Microwave PCB Fabrication: Materials, Tolerances and Tests

A microwave PCB looks like any other printed circuit until the requirements are read. The tolerances are tighter, the materials are different, the vias behave as circuit elements rather than as connections, and the verification is performed with instruments that measure reflection rather than continuity. Building one is therefore a joint exercise between the designer and a fabricator who has done it before.

What Makes a Board a Microwave Board

The dividing line is not a frequency but a behaviour. Once the wavelength in the dielectric becomes comparable with the length of the circuit, every discontinuity matters and the board is part of the circuit rather than a carrier for it. At that point the dielectric constant, the loss tangent, the copper surface and the precision of the geometry all enter the design equations.

Typical applications are radar front ends, satellite communication equipment, radio links and instrumentation. They share a requirement for a very low and repeatable dielectric loss and for a transmission line whose impedance is held close to its design value over the whole band of interest.

Choosing the Substrate

Polytetrafluoroethylene based laminates, with or without glass or ceramic reinforcement, remain the standard choice because their dielectric loss is very low and their dielectric constant is stable with frequency and temperature. Where the circuit is smaller or the requirement less severe, a ceramic filled composite or a low loss thermoset can be used at lower cost, with a small penalty in loss.

The choice has mechanical consequences as well. PTFE is soft and its coefficient of thermal expansion is high, which affects how it is drilled, pressed and assembled, and it makes the plated through hole the weak point in a thermal cycle. The implications for routing and for the transmission structures are described in high frequency trace and data bus routing.

Tolerances and the Etch Process

The impedance tolerance on a microwave board is often specified at plus or minus five percent or tighter. That tolerance is achieved by controlling the dielectric thickness, the copper thickness, the trace width and the shape of the etch profile. Because the trace is defined by etching, the finished width is narrower than the artwork and the cross section is not a rectangle, and both effects have to be predicted.

The design should also state the structure it refers to, because a microstrip and a stripline of the same nominal impedance have different geometries and different sensitivities to the same errors. The principles behind those structures are explained in microstrip and stripline routing.

Drilling and Vias

Vias are the second source of disturbance after the line itself. A plated through hole in a thick substrate leaves a stub whose length becomes a resonant element at the operating frequency, and the inductance of the via adds to the discontinuity of the transition. Where the design requires it, back drilling removes the stub; where the routing allows, a blind via avoids it altogether.

Drilling a soft PTFE laminate is itself a specialised process. The material tends to smear rather than cut, and a smeared barrel will not plate reliably, so tooling, feed rates and entry and exit materials are chosen for the material rather than for a general purpose board. Laser drilling is used for microvias where the structure calls for them, and the resulting geometry is described in blind and buried via stack selection.

Copper Foil and Surface Finish

Loss in the conductor comes from resistance and from surface roughness. At microwave frequencies the current concentrates near the surface, so a rough copper foil presents a longer effective path and adds loss that no amount of design effort can remove. Rolled annealed foil with a smooth surface is preferred for the critical layers, and the finish applied to the copper should be thin and uniform.

Immersion silver and immersion gold are common because they are thin, flat and solderable, while a hot air levelled tin is avoided because its thickness varies along the line and changes the impedance. Where the board must be stored before assembly, the shelf life of the finish and the packaging that preserves it become part of the specification.

Assembly and the Connector Transition

The transition from the board to a connector or a coaxial launch is often the largest discontinuity in the whole signal path. A well designed launch matches the field pattern of the line to the geometry of the connector, and it is usually modelled and measured rather than estimated. The same applies to the transition between a microstrip on the surface and a stripline in the interior.

Assembly itself is not difficult, but the thermal expansion of the laminate means the reflow profile and the support of the panel should be chosen for the material. Where the board is a hybrid with conventional laminate for the digital sections, the two materials must survive the same assembly cycle without the bond between them being disturbed.

Cost Drivers and What Actually Matters

The cost of a microwave PCB is driven by the material, the tolerance and the number of process steps rather than by the size of the board. A PTFE laminate costs several times what an equivalent epoxy glass laminate costs, the impedance requirement adds measurement and process control, and the low volume typical of radar and satellite work means the engineering effort is spread over few units.

What is worth paying for is repeatability. A cheaper material with a dielectric constant that varies from batch to batch will produce boards that meet the specification on the first article and miss it later, and the cost of that variation is measured in repeated tuning and in field returns rather than in the purchase order. Where the design allows a wider tolerance, a lower cost material can be a sensible choice; where it does not, the material and the process control are the design.

Verification

Time domain reflectometry is used on a coupon to confirm the impedance of the manufactured line, and a vector network analyser is used on the finished circuit or on a test structure to measure return loss, insertion loss and the frequency response of filters and matching networks. These measurements are usually recorded per design and retained with the coupon, because the board cannot be re-measured once it is assembled into the product.

Where the design includes an antenna or a filter, the measurement at the connector is not the same as the measurement at the radiating element, and the difference has to be accounted for. The verification plan should state what is measured, where the reference plane is, and what the pass criteria are, before the first prototypes are ordered rather than after they arrive.

FAQ

Can a microwave board be built on ordinary FR4? Only at the lower end of the range and for short runs. Loss and the variation of dielectric constant with frequency make FR4 unsuitable for critical microwave circuits.

Why is impedance tolerance tighter than on a digital board? Because the return loss of a microwave circuit depends directly on how closely the manufactured impedance matches the design value.

Is a hybrid stack common? Yes. Using the microwave material only on the radio frequency layers and a conventional laminate elsewhere controls both the cost and the board thickness.