In short: Microstrip puts the signal trace on the PCB surface over one reference plane for easy routing and lower loss, while stripline buries the trace between two planes for superior noise isolation and tighter controlled impedance — at the cost of a thicker board and slightly more dielectric loss.
Any time a signal edge is fast enough that its rise time rivals the flight time down the copper, the geometry of the trace stops being cosmetic and becomes an electrical component. Two transmission-line structures dominate high-speed digital and RF design on FR-4 and high-frequency laminates: microstrip and stripline. Choosing between them decides your impedance tolerance, crosstalk budget, EMI footprint, and even how many layers your stackup needs. This article unpacks the physics and the practical trade-offs with concrete numbers.
The two geometries at a glance
A microstrip is a conductor on the outer layer of the board, separated from a single reference plane by a dielectric (the prepreg/core). Its field lines live partly in the dielectric and partly in the air above it. A stripline is a conductor sandwiched symmetrically between two reference planes, so its field is fully enclosed in dielectric. That single structural difference — one plane and open air versus two planes and total enclosure — drives everything that follows.
Effective permittivity and propagation speed
Because a microstrip's field is shared between the board dielectric and the air above, the signal experiences an effective relative permittivity (εeff) that is a blend of the two — typically 2.8–3.3 on FR-4 (bulk εr ≈ 4.3, air = 1). A stripline, fully immersed in dielectric, sees the full εr. Propagation delay scales with √εeff, so a microstrip on FR-4 propagates at roughly 150 ps/inch while a stripline on the same material runs slower at about 180 ps/inch.
Timing consequence: On a 10-inch bus, that difference is about 300 ps of flight time — easily a third of a clock period on a 1 GHz interface. Length-matching groups must stay on the same layer type, because a trace that detours through a stripline layer picks up extra delay per inch that a surface microstrip of equal length never sees.
Impedance control
Controlled impedance is the whole reason these structures exist. For a 50 Ω single-ended microstrip on FR-4 with ~4 mil dielectric height, you need roughly 7–8 mil of trace width; a 50 Ω stripline with the same dielectric between planes typically wants a narrower ~5 mil trace because it is coupled to two planes instead of one. The symmetric, air-free environment of stripline makes its impedance less sensitive to etch variation and to solder-mask thickness (which only microstrip carries), so fabs routinely hold tighter tolerance — ±5–7% on stripline versus ±8–10% on microstrip. Field solvers in Cadence Sigrity, Siemens HyperLynx, and Synopsys tools compute these numbers from your exact stackup rather than relying on the closed-form IPC-2141 approximations, and for anything above a few hundred MHz you should trust the 2-D solver.
For a deeper treatment of stackup planning and fabricator tolerance negotiation, see our companion guide on controlled-impedance PCB design.
Loss: conductor vs dielectric
Total insertion loss splits into conductor loss (skin effect, scales with √f) and dielectric loss (scales linearly with f and with loss tangent tan δ). Microstrip, with half its field in near-lossless air, generally shows lower dielectric loss than stripline at the same frequency. Stripline, fully in FR-4 (tan δ ≈ 0.02), suffers more dielectric loss but is immune to the surface-roughness and solder-mask absorption effects that plague microstrip above ~10 GHz. For a 28 Gbps SerDes channel, designers often route the long, loss-critical segments as stripline on a smooth low-Dk laminate (Megtron 6, Tachyon) specifically to control the field environment, accepting the higher tan δ in exchange for predictability.
Noise, crosstalk and EMI
This is where stripline wins decisively. With two reference planes bracketing the trace, stripline fields are almost entirely contained, which slashes both radiated emissions and far-end crosstalk between adjacent traces. Microstrip's open top radiates and couples more, and its asymmetric field makes it prone to the "far-end crosstalk" that stripline largely cancels. If you are chasing a stubborn EMI failure or routing sensitive analog next to a noisy clock, moving the aggressor to an internal stripline layer is one of the highest-leverage fixes available.
| Property | Microstrip | Stripline |
|---|---|---|
| Trace location | Outer surface | Internal, between 2 planes |
| Reference planes | One | Two |
| εeff (FR-4) | ~2.8–3.3 (with air) | ~4.3 (full dielectric) |
| Propagation delay | ~150 ps/in | ~180 ps/in |
| Dielectric loss | Lower | Higher |
| EMI / crosstalk | Higher | Much lower |
| Routing/rework ease | Easy (surface access) | Hard (buried) |
| Layer cost | Low | Needs more layers |
Rule of thumb: Route DDR address/command, long SerDes, and anything EMI-sensitive as stripline; keep short breakout, impedance-forgiving nets, and anything that may need rework or probing on microstrip.
Practical routing strategy
Real boards use both. A typical 8-layer stackup places high-speed differential pairs on internal stripline layers (L3/L6) and uses the outer microstrip layers (L1/L8) for component breakout, test points, and less-critical signals. Watch three gotchas: (1) microstrip needs the solder-mask thickness in the impedance model — ignoring it shifts impedance by 1–2 Ω; (2) every stripline layer transition needs a nearby stitching via to give the return current an uninterrupted path, or you create an impedance discontinuity; (3) never route a high-speed trace across a plane split — on stripline the buried return has nowhere else to go. For teams building these skills hands-on, Avecas offers professional PCB and signal-integrity design services that complement structured training.
If you want to go further on crosstalk budgeting and return-path design, read our follow-up on PCB crosstalk and signal integrity, and browse the full catalog on our courses page.
Ready to master controlled-impedance routing, stackup design, and high-speed signal integrity end to end? Join the CourseTron High-Speed PCB & Signal Integrity course and design real 28 Gbps channels with field-solver verification.
Frequently asked questions
Is microstrip or stripline faster?
Microstrip is faster. Because part of its field travels in air, its effective permittivity is lower (~3 vs ~4.3 on FR-4), giving roughly 150 ps/inch versus 180 ps/inch for stripline. The denser dielectric around a stripline slows the wave.
Why does stripline have lower EMI?
Stripline is enclosed between two ground planes, so its electromagnetic field is almost fully contained within the board. That containment suppresses both radiated emissions and coupling to neighboring traces, which is why noise-sensitive and high-aggressor nets are routed internally.
Can I mix microstrip and stripline on one net?
You can, but be careful with timing and impedance. Each layer type has a different propagation delay per inch, so layer transitions change flight time and must be accounted for in length matching. Every transition also needs a stitching via to keep the return current continuous.
Which is cheaper to manufacture?
Microstrip, generally. It can exist on a simple 2-layer board (trace over one plane), whereas stripline requires the trace to be buried between two internal planes, pushing you toward a 4-layer-or-more stackup and higher fabrication cost.