Estimated reading time: 4 minutes

Regional Differences – a Voyage of Glass Reinforcement

Your experiences are colored by the region you inhabit, and when you travel through your local region you probably encounter threads of different cultural experience. Even on a local level a street may have an Italian or a Chinese culture woven within the fabric of the city… Threaded though printed circuit substrates are layers of glass cloth providing dimensional stability and strength to the composite as a whole and whilst discovering Italian, Indian or Chinese restaurants in a city is a pleasure, the discovery of a network of electrical variation in a substrate needs to be understood and catered for in the design. Back in the 1960s could the originators of the material have ever conceived that this glass “insulator” would need to be considered from an electrical perspective other than its simple lack of conductivity?

A number of factors are converging to the point where you do have to take the composite structure of base materials into account when you look at the high speed performance of a PCB. The two primary drivers are the increase in data rates to the point where transmission line characteristics need to be understood, and the decrease in line width, to a point where the traces are now small compared to the underlying bundles of fiberglass reinforcing the composite epoxy substrate.

Being composed of two quite differing materials, most woven glass substrates will exhibit quite different electrical characteristics on a “local” scale. Now that trace sizes are small compared with the fibre bundles, the material characteristic “seen” by a signal can vary depending on the type of structure used and its physical dimensions. In FR4 the epoxy may have an E_r of 3.6 while the glass E_r is typically more than 6. Figure 1 illustrates the cross section of a typical controlled impedance structure in FR-4.

So the answer to “What is the E_r of that material?” is “It depends”. Is it the bulk E_r, as seen across a sample in a split post resonator, or the actual E_r seen by the signal as it transits along a trace? Striplines will have fields propagating to both adjacent planes, but coplanar structures will have fields that extend to the sides of the trace. A coplanar structure may “see” far less glass than a stripline structure, not because of any change in material, but because the shape and density of the field is traveling through more resin than glass. In addition to the structure type, when considering edge-coupled differential structures, more tightly coupled designs (which have the pair closely spaced will have a more intense field between the lines, in a region which will likely be predominantly filled with resin – again reducing the effective dielectric constant of the structure for that particular scale and geometry.

None of the above result from problems with the base material, they are simply effects that are inherent when the substrate is a composite made from two or more quite differing electrical properties.

In microwave or power applications where the traces may be much larger to reduce skin effect, the size of the trace may be large compared with the underlying fibre bundles, and in this situation the structure may behave much more as if it is built on a homogeneous base.

You may have read about fibre weave mitigation in other articles; this is simply a technique of either rotating the PCB substrate so the glass weave is no longer parallel with the traces or routing critical traces at an angle of 11 degrees or so in order to avoid the traces lining up with a fibre bundle or a resin rich area. Some laminate suppliers are working hard on materials with a tighter weave to reduce this effect. Another possibility is the use of glass materials with a lower E_r to reduce the mismatch in electrical properties of the substrate.

Fortunately, impedance only varies as 1/√E_r, so this variation is a second order effect; however, when tolerances are tight it is important to know where sources of modelling error may creep in. When modelling any structure it is worth looking at the range of E_r your signal may experience to ensure that the fabricator can hold the line width and dielectric separation tightly enough to allow for the possible variation of structure dependent E_r. Figure 2 illustrates bulk and local E_r. Modelling of the design space can ensure you choose structures and dimensions that are least sensitive to process changes. 

In summary, you need to consider that whilst engineering tools may lead you to think your designs work with “ideal” materials obeying the laws of physics to the nth degree, the actual materials you work with are, as with all engineered products, a compromise trading off price / reliability / signal integrity. Being aware of the possible variations is the first step in ensuring you design in the headroom in your design to allow for normal process variations, this avoiding any unexpected surprises in your final design.