Radio PCB Design: Layout Rules for RF Boards

A radio board is judged by measurements that are invisible on a schematic. Sensitivity, spurious emission, phase noise and harmonic content all depend on the physical layout of the copper, and a design that simulates correctly can still fail compliance testing because of a ground plane that was interrupted under an amplifier. Radio PCB design is the discipline of keeping the physical structure as predictable as the circuit.

What Makes an RF Board Different

At radio frequencies the interconnect is no longer a wire but a transmission line, and its behaviour depends on geometry rather than on connectivity alone. A trace that connects the right pins but presents the wrong impedance will reflect energy, shift the tuning of the network attached to it and radiate, and the resulting fault appears as poor sensitivity rather than as a broken connection. That is why an RF board is measured rather than merely checked for continuity.

Components also stop behaving as their values suggest. A capacitor that is a short at low frequency becomes an inductor above its self-resonant point, and a resistor acquires a parallel capacitance that matters at gigahertz. The layout has to accommodate those realities, which is why RF boards are designed with a narrower set of standard practices than digital ones.

Transmission Line Structures

A microstrip is a trace on the outer layer with a ground plane beneath it, and it is the most common structure on a radio board because it is easy to build and easy to probe. Its impedance is set by the trace width, the dielectric thickness and the dielectric constant of the laminate, and it needs a continuous reference plane to behave as designed.

A grounded coplanar structure adds ground copper beside the trace on the same layer, stitched to the plane below with vias at close spacing. That arrangement reduces radiation from the trace and makes the impedance less sensitive to the substrate thickness, at the cost of tighter manufacturing tolerance on the gap. Stripline, buried between two planes, offers the best isolation where the layer count allows it, and is the usual choice for a mixer or a filter that has to be kept free of coupling from the layer above.

Impedance Matching Networks

Every active device has an input and output impedance that is not the system impedance, so matching networks are used to transform one into the other. Those networks are built from transmission line sections and discrete components, and their values depend on the parasitic elements introduced by the pads and the placement.

That is why matching is done at the end of the layout rather than at the beginning. The network is placed with a generous pad arrangement, the board is built, and the values are tuned on the first article with a network analyser. Designing the footprint to allow that tuning, with pads that accept a range of component values and a ground plane that stays continuous beneath them, saves a re-spin.

Ground Plane Discipline

On a radio board the ground plane is the reference for every transmission line, and interrupting it changes the impedance of everything that crosses the interruption. That means no slots under RF traces, no plane splits under a matching network, and no routing of digital signals through the RF region on any layer. Where a plane must be split for a legitimate reason, the split is routed around the RF section rather than through it.

Stitching vias are used to tie surface ground copper to the internal plane, spaced at a small fraction of a wavelength so that the structure behaves as a solid reference. The spacing that counts is the shortest wavelength present, not the carrier, because a harmonic or a switching edge with a fast rise time contains higher frequency content than the signal itself, as described in ground routing and power trace planning.

Isolation Between Stages

The most common way for a radio to fail its emission tests is coupling between stages that should be independent. A power amplifier output leaking into the local oscillator input produces spurious products that no amount of filtering after the fact will remove, and the coupling path is usually through the supply or across shared copper.

Physical separation, shielding in the form of a compartment or a metal can, and separate decoupled supply feeds for each stage all help, and the cheapest of the three is usually the separation, because it costs board area rather than components. Layout should follow the signal flow in a straight line where possible, so that the output of a stage is physically far from the input of the one before it. The general suppression principles apply here as much as anywhere, as the layout rules in radiated EMI and regulator layout make clear.

Materials and Loss

The laminate choice determines the loss of every passive structure on the board. A general purpose epoxy laminate has a dissipation factor that becomes significant above a few gigahertz, while a low loss material keeps the loss per unit length low enough to preserve the link budget, and the difference is measurable as a link margin rather than merely as a theoretical figure. The dielectric constant also sets the geometry, so a change of material changes every trace width on the board.

Thickness matters as well. Thin laminates make narrow traces for a given impedance, which suits dense routing, while thicker laminates give more mechanical stability and lower conductor loss. The trade-off is settled with the fabricator, because the stackup must be one the plant can actually build with consistent dielectric thickness.

Interference From the Digital Side

Most radio boards also carry digital circuitry, and the radio is usually the victim. The coupling mechanisms are shared ground impedance, shared supply impedance and radiation from long digital traces or their return paths. Separating the ground references at a single point, filtering the supply to the radio with a ferrite and a capacitor, and keeping the digital routing away from the RF region all address one of those paths.

The coupling mechanism most often overlooked is the clock, because it is periodic and its harmonics are predictable. A high speed clock harmonic can land inside the receive band and raise the noise floor across the whole channel, which appears as reduced sensitivity rather than as an obvious spurious signal. Keeping the clock trace short and its return path tight is as important as any RF specific measure.

FAQ

Does a radio board need a low loss laminate? Only when the frequency or the path length makes the loss significant. Below a few gigahertz with short traces, a standard laminate with careful geometry is often adequate, and the loss budget should be calculated rather than assumed.

Can a matching network be designed purely by simulation? Simulation gives a good starting point but the pads, the solder and the component tolerances are not fully captured. The final values should be confirmed on a built board with a network analyser.

How much ground stitching is enough? Space the vias at a small fraction of the shortest wavelength present, and place them in rows beside every RF trace and along the edges of the ground copper. More stitching never hurts the electrical performance, as the 3W guidance in crosstalk and the 3W rule illustrates for adjacent routing.