When FR-4 Is Not Enough: Choosing a Low-Loss Laminate

FR-4 is the default laminate for a reason. It is inexpensive, available in a wide range of thicknesses and copper weights, well characterised, and it behaves predictably in assembly. For the majority of digital boards it is the correct choice, and specifying something more exotic adds cost without adding margin. The question is where that stops being true.

This article sets out what a low-loss laminate actually changes, how dielectric constant and loss tangent translate into board behaviour, and how to tell whether a design has reached the point where FR-4 is no longer adequate.

What FR-4 Does Well

A woven glass and epoxy laminate has a dielectric constant around 4.3 and a loss tangent between 0.015 and 0.025 at a few gigahertz, depending on the resin content and the glass weave. Its mechanical properties are well understood, its coefficient of thermal expansion is matched to common assembly processes, and every fabricator stocks it.

The weaknesses only appear at frequency and at length. Loss rises with frequency, so a short trace at 5 GHz may be perfectly acceptable while a long one is not. And the dielectric constant is not a single number: it varies with the resin-to-glass ratio across the panel, which means two traces of identical geometry can have measurably different impedances. That is the problem a high-frequency PCB material is bought to solve.

Cost is the other side of the same coin. A low-loss laminate may cost several times more per panel, and it usually needs a different prepreg, a different drilling recipe and a slower lamination cycle. The decision is therefore never between good and bad material; it is between one set of trade-offs and another, and it should be made with a calculation rather than with a preference.

Dielectric Constant and Why Its Stability Matters

Dielectric constant sets the speed of the wave and the impedance of the trace. A variation of plus or minus 0.2 in the dielectric constant shifts a 50 ohm microstrip by roughly 2 ohms and changes the propagation delay enough to matter in a length-matched bus. On FR-4 that variation occurs both between panels and within a single panel, because the glass weave is not uniform.

A low-loss laminate is engineered for a tighter tolerance and a more uniform structure. Some use a different glass weave, some use a filler to smooth the resin distribution, and some avoid glass altogether. The result is impedance stability that FR-4 cannot match, which matters most for differential pairs, for long serial links and for any design with a strict skew budget.

Loss Tangent and Insertion Loss

Loss tangent measures how much energy the dielectric absorbs as the wave passes. It is the material property that dominates insertion loss at high frequency, above the point where conductor loss from the copper surface roughness has been controlled. Halving the loss tangent roughly halves the dielectric contribution to attenuation.

Whether that matters depends on the loss budget. A channel that must close with a few decibels of margin at 10 Gbps will often need a low-loss material; the same channel at half the length, or at a lower rate, may close on FR-4 with a well-controlled stack. The calculation should be done rather than assumed, because the material cost difference is substantial and grows with panel size.

Moisture Absorption and Process Sensitivity

All organic laminates absorb moisture, and absorbed water raises both the dielectric constant and the loss tangent. FR-4 absorbs more than most low-loss materials, so its electrical behaviour drifts with humidity unless the board is baked before measurement and protected afterwards.

The practical consequence appears in the assembly process. Absorbed moisture turns to steam during reflow, and a laminate with high absorption is more prone to delamination and to the popcorn effect under large packages. Low-loss materials with lower absorption are more forgiving of a marginal reflow profile, which is a reliability benefit that has nothing to do with insertion loss.

Thermal and Mechanical Trade-Offs

High-frequency laminates are not universally better in mechanical terms. Some have lower glass transition temperatures than a good FR-4, some have higher coefficients of thermal expansion in the z axis, and some are considerably more brittle. PTFE-based materials in particular are soft, prone to dimensional movement during processing, and require special handling at the drilling and plating stages.

The stack design therefore has to account for the material’s mechanical behaviour and not only its electrical parameters. A hybrid stack, with the high-frequency material used only on the layers that carry the critical signals, is a common compromise that captures most of the electrical benefit while keeping the rest of the board on a familiar laminate.

Hybrid Stacks and Where to Place the Material

In a hybrid build, the low-loss layers carry the high-speed routing and are bonded to ordinary FR-4 using a compatible prepreg. The challenge is the interface: the two materials have different dielectric constants, different expansion rates and different drilling behaviour, so the transition has to be planned.

Impedance controlled with a microstrip or stripline calculation must use the correct permittivity for each layer, and the via transitions between them should be modelled rather than assumed. Where the high-speed routing spans several layers, the comparison of FR-4 and high-frequency stack-ups is a useful reference for how the arrangement changes with the material.

Making the Cost Decision

Low-loss material typically costs several times more per unit area than FR-4, and the premium grows with the size of the panel and the number of layers that use it. Justifying it requires a loss budget that shows the channel failing on FR-4, or a skew or impedance requirement that FR-4 cannot hold.

Two cheaper alternatives are worth trying first. Shortening the critical channel, or moving it to a layer with a better reference plane, often recovers the same margin at no material cost. Adjusting the stack so that the high-speed layer is closer to its plane reduces radiation and improves impedance control. Only when those options are exhausted does the material change become the right answer.

A useful discipline is to write the loss budget before choosing the laminate and to state the margin the design requires at the worst-case corner. If the budget closes on FR-4 with a few decibels to spare, the cheaper material stays, and the effort goes into the stack arrangement instead. If it closes only on a low-loss laminate, the choice is justified by a number that can be shown to a reviewer.

FAQ

At what frequency does FR-4 stop being adequate? There is no single threshold, because length and loss budget matter as much as frequency. Many designs run comfortably at 10 Gbps on FR-4 over short channels, while others need a low-loss laminate at half that rate over a long backplane path.

Can I mix a low-loss laminate with ordinary FR-4? Yes, and hybrid stacks are common. The two materials must be bonded with a compatible prepreg, and the impedance calculation has to use the correct dielectric constant on each layer, since the transition between them is itself a discontinuity.

Does a low-loss laminate remove the need for impedance control? No. It makes the impedance more predictable, which makes control easier, but the geometry still has to be calculated against the actual stack. The high-frequency trace routing rules apply regardless of the material chosen.