PCB Routing Techniques: Microstrip, Stripline and Impedance

Every trace on a board is a transmission line, whether the designer treats it as one or not. Once the edge rate of a signal is fast enough that the rise time approaches the propagation delay of the route, the trace stops behaving as a wire and starts behaving as a distributed structure with an impedance. Managing that structure deliberately is what separates a design that works reliably from one that works until the timing margin is tested.

Microstrip and Stripline

A microstrip is a trace on an outer layer with a reference plane on the layer beneath it. It is the easiest structure to produce and the easiest to adjust, because the trace is exposed and its width can be changed without altering the stackup. Its impedance depends on the trace width, the copper thickness, the dielectric height between the trace and the plane, and the dielectric constant of the material.

A stripline is a trace buried between two reference planes, so it is enclosed by copper on both sides. That enclosure gives it two advantages. The field is contained within the dielectric, so the trace radiates very little and is correspondingly immune to external coupling. And the impedance is determined by the geometry to both planes, which makes it more stable with respect to the manufacturing tolerances than a microstrip, whose impedance varies with the solder mask covering it. The general rules governing which layer a signal should occupy are described in this article on layer stackup design.

The trade-off is that a stripline occupies two reference layers and cannot be adjusted after the board is built. Microstrip is the practical choice for most nets, and stripline is reserved for the fastest and most sensitive signals.

Controlling Characteristic Impedance

Impedance control means choosing the geometry so that the trace presents a defined impedance to the driver, usually 50 ohms for a single-ended net and a defined differential value for a pair. The designer”s free variables are limited: the dielectric heights are fixed by the stack, the copper thickness by the fabrication process, and the dielectric constant by the material. That leaves the trace width, and for a stripline the width relative to the two dielectric heights.

This is why the stackup has to be settled before routing begins. A trace width calculated for a 4 mil dielectric is the wrong width on a 6 mil dielectric, and a design that uses the wrong value will miss its impedance target by enough to matter at high frequency. The calculation should be performed for the specific stack the fabricator will build, using the actual material properties rather than nominal figures from a datasheet.

Impedance also has to remain consistent along the route. A trace that widens to pass a component pad, narrows through a via field or changes layer introduces a local discontinuity, and each discontinuity reflects part of the signal. Where a change is unavoidable, it should be made gradually, and the section of altered geometry should be as short as possible.

Routing Choices That Protect Impedance

The most common cause of impedance variation is not the trace width but the reference. A trace that runs over a continuous plane has a predictable impedance; the same trace crossing a plane boundary, a slot or an antipad region does not. Rapid changes in the return path below the trace change the effective impedance for that section, and they also create the radiated coupling discussed in this discussion of power plane splitting.

Layer transitions deserve the same attention. A signal via that changes layers without a nearby ground via forces the return current to find its own path, which introduces inductance at the transition point and a local impedance change. Placing a ground via adjacent to every signal via on a fast net is a low-cost improvement that keeps the transition transparent. The clearance requirements around those vias are set out in this article on via to trace clearance.

Differential Routing

A differential pair is routed as two traces with a constant spacing, matched in length, referencing the same plane. The pair carries the signal as the difference between the two conductors, and the receiver rejects what is common to both. That rejection depends on symmetry, so any deviation in spacing or length converts signal into common-mode noise.

Three practical rules follow. The spacing must be constant through the whole route, including through any tuning structures, because the differential impedance is a function of that spacing. The two traces must be the same electrical length, which is not the same as the same physical length when they occupy different layers. And both traces must reference the same plane throughout, since a change of reference for one of them changes the impedance of that trace alone.

Where the pair must be tuned to match another pair, the compensation should be applied to both traces so that symmetry is preserved, and the spacing between adjacent sections of the tuning structure should be large enough that they do not couple to each other. The rules for that geometry are described in this article on serpentine routing and length matching.

Applying the Techniques

PCB routing that respects these structures is mostly a matter of preparation, and the preparation is finished before the first trace is drawn. The sequence that works is to fix the stackup, define the impedance targets and the net classes, calculate the widths, and only then start routing. The fastest nets should be assigned to the layers with the best reference, and microstrip should be used where the adjustment flexibility is valuable. Every route should be reviewed for reference continuity, spacing consistency and the presence of companion ground vias at layer transitions.

Where a design combines both structures on one board, the microstrip nets should be checked for mask-related impedance shift while the stripline nets should be checked for the dielectric tolerances that the fabricator can hold, because the two structures are sensitive to different process variables.

`r`n

Where the design includes a fast bus, the same discipline extends to the group as a whole, and the techniques described in this discussion of high frequency traces and data buses should be applied to the group rather than to the individual nets.

FAQ

Should I use microstrip or stripline for a fast signal? Stripline gives better shielding and more stable impedance because the trace is enclosed by planes, so it suits the fastest and most sensitive nets. Microstrip is easier to route and to adjust, and it is adequate for most signals provided the reference plane is close.

Why does the stackup have to be fixed before routing? Because the impedance of a trace is determined as much by the dielectric height beneath it as by its width. Changing the stack changes every impedance calculation, so a design routed against the wrong stack has to be reworked rather than adjusted.

What is the most common cause of impedance variation along a route? A change in the reference beneath the trace. Crossing a plane boundary or a slot changes the effective impedance for that section even if the trace width never varies, and it also degrades the return path.