Still good bang for the buck: OI' faithful FR-4 shows no signs of fading into that good night.

I often hear the questions: When will copper reach its limits? When will it be necessary to switch to optical interconnects? Actually, we should be asking when FR-4 will reach its limits, and what will we do about it? The answer is, pretty." soon. Fortunately a new generation of materials with an attractive price/performance ratio is ready to give FR-4 a "mid-life kick."

The limitations of FR-4 are most noticeable in high-speed serial links with data rates in excess of 2.5 Gbps and in backplane traces longer than 12 inches. As a signal propagates down a transmission line built with FR-4 laminate, the higher-frequency components will be attenuated more than the lower-frequency components, due to the dielectric loss of the material.

This frequency-dependent loss results in the bandwidth of the transmitted signal being decreased and its rise time increased. Rise time degradation sets the ultimate limit to the data rate that can be transmitted in FR-4. It gives rise to effects such as intersymbol interference (ISI), collapse of the eye diagram and deterministic jitter.

All dielectrics show loss; it's a question that describes the amount of loss in a laminate is the material's dissipation factor. FR-4 has one of the highest dissipation factors of all laminates at roughly 0.02, and relatively constant with frequency. Lower-loss materials result in less rise time degradation, and the ability to transmit higher bit rate signals.

Nothing in life is free, and the lower the loss of a laminate, and the higher the bit rate it will support, in general, the more expensive it is. The proliferation of high-speed serial links, driven by Internet protocol, communications and higher speed processor busses, has accelerated the development and introduction of new materials. They fall into two classes: mid-range loss, with dissipation factors on the order of 0.008, and low-loss materials, with dissipation factors on the order of 0.004.

While lower-loss materials are more expensive, the copper-clad laminate version of the material is much more expensive than the prepreg version of the material used to "glue" layers together.

W. L. Gore and Associates was one of the first laminate suppliers to provide a cost-effective compromise. Instead of building an entire board with the lower loss and more expensive laminate and prepreg, their solution is to just use a low-loss prepreg for the layers that will carry the high data-rate signals. Their Speedboard C Prepreg (SBC) is a non-woven, expanded PTFE fiber-reinforced BT resin with a dielectric constant of 2.6 and a dissipation factor of 0.004. Its two key qualities are very low loss and compatible processing with FR-4.

When used in a stripline construction, the SBC layer is on one side of the signal line and standard FR-4 is on the other. This is called a mixed dielectric or "hybrid" stripline. A key innovation in this design is the ability to optimize the layer thicknesses not only to achieve a target impedance, but also to achieve a target data rate.

The thinner the SBC prepreg layer, the more the signal's electric fields are concentrated in the low-loss SBC layers, resulting in lower overall loss (FIGURE 1). Of course, as the laver becomes thinner, the signal trace line width must decrease in order to meet the impedance target. This results in higher resistive losses. There is an optimum SBC thickness that is design-dependent.

[FIGURE 1 OMITTED]

In simulation and test board evaluations, Gore has found the loss performance of hybrid stripline transmission lines to be as good as or better than homogeneous striplines fabricated with mid-range lossy materials with a dissipation factor of 0.008. FIGURE 2 shows the measured eye diagram for a pseudo-random bit stream at 10 Gbps, after traveling 13 inches down various striplines.

[FIGURE 2 OMITTED]

Not only is the signal quality better than that of mid-loss laminates, the total board cost can often be less. Though the SBC prepreg is on the order of 6X more expensive than an FR-4 prepreg, only one SBC layer is used per signal layer so the total board cost adder can be kept below 30% for some constructions. This compares to more than 50% higher cost if the entire board was made with a mid-loss material capable of the same data rate. With the integration of low-loss prepreg materials such as SBC, to paraphrase Mark Twam, "the reports of FR-4's death may be greatly exaggerated."

REFERENCE

Noel Hudson, "Potential Benefits of Mixed Dielectric Stripline," Proceedings of DesignCon 2004.

ERIC BOGATIN ([e-mail]) is the CTO at IDI and president of Bogatin Enterprises. Many of his papers are available on his Web site, www. Be TheSignal.com.