Glass Weave Skew. For most people it probably sounds like a progressive rock band. But we’re talking about Glass Weave Skew and differential signals.
Brandon Gore, Senior Staff Signal Integrity Engineer, and the Manager of Samtec’s Signal Integrity Group, R&D Hub recently presented “A Vehicle For Insitu Glass Fabric Characterization” at EDI CON USA 2017. The research was completed by Brandon and supported by Scott McMorrow, Samtec’s CTO of Signal Integrity Products.
In his presentation, Brandon explains how he and Scott characterized through measurement the impact of the “fiber weave skew” in a printed circuit board panel. Printed circuit boards are interdigitated layers of copper cladding and dielectric. This dielectric is typically an epoxy resin which is impregnated with fiber glass cloth.
A degradation to differential signaling can be realized because of the local dielectric constant difference between fiber glass (Dk ~ 5 to 6) and epoxy resin (Dk ~ 2.5-3). Since dielectric constant controls the signal propagation speed in a material, the “p” trace and “n” trace of differential pair could be electrically different lengths. Ideal differential signaling requires that the “p” and “n” traces be perfectly matched. Otherwise, differential signal attenuation and common mode generation become a concern. Like most signal integrity issues, the impact of this skew or offset between signal ended signals within a differential pair depends on the data rate.
Our goal was to determine the maximum delay for signal traces routed at several angles to the weave and incrementally spaced to “sample” the effective dielectric constant below the trace. This way we can relate back to the maximum delay that a differential signal could experience for a particular orientation to the glass weave.
We started from best practices in E-glass fabric selection in which we chose a tight weave that is mechanically spread, has a high resin content, as well as used multiple plys in our buildup layers. Namely, these fabrics are 1035 and 1078. The dielectric chosen was Panasonic Megtron 6. Our findings were that, even for a mechanically spread glass weave, routing in the grain direction of the fabric without any trace rotation can accumulate 1.7 ps/inch for 1035 glass and 5 ps/inch for 1078 glass. This contrasts to a skew of 0.2 ps/inch when routing at an angle to the weave or routing in the fill direction.
We also compared the “on-pitch” differential routing strategy (pairs having a signal trace pitch equal to the glass weave pitch) to that of having the differential routing “half-pitch” for the glass. We concluded that matching the pitch of the signal traces to the pitch of the glass is a good strategy for mitigating skew when routing in the grain direction. However, routing with a signal pitch that is “half-pitch” of the glass weave generated 3 ps/inch for 1035 and 6 ps/inch for 1078. This is more skew than we observed in the signal ended experiments because of the relationship between how narrow the trace widths are to how nestled between the glass bundles they can become.
Our study showed that if differential traces are routed in non-ideal spacing and directional configurations w.r.t. the PCB glass weave direction, that the skew developed can be unacceptable for high performance 28/56/112 Gbps systems. Samtec would propose that for higher performance systems, a Flyover™ twin-ax cable is the best choice for significantly reduced skew and attenuation mitigation. Samtec Eyespeed® twinax delivers less than 3.5 ps of skew per meter which is greater than a magnitude of skew mitigation compared to alternative substrate materials. Flyover Twinax is also an enabler for new modular system architecture options beyond the constraints of Cu traces laminated to dielectric substrates