Dielectric Constant Measurement Methods for Laminates

The dielectric constant of a laminate determines how fast a signal travels in a trace and how wide the trace has to be to present a given impedance. It is the number a designer uses most often when a controlled impedance stack is designed, and it is also the number that is most often quoted from a datasheet without asking how it was measured. Dielectric constant measurement is a subject with several accepted methods, and moving between them explains most of the disagreements about a material.

Why the Value Matters

A signal travels more slowly in a material with a higher dielectric constant, so the value sets the propagation delay of every line on the board. It also sets the capacitance per unit length, which is the quantity that a stack-up calculation uses to arrive at a trace width for a target impedance.

An error in the value therefore becomes an error in the trace width, and a trace that is the wrong width presents the wrong impedance. The mistake is difficult to detect after the fact because the board still functions; the symptom appears as a reflection or as a link that fails to meet its margin at the fastest data rate.

The Test Methods Compared

Several test methods are in common use, and each measures a slightly different thing. A parallel plate method measures a thin sample between two electrodes, which gives a value through the thickness at a relatively low frequency. A resonator measures a sample placed in a cavity or a strip line, which gives a value at a defined frequency close to the application.

Because the geometry and the frequency differ, the results differ. A supplier may quote a value from a parallel plate measurement at one megahertz while the design operates at several gigahertz, and the two figures are not interchangeable. The test method should always be quoted with the result, and the value used in a design should come from a method that matches the application.

Frequency Dependence of the Value

The dielectric constant of a resin system is not constant with frequency. It falls as frequency rises through the region where the polar groups in the polymer can no longer follow the field, and the change can be significant across the range a modern product uses.

The loss tangent behaves in the opposite way in relative terms: it usually rises with frequency in the useful range, which is why a material that performs well at one gigahertz may be marginal at ten. The two values together describe the material, and a design that uses one without the other is only half informed. The practical consequences are described in the guide to high frequency laminates.

Anisotropy and Measurement Direction

Laminates are filled with glass fabric, and the fabric has a different dielectric constant from the resin. The mixture is therefore not uniform, and the measured value depends on the direction of the electric field relative to the weave, as well as on the local position relative to a glass bundle.

The variation across a panel is small enough to ignore for most designs and large enough to matter for a very tight impedance tolerance. Where the requirement is tight, the measurement should be made on the same stack and in the same orientation as the product, and the result should be treated as an average rather than as a precise local value.

Sample Preparation and Thickness

The sample has to be clean, flat and of a known thickness, because the calculation for most methods depends on the geometry. A sample that has absorbed moisture or that has been contaminated by handling will give a value that is too high, and the error will be attributed to the material.

Thickness matters particularly for the through thickness methods, where a small error in the measurement of the sample becomes a proportional error in the result. The sample should be measured with a micrometer at several points, and the average should be used with the variation recorded.

Temperature, Moisture and Conditioning

Both the dielectric constant and the loss tangent change with temperature, and both change with moisture content. A board that is warm and humid has a higher dielectric constant than the same board in a dry, cool condition, which shifts impedances in a way that depends on the environment rather than on the design.

For this reason the test conditions should be stated, and a value quoted without them should be treated as approximate. Where the product operates over a wide temperature range, the design should be checked at both ends of the range rather than only at room temperature. The behaviour of specific materials is described in the guide to PTFE materials.

Reading a Material Datasheet

A material datasheet usually gives a dielectric constant and a loss tangent at one or two frequencies, a glass transition temperature and a decomposition temperature. The values are typical rather than guaranteed, and the test method is often described in a footnote that is easy to miss.

The useful reading is to compare the method and the frequency with the application before accepting the number. Where the datasheet does not state them, the supplier should be asked, because a value quoted at one megahertz cannot be used to design a line that operates at ten gigahertz without an adjustment.

Using the Value in Impedance Design

The value enters the impedance calculation together with the trace width, the trace thickness, the dielectric thickness and the presence of solder mask. Because several of those quantities are also uncertain, the calculation should be treated as a starting point that is verified on a coupon rather than as a final answer.

Verification is done with a test coupon and a time domain reflectometer or a network analyser, and the result is compared with the design intent. Where the measurement disagrees, the dielectric constant is one of the candidates, and it should be checked before the trace geometry is adjusted. The role of the coupon is described in the guide to the PCB test coupon.

Specification and Verification

The specification should state the value required, the frequency at which it applies, the direction of measurement and the test method. Without those four details the requirement is not a specification but a hope, and the supplier will meet it with whichever method gives a passing number.

Verification after delivery is possible by measuring a coupon from the production panel, and the result should be recorded with the lot. Where a design is sensitive to the dielectric constant, the coupling between the electrical requirement and the material choice should be reviewed with the fabricator before the stack is fixed, as described in the guide to laminate material properties.

Additional Considerations for This Build

Practical attention to resonant cavity pays for itself here, because it decides whether the finished board behaves as the drawing intended. Where the requirement is not stated on the fabrication drawing or in the assembly notes, the shop has to assume a default, and that default is rarely the value the design was simulated with. Stating resonant cavity explicitly, together with the tolerance that applies, removes the assumption and keeps the result predictable from batch to batch.

FAQ

Why do two datasheets give different values for the same material? Almost always because the measurement was made by different methods or at different frequencies. A value measured between parallel plates at a low frequency and one measured in a resonator at several gigahertz describe the same material under different conditions.

Does the dielectric constant change across a panel? It varies slightly because the glass weave and the resin are not uniformly distributed, and the variation is greater where the field is concentrated in the resin between glass bundles. For most designs the effect is small, and for very tight impedance control it has to be accounted for.

How should the value be used in a stack-up calculation? Use a value measured by a method and at a frequency close to the application, then verify the result on a coupon. Treating a datasheet figure as exact is the most common reason a stack-up calculation and a measurement disagree.