz as RF PCB designs is a reasonable working definition, with the decision about which model to use taken from the length to wavelength ratio rather than from the frequency alone.

Transmission Line Model and Material Choice

Because the dielectric constant of the substrate is higher than that of air, the wave travels more slowly in the board than in free space and the wavelength is correspondingly shorter. A trace therefore reaches the transmission line regime at a lower frequency than a free space calculation would suggest.

The requirements on the material follow from that. The substrate should have low dielectric loss, a dielectric constant that stays constant over the frequency and temperature range of the application, high thermal conductivity, and a smooth surface with good adhesion to the conductor. The conductor should have high conductivity, a low temperature coefficient of resistance, and good adhesion and solderability. Whether the trace runs as a microstrip on the surface or as a stripline between planes changes both the geometry needed and the loss, and the comparison is set out in microstrip and stripline routing.

<img src="https://www.gopcba.com/wp-content/uploads/2024/09/tupian4.png" alt="RF PCB design with controlled impedance microstrip traces” />

<img src="https://www.gopcba.com/wp-content/uploads/2026/08/91.png" alt="ground plane and via stitching under an RF trace” />

Component and Connector Considerations

RF components are less forgiving than their low frequency equivalents, and the differences are mechanical as much as electrical. Packages are often unusual, so a new component should be checked against the assembly process before it is designed in. The plating on the termination is worth inspecting, because a silver plated ceramic part can produce cold or open joints if the process is not matched to it. Flatness matters too, particularly for ceramic blocks such as voltage controlled oscillators, amplifiers and filters, where a surface flatness better than about 0.005 inch is the practical expectation. And the tolerance on passive components deserves attention, because once a component tolerance exceeds about five percent the distributed behaviour of the circuit becomes noticeably more variable.

Connectors are part of the RF path and are specified accordingly. The centre contact and the signal pins are usually gold plated at 30 to 50 microinches over 50 to 150 microinches of nickel, and where a surface mount connector meets the board the pad tolerance is held to a fraction of a pad size, because mechanical alignment is what determines whether the launch into the board is consistent.

Grounding and Routing Rules

Three rules cover most RF layout. Keep the digital circuitry physically away from the analogue RF section, because a digital edge is a broadband source and an RF receiver is a sensitive listener. Give every RF trace a solid, unbroken ground plane directly beneath it, so that the return current flows under the trace and the impedance is the one that was calculated. And keep the RF trace on the surface where it can be controlled, rather than burying it where the dielectric above and below both change the geometry.

Where an RF trace has to interact with other signals the order of preference is clear. Crossing a region at right angles is better than running alongside it, because a perpendicular crossing reduces the capacitive coupling to a minimum. Wherever an RF trace runs near another conductor, additional ground connected to the main ground plane reduces the coupling. Where two RF traces must run in parallel, a grounded trace between them, connected to the plane by via stitching, restores the isolation that the parallel run removed. RF differential pairs run as a pair with ground tracks outside them, stitched with vias in the same way. Where distance and ground cannot provide enough isolation, shielding a sensitive stage in a metal can is the remaining option, and the layout has to leave space and a good ground connection for it.

General rules for clearance between vias and traces in a multilayer stack are covered in via to trace clearance in multilayer boards, and the routing discipline for high frequency buses is described in high frequency traces and data bus routing.

Verification

An RF board is verified by measurement rather than by inspection, because the properties that matter are not visible. A vector network analyser measuring a test coupon from the same panel gives the impedance and the loss of the actual stack-up, which is the only way to confirm that the material and the lamination produced what the calculation assumed. On the assembled board, a spectrum analyser at the output confirms that the design meets its emission and spurious requirements, and a near field probe shows where energy is leaving the board if it does not.

The practical point is that these measurements belong to the design rather than to production test. A coupon measured early identifies a stack-up problem while the boards are still being fabricated; the same problem found after assembly costs the components as well as the boards. Where a design is expected to go into volume, agreeing the coupon layout and the test points with the fabricator before the first panel is built is the cheapest way to keep that information available.

FAQ

At what frequency does a trace become a transmission line? When its length reaches about five percent of the wavelength in the material. On a board with traces under half a metre that puts the lower end of practical RF work at roughly 30 megahertz.

Why does board thickness matter so much on an RF design? Because it is part of the impedance calculation. The thickness, the trace width and the dielectric constant together set the impedance, so the thickness is chosen from the calculation rather than from a stock size.

Why does an RF board need a solid ground plane under the traces? Because the return current flows in that plane directly beneath the trace. A broken or split plane forces the current to detour, which changes the impedance and radiates.

How should two RF traces be isolated when they must run in parallel? By placing a grounded trace between them and stitching it to the ground plane with vias. Running them in parallel without that ground trace allows them to couple.