Microwave PCB Materials and Fabrication: A Practical Overview

A microwave PCB is defined by controlled electrical properties rather than by a frequency label. At microwave frequencies, the laminate’s dielectric constant, the loss tangent and the roughness of the copper all contribute to insertion loss, and the geometry of every transition becomes electrically significant.

Fabrication for microwave work is therefore a materials and tolerance exercise as much as an imaging exercise. The design depends on properties that must be specified, held and verified rather than assumed.

Frequency Bands and Material Requirements

Below about one gigahertz, standard FR-4 performs adequately for many circuits, and the cost advantage is significant. Between one and six gigahertz, the loss becomes noticeable and the stability of the dielectric constant begins to matter.

Above roughly ten gigahertz, low-loss laminates become necessary, and above twenty gigahertz the copper surface roughness contributes a measurable share of the total loss. The material choice should follow the frequency and the allowed loss rather than a general rule.

Dielectric Constant and Its Tolerance

Dielectric constant sets the physical dimensions of every element on a microwave board: line widths for a given impedance, the length of a resonator, the spacing in a coupled structure. A change of a few percent moves the centre frequency of a filter by more than the passband width.

Suppliers therefore specify the dielectric constant with a tolerance and, ideally, its variation with frequency and temperature. A material with a nominal value of 3.0 but a tolerance of plus or minus 0.1 is less useful for a narrowband design than one specified more tightly.

Loss and Dissipation Factor

The dissipation factor describes how much energy the dielectric absorbs. It is small in absolute terms, around 0.001 to 0.004 for common microwave laminates, but it scales directly into insertion loss over a given length.

Loss also comes from the conductor. Copper surface roughness increases the effective path length for current at high frequency, which raises conductor loss. Smooth copper foils and treated foils with low profile are used where the loss budget is tight.

Surface Roughness and Copper Type

Standard copper foil has a roughness measured in microns, and at high frequency the current concentrates near the surface, so the roughness becomes an effective increase in resistance. The effect grows with frequency until it becomes the dominant conductor loss mechanism.

Low-profile and ultra-low-profile foils reduce that contribution. They are more expensive and require careful handling during lamination to maintain adhesion, so they are specified where the loss budget justifies them.

Plating and Structure Options

Microwave circuits often avoid plated through-holes in the signal path, because a via introduces inductance that cannot be compensated in a tuned structure. Where a via is necessary, its diameter and length are minimised and its position is included in the electromagnetic simulation.

Edge plating, cavities and shielding walls are used to create enclosed transmission lines and to suppress radiation. Each of those is a mechanical feature that must be planned with the fabricator, and each adds cost.

Fabrication Tolerances

Trace width tolerance translates into impedance tolerance, and impedance tolerance translates into a frequency shift. Etch compensation therefore has to be developed for the specific material and copper type rather than inherited from an FR-4 process.

Layer-to-layer registration matters for coupled structures, where the spacing between conductors on different layers defines the coupling. A small offset changes the response, and the tolerance should be stated in the design rules.

Panel-level thickness variation changes the dielectric height, which changes impedance. Where the design is sensitive, thickness tolerance is specified more tightly and verified on coupons.

Grounding and Shielding Practice

Microwave layout uses a continuous ground plane as the reference for every transmission line, with via fences along the edges of a line to suppress parallel plate modes. The spacing of the vias must be a small fraction of the wavelength at the highest frequency of interest.

Where two regions must be isolated, a shielded cavity or a row of vias is used rather than a plane split, because a split creates a slot antenna. The return current must be given a defined, short path in every case.

Connectors and Transitions

Connectors and transitions are often the largest source of loss and reflection on a microwave board. The transition from a coaxial connector to a microstrip line must maintain the impedance, and its geometry is usually modelled rather than drawn by eye.

The board edge and the connector footprint interact. Keeping the ground plane continuous under the launch and providing a via fence around the transition are the standard measures, and their absence shows up as a return loss that is worse than the simulation predicted.

Measurement and Verification

Microwave boards are verified electrically, not only geometrically. Coupons provide the impedance and loss data, and a network analyser measures the assembled circuit’s response against the design target.

Where the design is a filter or an antenna, the measurement is the acceptance criterion. The fabrication drawing should therefore specify what is to be measured and at which reference planes, so that the result can be compared with the simulation.

Cost Structure

Material is the largest item. Microwave laminates cost several times more than FR-4, and panel utilization determines how much of that cost each board carries.

Processing adds labour because the materials are softer, more sensitive to drilling parameters and often require special surface preparation. Low-profile copper and tight tolerances add inspection and process control.

Simulation and Its Limits

Microwave design relies on electromagnetic simulation, and the model is only as good as the material data and the geometry it contains. A simulation that uses a nominal dielectric constant and ignores copper roughness will predict less loss than the hardware delivers.

The practical approach is to validate the model against a measured coupon on the first build, then adjust the material parameters to match the measurement. From that point the model predicts the fabrication process rather than an idealised material.

Prototype and Production Consistency

A microwave design that works on the prototype must also work in production, and that requires the same material, the same foil type and the same tolerances from every lot. Material substitution, even to an apparently equivalent grade, changes the dielectric constant slightly and shifts the frequency.

Where a second source is required, the qualification should include a full electrical measurement of a representative circuit, not only a coupon impedance check, because the loss and the dielectric behavior are affected by factors a coupon does not capture.

Design Checklist

Confirm the frequency range and the allowed loss, choose the material from those two numbers, and verify that the dielectric constant tolerance supports the bandwidth required. Then check the transition design and the via fence spacing against the highest frequency.

Finally, confirm the fabrication tolerances with the supplier. A microwave design that assumes tighter tolerances than the process can hold will shift in frequency on the first build, and the correction is a redesign rather than an adjustment.

Related reading: high-frequency data bus routing, microstrip and stripline routing, and multilayer PCB advantages at high speed.

FAQ

Can FR-4 be used for microwave circuits? Below about one gigahertz for many designs, yes. Above that, the loss and the dielectric constant tolerance usually make a dedicated microwave laminate necessary.

Why does copper roughness matter at high frequency? Because the current concentrates near the conductor surface, so surface irregularities lengthen the effective current path and increase resistance and loss.

Should vias be avoided entirely? In tuned structures, yes wherever possible. Where a via is required, keep it short, keep it out of the signal path where possible, and include it in the electromagnetic simulation.