Glass Weave Effect Design Guide
The glass weave effect is a small variation in the dielectric constant of a laminate that comes from the fabric inside it. The weave is a periodic structure of glass bundles and resin rich gaps, and a trace runs over both. The dielectric constant under the trace therefore changes along its length, and on a high speed differential pair that change appears as a timing difference between the two lines.
What the Effect Is
A laminate is made of glass fabric impregnated with resin, and the two materials have different dielectric constants. The glass is around six and the resin is around three, so the local value depends on how much glass is under a given point.
The fabric is woven, so the glass forms bundles in two directions with gaps between them. A trace that runs across the weave sees a series of alternating regions, and the impedance of that trace varies along its length.
The scale of the weave is a few hundred microns, which is comparable with the width of a fine trace and with the spacing of a differential pair. That is why the effect matters on a modern high speed board and did not matter on an older one.
The effect is not a defect in the material. It is a property of any woven laminate, and the design has to accommodate it rather than assume it away.
Why It Causes Skew
Skew is the difference in arrival time between the two lines of a differential pair. It is caused by a difference in propagation delay, which depends on the dielectric constant under each line.
Two lines that are separated by one weave pitch can sit over different amounts of glass, so one sees a higher dielectric constant and propagates more slowly than the other. The delay difference accumulates along the length of the pair.
The effect is worst when the pair runs parallel to the weave for a long distance, because the two lines then stay in the same relative position over many weave periods. A pair that crosses the weave at an angle averages the variation out.
Skew also appears between the layers of a bus and between the traces of a parallel interface, where it reduces the timing margin of the whole link rather than of one pair.

Which Designs Are Affected
The effect matters where the unit interval is short and the skew budget is tight. A link of several gigabits per second has a budget of a few picoseconds, and the weave can consume a large part of it.
A narrow trace is affected more than a wide one, because a wide trace averages over more of the weave. A trace that is several times the weave pitch sees a nearly uniform dielectric, which is one reason wide traces are used for a low loss line.
The routing direction matters as much as the geometry. A pair that runs at an angle to the weave and changes direction several times is affected far less than one that runs straight along the weave for a long distance.
The effect also appears in a single ended line as a small impedance variation, which shows up as a reflection in a time domain measurement rather than as a timing error.
Reduction and Mitigation
The standard mitigation is a spread glass or a flattened weave, in which the bundles are spread so that the glass is more evenly distributed. The dielectric variation across the surface falls, and the skew falls with it.
The material is more expensive and less widely available, so it is used where the budget demands it. The high frequency laminate guide compares the materials that are available for this purpose.
Routing practice helps as well. A pair that is routed at a slight angle to the board axis, that changes direction, and that is kept short will accumulate far less skew than one that is routed straight.
A wider trace with a smaller gap is another option, since the wider line averages the weave. The trade is a larger via and a wider breakout, which may not fit in a dense area.
Measurement and Verification
The effect can be measured on a test coupon that carries pairs of different lengths and orientations. The skew is read from a time domain measurement, and the coupon shows how much the material contributes.
A coupon with pairs running parallel to the weave and at forty five degrees to it gives a direct comparison, and it is one of the most useful structures to include in a fabrication panel.
The measurement should be made on the material that will be used in production, since two laminates with the same dielectric constant can have very different weave structures.
The microstrip and stripline routing practice also affects the result, because the field distribution in the two structures is different and the weave affects them differently.

Practical Rules for a New Layout
Keep a pair short and let it change direction. A pair that is routed at an angle to the weave and that turns a few times averages the dielectric variation, while a long straight run parallel to the weave accumulates the worst case skew.
Widen the trace where the loss budget allows, because a wide line averages over more weave periods. The limit is the space available at the breakout, and a wide line also needs a wider via pad.
Consider the layer as well as the direction. A stripline is surrounded by two dielectric layers and sees an average of both, which reduces the effect compared with a microstrip that sees only one.
Finally, build a coupon. A skew structure costs almost nothing to add to a fabrication panel and it turns an argument about material into a measured number that the whole team can use.
Design Review
The review starts with the data rate and the skew budget, which decide whether the effect is significant for this product. A slow design can ignore it entirely.
The second item is the material, chosen for the weave structure as well as for the loss. The laminate properties table should be read with the weave in mind.
The third item is the routing, checked for direction, length and the number of layer changes. A pair that is spread across several layers accumulates skew from each of them, and the vias add their own contribution.
The last item is the coupon, which should include a skew structure if the budget is tight. The high speed design rules give the framework, and the coupon turns the framework into a number.
FAQ
Does the glass weave effect matter at one gigabit per second? It can, if the routing runs parallel to the weave for a long distance. The effect scales with the length and with the ratio of the weave pitch to the trace width.
What is spread glass? A laminate in which the glass bundles are spread or flattened so that the dielectric is more uniform across the surface. It reduces the skew at a higher material cost.
Can the effect be corrected in software? Partially, by adding delay to the faster line, but the correction is only as good as the measurement. Material and routing choices are more reliable.



