PCB Material Selection: Dielectric, Thermal and Structural Properties
Every thermal, mechanical and electrical behaviour of a board is set by the material it is built from. Copper geometry gets most of the attention during layout, but the substrate decides the impedance, the loss, the dimensional stability and the temperature at which the assembly stops behaving predictably. Serious PCB material selection means choosing the laminate for the environment and the signal, not defaulting to the cheapest grade that will pass a prototype.
Electrical Properties: Dielectric Constant and Loss
The dielectric constant of the substrate sets the propagation velocity of a trace and, together with the trace geometry, the characteristic impedance. It is not a single number. The value varies with frequency, with temperature and with the resin to glass ratio, and it differs between the woven glass and the resin that fills the weave. Two laminates described as having the same nominal dielectric constant can behave quite differently on a 25 GHz channel.
Loss follows the same pattern. The dissipation factor, usually quoted as loss tangent, sets how much energy the dielectric absorbs per unit length, and it rises with frequency. For a short low speed interconnect the difference between a standard grade and a low loss grade is invisible. For a 10 Gbps channel running 300 mm, it decides whether the link closes.
Dispersion and Signal Distortion
Because the dielectric constant is a function of frequency, different frequency components of the same signal travel at different speeds. A digital edge contains a broad spectrum of components, so dispersion spreads the edge and closes the eye diagram. The same mechanism makes attenuation frequency dependent, which is why a long channel needs equalisation.
For a narrowband analogue signal at a single frequency, dispersion is irrelevant. For a multi gigabit serial link it is one of the dominant design problems, and it is the reason a material with a flatter dielectric response commands a price premium. Measuring a coupon and modelling the channel with the supplier data is far cheaper than discovering the problem at compliance test.
Structural Properties: Glass Weave and Copper Roughness
The weave of the glass cloth leaves gaps between bundles that fill with resin, and the two materials have different dielectric constants. A trace that runs along a bundle sees a different effective dielectric constant from a trace that runs across the weave, and the difference appears as skew between the two traces of a differential pair. The effect is small below a few gigahertz and becomes significant above roughly 50 GHz, where it matters for radar, multi gigabit Ethernet and high speed serial channels.
Copper roughness is the second structural factor. Signal current crowds into the surface of the conductor at high frequency, and a rough surface makes the current path longer, which raises the effective resistance and the loss. The foil treatment, the plating process and the prepreg surface all influence how rough the finished conductor is, so the loss figure quoted for a laminate assumes a particular foil type.
Spread glass and flat weave laminates were developed to reduce the weave effect. Flattening the bundles or spreading the filaments makes the dielectric environment under a trace more uniform, which lowers the skew between the two halves of a differential pair without changing the nominal dielectric constant. The benefit is confined to the highest frequency channels, so the material is specified where the channel length or the data rate cannot be met any other way.
Thermal Properties: Conductivity, Specific Heat, CTE and Tg
Thermal conductivity sets how quickly heat moves through the substrate, and specific heat sets how much energy is required to raise its temperature. Together they determine where the board settles once it reaches equilibrium with its environment. A board that must dump heat into a chassis or a heat sink needs a substrate that conducts well, which is the reason metal core and ceramic substrates exist.
The CTE, or coefficient of thermal expansion, describes how much the material grows per degree of temperature rise, and it is anisotropic: the in plane expansion is governed largely by the glass, while the out of plane expansion is governed by the resin and is much larger. The glass transition temperature marks the point where the resin softens and the out of plane expansion rate jumps. A standard grade transitions near 130 degrees Celsius, and most suppliers offer a 170 degree grade that keeps the board in its stable region through lead free reflow.
The practical consequence is at the vias. The barrel expands with the plating while the surrounding resin expands much more, so the plating is stretched on every thermal cycle. High aspect ratio holes in a thick board with a low glass transition temperature are the worst combination, and that is exactly the case where a high Tg material or a different via structure earns its cost. The wider set of high speed considerations is covered in the notes on high frequency trace and data bus routing.
Mechanical Considerations
Stiffness, thickness tolerance and dimensional stability under heat all follow from the resin system and the reinforcement. A board that has to hold a fine pitch package through reflow without shrinking needs a laminate with predictable expansion in the plane, and a board that will be flexed needs a material and a construction chosen for that, not a rigid grade used in a thinner form. The routing method also depends on the construction, and microstrip and stripline behaviour is described in the notes on microstrip and stripline routing.
Soldermask and surface finish are part of the same decision chain. The finish determines the storage life and the number of reflow cycles the pads will survive, and it interacts with the surface roughness of the copper underneath. A finish that consumes copper, such as a thick immersion tin, changes the conductor geometry slightly and therefore the impedance of a fine trace.
Matching the Material to the Application
A useful way to decide is to write down the constraints in order: the highest signal frequency and the longest channel, the maximum ambient temperature, the number of thermal cycles expected over the product life, and the mechanical loads. Those four answers narrow the choice quickly. A consumer product with a short channel and moderate temperatures is well served by a standard grade. A base station line card, an automotive controller or an aerospace unit moves the requirement up the material ladder, and the stackup has to be balanced at the same time, as described in the notes on balanced stackup design.
Working With the Laminate Supplier
Published material data is a starting point, not a specification. Two laminates with the same generic description can differ in resin content, weave style and foil treatment, and the differences show up in impedance tolerance and in loss. Ask the supplier for the dielectric constant and loss data at the frequencies that matter, for the expansion figures in both axes and for the recommended process window. Where the design is sensitive, the data should be verified with a test coupon from the actual production laminate rather than accepted from a datasheet.
Cost should be considered last and proportionally. A low loss laminate can double the material cost of a board, which is significant on a large backplane and negligible on a small module where the alternative is a link that does not close. The right question is not whether the expensive material is better, but whether the cheaper one can meet the channel, the temperature and the lifetime that the product actually requires.
FAQ
Is FR4 good enough for high speed digital signals? For short channels at moderate data rates, yes. As the bit rate and the channel length rise, the loss and the weave effect move the requirement to a low loss or a spread glass material.
Why does the glass transition temperature matter? Above it the resin expands much faster out of plane, which loads plated holes on every thermal cycle and shortens the fatigue life of the via.
Does a higher dielectric constant mean a worse material? Not by itself. What matters is whether the value is stable over frequency and temperature and whether the loss is low enough for the channel.
