Microstrip, Stripline and Coplanar Waveguide for RF Transmission Line Design

Wireless capability is now expected in almost every product, and the radio section of the board is where analog layout discipline matters most. Above a few hundred megahertz the copper is no longer a connection; it is a transmission line whose width, dielectric spacing and reference plane set the impedance. Treating an RF transmission line as ordinary routing is the fastest way to lose range, gain and regulatory margin at the same time.

Why RF Layout Is Different

At low frequency the voltage on a trace is the same everywhere and a bend is an inconvenience. Once the wavelength of the signal approaches the physical size of the circuit, the pad on an RF pin, the trace, the passive components, the via and even the ground pour under the trace all become part of the electrical network. A quarter wavelength at 2.4 GHz in a typical laminate is only about 15 mm, so a modest layout error is electrically significant.

The consequence is that RF work needs controlled impedance, a defined reference plane and a simulation or measurement step. It also explains why an RF designer may hand over a layout with deliberate angled corners and compensation features that look like mistakes to someone reading the same DXF as an ordinary board.

This is also why the RF section of a board is often designed as a block and then integrated. The boundary is chosen at a point where the impedance is defined, usually a connector or a matching network, so the rest of the layout can be handled with ordinary rules without disturbing the radio path.

Microstrip

Microstrip is a trace on an outer layer with a solid ground plane on the layer directly below. Its impedance is set by the trace width, the copper thickness, the dielectric thickness and the dielectric constant, and 50 ohm is the usual target for a single ended line. Because the field is partly in the dielectric and partly in the air above the trace, the effective dielectric constant is lower than that of the laminate itself.

Microstrip and stripline cross sections on an RF printed circuit board

Microstrip is the most common choice for RF work because it takes fewer process steps than an inner layer structure, it is easy to probe and it allows components to be placed directly on the line. Its weaknesses are that the exposed top surface radiates and picks up interference, and that the single reference plane gives less isolation than a structure with ground on both sides.

Stripline

Stripline places the conductor between two ground planes, so the field is confined to the dielectric on both sides. That confinement gives two useful properties: good isolation between neighbouring lines, and no radiation from the trace itself. It also means the effective dielectric constant is essentially that of the laminate, so the wavelength is shorter and the same impedance requires a narrower trace than microstrip.

The narrower line is a mixed blessing. It supports higher routing density, but it also demands tighter etch tolerance and a laminate whose dielectric constant is consistent across the panel. A stripline layer is also buried, so it cannot be tuned with a probe or a knife once the board is laminated, and the structure costs more because it consumes two reference layers.

Offset stripline, where the conductor sits closer to one reference plane than the other, is a useful compromise. It keeps the field confined and the isolation high while allowing a wider conductor than a symmetric stripline of the same impedance, which relaxes the etch tolerance at the cost of a slightly asymmetric field.

Grounded Coplanar Waveguide

A grounded coplanar waveguide puts ground copper on the same layer as the signal, on both sides of it, as well as on the layer below. This surrounds the trace with return copper and gives a lower impedance to the return current than microstrip for the same footprint, which is why it is widely used for the feed line between a transceiver and an antenna.

Grounded coplanar waveguide with via fence on an RF board

The layer below is not optional. The coplanar ground on the top layer is stitched to the lower ground plane with a fence of vias along both sides of the line, and that fence is what keeps the return current where it belongs instead of letting it spread across the board. The via spacing should be a small fraction of the operating wavelength, and the fence should be continuous along the line rather than present only at the ends.

Comparing the Structures

All three structures perform well into the millimetre wave region, and the choice is usually decided by cost, isolation and manufacturability rather than by raw performance. Microstrip is the cheapest and the most tolerant of process variation, because the impedance depends on features that the fabricator controls well and is affected relatively little by etch variation. Stripline offers the best isolation and the densest routing. Grounded coplanar waveguide gives a wide impedance range, low dispersion and lower radiation loss at high frequency, at the cost of a more complex ground structure and a stricter set of layout rules.

Because the electrical behaviour of all three is dominated by the laminate, the material decision is made first. A low cost FR4 will limit performance at high frequency through its loss and through variation in dielectric constant, and choosing a material that matches the frequency, the channel length and the cost target is a bigger lever than the choice of line geometry. The interaction between geometry, material and stackup is covered in the notes on microstrip and stripline routing and on high frequency trace and data bus routing.

The surrounding layout matters as much as the line itself. The reference plane beneath a microstrip has to be solid with no slot crossing under the line, the ground return for a coplanar structure has to be continuous, and any component placed on the line, including a matching element or a test pad, becomes part of the impedance and must be accounted for in the simulation.

Bends, Corners and Compensation

When a controlled impedance line has to change direction, the bend radius should be at least three times the conductor width. A gentle radius spreads the current path and keeps the local impedance close to the design value. Where the geometry leaves no room for a radius, a right angle bend can be used, but it must be compensated.

The reason is that a square corner presents additional copper at the outside of the turn, which lowers the local impedance and reflects part of the signal. The standard fix is a mitre, a chamfer cut across the outer corner that removes the surplus copper and restores the impedance through the bend. Where two lines run in parallel for any distance, coupling is also a design variable, and the general rules are set out in the notes on EMI suppression design principles.

Test access deserves the same caution. A probe pad or a connector footprint added to a controlled impedance line introduces a discontinuity, so where the line has to be measured, the access point is normally designed as part of the line with its own compensation rather than tacked on afterwards.

Impedance Control in Practice

Impedance control starts with the stackup. The fabricator needs the target impedance, the layer, the copper weight and the line width so that the dielectric thickness can be fixed, and the design should be released with a coupon that allows the impedance to be measured after fabrication. Once the stackup is frozen, the trace geometry has to stay within the tolerance the process can hold, which is why narrow stripline is more demanding than microstrip. The coupon measurement is the only way to confirm that the board in hand matches the model.

FAQ

Is microstrip or stripline better for an RF trace? Microstrip is cheaper and easier to tune. Stripline gives better isolation and no radiation, at the cost of an extra reference layer and tighter tolerances.

Why does a right angle corner need compensation? The corner adds copper outside the turn, which lowers the local impedance and causes a reflection. A mitre removes that surplus.

Do coplanar ground vias really matter? Yes. Without a via fence the coplanar ground is a floating patch, and the return current takes a longer path with more radiation and more coupling.

Leave A Comment