
What Does Antenna Substrate Thickness Actually Control?
Antenna substrate thickness (h) is the dielectric spacing between the radiating element and its ground plane. A thicker substrate stores more energy in the fringing fields, which lowers the antenna’s radiation Q and widens impedance bandwidth—but it also launches more surface waves, which degrade efficiency and distort the radiation pattern. A thinner substrate suppresses surface waves and keeps the pattern clean, yet it raises Q and narrows bandwidth sharply. Every planar antenna design is a negotiation between these two effects, and h is the primary lever.
The Physics: Why Thickness Drives Bandwidth

For a rectangular microstrip patch, the classic relationship between fractional bandwidth (BW), substrate thickness h, and dielectric constant εr is approximately:
BW ≈ (3.77 × (εr − 1) / εr2) × (h / λ0) × (VSWR − 1) / √VSWR
where λ0 is the free-space wavelength. The key takeaway: bandwidth scales almost linearly with h/λ0. Double the substrate height relative to wavelength and you roughly double the impedance bandwidth. This relationship holds well for h/λ0 up to about 0.1; beyond that, higher-order effects and strong surface-wave coupling break the linear model.
The fundamental limit on bandwidth for a given antenna volume is set by the antenna Q factor. For an electrically small patch, Q drops as h increases, directly widening the achievable −10 dB return-loss bandwidth. The definition of antenna bandwidth you use (−6 dB vs −10 dB) shifts the numbers, but the trend with h is the same.
Surface Waves: The Cost of Going Thick
A grounded dielectric slab supports TM0 and TE1 surface-wave modes. The TM0 mode has no cutoff—it exists at any thickness—but its coupling strength grows with h. The TE1 mode turns on when:
h ≥ λ0 / (4 × √(εr − 1))
For Rogers RO4350B (εr = 3.48) at 5.8 GHz (λ0 ≈ 51.7 mm), the TE1 cutoff thickness is about 8.2 mm—well above typical substrates. But TM0 coupling on a 1.524 mm (60 mil) board at the same frequency already shunts a measurable fraction of radiated power into the substrate, reducing broadside gain by 0.3–0.8 dB depending on ground-plane size. At millimeter-wave frequencies the problem is far worse: at 28 GHz on the same material, a 0.508 mm (20 mil) substrate already corresponds to h/λ0 ≈ 0.047, and surface-wave losses can exceed 1 dB. Detailed loss budgets for high-frequency designs are covered in our guide to mmWave antenna loss mechanisms.
Practical Thickness Ranges by Frequency Band
| Frequency Band | Typical h Range | h/λ0 | Achievable BW (−10 dB) | Dominant Concern |
|---|---|---|---|---|
| 900 MHz – 2.4 GHz (IoT, Wi-Fi) | 1.524 – 3.175 mm (60 – 125 mil) | 0.005 – 0.025 | 1 – 5 % | Size, narrow BW |
| 5 – 6 GHz (Wi-Fi 6E, C-V2X) | 0.787 – 1.524 mm (31 – 60 mil) | 0.015 – 0.030 | 3 – 8 % | Balanced trade-off |
| 24 – 30 GHz (5G FR2, radar) | 0.254 – 0.508 mm (10 – 20 mil) | 0.025 – 0.050 | 5 – 12 % | Surface waves, fabrication tolerance |
| 60 – 79 GHz (WiGig, auto radar) | 0.127 – 0.254 mm (5 – 10 mil) | 0.030 – 0.065 | 8 – 18 % | Surface waves, conductor loss |
These ranges assume a single-layer patch with no bandwidth-enhancement techniques. Stacked patches, U-slot patches, or aperture-coupled feeds can push bandwidth 2–4× wider at the same h, but the surface-wave penalty still applies. For overall PCB thickness guidance beyond the antenna layer, see how thick an RF PCB should be.
Material Choice Interacts with Thickness
Substrate thickness and dielectric constant are coupled variables. A lower εr material (e.g., Rogers RT/duroid 5880 at εr = 2.20) gives wider bandwidth for a given h because the fringing fields extend further. It also pushes the TE1 surface-wave cutoff higher. The trade-off is a physically larger patch, which may be unacceptable in an array with tight element spacing.
Higher εr substrates (e.g., Rogers RO3010 at εr = 10.2) shrink the patch but severely narrow bandwidth and increase surface-wave coupling. They are sometimes used for GPS or L-band antennas where bandwidth requirements are modest (< 1 %) and miniaturization matters. For a low-loss mid-range option, Rogers RO3003G2 at Dk 3.00 offers a good balance at mmWave frequencies.
The Rogers RO4000 series laminate datasheets list available core thicknesses from 0.168 mm to 1.524 mm (6.6 to 60 mil), and the standard thickness steps constrain your design. Choosing a thickness between standard cores means bonding two thinner cores with prepreg, which shifts the effective Dk and adds a glue-line interface.
Using Prepreg to Hit a Non-Standard Thickness
When the ideal antenna substrate thickness falls between available core sizes, designers bond cores with prepreg. The pressed prepreg thickness depends on resin content, glass style, and copper coverage on adjacent layers. Use the tool below to look up pressed prepreg thickness for common glass styles and resin systems.
For example, a single sheet of 1080 prepreg with 65 % resin content presses to roughly 0.066 mm (2.6 mil). Bonding a 0.254 mm RO4350B core to a ground-plane layer through that prepreg yields a composite antenna substrate of about 0.320 mm—useful for a 28 GHz array where 0.254 mm alone is too thin for adequate bandwidth but 0.508 mm launches excessive surface waves. Always verify the composite Dk with the laminate supplier’s mixing-rule data, because the prepreg’s Dk may differ from the core’s by 0.1–0.3 units.
Bandwidth Enhancement Without Adding Thickness
If the maximum substrate thickness is constrained by the overall stackup, surface-wave budget, or standard core availability, several well-known techniques widen bandwidth at a fixed h:
- U-slot patch: A U-shaped slot cut into the patch introduces a second resonance, roughly doubling impedance bandwidth for h/λ0 ≈ 0.04–0.06.
- Stacked patches: A parasitic patch on a superstrate above the driven patch adds a coupled resonance. Bandwidth can reach 15–25 % at the cost of a second dielectric layer and tighter fabrication tolerance.
- Aperture-coupled feed: Decouples the feed network from the radiator through a slot in the ground plane, allowing independent optimization of feed-line substrate and antenna substrate thickness.
- Substrate integrated waveguide (SIW) cavity: Placing the patch over an SIW cavity suppresses surface waves and recovers some of the efficiency lost to a thick substrate.
- Metasurface superstrate: A periodic structure above the antenna can improve gain and shape bandwidth; see metasurface antenna design for implementation details.
Surface-Wave Mitigation Strategies

When a thick substrate is unavoidable for bandwidth, surface waves can be partially controlled:
- Via fences / substrate shielding: Rows of grounding vias around the patch perimeter act as a metallic wall, blocking lateral surface-wave propagation. Via pitch must be ≤ λg/10 to be effective.
- Electromagnetic bandgap (EBG) structures: Periodic mushroom-like unit cells in the ground plane create a stopband for surface waves at the operating frequency. Effective but consumes board area.
- Air-cavity or partially hollowed substrate: Removing dielectric beneath the patch (via milling or LTCC cavity techniques) lowers the effective εr toward 1, dramatically reducing surface-wave coupling while preserving the physical height for bandwidth.
Thickness Tolerance and Its Effect on Performance
Standard PCB core thickness tolerance per IPC-6012 Class 2 is ±10 % for cores under 0.250 mm and ±0.025 mm for thicker cores. A ±10 % swing on a 0.254 mm (10 mil) core at 28 GHz shifts the patch resonant frequency by approximately ±1.5 %, which can move the return-loss null outside a narrow 5 % bandwidth window. For production antennas, specifying Class 3 tolerance (tighter) or adding a post-fabrication tuning element (e.g., a varactor-loaded matching stub) is common. Our article on taking an antenna from prototype to production covers tolerance budgeting in detail.
Assembly Considerations: Stencil Thickness
Antenna substrate thickness also affects the overall stackup height, which in turn influences component assembly. Thicker antenna boards may require adjusted stencil apertures and thicknesses for solder paste deposition on connector pads and matching-network components. Use the tool below to select an appropriate stencil thickness based on your smallest component pitch and pad geometry.
A typical result: for 0402 passives on a matching network next to the antenna, a 0.100 mm (4 mil) stencil with a 0.70 area ratio gives reliable paste transfer. If you’re also placing a U.FL connector with 1.0 mm pitch pads, the same stencil works without step-downs. For fine-pitch mmWave ICs (0.4 mm BGA), you may need a thinner 0.080 mm stencil or localized step-down, and the extra antenna substrate height must be accounted for in the reflow profile to avoid shadowing effects.
Design Workflow: Choosing h Step by Step
- Define bandwidth requirement in absolute MHz and as a fraction of center frequency. Decide whether −6 dB or −10 dB return loss is the specification.
- Estimate minimum h/λ0 from the bandwidth formula or a full-wave quick sweep. For a simple patch, BW(%) ≈ 3.77 × (h/λ0) × 100 at εr ≈ 3.5 is a useful first pass.
- Check surface-wave budget. Compute TE1 cutoff thickness. If h is above 60 % of that value, plan mitigation (via fence, EBG, or cavity).
- Select a standard core thickness from your chosen laminate family. If no single core matches, plan a core-plus-prepreg bonded stackup.
- Simulate. Run a full-wave solver (Ansys HFSS or CST) with the exact stackup, including copper roughness and loss tangent at the operating frequency. Confirm bandwidth, gain, and surface-wave loss.
- Tolerance-check. Re-simulate at h ± the fabrication tolerance. If the bandwidth margin disappears, either tighten the tolerance spec or widen bandwidth with a slot or stacked element.
- Prototype and measure. Validate S11 and radiation pattern on the first build. Thickness-related shifts show up as a frequency offset of the return-loss null—measure it and correlate with actual board thickness from a cross-section coupon.
Frequently Asked Questions
What is the ideal antenna substrate thickness for a 5.8 GHz patch?
For a single rectangular patch on εr ≈ 3.5 material targeting 3–5 % bandwidth at −10 dB, a substrate thickness of 1.524 mm (60 mil) is a well-proven starting point. This gives h/λ0 ≈ 0.029, with manageable surface-wave loss. Going to 0.787 mm (31 mil) halves the bandwidth to about 1.5–2 % but improves pattern purity.
Does a thicker substrate always mean more loss?
Not always in total. A thicker substrate increases surface-wave loss and dielectric loss (longer path through the material), but it also reduces conductor loss relative to radiated power because the radiation resistance rises. At low frequencies where surface waves are weak, a thicker substrate can actually improve total efficiency. At mmWave frequencies, surface-wave loss usually dominates and a thicker substrate does mean lower efficiency.
Can I use FR-4 as an antenna substrate?
FR-4 works for non-critical antennas below about 3 GHz where its high loss tangent (Df ≈ 0.020 at 1 GHz, rising with frequency) and poorly controlled Dk (±5–10 %) are tolerable. Above 5 GHz, the losses and Dk variation make it impractical for anything beyond a quick proof-of-concept.
How does substrate thickness affect an antenna array’s scan performance?
Thicker substrates increase surface-wave coupling between elements in an array, which raises mutual coupling and can cause scan blindness at certain angles. The scan-blindness angle depends on element spacing, εr, and h. Keeping h/λ0 below 0.04 and element spacing near 0.5 λ0 generally avoids scan blindness in the ±60° scan volume.
What thickness should I use for a 77 GHz automotive radar antenna?
Most 77 GHz automotive radar front-ends use 0.127–0.254 mm (5–10 mil) substrates on low-loss PTFE-based laminates (e.g., Rogers RO3003 at εr = 3.00, Df = 0.0010). The thin substrate keeps surface waves in check at a wavelength of only 3.9 mm, while the low Dk widens bandwidth enough for FMCW chirps spanning 76–81 GHz.
Ready for a quotation?
Send your requirements for a material, stackup and manufacturing review. Price and schedule are confirmed for your project.
