
What Is UWB Antenna Design?
UWB antenna design is the process of creating a printed antenna that maintains acceptable impedance match (typically VSWR < 2:1), stable radiation pattern, and adequate gain across the entire 3.1–10.6 GHz band allocated by the FCC under Part 15 UWB rules. That is a fractional bandwidth exceeding 109 %, which makes conventional resonant approaches—tuned to a single frequency—useless. Instead, UWB antennas rely on traveling-wave, self-complementary, or multi-resonance geometries that spread current paths across a wide spectrum.
The challenge is not just bandwidth. UWB systems used in indoor positioning, automotive keyless entry, and short-range radar also need low group-delay variation (under ±1 ns across the band) to preserve pulse shape. Every layout choice—substrate Dk, ground-plane size, feed transition—directly affects both impedance bandwidth and time-domain fidelity. This guide walks through the topologies, materials, feed methods, and verification steps that make a UWB antenna work on a real PCB.
Key UWB PCB Antenna Topologies

Three families dominate printed UWB antenna design. Each trades off size, directivity, and manufacturing complexity differently. The table below compares them at a glance; the sections that follow go deeper.
| Topology | Typical Size (at 6.5 GHz center) | Peak Gain | Radiation Pattern | Best Use Case |
|---|---|---|---|---|
| Planar monopole (disc, elliptical, leaf) | 20 × 25 mm | 2–5 dBi | Omni-directional in H-plane | Positioning tags, IoT nodes |
| Vivaldi / tapered-slot | 40 × 50 mm | 4–10 dBi | End-fire, directive | Ground-penetrating radar, imaging |
| CPW-fed slot | 25 × 30 mm | 2–4 dBi | Bi-directional broadside | Compact modules, wearables |
Planar Monopole Antennas
A planar monopole is the workhorse of compact UWB devices. The radiating element—an elliptical, circular, or leaf-shaped copper patch—is printed on one side of the substrate, while a partial ground plane occupies the opposite side. Bandwidth is controlled primarily by the shape of the radiator’s lower edge and the gap between that edge and the ground plane, typically 0.3–0.5 mm for a 0.8 mm thick FR-4 board. Smoothing the transition from feedline to radiator (beveling corners, adding a tapered feed stub) suppresses higher-order modes that would otherwise create notches inside the band.
For UWB positioning tag designs, the planar monopole is the default choice because it fits inside a credit-card-sized module and provides near-omnidirectional coverage in the azimuth plane.
Vivaldi and Tapered-Slot Antennas
When directive, end-fire radiation is needed—think through-wall radar or medical imaging—a Vivaldi antenna delivers. The exponential taper profile is defined by the equation y = c₁ · eRz + c₂, where R is the taper rate. Higher R values produce narrower beamwidths but require longer boards. A typical 3.1–10.6 GHz Vivaldi on 0.508 mm Rogers RO4003C (Dk 3.38) measures roughly 40 × 50 mm and achieves 7–10 dBi above 6 GHz. For more on end-fire geometries, see the guide to end-fire antenna design on PCB.
CPW-Fed Slot Antennas
Coplanar-waveguide (CPW) fed slots keep all metallization on a single layer, which simplifies fabrication and eliminates via transitions. The slot width and shape (rectangular with rounded ends, or a tapered keyhole) set the lower cutoff frequency. Adding a tuning stub behind the slot extends the upper band edge past 10.6 GHz. The trade-off is a bi-directional pattern; if back radiation is unacceptable, a reflector plane can be added at λ/4 spacing at the geometric center frequency (~6.5 GHz ≈ 11.5 mm), at the cost of increased profile.
Substrate and Material Selection
UWB antennas are less material-sensitive than narrowband designs because the wide bandwidth masks small Dk shifts. That said, the substrate still matters for loss, dimensional stability, and group-delay flatness.
| Material | Dk (10 GHz) | Df (10 GHz) | Typical Thickness | Notes |
|---|---|---|---|---|
| FR-4 (mid-Tg) | 4.2–4.5 | 0.018–0.025 | 0.8 / 1.0 mm | Adequate for tags and consumer IoT; Dk variation across band causes ~0.4 ns group-delay ripple |
| Rogers RO4350B | 3.48 | 0.0037 | 0.508 / 0.762 mm | Good balance of loss and cost for automotive UWB |
| Rogers RO3003 | 3.00 | 0.0013 | 0.254 / 0.508 mm | Lowest loss; justified for radar front-ends above 8 GHz |
| Isola Astra MT77 | 3.00 | 0.0017 | Various | Thermoset alternative to PTFE, easier to process |
For most UWB PCB projects below channel 9 (centered at 7.9872 GHz), standard FR-4 performs acceptably if the designer accounts for the Dk spread. Above channel 9, or when the link budget is tight, switching to RO4350B is the pragmatic move. The Dk selection trade-offs for antennas versus transmission lines are worth reviewing because a UWB board often carries both the antenna and a 50 Ω feed network on the same laminate.
Substrate thickness directly sets the impedance bandwidth of planar monopoles: thinner boards (0.4–0.8 mm) generally yield wider bandwidth because the ground-plane coupling is weaker, but they also reduce gain at the low end. A 0.8 mm FR-4 or 0.508 mm RO4350B is the sweet spot for most 3.1–10.6 GHz monopole designs.
Feed Techniques and Impedance Matching
Microstrip-to-Radiator Transition
The 50 Ω microstrip feed must transition smoothly into the wide radiating element. An abrupt width change creates a capacitive discontinuity that shows up as a notch around 5–6 GHz in many first-pass designs. Three proven fixes:
- Tapered microstrip: Linearly widen the trace over 3–5 mm before it meets the radiator. Simple, effective, costs no extra area.
- Stepped impedance transformer: Two quarter-wave sections (at ~6.5 GHz, each ~5.5 mm on FR-4) transform 50 Ω to the radiator’s feed-point impedance, typically 80–120 Ω.
- Beveled ground plane: Chamfering the top corners of the partial ground plane by 2–4 mm on each side smooths the current transition and can add 1–2 GHz of usable bandwidth at the high end.
CPW Feeding
CPW feeds avoid the ground-plane discontinuity problem entirely because the ground conductors are coplanar with the signal trace. The gap width (typically 0.15–0.3 mm) and the 50 Ω center conductor width are calculated together using standard CPW equations. A CPW feed also makes SMA connector attachment straightforward: the connector’s center pin solders to the center conductor, and the ground tabs solder directly to the coplanar grounds without vias.
Band-Notch Techniques
Many UWB systems must reject the 5.15–5.825 GHz WLAN band to meet coexistence regulations. The most common approach is etching a half-wavelength C-shaped or U-shaped slot into the radiator or the feed line. The slot length is approximately λ_eff / 2 at 5.5 GHz (roughly 14 mm on FR-4). Slot width controls notch depth; 0.2 mm gives about 10 dB rejection, while 0.5 mm gives 15–20 dB. Multiple slots at different lengths can create dual- or triple-notch responses.
Ground-Plane Design and Size Effects
The partial ground plane is not a passive element—it is half the antenna. Its length, width, and edge profile directly set the lower cutoff frequency and pattern symmetry.
- Ground-plane length: Should be roughly λ/4 at the lowest operating frequency. At 3.1 GHz on FR-4 (Dk ≈ 4.3), that is about 12 mm. Shorter ground planes push the lower cutoff upward; longer ones improve low-band match but may introduce pattern nulls at higher frequencies.
- Ground-plane width: Wider than the radiator by at least 2 mm on each side to maintain pattern symmetry. Narrower boards (as in USB-dongle form factors) tilt the H-plane pattern and reduce peak gain by 1–2 dBi.
- Edge treatment: Rounding or stepping the top edge of the ground plane (the edge closest to the radiator) is the single most effective bandwidth-enhancement technique for planar monopoles. A 2 mm radius round on each top corner can add 500 MHz of usable bandwidth at the low end.
UWB Antenna Design Checklist
Before committing to fabrication, walk through every parameter that affects UWB antenna performance. Enter your design specifics into the checklist builder below—it generates a point-by-point review covering substrate, geometry, feed, ground plane, band-notch, and test plan items so nothing is missed between simulation and prototype.
A completed checklist for a typical UWB positioning tag on 0.8 mm FR-4 might flag items like: confirm Dk tolerance band from the laminate vendor (±0.15 for generic FR-4), verify ground-plane chamfer dimensions match the EM model, and ensure the SMA-to-microstrip transition is included in the simulation. Addressing every flagged item before ordering panels avoids the most common first-spin failures—shifted lower cutoff, mid-band notches, and connector-induced mismatch. For a broader set of RF-specific manufacturability rules, review the RF DFM checklist.
Simulation and Measurement Workflow

Full-Wave EM Simulation
UWB antennas demand 3D full-wave solvers (FDTD or FEM) because the wide bandwidth means lumped-element or transmission-line models cannot capture the frequency-dependent current distribution. Key simulation practices:
- Use frequency-dependent Dk and Df values, not a single-frequency number. Rogers RO4350B datasheets provide Dk at multiple frequencies; interpolate for the solver.
- Model the SMA connector or U.FL footprint explicitly. Connector parasitics shift the upper-band S11 by 0.5–1.5 GHz.
- Mesh the feed gap and slot edges at λ/20 at 10.6 GHz (~1.0 mm in free space, ~0.5 mm in substrate). Coarser mesh hides the notches that will appear on the bench.
- Extract group delay from the simulated S21 of a two-antenna link, not just S11. A flat S11 does not guarantee flat group delay.
Prototype Measurement
Measure return loss with a calibrated VNA from 2 to 12 GHz to capture both band edges and any out-of-band resonances. For radiation pattern and gain, an anechoic chamber or compact-range setup is needed; the antenna measurement setup guide covers chamber, OTA, and near-field options. Time-domain performance is verified by transmitting a short Gaussian pulse and comparing the received pulse shape to the transmitted one; a correlation coefficient above 0.90 is typically acceptable for IEEE 802.15.4z ranging.
Evaluating Design Changes: Cost vs. Performance ROI
UWB antenna development often involves iterating between FR-4 and low-loss laminates, adding band-notch features, or resizing the ground plane. Each change carries a cost—new tooling, longer lead times, higher material price—and a performance return. Use the design-change ROI calculator below to quantify whether a proposed change (for example, switching from FR-4 to RO4350B to recover 1.5 dB of gain at 9 GHz) justifies the added cost at your production volume.
For a 10 k-unit automotive keyless-entry UWB module, upgrading from FR-4 to RO4350B might add $0.80–$1.50 per board but eliminate the need for an external LNA, saving $0.60 in BOM plus board area. The calculator helps frame that trade-off numerically so the decision is engineering-driven, not gut-driven.
Common Pitfalls and How to Avoid Them
- Ignoring the cable in measurement: A semi-rigid cable routed along the ground plane acts as a parasitic sleeve balun and shifts the pattern. Use a ferrite-choked cable or measure with the cable perpendicular to the ground plane.
- Designing to the full 3.1–10.6 GHz when only one channel is used: IEEE 802.15.4z devices often operate on channel 5 (6.4896 GHz, 499.2 MHz BW) or channel 9 (7.9872 GHz). Designing for only the required channel relaxes size and loss constraints significantly.
- Neglecting the enclosure: A plastic radome with Dk of 2.8–3.2 and 1–2 mm wall thickness detunes the antenna by 100–300 MHz downward. Always simulate with the radome included or add 5 % frequency margin.
- Via fencing overkill: Stitching vias around the ground plane improves isolation on multi-antenna boards but can create waveguide-below-cutoff effects if spacing exceeds λ/10 at 10.6 GHz (~2.8 mm). Keep via pitch at 2.0–2.5 mm.
- Solder-mask over the radiator: Solder mask (Dk ~3.5, 20–30 µm thick) adds a dielectric loading that shifts resonance downward by 50–150 MHz. Either remove the mask from the antenna area or model it in simulation.
Compliance and Standards
UWB transmitters in the US must meet FCC Part 15 Subpart F emission limits: –41.3 dBm/MHz EIRP across 3.1–10.6 GHz. In Europe, ETSI EN 302 065 applies, with additional low-duty-cycle or detect-and-avoid requirements in certain sub-bands. The antenna’s gain directly determines how much power the transmitter can emit, so a higher-gain Vivaldi design reduces the allowed conducted power proportionally. Always verify total EIRP with a calibrated measurement, not just simulated gain.
Frequently Asked Questions
Can I use FR-4 for a UWB antenna?
Yes. FR-4 works well for UWB antennas operating below channel 9 (~8 GHz) where its dielectric loss (Df ~0.02) causes only 0.5–1.0 dB of efficiency reduction compared to RO4350B. Above 8 GHz, or when group-delay flatness is critical, a low-loss laminate is preferable.
What is the minimum ground-plane size for a 3.1 GHz UWB antenna?
The ground plane should be at least λ/4 long at the lowest frequency. On FR-4 (effective Dk ~4.3), that is approximately 12 mm. Width should exceed the radiator width by at least 4 mm total to maintain pattern symmetry.
How do I add a 5 GHz WLAN notch to a UWB antenna?
Etch a half-wavelength (≈14 mm on FR-4) C-shaped or U-shaped slot into the radiator or feed line, centered at 5.5 GHz. Adjust slot width for notch depth: 0.2 mm for ~10 dB, 0.5 mm for 15–20 dB rejection.
What group-delay variation is acceptable for UWB ranging?
For IEEE 802.15.4z compliant ranging (±10 cm accuracy), group-delay variation should stay below ±1 ns across the channel bandwidth (typically 499.2 MHz). For full-band impulse radar, ±0.5 ns is a tighter but achievable target on low-loss substrates.
How does the radome affect a UWB antenna?
A 1.5 mm ABS radome (Dk ~2.9) typically shifts the antenna’s resonant frequencies downward by 100–300 MHz and can degrade return loss at the upper band edge. Always co-simulate the radome or add 5 % frequency margin to the bare-board design.
Ready for a quotation?
Send your requirements for a material, stackup and manufacturing review. Price and schedule are confirmed for your project.
