That is the question.
Is it better to route signals in microstrip or stripline traces? Most designers would say "It depends."
The answer really depends upon your definition of "better." So much of the design process is balancing tradeoffs, usually between acceptable performance and on-schedule delivery with acceptable risk and lowest cost.
This is why it is so dangerous to base designs on rules. Every design is custom. Design guidelines and rules of thumb are powerful tools to help zero in on possible design options, but ultimately, as design margins tighten up, every engineer needs to become proficient with the tools required to reach the optimum custom design as quickly as possible.
All we can do is point out the pros and cons of microstrip and stripline, illustrate the tools for exploring design space and ssuggest examples so you can decide what is "better."
Obviously, fewer traces can be routed in microstrip than in a stripline multilayer board, especially if the surface is already cluttered with components. This analysis applies to a situation where we have the opportunity to route a surface trace or a stripline trace, and we wish to evaluate which path is preferred.
For performance, what's important in a signal trace? In most high-speed designs, the first-order factors are controlled impedance, crosstalk and attenuation, which influences interconnect bandwidth.
What's not on the list is EMI. While it is correct that a microstrip signal trace will radiate more than a stripline, due to the finite total inductance in the return plane, very rarely do products fail FCC certification tests because of microstrip traces. They fail because of common signals on external cables; either shielded cables or twisted pair. If you have this problem nailed and do not have a shielded enclosure but still have an EMI issue, then consider stripline over microstrip.
Most fab houses don't do a very good job of controlling the impedance of outer surface layers, but there is nothing inherent about microstrip that would make it a poor controlled impedance interconnect. It is more an issue of risk with the fab house, which can be reduced by qualifying your supplier.
One advantage of microstrip is that to achieve an impedance of 50 Ω in a single-ended line, the dielectric thickness is about half the line width when using FR-4. For stripline, to achieve 50 Ω for a single-ended line requires a total dielectric thickness between the two reference planes of about twice the line width.
In a differential pair, the difference is even more dramatic. You can use a pair of traces with tighter coupling and thinner dielectric in microstrip than in stripline traces. For example, with .005" wide, half-ounce traces, in edge-coupled microstrip with a spacing equal to the line width, the dielectric thickness is .0035". In edge-coupled stripline, with the same traces, you can't even get as high as 100 Ω with a space as tight as the line width. If we go to a space twice the line width, the total dielectric thickness is .028".
Stripline traces will have very little far-end crosstalk, while microstrip can have enough to easily exceed noise budgets. This is a real issue in the decision process. If far-end noise is a problem, don't use microstrip.
But when the trace-to-trace spacing is less than twice the line width, stripline actually has more near-end noise than microstrip. For spacings larger than twice the line width, stripline has lower near-end noise.
FIGURE 1. Electric field lines in stripline and microstrip showing the contribution of low loss air in microstrip.
FIGURE 2. Attenuation in equal line width edge-coupled differential pairs in microstrip and stripline, calculated with the Polar Instruments SI9000 field solver.
|
Attenuation is subtler. For the same 100 Ω differential impedance pairs, with .005" wide lines, the conductor loss is almost comparable between microstrip and stripline. However, the dielectric loss is lower in microstrip than stripline for FR-4. FIGURE 1 illustrates why. Some of the field lines in microstrip are in air, where they see a lower dissipation factor than the bulk laminate. This gives microstrip about 30% lower attenuation than stripline, which means, potentially, a higher interconnect bandwidth by 30%.
FIGURE 2 shows the attenuation per length for microstrip and stripline. Above about 1 GHz bandwidth, there is an attenuation advantage for microstrip lines.
No decision is based on one answer. There are too many factors to be balanced in the tradeoffs. PCD&M
Dr. Eric Bogatin is the CTO of IDI and president of Bogatin Enterprises; This email address is being protected from spambots. You need JavaScript enabled to view it..
