Sub6 vs mmWave: 5G PCB Design Differences

Sub6 vs mmWave: What Changes on the PCB?

Sub-6 GHz 5G (typically 0.6–6 GHz, covering bands n77, n78, n79) and mmWave 5G (24–47 GHz, bands n257–n261) impose fundamentally different demands on PCB materials, geometry and fabrication tolerances. Understanding sub6 vs mmWave trade-offs is essential before committing to a stackup. Sub-6 GHz boards can often use modified FR-4 or mid-tier laminates with standard etching; mmWave boards require low-loss substrates, tighter trace control and surface finishes that stay flat through tens of gigahertz. The gap is not just frequency—it cascades into every layer of the stackup, the bill of materials and the factory capability you need to specify.

This article walks through each design dimension, gives you the numbers that matter, and includes tools to check whether your fabricator can actually build what you draw. For a broader overview of what 5G demands from the bare board, see the full 5G PCB requirements guide.

Frequency Bands and Why the Sub6 vs mmWave Gap Matters

The 3GPP NR spectrum splits into two frequency ranges:

Because wavelength shrinks roughly 8× going from 3.5 GHz to 28 GHz, every physical dimension on the board—trace width, via pad, ground clearance, antenna element pitch—becomes proportionally more critical. A 1 mil (0.0254 mm) etch variation that barely registers at 3.5 GHz can shift a mmWave antenna’s center frequency by hundreds of megahertz.

Material Selection: Where the Sub6 vs mmWave Cost Split Begins

Sub-6 GHz Material Options

At frequencies up to about 6 GHz, designers have a wide material palette. Standard high-Tg FR-4 (Dk ≈ 4.2–4.5, Df ≈ 0.020 at 1 GHz) works for digital and power sections. For the RF signal path, a mid-range laminate such as Rogers RO4350B (Dk 3.48, Df 0.0037 at 10 GHz) or Isola Astra MT77 delivers low loss without exotic processing. Hybrid stackups—one or two RF-grade cores bonded to FR-4 cores for the digital layers—are routine and keep cost in check.

mmWave Material Requirements

Above 24 GHz, dielectric loss rises sharply with frequency, so Df must be well below 0.004 at the operating band. PTFE-based laminates such as Rogers RO3003 (Dk 3.00, Df 0.0013 at 10 GHz) or liquid-crystal polymer (LCP) films become the baseline for antenna and feed layers. Dk tolerance tightens to ±0.02 or better, because a Dk shift of 0.05 at 28 GHz moves a patch antenna’s resonance by roughly 1–2 % of center frequency—enough to miss the band edge.

For a side-by-side look at high-frequency laminates across vendors, the RF laminate comparison tool lets you filter by Dk, Df and frequency range.

Copper roughness deserves special attention. At mmWave frequencies the skin depth in copper falls below 0.5 µm, so surface roughness on the order of standard electrodeposited copper (Rz ≈ 5–8 µm) adds measurable conductor loss. Specifying very-low-profile (VLP) or hyper-VLP foil (Rz < 1.5 µm) is standard practice above 24 GHz. The Rogers RO3003 datasheet publishes insertion-loss data with different foil types to quantify this effect.

Stackup Architecture

Sub-6 GHz Stackups

A typical sub-6 GHz 5G radio board uses 6–10 layers. The RF front-end occupies one or two microstrip or stripline layers on a low-loss core, with the remaining layers in FR-4 for power distribution and digital I/O. Hybrid bonding (low-loss prepreg between the RF core and FR-4 cores) is well-understood, and most mid-volume fabricators handle it without special qualification.

mmWave Stackups

mmWave boards are thinner (often 0.5–1.0 mm total) and may use only 4–6 layers, because every via transition is a potential source of parasitic inductance and radiation. Antenna layers sit on the outermost surface to avoid via-feed loss; stripline feeds are reserved for distribution networks where shielding matters. The dielectric thickness under a 50 Ω microstrip at 28 GHz on RO3003 (Dk 3.00) is roughly 5 mil (0.127 mm), demanding ±0.5 mil lamination tolerance. For detailed rules on laying out boards in this regime, the mmWave PCB design guide covers trace geometry, via fencing and transition structures.

Trace Geometry and Tolerances

Trace geometry is one of the most tangible sub6 vs mmWave differences an engineer encounters during layout. At 3.5 GHz on RO4350B (Dk 3.48) with a 10 mil dielectric, a 50 Ω microstrip is approximately 22 mil (0.56 mm) wide—comfortable for standard etching with ±1 mil tolerance. At 28 GHz on RO3003 (Dk 3.00) with a 5 mil dielectric, the same 50 Ω line narrows to roughly 12 mil (0.30 mm), and the tolerance budget shrinks to ±0.5 mil to keep impedance within ±5 %.

Grounded coplanar waveguide (GCPW) is the dominant transmission-line style at mmWave because the coplanar ground strips suppress higher-order modes and provide a well-defined return path. The gap between signal and coplanar ground is typically 4–6 mil, requiring laser-direct-imaging (LDI) lithography rather than conventional contact exposure.

Via Transitions

Signal vias at mmWave need back-drilling or controlled-depth drilling to remove the stub below the target layer. A 10 mil via stub that is invisible at 3.5 GHz creates a quarter-wave resonance near 28 GHz, producing a deep notch in S21. Via anti-pads must be tightly controlled (typically 20–24 mil diameter for a 10 mil drill) and surrounded by a ring of ground vias at λ/10 spacing to suppress parallel-plate modes.

Surface Finish: Sub6 vs mmWave Considerations

Surface finish affects both solderability and RF performance. At sub-6 GHz, HASL, OSP and ENIG all work, and the choice is driven by assembly needs. At mmWave, the finish on antenna pads and transmission lines directly influences conductor loss and impedance consistency.

For mmWave antenna arrays, immersion silver or electrolytic gold are preferred because they avoid the lossy nickel barrier present in ENIG. The nickel layer (typically 3–5 µm) has a resistivity roughly 4× that of copper, and at 28 GHz the current crowds into the top 0.5 µm of metal—right where the nickel sits. Use the tool below to compare finish options against your frequency and assembly requirements:

If the tool flags ENIG as marginal for your frequency, consider specifying immersion silver for the RF layers while keeping ENIG on the BGA/connector side through selective finishing—a common split-finish approach on 5G boards.

Antenna Integration: On-Board vs On-Package

At sub-6 GHz, discrete antenna elements or external antenna modules connect to the board through coaxial connectors or board-edge launches. The antenna itself may be a separate stamped or PCB structure. At mmWave, the short wavelength makes it practical—and often necessary—to integrate the antenna array directly onto the PCB or into the IC package (antenna-in-package, AiP). The trade-offs between these two architectures are covered in the antenna on package vs on board comparison.

For a 39 GHz n260 board, a 4×4 patch array occupies roughly 16 × 16 mm—small enough to fit on a module, but the PCB under it must maintain Dk uniformity across that aperture to within ±1 %. Any pocket milling, cavity routing or embedded-component cutout in that zone risks delamination or Dk variation.

Manufacturing Capability: The Sub6 vs mmWave Fab Gap

The gap between sub-6 GHz and mmWave is ultimately a manufacturing capability gap. The table below summarizes the fabrication parameters that tighten most sharply:

Before committing to a design, verify that your fabricator’s process limits cover every row above. Use the capability matrix below to cross-check minimum feature sizes, layer counts and material compatibility against your design’s requirements:

If the matrix shows your mmWave stackup exceeds the fab’s standard capability—for example, requiring 3/3 mil trace/space on PTFE—you will need to confirm that the factory has qualified that combination on the specific laminate you chose. PTFE materials are softer than FR-4 and prone to smearing during drilling and dimensional instability during lamination, so process qualification is not transferable from one substrate to another.

Inspection and Test

Sub-6 GHz boards are typically verified with time-domain reflectometry (TDR) for impedance and standard automated optical inspection (AOI) for trace defects. mmWave boards add several requirements: coupon-based insertion-loss testing per IPC-TM-650, cross-section analysis of via stubs after back-drilling, and in some cases X-ray inspection of embedded or blind vias to confirm barrel fill and registration.

Antenna-on-board designs above 24 GHz may also require over-the-air (OTA) pattern measurements in an anechoic chamber, which is a system-level test rather than a bare-board test. Budget for this early—chamber time and fixturing are non-trivial cost items.

Design Tool Considerations

Full-wave 3D electromagnetic simulation (HFSS, CST, openEMS) is essential for mmWave transitions, antenna elements and via arrays. At sub-6 GHz, 2.5-D tools and rule-based impedance calculators cover most routing tasks. The choice of PCB EDA tool also matters: both Altium and KiCad support impedance-controlled routing, but their 3D EM integration depth differs—see the Altium vs KiCad RF comparison for specifics. Whichever tool you use, export Gerbers with enough decimal precision (at least 2:5 format, i.e., 5 decimal places in inches) to preserve mmWave feature resolution.

Quick Decision Guide

Choose a sub-6 GHz–optimized board approach when:

• Your NR band is below 7.125 GHz (FR1).
• Budget favors hybrid FR-4 / mid-tier laminate stackups.
• Standard etch tolerances (±1 mil) and ENIG finish are acceptable.
• The antenna is a discrete component or a separate PCB.

Choose a mmWave-optimized board approach when:

• Your NR band is in FR2 (24.25 GHz and above).
• Antenna integration on the PCB or in-package is required.
• You can specify PTFE or LCP laminates and VLP copper.
• Your fabricator is qualified for ≤ 3 mil features on low-loss substrates with ±5 % impedance control.
• You have access to full-wave EM simulation and OTA test facilities.

The sub6 vs mmWave decision is not binary on many 5G platforms—a single product may carry both an FR1 radio and an FR2 module, each with its own stackup and material strategy. Treat them as separate PCB design problems that happen to share a system architecture.

FAQ

Can I use FR-4 for mmWave 5G?

Standard FR-4 has a Df of roughly 0.020 at 10 GHz, which translates to prohibitive insertion loss at 28 GHz (often > 1 dB/cm). It is not suitable for mmWave signal paths. It can still serve as the digital/power core in a hybrid stackup, but the RF layers must use a low-loss laminate.

Is ENIG acceptable at 28 GHz?

ENIG introduces a 3–5 µm nickel barrier whose resistivity degrades conductor loss at mmWave frequencies. Immersion silver or electrolytic gold are preferred for antenna and feed-line pads. ENIG remains acceptable for BGA lands and low-frequency I/O on the same board.

How much more does a mmWave board cost than a sub-6 GHz board?

Material alone can be 5–15× the cost of FR-4, and tighter process tolerances reduce panel yield. Total bare-board cost for a mmWave design is typically 3–8× that of an equivalent-size sub-6 GHz board, depending on layer count, material choice and volume. Exact pricing depends on the specific design; request a quote with your actual stackup to get a meaningful number.

What IPC class should I specify for mmWave 5G?

IPC Class 3 per IPC-6012E is recommended for mmWave boards because it tightens allowable defects in plating, registration and dielectric thickness—all parameters that directly affect mmWave performance. Class 2 is often sufficient for sub-6 GHz consumer-grade radios.

Do I need back-drilling for sub-6 GHz boards?

Rarely. A 10 mil via stub resonates near 28 GHz, well above FR1 frequencies. At sub-6 GHz, stub effects are usually negligible unless the via is exceptionally long (> 60 mil) or the signal bandwidth extends above 10 GHz.