Radio PCB: Designing for RF and Wireless Links

What a Radio Board Has to Do

A radio board takes a signal from a digital interface, converts it to a radio frequency, amplifies it and delivers it to an antenna, and it does the reverse on the way in. Every stage of that chain is sensitive to something different: the transceiver to its reference clock and its supply noise, the amplifier to its bias and its thermal path, the filter to the impedance it sees at both ports, and the antenna to everything within a few centimetres of it.

What makes the board a radio board rather than an ordinary digital board is that geometry becomes part of the circuit. A trace is no longer a wire but a transmission line whose width, spacing and distance to the reference plane set its characteristic impedance. That single change is the reason RF layout is treated as a separate discipline.

The Blocks on the Board

• Transceiver. The device that modulates and demodulates. It is usually a packaged part with a differential or single ended interface to the antenna path.
• Reference oscillator. A crystal or a temperature compensated oscillator. Its phase noise sets the purity of the transmitted signal and the sensitivity of the receiver.
• Matching network. Discrete inductors and capacitors that transform the impedance of the transceiver to the impedance of the following stage, usually 50 ohms.
• Filter. A surface acoustic wave, bulk acoustic wave or ceramic filter that passes the wanted band and rejects the rest.
• Power amplifier. Where the output power is high enough to need one, with its own supply, bias and thermal path.
• Low noise amplifier. On the receive path, placed as close to the antenna as the filter allows.
• Antenna and feed. A printed structure, a chip antenna or a connector, with the keep-out area it requires.
• Shielding. A can or a fence over the sensitive sections where the product’s own emissions demand it.

Impedance Control

Every RF trace is designed for a target impedance, commonly 50 ohms single ended and 100 ohms differential. That value is reached by choosing the trace width and its distance to the reference plane for the dielectric constant of the laminate, and the fabricator builds the stack to those numbers.

Four practical consequences follow.

The stack is fixed before the routing. Changing the dielectric thickness after the layout moves every impedance on the board, so the stack is agreed with the fabricator first.

The reference plane is continuous. Every RF trace runs over an unbroken ground plane. Where the trace changes layer, a return via is placed next to the signal via, so that the return current has somewhere to go.

The geometry is uniform. The width, the spacing and the reference distance stay constant along a trace. A short section that is wider, or a place where the plane is interrupted, is an impedance discontinuity that reflects part of the signal.

The impedance is verified. A test coupon is built on the production panel and measured, and the result is compared with the design intent. Our notes on PCB design and layout cover the transmission line design and our notes on PCB manufacturing describe the coupon and the process control behind it.

Grounding and Shielding

An RF design lives or dies by its ground. The ground plane is the reference for every transmission line, the return path for every signal and the shield between the sections of the board.

One ground, referenced everywhere. Splitting a ground plane into an analogue ground and a digital ground creates a discontinuity that any signal crossing it has to negotiate. The better practice is a single plane with the noisy sections placed in a defined area and the sensitive sections physically separated.

Stitch the plane with vias. Where the board has ground on more than one layer, vias spaced a small fraction of a wavelength apart tie the planes together and prevent them from resonating.

Keep the digital away from the RF. The transceiver’s digital interface carries fast edges, and those edges couple into the RF path if the traces run in parallel. Route the digital section on the far side of the board, or behind a ground plane, and keep the crossing points perpendicular.

Shield what needs shielding. A metal can over the transceiver reduces both the emissions from the board and the susceptibility of the board, and a fence with a soldered lid does the same job at lower cost where the layout can accept it.

Materials

The material is chosen from the frequency and the length of the RF traces. Standard FR-4 is usable below a few hundred megahertz and for short traces at higher frequencies, and it is what a simple sub-gigahertz product will use. Above that, the loss becomes measurable and the dielectric constant of FR-4 varies enough between batches to move the impedance and the resonant frequency of a printed filter.

The next step is a modified FR-4 with a lower dissipation factor, then a mid loss laminate, and finally a PTFE based or ceramic filled material for the highest frequencies and the most demanding filters. The cost rises with each step, so the usual approach is a hybrid stack, with the speciality laminate used only for the RF layers and ordinary laminate for the digital and power layers.

The finish also matters at high frequency. A thick tin or hot air levelled coating is uneven and lossy, so electroless nickel immersion gold or immersion silver is used where the RF performance is critical, and the plating thickness is controlled. Our notes on telecommunications PCBA describe the class of equipment these boards belong to.

Testing

Two kinds of test matter. The first is the coupon on the production panel, which confirms the impedance and the plating thickness on the actual panel rather than on a sample made separately. The second is the functional test of the finished assembly, which confirms that the transmit power, the receive sensitivity and the frequency accuracy are inside specification.

A vector network analyser measures the impedance and the insertion loss of the RF paths, and a spectrum analyser or a radio test set measures the transmitted power and the spectrum. Where the product is produced in volume, a purpose built test fixture with a calibrated path to the antenna port makes the measurement repeatable, and the results are recorded against the serial number.

One caution applies to every RF assembly. Touching the board, or testing it without the shield fitted, changes the result; a board that passes on an open bench and fails in the enclosure was characterised in the wrong condition. Our notes on PCBA testing describe the test methods, and our notes on PCB assembly cover the assembly of the finished board.

Types of Radio Board

• Single band. One frequency, such as a 433 megahertz remote control, with a simple matching network and a printed or wire antenna.
• Multi band. Wi-Fi and Bluetooth sharing one module, or a cellular modem with several bands, which needs filtering and an antenna that works across the range.
• Analogue radio. Traditional amplitude and frequency modulation receivers, where the layout keeps the oscillator and the intermediate frequency sections away from the audio path.
• Digital radio. GPS, software defined radio and digital broadcast, where the processing is digital and the RF section is a defined block on the board.
• Power radio. Transmitters with a power amplifier, where the thermal path, the supply decoupling and the harmonic filtering dominate the layout.

What It Costs

Indicative prices in small quantity run around 50 to 80 US dollars for a simple single layer FR-4 radio board, 120 to 180 dollars for a double sided board on a Rogers material, and 200 to 350 dollars and above for a four to six layer high frequency board with shielding. Above a thousand pieces the unit price can fall by 30 to 60 percent.

The cost drivers are the frequency, the board size, the material, the shielding and the test. The test is the one that surprises people: a board that has to be measured on a network analyser individually costs far more to test than one that only needs a functional check, and the test plan is worth agreeing before the design is released.

FAQ

Can a radio be built on ordinary FR-4? Below about 500 megahertz it is usually adequate, and it is the normal choice for short range devices. Above that the loss and the variation in dielectric constant become significant.

What is the difference between an RF PCB and a radio PCB? Radio PCB usually refers to a wireless communication circuit such as a transceiver or a module, while RF PCB is the broader term for any board that carries radio frequency signals.

Why does impedance control matter so much? Because a mismatch at any point in the chain reflects power, reduces the radiated output and degrades the receiver noise figure.

What software is used to design one? A general PCB tool for the layout and a field solver or an RF simulator such as a circuit and electromagnetic analysis package for the transmission lines, the matching and the filters.

Conclusion

A radio board is an ordinary circuit board in which the geometry carries the signal. Agree the stack with the fabricator before the layout, keep every RF trace over a continuous ground plane with a constant geometry, place the matching network and the filter from a simulation rather than by eye, and test both the coupon and the finished assembly in the enclosure it will be used in. The material and the shielding are chosen last, because the structure of the board decides most of the performance.