High Frequency PCB Laminate: Materials and Selection Guide
Above a few hundred megahertz, the laminate stops being a passive insulator and becomes part of the circuit. Its dielectric constant sets the geometry required for a given impedance, and its dissipation factor, together with the roughness of the copper beneath it, sets how much of the signal survives the journey. A high frequency PCB laminate is therefore selected on electrical grounds first, and mechanical and cost considerations follow, which is the reverse of how materials are chosen for ordinary digital boards.
Why Standard FR-4 Runs Out of Road
FR-4 is a woven glass fabric in an epoxy resin, and its dielectric constant is the average of those two materials plus the air trapped in the weave. That structure makes the dielectric constant vary from point to point across the board, which is tolerable at low frequency and becomes a problem when the wavelength approaches the dimensions of the variation. The resin also absorbs moisture, and absorbed water raises both the dielectric constant and the loss, so the electrical performance of an FR-4 board depends on the humidity it has seen.
Loss is the more fundamental limit. The epoxy resin has a dissipation factor in the region of 0.02, which is more than an order of magnitude worse than the materials used for radio frequency work. At low frequency that difference is invisible; at ten gigahertz it is the difference between a usable link and a link that cannot close. Copper surface roughness compounds the problem, because a rough foil forces the current to follow a longer path and adds a loss that grows with frequency.
PTFE-Based Materials
Polytetrafluoroethylene was the first material adopted for microwave work and it remains the reference point. Its dielectric constant is low and remarkably stable with frequency and temperature, and its dissipation factor is well under 0.001. Those properties come with mechanical penalties: PTFE is soft, it cold flows under pressure, it is difficult to drill cleanly and it has a very high coefficient of thermal expansion, which makes it a poor match to the copper foil attached to it.
Practical PTFE laminates are therefore composites, filled with glass microfibre or ceramic powder to control expansion and stiffness, and offered with different filler fractions that trade dielectric constant against loss. A low filler content keeps the dielectric constant near 2.2 and gives the lowest loss, while a higher filler content raises it towards 3 and improves dimensional stability. The choice is driven by the frequency and by how much the design depends on the board staying flat through assembly.

Hydrocarbon and Ceramic-Filled Laminates
Hydrocarbon laminates with ceramic fillers occupy the middle ground. They have a dielectric constant between that of PTFE and FR-4, a dissipation factor around 0.004, and mechanical behaviour much closer to FR-4, which means they can use standard drilling, plating and lamination processes. That last point is what has made them the default for automotive radar, 5G infrastructure and high speed digital boards, where the electrical performance of PTFE would be welcome but the cost and processing difficulty would not.
They are also produced as thermoset systems that can be laminated in the same press as ordinary FR-4, which makes mixed stack-ups practical. A board can have a high frequency material on the layers that carry the radio routing and FR-4 on the digital layers, laminated together into a single construction. Our layer assignment notes describe how that split is planned.
Dielectric Constant, Dissipation Factor and Copper Roughness
Three numbers dominate the selection. The dielectric constant determines the trace width needed for a fifty ohm line and the amount of phase shift a given physical length introduces, so a material whose dielectric constant drifts with temperature or frequency will make a filter or a phased array behave differently at the band edges than in the middle. The dissipation factor determines how much of the signal is converted to heat per unit length.
Copper surface roughness is the number that is most often overlooked. At low frequency the current flows uniformly through the foil, but as frequency rises it concentrates near the surface, and if that surface is rough the current path becomes longer than the geometric length. The resulting loss can exceed the dielectric loss in some constructions. Foils with low roughness profiles are available, and they are worth specifying where the link budget is tight, though they adhere less well to the laminate and need the fabricator to use an appropriate bonding treatment.
Stack-Up and Design Practice
A high frequency stack-up is built around one principle: keep the return path directly beneath the signal and keep the dielectric thickness constant along the whole trace. That means a solid ground plane on the adjacent layer, no plane splits under radio traces, and a controlled dielectric thickness achieved with the right prepreg combination rather than by chance.
Where a mixed stack-up is used, the transition between materials deserves attention. The dielectric constant changes at the interface, so the impedance of a trace crossing it changes too unless the trace width is adjusted. Grounded coplanar waveguide is common on radio layers because it allows a wider trace for the same impedance, which reduces conductor loss, and because it is less sensitive to the exact dielectric thickness. Via structures that connect through a thick RF stack should be accompanied by ground vias around them to keep the return path continuous, which is the same discipline used on high density digital boards. Our design release checklist places these checks in the review sequence.

Fabrication, Cost and Sourcing
These materials cost several times more than FR-4 per unit area, and the processing adds more. PTFE requires specialised drilling parameters, plasma treatment before plating and careful handling to avoid contamination. Ceramic-filled hydrocarbon materials are closer to FR-4 in behaviour but still need their own lamination profiles, and panel sizes are often smaller, which reduces the number of boards per panel and raises the unit cost again.
Lead time is the other practical consideration. Radio frequency laminates are stocked in fewer thicknesses and copper weights than FR-4, so a stack-up designed around a specific combination may need to be built to order. Checking availability before the layout is frozen, and being prepared to adjust a dielectric thickness by a small amount, avoids a redesign driven purely by material supply. Our thermal management notes describe how the thermal behaviour of a low-loss material compares with the FR-4 it replaces.
Applications
Millimetre wave radar for automotive safety systems, phased array antennas for satellite links, 5G base station radios and high speed backplanes all use these materials, as do test fixtures and calibration standards where the loss of the fixture must not mask the behaviour of the device under test. Below roughly one gigahertz, well designed FR-4 with a controlled stack-up is usually sufficient, and specifying a low loss material there adds cost without measurable benefit.
FAQ
When should I move off FR-4? When the loss budget or the required impedance tolerance can no longer be met at the working frequency and trace length. Above roughly three gigahertz, or where a long trace carries a signal in the gigahertz range, the move is usually justified.
Is a lower dielectric constant always better? A low dielectric constant allows wider traces for the same impedance, which reduces conductor loss, so it is usually an advantage. It also makes the circuit more sensitive to mechanical tolerances, so the stack-up has to be controlled more tightly.
Can high frequency and FR-4 layers be mixed in one board? Yes, and it is one of the most cost effective approaches. The radio layers use the low loss material while the digital and power layers use FR-4, laminated into a single construction.



