Access Official Technical PDF

Share Summary:LinkedInEmail

For engineers developing millimeter-wave integrated circuits, mm-wave CMOS transmission lines are not simply connections between circuit blocks. Their geometry and placement within the back end of line (BEOL) affect characteristic impedance, attenuation, phase shift and the silicon area required to achieve a target electrical length. Comparing line structures and process limits early can help designers choose a suitable implementation before committing to a layout or fabrication run.

Coplanar waveguide (CPW), coplanar waveguide with ground (CPWG) and shielded coplanar waveguide (SCPW) each use a different arrangement of return paths and shielding. Those differences change the distributed inductance and capacitance that govern line behavior. A semilumped CPW, which adds capacitive loading, offers another route to reducing physical length.

Why CMOS transmission line design is process-dependent

CMOS transmission lines are generally built from upper BEOL metal layers, where conductor properties and distance from the silicon substrate influence electrical performance. The metals are embedded in dielectric material, and each layer has its own thickness, conductivity and vertical position. Layout rules also constrain trace width, spacing and metal density, including requirements for dummy fill.

These constraints limit the geometries available in any given process. They also make it difficult to transfer a line design directly between technology nodes: a change in metal stack or substrate interaction can alter impedance, loss and phase behavior. Because silicon area is costly, designers also have reason to seek greater phase shift per unit length, so a shorter line can provide a required electrical length.

A useful comparison combines electromagnetic simulation with on-silicon measurements. In the reported approach, CPW, CPWG and SCPW test structures fabricated in 40-nm CMOS were characterized using two-port S-parameters. Probe-pad parasitics were removed through L-2L de-embedding, and the measured results were compared with simulations of the same structures. Agreement between the methods supported extending the analysis to additional simulated geometries and processes.

How CPW, CPWG and SCPW differ

CPW: a practical reference structure

A CPW has a central signal conductor flanked by coplanar ground conductors. This arrangement suits ground-signal-ground probing and allows convenient shunt connections to ground. Its signal and return paths are also influenced by the underlying substrate, with the degree of interaction depending on factors such as substrate resistivity and conductor position.

Changing the signal width-to-gap ratio is a direct way to influence characteristic impedance. The selected metal layer also matters: changing its thickness or height above the substrate affects the line’s distributed properties and, in turn, its impedance, attenuation and phase constant.

CPWG: adding a ground plane below the signal

CPWG adds a dedicated ground plane beneath the signal trace. This plane becomes part of the return path and changes both distributed inductance and capacitance through electromagnetic coupling with the signal conductor. Its presence can also influence the range of impedance values available from a given combination of top-layer width and spacing.

Because the lower plane has a substantial effect on the line’s electrical behavior, changing the coplanar width-to-gap ratio alone may not produce the same range of results as it would in a CPW. The practical design must account for the full cross-section, not just the visible top-layer geometry.

SCPW: shielding with a slow-wave effect

An SCPW uses a grid of metal segments beneath the signal path to connect the two coplanar ground regions. Unlike a solid ground plane directly under the signal, this shield does not provide the same magnetic interaction with the signal path. It does, however, increase distributed capacitance. The resulting lower characteristic impedance and higher phase constant are associated with the structure’s slow-wave behavior.

In the analyzed designs, SCPW provided greater phase shift per unit length than regular CPW and generally lower loss. A higher phase constant can allow a shorter line to deliver the same electrical length, while shielding can reduce unwanted interaction with the substrate. These outcomes depend on the process and geometry; the structure still needs to be evaluated for the intended design.

Use BEOL limits to assess process capability

A process design kit (PDK) can be assessed by combining known BEOL dimensions with layout design-rule limits. For the selected upper metal, relevant inputs include its thickness and height above the substrate, the minimum and maximum permitted signal widths, and the minimum spacing between conductors. A practical maximum spacing can also be defined for the intended coplanar structure.

Simulating carefully chosen width-and-spacing combinations can provide an initial estimate of the achievable characteristic-impedance and phase-shift ranges. The analysis described used combinations based on minimum and maximum width and spacing limits, rather than requiring a large number of fabricated structures. This gives designers a screening method for understanding whether a process is likely to support a desired line geometry and electrical target.

Such estimates are not a substitute for validating the final implementation. They are a way to focus further modeling and measurement on promising structures, while recognizing that BEOL stack, substrate conductivity and metal properties all contribute to performance.

Reducing line length with semilumped CPW

Another approach is to replace a quarter-wavelength transmission line segment with a shorter low-pass equivalent structure. In the described method, the replacement line has a higher characteristic impedance and uses shunt capacitors as loading elements. The required capacitance is calculated by matching the network behavior of the original quarter-wave segment to the equivalent low-pass network.

At millimeter-wave frequencies, a small integrated capacitance needs careful treatment. Rather than combining a transmission-line electromagnetic model with a potentially inaccurate capacitor model, the complete layout—including the capacitors—can be modeled in a single extraction. The reported design also used two parallel metal-oxide-metal (MOM) capacitors to maintain symmetry in the signal path.

A semilumped CPW fabricated in 40-nm CMOS was measured and modeled around 78 GHz. The reported device had a characteristic impedance near 50 ohms and a phase shift of about 99 degrees at that frequency, compared with an ideal quarter-wave value of 90 degrees. At 60 GHz, its phase constant was reported as 290 degrees per millimeter, nearly twice that of the compared regular CPW. These results demonstrate the potential for increased phase shift per length; actual performance remains dependent on the layout and process.

Practical implications for mm-wave designers

Choosing a CMOS transmission line involves balancing electrical performance, physical area, process rules and modeling confidence. CPW offers a straightforward reference geometry, while CPWG and SCPW alter the field distribution through lower metal structures. Semilumped loading can further increase phase shift per length, but requires the capacitors and line to be treated as a complete electromagnetic structure.

For design teams, the practical sequence is to understand the BEOL stack and design-rule envelope, compare candidate line types through simulation, and validate modeling against measurement where possible. This process-based evaluation can help identify suitable geometries, estimate achievable impedance and phase behavior, and avoid assuming that a design will transfer unchanged between CMOS nodes. The result is a more informed basis for balancing mm-wave performance with silicon-area requirements.

Access Official Technical PDF

About The Author