Microwave PCB: Materials, Loss and Fabrication Limits

Below a few gigahertz, a board behaves in a way that most engineers find intuitive: impedance depends on geometry, loss is modest, and a simulation that matches the measurement to within ten percent is considered good. Above that, the rules change. Conductor roughness, dielectric anisotropy, via geometry and the thickness of the plating all enter the result, and a design that ignores them will not match its simulation.

This article covers what a microwave PCB design has to account for, how material choices affect the outcome, and which fabrication tolerances should be confirmed before release.

What Changes Above a Few Gigahertz

The first change is that the current stops using the whole conductor. Skin depth shrinks as frequency rises, so current concentrates in a thin layer at the surface of the trace. At 10 GHz the skin depth in copper is under a micron, which means the effective conductor is a film whose resistance depends on how smooth the surface is.

The second change is that the wavelength becomes comparable with the features. At 10 GHz a wavelength in a dielectric with a permittivity of four is roughly 15 mm, so a 3 mm via stub or a 5 mm connector pin is a significant electrical length and cannot be treated as a lumped element.

<img src="https://www.gopcba.com/wp-content/uploads/2025/09/10层2阶HDI板.jpg" alt="Microwave PCB with microstrip traces and ground stitching” />

Material Selection

The dielectric constant sets the geometry and the loss tangent sets the attenuation. For microwave work the important properties are a low and stable loss tangent, a dielectric constant that varies little with frequency and temperature, and low moisture absorption. Standard FR-4 fails on the first two, which is why low-loss laminates and PTFE-based materials dominate at these frequencies.

The choice within that family depends on the application. PTFE materials offer the lowest loss but are soft, move during processing and are difficult to drill. Ceramic-filled hydrocarbon laminates are stiffer, more dimensionally stable and easier to process, at a slight penalty in loss. The decision should follow from the loss budget rather than from a preference for the lowest loss figure available.

Surface Roughness and Conductor Loss

Above a few gigahertz, conductor loss depends on surface roughness. A rough copper surface increases the path length the current has to travel and adds loss well beyond what a smooth conductor of the same cross-section would produce. Rolled copper is smoother than electrodeposited copper, and the difference can be measured as a fraction of a decibel per centimetre at 10 GHz.

The roughness also affects the effective dielectric constant, because the copper profile is partly embedded in the laminate. This is why some materials specify a roughness correction factor to be used in the impedance and loss calculation. Ignoring it produces an insertion loss prediction that is optimistic by a margin that grows with frequency.

Dimensional Tolerance and Impedance

Impedance tolerance is set by the dielectric thickness, the trace width and the dielectric constant. At microwave frequencies the requirement is usually tighter than on a digital board, and the fabricator has to adjust the artwork so that the finished geometry lands on target. The tolerance that matters is the one on the finished board, not on the artwork.

Dielectric constant variation is the harder term to control. A material specified as 3.0 plus or minus 0.04 gives a much tighter impedance distribution than one specified as 3.0 plus or minus 0.15, and that difference shows up directly in the measured performance of a filter or a coupler. Where the design is a resonant structure, this tolerance often decides whether the part works at all.

Plating, Finish and Skin Effect

Surface finish matters more at microwave frequencies than elsewhere. A thin nickel layer under a gold finish has higher resistivity than copper and, because the current is concentrated near the surface at these frequencies, that layer carries a significant share of the current and adds loss. Where loss is critical, finishes without a nickel barrier are preferred.

Plating tolerance on the trace itself also matters, because the finished width includes the plated thickness. A trace that is plated up after etching is wider than the artwork, and the difference has to be included in the impedance calculation. This is one of the reasons a fabricator who builds microwave boards routinely will ask for the target impedance rather than a trace width.

Vias, Transitions and Grounding

A via at microwave frequencies is a discontinuity, not a connection. Its inductance depends on its length, so a through via that passes from an outer layer to an inner layer and continues to the far side has a stub below the transition that resonates at a frequency determined by its length. Back-drilling removes the stub, and where it cannot be removed, the via should be shortened by routing on the layer nearest the surface.

Grounding has to be dense and continuous, because the return current follows the path of least inductance and any break in the ground allows it to spread. Stitching vias around a microwave trace, spaced at a fraction of a wavelength, keep the ground potential uniform and suppress the parallel-plate modes that can otherwise propagate in the substrate.

Fabrication Limits to Confirm

Before releasing a microwave design, confirm the minimum trace width and gap the fabricator can hold, the tolerance on the dielectric thickness, the copper surface roughness of the material in stock, the achievable registration between layers, and whether controlled-impedance coupons will be measured on the panel. The microstrip and stripline routing rules give the geometry that these limits apply to.

For the transitions between layers and around the RF section, the high-frequency routing guidance covers how traces should be arranged to keep the return path intact. Where a via has to carry an RF signal between layers, the via design rules set out the dimensions and the ground arrangement that keep the discontinuity small.

Test Coupons and Verification

Microwave boards should be verified with coupons built on the same panel, not with measurements taken on the finished product alone. A coupon provides a controlled environment for measuring the dielectric constant, the loss and the impedance, and it can be cross-sectioned to confirm the dielectric thickness. When a design fails to meet its specification, the coupon tells you whether the problem is the layout or the process.

The usual coupon set includes a straight line for insertion loss, a set of impedance test structures at the design geometry, and a resonator for measuring the dielectric constant at the frequency of interest. Insisting on that set, and reading the results before approving the panel for assembly, is what keeps a microwave design from being debugged on the finished assembly.

FAQ

Can FR-4 be used for a microwave design? It can be used up to a few gigahertz for short, non-critical runs, provided the loss and the dielectric constant variation are acceptable. For resonant structures, filters or any design with a tight loss budget, a low-loss material is required.

How tight should the impedance tolerance be? A common specification is plus or minus five percent for microwave work, tighter than the ten percent used on digital boards. Whether that is achievable depends on the dielectric constant tolerance of the material as much as on the fabricator.

Does the surface finish really change the loss? Yes, and measurably. A finish with a nickel barrier places a higher-resistivity layer exactly where the current concentrates, so it adds attenuation. Where the loss budget is tight, the finish should be chosen with the same care as the laminate.