Microstrip vs. Grounded Coplanar Waveguide

Choosing between microstrip and grounded coplanar waveguide (GCPW) is often reduced to a frequency question: use microstrip at lower microwave frequencies and move to GCPW as the design approaches millimeter wave. That rule is convenient, but it is not a reliable design method.

The two structures differ in more than where the ground copper is placed. They establish different electromagnetic field distributions, different return-current paths, different sensitivities to conductor geometry, and different mechanisms for radiation, dispersion, and parasitic-mode excitation. Those differences become increasingly important when the transmission line is no longer electrically isolated from connector launches, package transitions, ground-via arrays, solder mask, plating, and normal PCB fabrication tolerances.

The relevant question is therefore not simply which structure has lower loss or supports higher frequency. It is which electromagnetic environment remains controllable after the ideal transmission line is turned into a fabricated interconnect.

The Important Difference Is Field Topology

A conventional microstrip consists of a signal conductor above a continuous reference plane. Its dominant propagation mode is quasi-TEM: the electric and magnetic fields are not perfectly transverse because the fields occupy two media – substrate and air – but the transmission line can still be treated approximately through distributed capacitance and inductance over a useful frequency range.

In a GCPW, the signal conductor is bordered by ground conductors on the same layer while a reference plane remains underneath the dielectric. The top grounds are normally connected to the lower plane with vias. The resulting structure is not simply a microstrip with extra copper beside the trace. The lateral gaps become major capacitive regions, and a substantial fraction of the field can be concentrated between the center conductor and the coplanar grounds.

That distinction changes the relative importance of the primary geometric parameters. For microstrip, conductor width W, dielectric height h, copper thickness t, and dielectric properties dominate first-order impedance behavior. GCPW adds another highly influential dimension: the signal-to-ground gap S. Ground width, via position, via pitch, copper thickness, and the geometry beyond the nominal coplanar gap can also become relevant as frequency increases.

This is why a GCPW should not be designed by taking a known microstrip width, adding ground pour beside it, and adjusting the gap until a transmission-line calculator reports 50 ohms. The altered field topology means that the complete cross section has changed.

Characteristic Impedance Is Only the Beginning

For a general lossy transmission line, characteristic impedance can be written as:

Z0 = sqrt[(R + jωL) / (G + jωC)]

where R, L, G, and C are the per-unit-length resistance, inductance, conductance, and capacitance.

At sufficiently low loss, the familiar approximation Z0 ≈ sqrt(L/C) becomes useful. But two transmission lines that both calculate to 50 ohms can still have very different current density, effective permittivity, attenuation, dispersion, and sensitivity to manufacturing variation.

That distinction becomes important when comparing microstrip with GCPW. In microstrip, much of the electric field couples vertically to the lower ground plane. In tightly coupled GCPW, more field energy is concentrated laterally across the coplanar gaps. Because a larger fraction of that field can occupy air, the effective dielectric constant of a GCPW may be lower than that of a microstrip fabricated on the same laminate.

The practical consequence is that impedance equivalence does not imply propagation equivalence. Two nominally 50 ohm lines can exhibit different phase constants:

β ≈ (ω / c) sqrt(εeff)

and therefore different electrical lengths for the same physical distance.

This matters in phase-matched networks, couplers, filters, phased-array feeds, calibration structures, and any design in which phase error is as significant as insertion loss.

Loss: Why “GCPW Has Lower Loss” Is an Incomplete Statement

A useful first-order loss budget separates attenuation into several mechanisms:

αtotal ≈ αc + αd + αr + αmode

where the terms represent conductor loss, dielectric loss, radiation loss, and losses associated with unwanted mode conversion or resonant behavior.

This decomposition immediately explains why there is no universal low-loss winner.

Conductor Loss

Microstrip can have an advantage in conductor loss because its current distribution is not forced as strongly into two narrow lateral edge regions. In GCPW, especially when the coplanar gaps are small, current crowding develops along the signal edges and the adjacent ground edges. At microwave frequencies, where skin effect already confines current to a shallow region of the conductor, this concentration can increase effective resistance.

Consequently, a tightly coupled GCPW can exhibit greater conductor loss than a comparable microstrip even though its radiation behavior is better.

Copper roughness further complicates the comparison. As skin depth becomes small relative to the conductor surface profile, the actual RF current path becomes electrically longer and more resistive than an ideal smooth-conductor model predicts. Copper profile, plating, and surface finish should therefore be included in loss analysis rather than represented solely through bulk copper conductivity.

Dielectric Loss

Dielectric attenuation depends on the material loss tangent and on how much electromagnetic energy resides in the dielectric. Because the field distribution differs between microstrip and GCPW, the same laminate does not necessarily contribute the same dielectric-loss fraction in both structures.

This is one reason a laminate datasheet Dk or Df should not be treated as a complete transmission-line model. Effective dielectric constant is structure dependent, and the appropriate material values for simulation should correspond as closely as possible to the test method, frequency range, copper system, and transmission-line geometry being modeled.

Radiation Loss

The balance changes at sufficiently high frequency. Microstrip becomes progressively more susceptible to radiation as electrical dimensions increase. Substrate thickness and dielectric constant affect this behavior: an electrically thick structure permits stronger fringing and can support radiation more readily than a thin, well-confined line.

GCPW provides lateral ground conductors that help confine the field around the signal path. Properly implemented, this can reduce radiation and coupling to neighboring structures, particularly as the operating frequency moves into the millimeter-wave regime.

This is one of the strongest reasons GCPW becomes attractive at high frequency – but it should not be confused with saying its conductor attenuation is automatically lower.

Dispersion and Phase Stability

Both structures are commonly described as quasi-TEM transmission lines, but neither is perfectly nondispersive. Their effective permittivity can change with frequency because the field distribution between dielectric and air changes as the electromagnetic dimensions of the structure increase.

Microstrip dispersion becomes more relevant as frequency rises and the substrate becomes electrically thicker. The resulting frequency-dependent phase velocity can matter well before attenuation alone becomes unacceptable.

A properly designed GCPW can exhibit lower dispersion at millimeter-wave frequencies because the fields are more tightly controlled around the coplanar conductor system. This makes GCPW attractive in broadband and phase-sensitive networks.

However, phase stability must be considered together with fabrication sensitivity. A topology that has favorable intrinsic dispersion but shifts significantly with gap etch variation may not deliver better phase repeatability in volume production.

For an engineering comparison, the important quantity is therefore not merely simulated effective permittivity. It is the distribution of effective permittivity and phase delay after realistic geometry and material tolerances are applied.

GCPW Trades Field Confinement for Geometric Sensitivity

One of microstrip’s strongest practical advantages is its relative tolerance to fabrication variation.

For a given laminate construction, microstrip impedance is primarily controlled by trace width, dielectric thickness, copper thickness, and dielectric constant. All of these require control, but the geometry is comparatively simple.

GCPW introduces at least one additional high-sensitivity fabrication variable: the coplanar gap.

If the nominal signal-to-ground spacing is small, an absolute etch error that would be modest relative to the microstrip trace width may represent a significant percentage of the GCPW gap. Because that gap contributes strongly to capacitance, the impedance shift can be substantial.

Copper thickness also affects the two structures differently. In a thin-conductor analytical model, conductors are often treated as two-dimensional sheets. Real PCB copper has finite thickness and etched sidewalls. In tightly coupled GCPW, those vertical conductor walls participate directly in the lateral field. Increasing finished copper thickness can therefore modify both capacitance and effective dielectric constant more strongly than expected from a zero-thickness model.

The etched conductor is also not perfectly rectangular. Depending on the foil, plating process, resist, etching chemistry, and etch compensation, the top and bottom widths can differ. At millimeter-wave frequencies, assuming an ideal rectangular conductor can introduce meaningful error.

For production design, the geometry passed to the EM solver should therefore approximate the geometry the fabricator will actually build – not simply the nominal Gerber width.

A Useful Sensitivity Exercise

For a serious GCPW design, nominal impedance simulation is not enough. Sweep at least:

• signal width,
• signal-to-ground gap,
• finished copper thickness,
• dielectric thickness,
• dielectric constant,
• etch profile where relevant,
• via diameter and via location, and
• solder mask geometry if mask exists near the RF line.

The result should be viewed as a sensitivity space rather than as a single impedance number.

This is especially important when the nominal design obtains 50 ohms by using an extremely narrow coplanar gap. A geometry that produces an elegant simulation but leaves almost no manufacturing margin may be inferior to a slightly larger structure whose impedance is less sensitive to process variation.

The Ground Vias Are Part of the Transmission Line

This is where many practical GCPW layouts become electromagnetically different from the topology that was simulated.

The top ground conductors cannot simply be assumed to remain at the same RF potential as the lower reference plane. Their connection is established through a distributed array of vias, and each via has finite inductance. The geometry of that array therefore contributes to the boundary conditions of the transmission line.

Via fences serve several purposes. They reduce the effective inductive distance between the top and bottom grounds, help maintain symmetry between the two coplanar ground regions, limit field spreading, and suppress parasitic propagation in the cavity formed by the grounded structure.

But the usual statement that vias should be spaced at ‘lambda/10’ is only a starting heuristic.

The relevant wavelength is not necessarily the free-space wavelength, and the acceptable pitch depends on the structure being suppressed, dielectric system, via diameter, ground geometry, and required operating bandwidth. Published engineering guidance uses values around lambda/8 to lambda/10 in some structures, while experimental work at submillimeter-wave frequencies has required still tighter spacing. These are design references, not universal physical constants.

Via location relative to the coplanar gap matters as well. Moving the vias too close to the lateral ground edge perturbs the desired field and changes impedance. Moving them too far away weakens control over the local ground structure and increases the region in which unwanted fields can develop.

For broadband or millimeter-wave work, via position and pitch belong in the full-wave model.

GCPW Has More Than One Possible Mode

A particularly important distinction between idealized GCPW analysis and a fabricated grounded structure is the possibility of unwanted modes.

Adding the backside ground plane creates a region in which fields can propagate in ways that are not represented by the desired dominant GCPW mode. Parallel-plate-like and common-mode behavior can be excited, particularly at discontinuities, transitions, asymmetric structures, and inadequately grounded sections.

This is not simply additional ‘loss.’ Energy transferred into another mode can propagate with a different phase constant, reflect at a later discontinuity, and recombine with the intended mode. The result can appear as narrow resonances, unexpected return-loss degradation, or strong frequency-dependent behavior that an ideal transmission-line model never predicted.

Recent modal analyses of microstrip-to-GCPW structures have shown that resonant behavior can be understood through the relative phase accumulation of simultaneously excited modes. That observation is important because it explains why a line may appear well matched over much of the band and still develop a sharp resonance at a particular frequency.

Via fencing helps move or suppress these effects, but it should not be treated as a substitute for understanding where mode conversion occurs.

Connector Launches Often Favor GCPW – But the Launch Must Be Designed as a 3D Structure

One of GCPW’s most practical advantages appears at the transition from coaxial connectors or RF packages to the PCB.

A coaxial transmission line brings both signal and return conductors to the same physical region. Pure microstrip, by contrast, places the signal on the top surface while the principal ground plane is below the substrate. If the connector body or ground contacts cannot establish a short RF return path to that plane, the launch develops additional inductance.

A local GCPW launch can provide top-layer ground immediately beside the signal conductor, reducing the distance over which the return current must travel before reaching the connector ground.

This is why an otherwise microstrip circuit may intentionally use GCPW only in the connector-launch region.

However, the launch cannot be optimized by matching the nominal connector-pin width to a 50 ohm PCB trace.

At microwave and millimeter-wave frequencies, the launch is a three-dimensional field transition involving:

• connector center conductor geometry,
• connector ground-contact spacing,
• local PCB signal width,
• coplanar gap,
• ground-pad geometry,
• edge plating or ground contact where applicable,
• ground-via position and pitch,
• solder volume, and
• the transition from the launch geometry into the uniform transmission line.

A launch may therefore use dimensions that would not independently calculate to exactly 50 ohms. Local capacitive and inductive features can deliberately compensate one another over a finite physical length to produce the required broadband response.

For critical RF launches, PCB prototype manufacturing provides a practical way to verify the simulated transition on real hardware before the geometry is released for volume production.

This is an area where 3D full-wave optimization is far more meaningful than a two-dimensional transmission-line calculator.

Microstrip Still Has an Important Production Advantage

The increasing use of GCPW in high-frequency modules can create the impression that microstrip is an older topology that should be replaced whenever board space permits. That conclusion misses one of microstrip’s strongest engineering attributes: simplicity.

A well-designed microstrip has fewer geometric degrees of freedom, requires no continuous coplanar gap, and does not depend on a distributed via fence to establish its fundamental reference structure. This generally makes the relationship between fabrication tolerance and electrical behavior easier to predict.

At frequencies where radiation, isolation, and launch geometry are still manageable, that robustness can outweigh the additional field confinement of GCPW.

This is particularly relevant for filters, couplers, matching networks, and transmission-line sections whose dimensions are already sensitive to manufacturing variation. Adding another strongly coupled dimensional variable is not automatically beneficial.

In other words, GCPW should be selected because its electromagnetic boundary conditions solve a specific problem – not because ground copper beside an RF trace appears inherently safer.

Solder Mask and Surface Finish Affect the Two Structures Differently

Both microstrip and GCPW have substantial electric fields at the top surface of the board. Any dielectric placed over that region becomes part of the electromagnetic environment.

Solder mask therefore changes effective permittivity and adds dielectric loss. The effect becomes especially relevant where fields are strongly concentrated close to the conductor edges.

For GCPW, this includes the lateral gaps – the region that often carries some of the strongest electric fields in the entire cross section. Filling or partially covering those gaps with solder mask changes the capacitance and can alter both impedance and attenuation.

If production boards will contain solder mask close to the line, modeling an exposed-copper geometry and then fabricating a masked structure is not a controlled comparison.

Surface finish requires similar care. Nickel-containing finishes can increase RF conductor loss compared with bare or low-resistance copper surfaces because the RF current interacts with the plated conductor system rather than with bulk copper alone. The impact depends on frequency, finish thickness, line geometry, and current distribution.

Because GCPW concentrates current along conductor edges, tightly coupled structures can be particularly sensitive to such changes.

Do Not Compare the Structures Using Trace Loss Alone

A straight-line coupon is useful, but it can make both structures look better than the final circuit.

Real RF layouts contain bends, component pads, DC-blocking capacitors, shunt elements, package launches, tees, transitions, connectors, and ground discontinuities. Each one changes the local field distribution.

The comparison should therefore include the discontinuities that actually define system performance.

For example, a microstrip line may show slightly lower attenuation per unit length than a GCPW coupon but require a more difficult coaxial launch. Conversely, a GCPW line may provide an excellent launch and strong isolation while developing an avoidable resonance because its ground-via arrangement was not designed for the full operating band.

Neither result is captured by comparing only dB/mm of uniform transmission line.

Measurement: Separate the Line From the Fixture

When measured performance does not match simulation, the first suspect should not automatically be the laminate Dk.

A meaningful validation plan distinguishes the uniform transmission line from its launches and fixtures.

For impedance, TDR can locate discontinuities along the physical path rather than reducing the entire structure to one averaged 50 ohm number. For frequency-domain performance, a VNA provides insertion and return loss, but fixture effects must be removed or accounted for if the goal is to characterize the line itself.

Techniques such as TRL calibration, multiline TRL, or appropriate de-embedding structures are particularly valuable when the launch occupies a non-negligible electrical length.

For loss extraction, using two otherwise identical lines with different physical lengths can help separate distributed line attenuation from common launch loss. The phase difference between those lines can also be used to estimate propagation constant and effective permittivity.

At millimeter-wave frequencies, this type of correlation between fabricated geometry, EM simulation, and measured propagation data is considerably more useful than asking whether a finished board merely ‘passes 50 ohm impedance.’

Microstrip vs. GCPW: Engineering Trade-Offs

When Microstrip Is the Better Engineering Choice

Microstrip remains compelling when the required bandwidth and frequency do not make radiation or dispersion dominant limitations, and when manufacturing repeatability is a higher priority than maximum field confinement. It is often the cleaner choice when:

• the transmission line is relatively short and radiation is controlled,
• the connector or package transition already has a proven microstrip launch,
• the design requires generous fabrication margin,
• conductor loss is a major part of the loss budget,
• the line participates in a structure whose dimensions must remain highly repeatable, or
• additional coplanar ground metal would interfere with tuning, coupling, or component placement.

A straightforward electromagnetic structure is often easier to manufacture repeatedly than a theoretically superior structure with narrow process margins.

When GCPW Is the Better Engineering Choice

GCPW becomes particularly valuable when the designer needs direct control over the field near the top surface of the PCB. Typical motivations include:

• millimeter-wave interconnects where radiation and dispersion require tighter control,
• dense RF layouts where isolation from adjacent circuitry is important,
• coaxial or package launches that benefit from immediate top-layer ground return,
• RFIC and MMIC interfaces with ground-signal-ground pad arrangements,
• broadband transitions where local return-current geometry dominates performance, and
• structures in which suppression of specific unwanted propagation paths can be engineered through the ground and via geometry.

The cost of those advantages is a larger set of parameters that must remain controlled through layout, EM simulation, fabrication, and measurement.

The Best Design May Use Both

Microstrip and GCPW should not necessarily be treated as mutually exclusive board-level architectures.

A practical RF PCB may use GCPW around an edge-launch connector, transition to microstrip through a controlled taper, route the longer low-loss section as microstrip, and return to GCPW around a device with ground-signal-ground terminals.

This hybrid approach can be more efficient than forcing one topology through every part of the signal path.

The transitions themselves must, of course, be designed as electromagnetic structures. Abruptly terminating the coplanar grounds does not automatically create a transparent microstrip transition. The lateral field must evolve into the vertical microstrip field while maintaining a continuous return path and avoiding excitation of unwanted modes.

What Changes at Millimeter-Wave Frequencies

At millimeter wave, manufacturing geometry stops behaving like a small correction to the schematic design and increasingly becomes the design.

A copper thickness that was negligible compared with wavelength and trace width at a few gigahertz becomes electromagnetically meaningful. Connector solder, pad shape, via diameter, coplanar gap, etch taper, conductor roughness, dielectric thickness, and local ground geometry all occupy a larger fraction of the guided wavelength.

This is also where simplistic frequency thresholds become misleading.

There is no universal frequency above which microstrip ‘stops working,’ just as there is no frequency at which GCPW automatically becomes the correct choice. Substrate thickness, dielectric constant, line dimensions, loss requirements, allowed radiation, phase tolerance, connector geometry, and production capability determine where the trade changes for a particular design.

A thin microstrip on a suitable low-loss laminate can remain effective at frequencies where an electrically thick microstrip would radiate excessively. A poorly grounded GCPW can perform worse than either.

Design the Structure the Fabricator Will Actually Build

The most important distinction between a schematic-level RF design and a production microwave PCB is that the physical structure has tolerances.

Before releasing either microstrip or GCPW, the RF designer and PCB manufacturer should agree on the parameters that meaningfully affect the model: dielectric construction, finished copper thickness, conductor width definition, relevant etch tolerance, controlled-impedance methodology, copper profile where loss is critical, surface finish, solder mask conditions, and the geometry of high-frequency ground vias.

For GCPW in particular, specifying only ’50 ohm controlled impedance’ may not be enough. A fabricator can theoretically compensate a process variation by changing trace width, while the RF designer may need the original width and gap relationship preserved because it controls field distribution, coupling, launch behavior, or device-pad transition.

The impedance target and the geometry constraints therefore need to be distinguished.

That distinction becomes increasingly important for microwave PCB production because the electromagnetic response is determined by the fabricated cross section – not by the nominal CAD values alone.

Conclusion

Microstrip and grounded coplanar waveguide are not competing versions of the same 50 ohm trace.

Microstrip offers a comparatively simple field structure, strong manufacturing robustness, and potentially lower conductor loss. GCPW adds lateral grounding that can improve field confinement, isolation, launch behavior, dispersion, and radiation performance, particularly as designs move toward millimeter-wave frequencies.

Those advantages are purchased with additional sensitivity to gap geometry, copper thickness, ground-via implementation, discontinuities, and parasitic modes.

The decision should therefore be made from the complete electromagnetic problem rather than from frequency alone. Loss needs to be decomposed rather than summarized. Ground vias need to be modeled rather than treated as decoration. Connector launches need to be analyzed as three-dimensional transitions. Fabrication tolerances need to be included before the design is considered finished.

For professional RF work, the better transmission line is not the one that produces the cleanest nominal 50 ohm calculation.

It is the one whose fields, modes, tolerances, transitions, and return-current path can still be controlled after the design leaves the simulator and becomes a physical circuit.